{"count": 501, "next": null, "previous": null, "results": [{"id": 13330, "uid": "0ed9422357395a0d4879191c66f4faa2", "name": "Near-optimal Rank Adaptive Inference of High Dimensional Matrices", "authors": [{"id": 19823, "fullname": "Fr\u00e9d\u00e9ric Zheng", "url": "http://virtual.aistats.org/api/miniconf/users/19823?format=json", "institution": "KTH University"}, {"id": 21879, "fullname": "Yassir Jedra", "url": "http://virtual.aistats.org/api/miniconf/users/21879?format=json", "institution": "Imperial College London"}, {"id": 4579, "fullname": "Alexandre Proutiere", "url": "http://virtual.aistats.org/api/miniconf/users/4579?format=json", "institution": "KTH Royal Institute of Technology"}], "abstract": "We address the problem of estimating a high-dimensional matrix from linear measurements, with a focus on designing optimal rank-adaptive algorithms. These algorithms infer the matrix by estimating its singular values and the corresponding singular vectors up to an effective rank, adaptively determined based on the data. We establish, for the first time, instance-specific lower bounds for the sample complexity of such algorithms. We uncover fundamental trade-offs in selecting the effective rank: balancing the precision of estimating a subset of singular values against the approximation cost incurred for the remaining ones. Our analysis identifies how the optimal effective rank depends on the matrix being estimated, the sample size, and the noise level. We propose an algorithm that combines a Least-Squares estimator with a universal singular value thresholding procedure. We provide finite-sample error bounds for this algorithm, that are tighter than those of existing rank-adaptive algorithms. Furthermore, our bounds nearly match the derived fundamental limits.  Finally, we confirm experimentally that our algorithm outperforms existing rank-adaptive algorithms.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13330", "url": null, "sourceid": 1422, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=zKuA2nxqQF", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11023, "modified": "2026-03-29T20:42:54.391596-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=zKuA2nxqQF", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "110", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13332, "uid": "25e2a30f44898b9f3e978b1786dcd85c", "name": "Robust Learning of A Group DRO Neuron", "authors": [{"id": 21880, "fullname": "Guyang Cao", "url": "http://virtual.aistats.org/api/miniconf/users/21880?format=json", "institution": "Department of Computer Science, University of Wisconsin - Madison"}, {"id": 21881, "fullname": "Shuyao Li", "url": "http://virtual.aistats.org/api/miniconf/users/21881?format=json", "institution": "Meta"}, {"id": 21882, "fullname": "Sushrut Karmalkar", "url": "http://virtual.aistats.org/api/miniconf/users/21882?format=json", "institution": "Research, Microsoft"}, {"id": 1157, "fullname": "Jelena Diakonikolas", "url": "http://virtual.aistats.org/api/miniconf/users/1157?format=json", "institution": "University of Wisconsin-Madison"}], "abstract": "We study the problem of learning a single neuron under standard squared loss in the presence of arbitrary label noise and group-level distributional shifts, for a broad family of covariate distributions. Our goal is to identify a \"best-fit\" neuron parameterized by ${\\boldsymbol w}^{\\star}$ that performs well under the most challenging reweighting of the groups. Specifically, we address a Group Distributionally Robust Optimization problem: given sample access to $K$ distinct distributions ${\\mathcal p_{[1]}},\\dots, {\\mathcal p_{[K]}}$, we seek to approximate ${\\boldsymbol w}^*$ that minimizes the worst-case objective over convex combinations of group distributions ${\\boldsymbol \\lambda} \\in \\Delta_K$, where the objective is $\\sum_{i \\in [K]}\\lambda_{[i]},\\mathbb E_{(\\mathbf x,y)\\sim{\\mathcal p_{[i]}}}(\\sigma(\\boldsymbol w\\cdot\\boldsymbol x)-y)^2 - \\nu d_f(\\boldsymbol\\lambda,\\tfrac1K\\boldsymbol1)$ and $d_f$ is an $f$-divergence that imposes (optional) penalty on deviations from uniform group weights, scaled by a parameter $\\nu \\geq 0$.   We develop a computationally efficient primal-dual algorithm that outputs a vector $\\widehat{\\boldsymbol w}$ that is constant-factor competitive with ${\\boldsymbol w}^{*}$ under the worst-case group weighting.  Our analytical framework directly confronts the inherent nonconvexity of the loss function, providing robust learning guarantees in the face of arbitrary label corruptions and group-specific distributional shifts. The implementation of the dual extrapolation update motivated by our algorithmic framework shows promise on LLM pre-training benchmarks.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13332", "url": null, "sourceid": 1729, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=zAU4aKo7Ry", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11025, "modified": "2026-03-29T20:42:54.504324-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=zAU4aKo7Ry", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "159", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13333, "uid": "3cec07e9ba5f5bb252d13f5f431e4bbb", "name": "Neural Variance-aware Dueling Bandits with Deep Representation and Shallow Exploration", "authors": [{"id": 19292, "fullname": "Youngmin Oh", "url": "http://virtual.aistats.org/api/miniconf/users/19292?format=json", "institution": "InfiniTree"}, {"id": 21883, "fullname": "Jinje Park", "url": "http://virtual.aistats.org/api/miniconf/users/21883?format=json", "institution": "Samsung Eletronics"}, {"id": 21884, "fullname": "Taejin Paik", "url": "http://virtual.aistats.org/api/miniconf/users/21884?format=json", "institution": "Samsung Advanced Institute of Technology"}], "abstract": "We introduce the first variance-aware algorithms for contextual dueling bandits that leverage shallow exploration strategies with neural networks for nonlinear utility approximation. A key theoretical challenge is the absence of a closed-form estimator, which led prior work to require an extremely large network width $m$ (i.e., $m = \\widetilde{\\Omega}(T^{14})$). We address this constraint with a novel analytical approach that combines iterative self-improvement with spectral analysis. Our analysis significantly reduces the network width requirement to $m = \\widetilde{\\Omega}(T^{6})$, and shows that our algorithms achieve a sublinear regret of  $ \\widetilde{\\mathcal{O}}\\left(d\\sqrt{\\sum_{t=1}^{T} \\sigma_t^2} + \\sqrt{dT}\\right) $ under both UCB and TS frameworks. Empirical results show that the proposed algorithms are not only computationally efficient and exhibit sublinear regret in practical settings, but also achieve state-of-the-art performance on both synthetic and real-world tasks.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13333", "url": null, "sourceid": 247, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=yzqazIZbIQ", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11026, "modified": "2026-03-29T20:42:54.548531-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=yzqazIZbIQ", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "115", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13504, "uid": "26588e932c7ccfa1df309280702fe1b5", "name": "Gradient Regularized Natural Gradients", "authors": [{"id": 22256, "fullname": "Satya Dash", "url": "http://virtual.aistats.org/api/miniconf/users/22256?format=json", "institution": "University of Manchester, University of Manchester"}, {"id": 22257, "fullname": "Hossein Abdi", "url": "http://virtual.aistats.org/api/miniconf/users/22257?format=json", "institution": "University of Manchester"}, {"id": 23262, "fullname": "Wei Pan", "url": "http://virtual.aistats.org/api/miniconf/users/23262?format=json", "institution": "Newcastle University, UK"}, {"id": 3700, "fullname": "Samuel Kaski", "url": "http://virtual.aistats.org/api/miniconf/users/3700?format=json", "institution": "Aalto University and University of Manchester"}, {"id": 22259, "fullname": "Mingfei Sun", "url": "http://virtual.aistats.org/api/miniconf/users/22259?format=json", "institution": "University of Manchester"}], "abstract": "Gradient regularization (GR) has been shown to improve the generalizability of trained models. While Natural Gradient Descent has been shown to accelerate optimization in the initial phase of training, little attention has been paid to how the training dynamics of second-order optimizers can benefit from GR. In this work, we propose Gradient-Regularized Natural Gradients (GRNG), a family of scalable second-order optimizers that integrate explicit gradient regularization with natural gradient updates. Our framework introduces two frequentist algorithms: Regularized Explicit Natural Gradient (RENG), which utilizes double backpropagation to explicitly minimize the gradient norm, and Regularized Implicit Natural Gradient (RING), which incorporates regularization implicitly into the update direction. We also propose a Bayesian variant based on a Regularized-Kalman formulation that eliminates the need for FIM inversion entirely. We establish convergence guarantees for GRNG, showing that gradient regularization improves stability and enables convergence to global minima. Empirically, we demonstrate that GRNG consistently enhances both optimization speed and generalization compared to first-order methods (SGD, AdamW) and second-order baselines (K-FAC, Sophia), with strong results on vision and language benchmarks.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13504", "url": null, "sourceid": 1258, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=fffAxw0cPm", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11197, "modified": "2026-03-29T20:43:01.454231-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=fffAxw0cPm", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "69", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13334, "uid": "9188905e74c28e489b44e954ec0b9bca", "name": "Meta Sparse Principal Component Analysis", "authors": [{"id": 21885, "fullname": "Imon Banerjee", "url": "http://virtual.aistats.org/api/miniconf/users/21885?format=json", "institution": "Northwestern University"}, {"id": 827, "fullname": "Jean Honorio", "url": "http://virtual.aistats.org/api/miniconf/users/827?format=json", "institution": "University of Melbourne"}], "abstract": "We study the meta-learning for support recovery (i.e., non-zero coordinates of the eigenvectors) in high-dimensional Principal Component  Analysis. We reduce the sufficient sample complexity in a novel task, with the information that is learned from auxiliary tasks, where a task is defined as a random Principal Component (PC) matrix with its own support. We pool data from all the tasks to execute an improper estimation of a single PC matrix, by maximising the $\\ell_1$-regularised predictive covariance. With $m$ tasks for $p$-variate sub-Gaussian random vectors, we establish the sufficient sample complexity for each task to be of the order $O(\\sqrt{m^{-1}\\log p})$, with high probability. This is very relevant for meta-learning where there are many tasks $m = O(\\log p)$, each with very few samples, i.e., $n = O(1)$, in an scenario where multi-task learning fails. For a novel task,  we prove that the sufficient sample complexity of successful support recovery can be reduced to $O(\\log |J|)$, under an additional constraint that the support of the novel task is a subset of the estimated support union ($J$) from the auxiliary tasks. This reduces the original sample complexity of $O(\\log p)$ for learning a single task. Theoretical claims are validated with numerical simulations and the problem of true covariance estimation in brain-imaging and cancer genetics data sets are considered to validate the proposed methodology.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13334", "url": null, "sourceid": 244, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=ywzOo5YuF2", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11027, "modified": "2026-03-29T20:42:54.592050-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=ywzOo5YuF2", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "107", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13870, "uid": "550a141f12de6341fba65b0ad0433500", "name": "RoseCDL: Robust and Scalable Convolutional Dictionary Learning for rare-event and anomaly detection", "authors": [{"id": 23015, "fullname": "Jad Yehya", "url": "http://virtual.aistats.org/api/miniconf/users/23015?format=json", "institution": "INRIA"}, {"id": 23016, "fullname": "Mansour Benbakoura", "url": "http://virtual.aistats.org/api/miniconf/users/23016?format=json", "institution": "INRIA"}, {"id": 23017, "fullname": "C\u00e9dric Allain", "url": "http://virtual.aistats.org/api/miniconf/users/23017?format=json", "institution": "INRIA"}, {"id": 23018, "fullname": "Beno\u00eet Mal\u00e9zieux", "url": "http://virtual.aistats.org/api/miniconf/users/23018?format=json", "institution": "CEA"}, {"id": 23019, "fullname": "Matthieu Kowalski", "url": "http://virtual.aistats.org/api/miniconf/users/23019?format=json", "institution": "Universit\u00e9 Paris-Saclay"}, {"id": 12452, "fullname": "Thomas Moreau", "url": "http://virtual.aistats.org/api/miniconf/users/12452?format=json", "institution": "Inria"}], "abstract": "Detecting rare events and anomalies in large-scale signals is essential in fields such as astronomy, physical simulations, and biomedical science. In many cases, this problem naturally decomposes into identifying common local patterns and detecting deviations that correspond to anomalies. Convolutional Dictionary Learning (CDL) is a powerful tool for modeling local structures, but its adoption for this task has been limited by computational demands and sensitivity to outliers. We introduce RoseCDL, a novel CDL algorithm designed for robust and scalable modeling  of signal pattern distribution. RoseCDL leverages stochastic windowing for efficient training and incorporates inline outlier detection to enhance robustness. This enables unsupervised identification of anomalous and rare patterns in long signals based on the local reconstruction loss. Experiments on real-world datasets show that RoseCDL delivers improved detection accuracy and computational efficiency, making CDL practical for challenging detection tasks in large-scale signal analysis.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13870", "url": null, "sourceid": 444, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=4XMkOFxxfb", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11563, "modified": "2026-03-29T20:43:16.672832-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=4XMkOFxxfb", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "162", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13335, "uid": "25b2822c2f5a3230abfadd476e8b04c9", "name": "Spectral Clustering for Directed Graphs via Likelihood Estimation on Stochastic Block Models", "authors": [{"id": 21886, "fullname": "Ning Zhang", "url": "http://virtual.aistats.org/api/miniconf/users/21886?format=json", "institution": "Oxofrd, University of Oxford"}, {"id": 4024, "fullname": "Xiaowen Dong", "url": "http://virtual.aistats.org/api/miniconf/users/4024?format=json", "institution": "University of Oxford"}, {"id": 21887, "fullname": "Mihai Cucuringu", "url": "http://virtual.aistats.org/api/miniconf/users/21887?format=json", "institution": "University of California, Los Angeles"}], "abstract": "Graph clustering is a fundamental task in unsupervised learning with broad real-world applications. While spectral clustering methods for undirected graphs are well-established and guided by a minimum cut optimization consensus, their extension to directed graphs remains relatively underexplored due to the additional complexity introduced by edge directions. In this paper, we leverage statistical inference on stochastic block models to guide the development of a spectral clustering algorithm for directed graphs. Specifically, we study the maximum likelihood estimation under a widely used directed stochastic block model, and derive a global objective function that aligns with the underlying community structure. Based on the spectral relaxation, we introduce a novel, self-adaptive spectral clustering method for directed graphs, and provide theoretical guarantees on the miscluster error. Extensive experiments on synthetic and real-world datasets demonstrate significant performance gains over existing baselines.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13335", "url": null, "sourceid": 425, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=ysFHjFNWyB", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11028, "modified": "2026-03-29T20:42:54.644025-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=ysFHjFNWyB", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "173", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13872, "uid": "36a16a2505369e0c922b6ea7a23a56d2", "name": "Demystifying Transition Matching: When and Why It Can Beat Flow Matching", "authors": [{"id": 19875, "fullname": "Jaihoon Kim", "url": "http://virtual.aistats.org/api/miniconf/users/19875?format=json", "institution": "KAIST"}, {"id": 23022, "fullname": "Rajarshi Saha", "url": "http://virtual.aistats.org/api/miniconf/users/23022?format=json", "institution": "Amazon"}, {"id": 18112, "fullname": "Youngsuk Park", "url": "http://virtual.aistats.org/api/miniconf/users/18112?format=json", "institution": "Amazon, AWS AI Labs"}, {"id": 23023, "fullname": "Minhyuk Sung", "url": "http://virtual.aistats.org/api/miniconf/users/23023?format=json", "institution": "Korea Advanced Institute of Science &amp; Technology"}], "abstract": "Flow Matching (FM) underpins many state-of-the-art generative models, yet recent results indicate that Transition Matching (TM) can achieve higher quality with fewer sampling steps. This work answers the question of when and why TM outperforms FM. First, when the target is a unimodal Gaussian distribution, we prove that TM attains strictly lower KL divergence than FM for finite number of steps. The improvement arises from stochastic difference latent updates in TM, which preserve target covariance that deterministic FM underestimates. We then characterize convergence rates, showing that TM achieves faster convergence than FM under a fixed compute budget. Second, we extend the analysis to Gaussian mixtures and identify local\u2013unimodality regimes in which the sampling dynamics approximate the unimodal case, where TM can outperform FM. The approximation error decreases as the minimal distance between component means increases, highlighting that TM is favored when the modes are well separated. However, when the target variance approaches zero, each TM update converges to the FM update, and the performance advantage of TM diminishes. In summary, we show that TM outperforms FM when the target distribution has well-separated modes and non-negligible variances. We validate our theoretical results with controlled experiments on Gaussian distributions, and extend the comparison to real-world applications in image and video generation.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13872", "url": null, "sourceid": 1081, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=4NHbd0mQga", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11565, "modified": "2026-03-29T20:43:16.778298-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=4NHbd0mQga", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "38", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13874, "uid": "29530de21430b7540ec3f65135f7323c", "name": "Open Multi-agent Multi-armed Bandit with Applications in Permissionless Blockchain", "authors": [{"id": 18368, "fullname": "Mengfan Xu", "url": "http://virtual.aistats.org/api/miniconf/users/18368?format=json", "institution": "University of Massachusetts at Amherst"}, {"id": 18356, "fullname": "Diego Klabjan", "url": "http://virtual.aistats.org/api/miniconf/users/18356?format=json", "institution": "Northwestern University"}], "abstract": "We study a multi-agent multi-armed bandit problem (MA-MAB) in open systems, where multiple agents can enter and leave at any time and face multiple bandit problems to minimize the group-wise cumulative regret. To our knowledge, this is the first work to consider a dynamic set of agents that arrive and depart according to stochastic processes, systematically evolving over time. We also extend to a permissionless blockchain-based MA-MAB (PB-MA-MAB) problem, where agents may behave either honestly or maliciously depending on compliance with the mechanism, and malicious agents may disrupt honest ones. These formulations pose new challenges, as regret grows with the increasing number of agents.  To this end, we design new UCB-based methodologies for both MA-MAB and PB-MA-MAB, introducing information-integration rules for existing agents and information-access mechanisms for new agents to fully leverage available information. We derive regret bounds for our algorithms and characterize the complexity of the formulation via regret lower bounds in both settings. We establish regret upper bounds of order $\\max\\{O(M_0), O(\\log T), O(\\tfrac{\\log^2 T}{(M_0)})1_{\\{\\lambda > 0\\}}\\}$ (a significant improvement over the na\u00efve bound $(M_0 + T)\\log T$), where $M_0$ is the initial number of agents and $C$ reflects the arrival/departure rate. We also prove lower bounds of $O(\\log T)$ and $O(M_0)$ for all consistent algorithms, and tighter bounds of $O(\\log T + M_0)$ or $O(\\log^2 T)$ for a subset including ours. These imply that our algorithm is nearly optimal in general and optimal in certain cases.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13874", "url": null, "sourceid": 1930, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=43AwLfHfXp", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11567, "modified": "2026-03-29T20:43:16.851417-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=43AwLfHfXp", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "127", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13348, "uid": "d88518acbcc3d08d1f18da62f9bb26ec", "name": "Adaptive Candidate Point Thompson Sampling for High-Dimensional Bayesian Optimization", "authors": [{"id": 21912, "fullname": "Donney Fan", "url": "http://virtual.aistats.org/api/miniconf/users/21912?format=json", "institution": "University of British Columbia"}, {"id": 13157, "fullname": "Geoff Pleiss", "url": "http://virtual.aistats.org/api/miniconf/users/13157?format=json", "institution": "University of British Columbia, Vector Institute"}], "abstract": "In Bayesian optimization, Thompson sampling selects the evaluation point by sampling from the posterior distribution over the objective function maximizer. Because this sampling problem is intractable for Gaussian process (GP) surrogates, the posterior distribution is typically restricted to \ufb01xed discretizations (i.e., candidate points) that become exponentially sparse as dimensionality increases. While previous works aim to increase candidate point density through scalable GP approximations, our orthogonal approach increases density by adaptively reducing the search space during sampling. Specifically, we introduce Adaptive Candidate Thompson Sampling (ACTS), which generates candidate points in subspaces guided by the gradient of a surrogate model sample. ACTS is a simple drop-in replacement for existing TS methods\u2014including those that use trust regions or other local approximations\u2014producing better samples of maxima and improved optimization across synthetic and real-world benchmarks.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13348", "url": null, "sourceid": 1595, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=xZ2R2lajvG", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11041, "modified": "2026-03-29T20:42:55.268849-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=xZ2R2lajvG", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "14", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13876, "uid": "160c88652d47d0be60bfbfed25111412", "name": "Lag Operator SSMs: A Geometric Framework for Structured State Space Modeling", "authors": [{"id": 23031, "fullname": "Sutashu Tomonaga", "url": "http://virtual.aistats.org/api/miniconf/users/23031?format=json", "institution": "Okinawa Institute of Science and Technology (OIST)"}, {"id": 23032, "fullname": "Kenji Doya", "url": "http://virtual.aistats.org/api/miniconf/users/23032?format=json", "institution": "Okinawa Institute of Science and Technology Graduate University"}, {"id": 23033, "fullname": "Noboru Murata", "url": "http://virtual.aistats.org/api/miniconf/users/23033?format=json", "institution": "Waseda University"}], "abstract": "Structured State Space Models (SSMs), which are at the heart of the recently popular Mamba architecture, are powerful tools for sequence modeling. However, their theoretical foundation relies on a complex, multi-stage process of continuous-time modeling and subsequent discretization, which can obscure intuition. We introduce a direct, first-principles framework for constructing discrete-time SSMs that is both flexible and modular. Our approach is based on a novel lag operator, which geometrically derives the discrete-time recurrence by measuring how the system's basis functions \"slide\" and change from one timestep to the next. The resulting state matrices are computed via a single inner product involving this operator, offering a modular design space for creating novel SSMs by flexibly combining different basis functions and time-warping schemes. To validate our approach, we demonstrate that a specific instance exactly recovers the recurrence of the influential HiPPO model. Numerical simulations confirm our derivation, providing new theoretical tools for designing flexible and robust sequence models.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13876", "url": null, "sourceid": 1190, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=3woUBLisIm", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11569, "modified": "2026-03-29T20:43:16.919532-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=3woUBLisIm", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "86", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13499, "uid": "faafda66202d234463057972460c04f5", "name": "CAWI: Copula-Aligned Weight Initialization for Randomized Neural Networks", "authors": [{"id": 19251, "fullname": "Mushir Akhtar", "url": "http://virtual.aistats.org/api/miniconf/users/19251?format=json", "institution": "Indian Institute of Technology Indore, India"}, {"id": 22245, "fullname": "M. Tanveer", "url": "http://virtual.aistats.org/api/miniconf/users/22245?format=json", "institution": "IIT Indore"}, {"id": 22246, "fullname": "Mohd. Arshad", "url": "http://virtual.aistats.org/api/miniconf/users/22246?format=json", "institution": "Indian Institute of Technology, Indore"}], "abstract": "Randomized neural networks (RdNNs) enable efficient, backpropagation-free training by freezing randomly initialized input-to-hidden weights, which permits a closed-form solution for the output layer. However, conventional random initialization is blind to inter-feature dependence\u2014ignoring correlations, asymmetries, and tail dependence in the data\u2014which degrades conditioning and predictive performance. To the best of our knowledge, this limitation remains unaddressed in the RdNN literature. To close this gap, we propose CAWI (Copula-Aligned Weight Initialization), a framework that draws input-to-hidden weights from a data-fitted copula that matches empirical dependence, ensuring the frozen projections respect inter-feature dependence without sacrificing closed-form solution. CAWI (i) maps each feature to the unit interval using empirical CDFs, (ii) fits a multivariate copula that captures rank-based dependence among features, and (iii) samples each weight column $w_j$ from the fitted copula and applies a fixed inverse marginal transform to set scale. The objective, solver, and ``freeze-once'' paradigm remain unchanged; only the sampling law for $W$ becomes dependence-aware. For dependence modeling, we consider two copula families: elliptical (Gaussian, t) and Archimedean (Clayton, Frank, Gumbel). This enables CAWI to handle diverse dependence, including tail dependence. We evaluate CAWI across 83 diverse classification benchmarks (binary and multiclass) and two biomedical datasets, BreaKHis and the Schizophrenia dataset, using standard shallow and deep RdNN architectures. CAWI consistently delivers significant improvements in predictive performance over conventional random initialization. Codes are provided at \\url{https://github.com/mtanveer1/CAWI}.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13499", "url": null, "sourceid": 1608, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=fwqydIjLn6", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11192, "modified": "2026-03-29T20:43:01.267120-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=fwqydIjLn6", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "39", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13362, "uid": "2cd4e8a2ce081c3d7c32c3cde4312ef7", "name": "Learning Equivariant Functions via Quadratic Forms", "authors": [{"id": 5614, "fullname": "Pavan Karjol", "url": "http://virtual.aistats.org/api/miniconf/users/5614?format=json", "institution": "Indian Institute of Science, Bengaluru, Karnataka"}, {"id": 21939, "fullname": "Vivek Kashyap", "url": "http://virtual.aistats.org/api/miniconf/users/21939?format=json", "institution": "Indian Institute of Science, Indian institute of science, Bangalore"}, {"id": 21940, "fullname": "Rohan Kashyap", "url": "http://virtual.aistats.org/api/miniconf/users/21940?format=json", "institution": "Carnegie Mellon University"}, {"id": 21941, "fullname": "Prathosh AP", "url": "http://virtual.aistats.org/api/miniconf/users/21941?format=json", "institution": "Indian Institute of Science, Indian institute of science, Bangalore"}], "abstract": "In this study, we introduce a method for learning group equivariant functions by learning the associated quadratic form $x^TAx$  corresponding to the group from the data. Certain groups, known as generalised orthogonal groups, preserve a specific quadratic form, and we leverage this property to uncover the underlying symmetry group under the assumption that it is generalised orthogonal group. By utilising the corresponding unique symmetric matrix, we incorporate suitable inductive biases into the neural network architecture, leading to models that are both simplified and efficient. Our approach results in an invariant model that preserves norms, while the equivariant model is represented as a product of a norm-invariant model and a scale-invariant model, where the \u201cproduct\u201d refers to the group action. Moreover, we extend our framework to a more general setting where the function acts on tuples of input vectors via a diagonal group action. In this extension, the equivariant function is decomposed into an angular component extracted solely from the normalised first vector and a scale-invariant component that depends on the full Gram matrix of the tuple. This decomposition captures the inter-dependencies between multiple inputs while preserving the underlying group symmetry. We assess the effectiveness of our framework across multiple tasks, including polynomial regression, top quark tagging, and moment of inertia matrix prediction. Comparative analysis with baseline methods demonstrates that our model consistently excels in both discovering the underlying symmetry and efficiently learning the corresponding equivariant function.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13362", "url": null, "sourceid": 2095, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=wLZ2z50W6T", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11055, "modified": "2026-03-29T20:42:55.826551-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=wLZ2z50W6T", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "90", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13367, "uid": "84d9ee44e457ddef7f2c4f25dc8fa865", "name": "RL-finetuning LLMs from on- and off-policy data with a single algorithm", "authors": [{"id": 148, "fullname": "Yunhao Tang", "url": "http://virtual.aistats.org/api/miniconf/users/148?format=json", "institution": "DeepMind"}, {"id": 21951, "fullname": "Taco Cohen", "url": "http://virtual.aistats.org/api/miniconf/users/21951?format=json", "institution": "Meta"}, {"id": 21952, "fullname": "David Zhang", "url": "http://virtual.aistats.org/api/miniconf/users/21952?format=json", "institution": "Facebook"}, {"id": 23259, "fullname": "Gabriel Synnaeve", "url": "http://virtual.aistats.org/api/miniconf/users/23259?format=json", "institution": "Facebook"}, {"id": 21878, "fullname": "R\u00e9mi Munos", "url": "http://virtual.aistats.org/api/miniconf/users/21878?format=json", "institution": "Meta"}], "abstract": "We introduce a novel reinforcement learning algorithm (AGRO, for Any-Generation Reward Optimization) for finetuning Large Language Models. AGRO leverages the concept of response consistency, which states that the optimal policy satisfies a notion of consistency across any possible generation of the model. We derive algorithms that find optimal solutions via sample-based policy gradient and provide theoretical guarantees on their convergence. Our experiments demonstrate the effectiveness of AGRO in both on-policy and off-policy settings, showing improved performance on the MATH dataset over baseline methods.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13367", "url": null, "sourceid": 199, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=v8n0UHCu8R", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11060, "modified": "2026-03-29T20:42:55.991882-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=v8n0UHCu8R", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "158", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13368, "uid": "1113d7a76ffceca1bb350bfe145467c6", "name": "Q-ShiftDP: A Differentially Private Parameter-Shift Rule for Quantum Machine Learning", "authors": [{"id": 19534, "fullname": "Hoang Ngo", "url": "http://virtual.aistats.org/api/miniconf/users/19534?format=json", "institution": "University of Florida"}, {"id": 21954, "fullname": "Nhat Hoang-Xuan", "url": "http://virtual.aistats.org/api/miniconf/users/21954?format=json", "institution": "University of Florida"}, {"id": 21955, "fullname": "Quan Nguyen", "url": "http://virtual.aistats.org/api/miniconf/users/21955?format=json", "institution": "University of Florida"}, {"id": 12751, "fullname": "Nguyen Do", "url": "http://virtual.aistats.org/api/miniconf/users/12751?format=json", "institution": "Posts and Telecommunications Institute of Technology"}, {"id": 21956, "fullname": "Incheol Shin", "url": "http://virtual.aistats.org/api/miniconf/users/21956?format=json", "institution": "Pukyong National University"}, {"id": 9841, "fullname": "My T. Thai", "url": "http://virtual.aistats.org/api/miniconf/users/9841?format=json", "institution": "University of Florida"}], "abstract": "Quantum Machine Learning (QML) promises significant computational advantages, but preserving training data privacy remains challenging. Classical approaches like differentially private stochastic gradient descent (DP-SGD) add noise to gradients but fail to exploit the unique properties of quantum gradient estimation. In this work, we introduce the Differentially Private Parameter-Shift Rule (Q-ShiftDP), the first privacy mechanism tailored to QML. By leveraging the inherent boundedness and stochasticity of quantum gradients computed via the parameter-shift rule, Q-ShiftDP enables tighter sensitivity analysis and reduces noise requirements. We combine carefully calibrated Gaussian noise with intrinsic quantum noise to provide formal privacy and utility guarantees, and show that harnessing quantum noise further improves the privacy\u2013utility trade-off. Experiments on benchmark datasets demonstrate that Q-ShiftDP consistently outperforms classical DP methods in QML.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13368", "url": null, "sourceid": 1952, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=usAR9HRSI1", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11061, "modified": "2026-03-29T20:42:56.027988-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=usAR9HRSI1", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "146", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13426, "uid": "97d98119037c5b8a9663cb21fb8ebf47", "name": "Efficient Bilevel Optimization with KFAC-Based Hypergradients", "authors": [{"id": 22067, "fullname": "Disen Liao", "url": "http://virtual.aistats.org/api/miniconf/users/22067?format=json", "institution": "University of Waterloo"}, {"id": 22068, "fullname": "Felix Dangel", "url": "http://virtual.aistats.org/api/miniconf/users/22068?format=json", "institution": "Vector Institute, Toronto"}, {"id": 12766, "fullname": "Yaoliang Yu", "url": "http://virtual.aistats.org/api/miniconf/users/12766?format=json", "institution": "University of Waterloo"}], "abstract": "Bilevel optimization (BO) is widely applicable to many machine learning problems. Scaling BO, however, requires repeatedly computing hypergradients, which involves solving inverse Hessian-vector products (IHVPs). In practice, these operations are often approximated using crude surrogates such as one-step gradient unrolling or identity/short Neumann expansions, which discard curvature information.  We build on implicit function theorem-based algorithms and propose to incorporate Kronecker-factored approximate curvature (KFAC), yielding curvature-aware hypergradients with a better performance efficiency trade-off than Conjugate Gradient (CG) or Neumann methods and consistently outperforming unrolling. We evaluate this approach across diverse tasks, including meta-learning and AI safety problems. On models up to BERT, we show that curvature information is valuable at scale, and KFAC can provide it with only modest memory and runtime overhead. Our implementation is available at \\url{https://github.com/liaodisen/NeuralBo}.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13426", "url": null, "sourceid": 1745, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=nao79LGFCr", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11119, "modified": "2026-03-29T20:42:58.355269-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=nao79LGFCr", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "61", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13371, "uid": "6766aa2750c19aad2fa1b32f36ed4aee", "name": "Balanced and Robust Multi-Treatment Experimental Designs via Randomized Differencing", "authors": [{"id": 21959, "fullname": "Qing Chen", "url": "http://virtual.aistats.org/api/miniconf/users/21959?format=json", "institution": "Rutgers University"}, {"id": 21960, "fullname": "Jing Jia", "url": "http://virtual.aistats.org/api/miniconf/users/21960?format=json", "institution": ", Rutgers University"}, {"id": 14879, "fullname": "Peng Zhang", "url": "http://virtual.aistats.org/api/miniconf/users/14879?format=json", "institution": "Rutgers University"}], "abstract": "We introduce GKK+, a new design for multi-arm randomized controlled trials. Standard Bernoulli randomization is robust but often yields poor covariate balance, while existing restricted-randomness designs mainly address two-arm settings. GKK+ extends the Karmarkar\u2013Karp (KK) differencing method to multiple arms. When covariates are smooth and well-behaved, GKK+ achieves an exponentially better covariate balance than the standard Bernoulli design while preserving sufficient randomness. GKK+ improves efficiency in treatment effect estimation while supporting standard asymptotic inference. Simulations on synthetic and real datasets demonstrate improved balance and lower estimator variance compared to existing methods.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13371", "url": null, "sourceid": 634, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=uKBcyB9ka5", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11064, "modified": "2026-03-29T20:42:56.139835-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=uKBcyB9ka5", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "19", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13884, "uid": "8f19793b2671094e63a15ab883d50137", "name": "Amortized Structural Variational Inference", "authors": [{"id": 19783, "fullname": "Shitao Fan", "url": "http://virtual.aistats.org/api/miniconf/users/19783?format=json", "institution": "university of maryland"}, {"id": 23051, "fullname": "Carlos Misael Madrid Padilla", "url": "http://virtual.aistats.org/api/miniconf/users/23051?format=json", "institution": "Washington University, Saint Louis"}, {"id": 21918, "fullname": "Yun Yang", "url": "http://virtual.aistats.org/api/miniconf/users/21918?format=json", "institution": "University of Maryland, College Park"}, {"id": 23052, "fullname": "Lizhen Lin", "url": "http://virtual.aistats.org/api/miniconf/users/23052?format=json", "institution": "University of Maryland, College Park"}], "abstract": "Variational inference (VI) is widely used for approximate Bayesian inference, but it can scale poorly and often requires re-optimization when new data arrive. Amortized variational inference (AVI) learns a global inference map, yet standard mean-field AVI can suffer from large variational and amortization gaps because of independence assumptions. We propose amortized structural variational inference (ASVI), which injects structural dependencies among latent variables through neural architectures that encode local neighborhood information. ASVI reduces both gaps while retaining scalability. Simulations and real-data experiments show that ASVI improves predictive accuracy and posterior fidelity over AVI, and matches structured VI at lower computational cost.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13884", "url": null, "sourceid": 1586, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=3PYMQEqn52", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11577, "modified": "2026-03-29T20:43:17.214828-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=3PYMQEqn52", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "21", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13372, "uid": "c8fbbc86abe8bd6a5eb6a3b4d0411301", "name": "Efficient Logistic Regression with Mixture of Sigmoids", "authors": [{"id": 21961, "fullname": "Federico Di Gennaro", "url": "http://virtual.aistats.org/api/miniconf/users/21961?format=json", "institution": "ETH Z\u00fcrich"}, {"id": 5584, "fullname": "Saptarshi Chakraborty", "url": "http://virtual.aistats.org/api/miniconf/users/5584?format=json", "institution": "UC Berkeley"}, {"id": 21962, "fullname": "Nikita Zhivotovskiy", "url": "http://virtual.aistats.org/api/miniconf/users/21962?format=json", "institution": "University of California, Berkeley"}], "abstract": "This paper explores the Exponential Weights Algorithm (EWA) with an isotropic Gaussian prior for online Logistic Regression. Our analysis establishes a worst-case regret bound of order $O\\left(d \\log(Bn)\\right)$ against the best linear predictor of norm at most $B$, together with a practical implementation whose total runtime is $\\tilde O(B^3 n^5)$ -- a substantial improvement over the $O(B^{18} n^{37})$ complexity of prior work (Foster et al., 2018) that achieved the same rate as Kakade and Ng (2005). Beyond efficiency, our analysis also reveals new geometric insights in the regime of linearly separable data. As the comparator norm bound $B\\rightarrow\\infty$, we show EWA's predictions converge to a solid-angle vote over separating directions. Indeed, this voting classifier is weighted by a truncated Gaussian distribution on the cone of directions that separates the data. Moreover, on any fixed-margin slice, the mode of such a distribution has the same direction as the hard-margin SVM solution (up to scale). Non-asymptotically, we prove that for a sufficiently large $B$, the regret becomes independent of $B$ and scales logarithmically with the margin $\\gamma$. This yields a rate of $O(d\\log(n))$ for well-specified logistic models with Gaussian design.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13372", "url": null, "sourceid": 906, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=uIrXC4fJIv", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11065, "modified": "2026-03-29T20:42:56.171548-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=uIrXC4fJIv", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "45", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13886, "uid": "bd0cc810b580b35884bd9df37c0e8b0f", "name": "The Reasoning-Creativity Trade-off: Toward Creativity-Driven Problem Solving", "authors": [{"id": 22882, "fullname": "Max Ruiz Luyten", "url": "http://virtual.aistats.org/api/miniconf/users/22882?format=json", "institution": "University of Cambridge"}, {"id": 863, "fullname": "Mihaela van der Schaar", "url": "http://virtual.aistats.org/api/miniconf/users/863?format=json", "institution": "University of Cambridge"}], "abstract": "State-of-the-art large language model (LLM) pipelines rely on bootstrapped reasoning loops\u2014sampling diverse chains of thought and reinforcing the highest-scoring ones\u2014primarily optimizing a scalar reward such as correctness. We analyze how this design choice is sensitive to the collapse of the model\u2019s distribution over reasoning paths, slashing semantic entropy and undermining creative problem-solving. To diagnose this failure, we introduce Distributional Creative Reasoning (DCR), a unified variational objective that casts training as gradient flow in the space of probability measures on solution traces. STaR, GRPO, and DPO, as well as entropy bonuses, novelty search, and quality\u2013diversity objectives, all emerge as special cases of the same loss. The framework delivers three core results: (i) a diversity decay theorem detailing how scalar-only objectives lead to distinct modes of diversity collapse for STaR, GRPO, and DPO; (ii) diversity-enhancing designs that, by sufficiently incorporating our DCR functional, ensure convergence to a unique, stable, and diverse policy, effectively counteracting collapse; and (iii) simple, actionable recipes to achieve this in practice. DCR thus offers the first principled recipe for LLMs that remain both correct \\emph{and} creative.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13886", "url": null, "sourceid": 2169, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=2xNpScK097", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11579, "modified": "2026-03-29T20:43:17.282550-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=2xNpScK097", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "178", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13890, "uid": "b5f1e8fb36cd7fbeb7988e8639ac79e9", "name": "Meta-probabilistic Modeling", "authors": [{"id": 23061, "fullname": "Kevin Zhang", "url": "http://virtual.aistats.org/api/miniconf/users/23061?format=json", "institution": "Massachusetts Institute of Technology"}, {"id": 12393, "fullname": "Yixin Wang", "url": "http://virtual.aistats.org/api/miniconf/users/12393?format=json", "institution": "University of Michigan"}], "abstract": "Probabilistic graphical models (PGMs) are widely used to discover latent structure in data, but their success hinges on selecting an appropriate model design. In practice, model specification is difficult and often requires iterative trial-and-error. This challenge arises because classical PGMs typically operate on individual datasets. In this work, we consider settings involving collections of related datasets and propose meta-probabilistic modeling (MPM) to learn the generative model structure itself. MPM uses a hierarchical formulation in which global components encode shared patterns across datasets, while local parameters capture dataset-specific latent structure. For scalable learning and inference, we derive a tractable VAE-inspired surrogate objective together with a bi-level optimization algorithm. Our methodology supports a broad class of expressive probabilistic models and has connections to existing architectures, such as Slot Attention. Experiments on object-centric representation learning and sequential text modeling demonstrate that MPM effectively adapts generative models to data while recovering meaningful latent representations.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13890", "url": null, "sourceid": 2067, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=2bOpsanMFj", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11583, "modified": "2026-03-29T20:43:17.428658-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=2bOpsanMFj", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "98", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13375, "uid": "673271cc47c1a4e77f57e239ed4d28a7", "name": "Leveraging Machine-Learned Advice in Strategic Interactions with No-Regret Learners", "authors": [{"id": 21965, "fullname": "Tinashe Handina", "url": "http://virtual.aistats.org/api/miniconf/users/21965?format=json", "institution": "California Institute of Technology"}, {"id": 21966, "fullname": "Tongxin Li", "url": "http://virtual.aistats.org/api/miniconf/users/21966?format=json", "institution": "The Chinese University of Hong Kong, Shenzhen"}, {"id": 21967, "fullname": "Kishan Panaganti", "url": "http://virtual.aistats.org/api/miniconf/users/21967?format=json", "institution": "Tencent AI Lab"}, {"id": 3643, "fullname": "Eric Mazumdar", "url": "http://virtual.aistats.org/api/miniconf/users/3643?format=json", "institution": "University of California Berkeley"}, {"id": 9606, "fullname": "Adam Wierman", "url": "http://virtual.aistats.org/api/miniconf/users/9606?format=json", "institution": "California Institute of Technology"}], "abstract": "As machine learning becomes increasingly integrated into decision-making across domains, understanding how machine-learned advice can be leveraged in strategic environments is of growing importance. In this work, we study how an agent in a two-player repeated game can effectively utilize potentially imperfect advice when interacting with a no-regret learner (i.e., satisfying a no-external or no-swap regret condition). We characterize the advice landscape by introducing a pseudo-metric to quantify the usefulness of an advice instance. We demonstrate the pseudo-metric's applicability through two forms of advice: simulators and payoff matrix predictions. We then show how an optimizing player, equipped with correctness guarantees on the advice, could leverage simulators to compute approximate Stackelberg strategies more efficiently, reducing the interaction complexity traditionally required and illustrating the power of good advice. Finally, we extend our analysis to settings where the advice does not have any guarantee of correctness. We find that, in general, a player cannot simultaneously guarantee near Stackelberg performance when the advice is approximately accurate and a no-regret condition when the advice is inaccurate. We do show, however, that it is possible for an advice-aided player to weakly dominate their utility in some (coarse)-correlated equilibria.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13375", "url": null, "sourceid": 1697, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=tznf37nYzM", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11068, "modified": "2026-03-29T20:42:56.289016-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=tznf37nYzM", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "95", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13468, "uid": "1cd3882394520876dc88d1472aa2a93f", "name": "Interpreting and Controlling Model Behavior via Constitutions for Atomic Concept Edits", "authors": [{"id": 22167, "fullname": "Neha Kalibhat", "url": "http://virtual.aistats.org/api/miniconf/users/22167?format=json", "institution": "Google DeepMind"}, {"id": 22168, "fullname": "Zi Wang", "url": "http://virtual.aistats.org/api/miniconf/users/22168?format=json", "institution": "Google DeepMind"}, {"id": 22169, "fullname": "Prasoon Bajpai", "url": "http://virtual.aistats.org/api/miniconf/users/22169?format=json", "institution": "Indian Institute of Technology, Delhi"}, {"id": 22170, "fullname": "Drew Proud", "url": "http://virtual.aistats.org/api/miniconf/users/22170?format=json", "institution": "DeepMind"}, {"id": 19627, "fullname": "Wenjun Zeng", "url": "http://virtual.aistats.org/api/miniconf/users/19627?format=json", "institution": "Google Deepmind"}, {"id": 22171, "fullname": "Been Kim", "url": "http://virtual.aistats.org/api/miniconf/users/22171?format=json", "institution": "Google DeepMind"}, {"id": 22172, "fullname": "Mani Malek", "url": "http://virtual.aistats.org/api/miniconf/users/22172?format=json", "institution": "Google"}], "abstract": "We introduce a black-box interpretability framework that learns a verifiable constitution: a natural language summary of how  changes to a prompt affect a model's specific behavior, such as its alignment, correctness, or adherence to constraints. Our method leverages atomic concept edits (ACEs), which are targeted operations that add, remove, or replace an interpretable concept in the input  prompt. By systematically applying ACEs and observing the resulting effects on model behavior across various tasks, our framework learns a causal mapping from edits to predictable outcomes. This learned constitution provides deep, generalizable insights into the model. Empirically, we validate our approach across diverse tasks, including mathematical reasoning and text-to-image alignment, for controlling and understanding model behavior. We found that for text-to-image generation, GPT-Image tends to focus on grammatical adherence, while Imagen 4 prioritizes atmospheric coherence. In mathematical reasoning, distractor variables confuse GPT-5 but leave Gemini 2.5 models and o4-mini largely unaffected. Moreover, our results show that the learned constitutions are highly effective for controlling model behavior, achieving an average of $1.86$ times boost in success rate over methods that do not use constitutions.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13468", "url": null, "sourceid": 1509, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=jEhTx67C6L", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11161, "modified": "2026-03-29T20:43:00.112062-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=jEhTx67C6L", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "175", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13498, "uid": "e94550c93cd70fe748e6982b3439ad3b", "name": "Private Synthetic Graph Generation and Fused Gromov-Wasserstein Distance", "authors": [{"id": 22244, "fullname": "Leoni Carla Wirth", "url": "http://virtual.aistats.org/api/miniconf/users/22244?format=json", "institution": "University of Oxford"}, {"id": 5701, "fullname": "Gholamali Aminian", "url": "http://virtual.aistats.org/api/miniconf/users/5701?format=json", "institution": "The Alan Turing Institute"}, {"id": 415, "fullname": "Gesine Reinert", "url": "http://virtual.aistats.org/api/miniconf/users/415?format=json", "institution": "University of Oxford"}], "abstract": "Networks are popular  representations of complex data. In particular, differentially private synthetic networks are much in demand.  Here, instead of  starting from a network,  we start with the complex data set itself and construct both a network representation and a corresponding synthetic network generator. We build a network model directly based on the underlying complex system data, capturing its structure and attributes. Using a random connection model, we  devise  an effective algorithmic approach for generating attributed synthetic networks  which is $\\epsilon$-differentially private at the vertex level, while preserving  utility.  We provide theoretical guarantees for the  accuracy of the private synthetic networks using the  fused Gromov-Wasserstein distance.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13498", "url": null, "sourceid": 742, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=g5QpPIwSst", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11191, "modified": "2026-03-29T20:43:01.228613-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=g5QpPIwSst", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "129", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13580, "uid": "cdc0d6e63aa8e41c89689f54970bb35f", "name": "Influence Attributions can be Systematically Altered by Model Manipulation", "authors": [{"id": 181, "fullname": "Chhavi Yadav", "url": "http://virtual.aistats.org/api/miniconf/users/181?format=json", "institution": "UCSD"}, {"id": 9697, "fullname": "Ruihan Wu", "url": "http://virtual.aistats.org/api/miniconf/users/9697?format=json", "institution": "Cornell University"}, {"id": 548, "fullname": "Kamalika Chaudhuri", "url": "http://virtual.aistats.org/api/miniconf/users/548?format=json", "institution": "University of California, San Diego"}], "abstract": "Influence Functions are a standard tool for attributing predictions to training data in a principled manner and are widely used in applications such as data valuation and fairness. In this work, we present realistic incentives to manipulate influence-based attributions and investigate whether these attributions can be \\textit{systematically} altered by an adversary. We show that small systemic perturbations to models can indeed alter influence-based attributions \\textit{as desired}. We work on logistic regression models trained on ResNet feature embeddings and standard tabular fairness datasets and provide efficient attacks with backward-friendly implementations. Our work raises questions on the reliability of influence-based attributions in adversarial circumstances.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13580", "url": null, "sourceid": 619, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=X7UA3hrmUs", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11273, "modified": "2026-03-29T20:43:04.358464-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=X7UA3hrmUs", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "89", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13595, "uid": "08c5433a60135c32e34f46a71175850c", "name": "Conformal Prediction in Hierarchical Classification with Constrained Representation Complexity", "authors": [{"id": 10105, "fullname": "Thomas Mortier", "url": "http://virtual.aistats.org/api/miniconf/users/10105?format=json", "institution": "Ghent University"}, {"id": 22454, "fullname": "Alireza Javanmardi", "url": "http://virtual.aistats.org/api/miniconf/users/22454?format=json", "institution": "Ludwig-Maximilians-Universit\u00e4t M\u00fcnchen"}, {"id": 22455, "fullname": "Yusuf Sale", "url": "http://virtual.aistats.org/api/miniconf/users/22455?format=json", "institution": "LMU Munich"}, {"id": 10107, "fullname": "Eyke H\u00fcllermeier", "url": "http://virtual.aistats.org/api/miniconf/users/10107?format=json", "institution": "Ludwig-Maximilians-Universit\u00e4t Munich"}, {"id": 10109, "fullname": "Willem Waegeman", "url": "http://virtual.aistats.org/api/miniconf/users/10109?format=json", "institution": "Ghent University"}], "abstract": "Conformal prediction has emerged as a widely used framework for constructing valid prediction sets in classification and regression tasks. In this work, we extend the split conformal prediction framework to hierarchical classification, where prediction sets are commonly restricted to internal nodes of a predefined hierarchy, and propose two computationally efficient inference algorithms. The first algorithm returns internal nodes as prediction sets, while the second one relaxes this restriction. Using the notion of representation complexity, the latter yields smaller set sizes at the cost of a more general and combinatorial inference problem. Empirical evaluations on several benchmark datasets demonstrate the effectiveness of the proposed algorithms in achieving nominal coverage.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13595", "url": null, "sourceid": 597, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=VQK48VVYrg", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11288, "modified": "2026-03-29T20:43:04.967139-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=VQK48VVYrg", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "34", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13377, "uid": "84d2004bf28a2095230e8e14993d398d", "name": "The Role of Causal Features in Strategic Classification for Robustness and Alignment", "authors": [{"id": 14427, "fullname": "Antonio Gois", "url": "http://virtual.aistats.org/api/miniconf/users/14427?format=json", "institution": "Mila &amp; University of Montreal"}, {"id": 19925, "fullname": "Sophia Gunluk", "url": "http://virtual.aistats.org/api/miniconf/users/19925?format=json", "institution": "Mila - Quebec AI Institute"}, {"id": 21968, "fullname": "Nir Rosenfeld", "url": "http://virtual.aistats.org/api/miniconf/users/21968?format=json", "institution": "Technion"}, {"id": 21969, "fullname": "Nidhi Hegde", "url": "http://virtual.aistats.org/api/miniconf/users/21969?format=json", "institution": "University of Alberta"}, {"id": 533, "fullname": "Simon Lacoste-Julien", "url": "http://virtual.aistats.org/api/miniconf/users/533?format=json", "institution": "Mila, Universit\u00e9 de Montr\u00e9al"}, {"id": 4260, "fullname": "Dhanya Sridhar", "url": "http://virtual.aistats.org/api/miniconf/users/4260?format=json", "institution": "University of Montreal"}], "abstract": "In strategic classification, an institution (e.g., a bank) anticipates adaptation from users who change their features to increase utility in a classification task (e.g., loan repayment). Since a key challenge is the distribution shift induced by users, we turn to causal models, which have been shown to bound the worst-case out-of-distribution (OOD) risk, and establish several new results that link causality and strategic classification. First, we show that causal classification leads to optimal classification error after any sufficiently large adaptation, when the noise is bounded in a certain way. Second, when these assumptions do not hold, we show OOD cross-entropy risk of optimal classifiers decomposes into an OOD bias term and a term arising from not using all observable features, allowing us to understand when causal classifiers have an advantage. Finally, we show that the use of causal features can allow alignment of long-term incentives between institutions and users, contrasting with previous work that highlights social costs of such approaches. We validate our theory empirically on synthetic data, finding that our results predict behavior in practice.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13377", "url": null, "sourceid": 1029, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=tgnnCHVGxd", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11070, "modified": "2026-03-29T20:42:56.388860-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=tgnnCHVGxd", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "159", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13628, "uid": "5ca3e9b122f61f8f06494c97b1afccf3", "name": "EventFlow: Forecasting Temporal Point Processes with Flow Matching", "authors": [{"id": 21964, "fullname": "Gavin Kerrigan", "url": "http://virtual.aistats.org/api/miniconf/users/21964?format=json", "institution": "University of Oxford"}, {"id": 22520, "fullname": "Kai Nelson", "url": "http://virtual.aistats.org/api/miniconf/users/22520?format=json", "institution": "University of California, Berkeley"}, {"id": 9562, "fullname": "Padhraic Smyth", "url": "http://virtual.aistats.org/api/miniconf/users/9562?format=json", "institution": "University of California, Irvine"}], "abstract": "Continuous-time event sequences, in which events occur at irregular intervals, are ubiquitous across a wide range of industrial and scientific domains. The contemporary modeling paradigm is to treat such data as realizations of a temporal point process, and in machine learning it is common to model temporal point processes in an autoregressive fashion using a neural network. While autoregressive models are successful in predicting the time of a single subsequent event, their performance can degrade when forecasting longer horizons due to cascading errors and myopic predictions. We propose EventFlow, a non-autoregressive generative model for temporal point processes. The model builds on the flow matching framework in order to directly learn joint distributions over event times, side-stepping the autoregressive process. EventFlow is simple to implement and achieves a 20\\%-53\\% lower forecast error than the nearest baseline on standard TPP benchmarks while simultaneously using fewer model calls at sampling time.", "topic": null, "keywords": [], "decision": "Accept (Oral)", "session": "Paper Award Talks", "eventtype": "Oral", "event_type": "Oral", "room_name": null, "virtualsite_url": "/virtual/2026/oral/13628", "url": null, "sourceid": -1384, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=2026-Oral--1384-84521f62", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Paper%20Award%20Talks?format=json", "parent_id": 11481, "eventmedia": [{"id": 11321, "modified": "2026-03-29T20:43:06.339893-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=2026-Oral--1384-84521f62", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": null, "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13746, "uid": "99c5e07b4d5de9d18c350cdf64c5aa3d", "name": "Identifiability of Potentially Degenerate Gaussian Mixture Models With Piecewise Affine Mixing", "authors": [{"id": 22756, "fullname": "Danru Xu", "url": "http://virtual.aistats.org/api/miniconf/users/22756?format=json", "institution": "University of Amsterdam"}, {"id": 18153, "fullname": "Sebastien Lachapelle", "url": "http://virtual.aistats.org/api/miniconf/users/18153?format=json", "institution": "Samsung"}, {"id": 2119, "fullname": "Sara Magliacane", "url": "http://virtual.aistats.org/api/miniconf/users/2119?format=json", "institution": "Saarland University, University of Amsterdam"}], "abstract": "Causal representation learning (CRL) aims to identify the underlying latent variables from high-dimensional observations, even when variables are dependent with each other. We study this problem for latent variables that follow a potentially degenerate Gaussian mixture distribution and that are only observed through the transformation via a piecewise affine mixing function. We provide a series of progressively stronger identifiability results for this challenging setting in which the probability density functions are ill-defined because of the potential degeneracy. For identifiability up to permutation and scaling, we leverage a sparsity regularization on the learned representation. Based on our theoretical results, we propose a two-stage method to estimate the latent variables by enforcing sparsity and Gaussianity in the learned representations. Experiments on synthetic and image data highlight our method\u2019s effectiveness in recovering the ground-truth latent variables.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13746", "url": null, "sourceid": 567, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=EXUX2g9XVu", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11439, "modified": "2026-03-29T20:43:11.349477-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=EXUX2g9XVu", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "83", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13754, "uid": "5cf21ce30208cfffaa832c6e44bb567d", "name": "Confidence-Guided Self-Training for Gradual Domain Adaptation", "authors": [{"id": 22774, "fullname": "Akram Heidarizadeh", "url": "http://virtual.aistats.org/api/miniconf/users/22774?format=json", "institution": "University of Central Florida"}, {"id": 22775, "fullname": "Akram Awad", "url": "http://virtual.aistats.org/api/miniconf/users/22775?format=json", "institution": "University of Central Florida"}, {"id": 22776, "fullname": "HanQin Cai", "url": "http://virtual.aistats.org/api/miniconf/users/22776?format=json", "institution": "University of Central Florida"}, {"id": 20591, "fullname": "George Atia", "url": "http://virtual.aistats.org/api/miniconf/users/20591?format=json", "institution": "University of Central Florida"}], "abstract": "Domain adaptation addresses the challenge of distributional shift between a labeled source domain and an unlabeled target domain. In gradual domain adaptation (GDA), the shift is assumed to occur through a sequence of intermediate domains, enabling smoother adaptation. A popular approach in this setting is self-training, where a model iteratively generates pseudo-labels for unlabeled data. However, pseudo-labeling errors can accumulate across rounds, especially under large shift, undermining generalization.  We develop a theoretical framework for self-training under gradual domain shift that explicitly quantifies and controls the pseudo-labeling error incurred at each round. Our first result is a modular generalization bound that decomposes the excess target risk into *coverage*, *pseudo-label error*  $(\\varepsilon_k)$ on the accepted set, domain shift, sample complexity, and regularization. Unlike prior bounds, our analysis separates the coverage penalty (due to rejecting inputs) from the pseudo-label error (controlled by confidence calibration or margin filtering, including Tsybakov-type noise via margin decay or calibration assumptions). We also provide the first theoretical justification for percentile (quantile) thresholding schemes used in practice: such schedules directly control coverage while tightening $\\varepsilon_k$, yielding a principled coverage--noise tradeoff. Under mild conditions, both terms accumulate only logarithmically, leading to improved generalization. We validate these insights across multiple GDA benchmarks, using both observed and OT-generated intermediate domains.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13754", "url": null, "sourceid": 1696, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=E7GKOyRPRl", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11447, "modified": "2026-03-29T20:43:11.726969-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=E7GKOyRPRl", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "42", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13380, "uid": "d240e3d38a8882ecad8633c8f9c78c9b", "name": "Unified Causal Discovery and Missing Data Imputation", "authors": [{"id": 21972, "fullname": "Osman Ali Mian", "url": "http://virtual.aistats.org/api/miniconf/users/21972?format=json", "institution": "IKIM - the Institute for Artificial Intelligence in Medicine"}, {"id": 21973, "fullname": "Jens Kleesiek", "url": "http://virtual.aistats.org/api/miniconf/users/21973?format=json", "institution": "Institute for AI in Medicine (IKIM), University Medicine Essen"}, {"id": 9978, "fullname": "Michael Kamp", "url": "http://virtual.aistats.org/api/miniconf/users/9978?format=json", "institution": "Institute for AI in Medicine (IKIM), Ruhr-University Bochum, and Monash University"}], "abstract": "Causal discovery and data imputation are often treated separately, yet both face challenges when data are missing. Existing causal discovery methods discard incomplete samples, leading to significant information loss, while standard imputation approaches rely on spurious correlations that distort the underlying causal signal. We introduce LOGIC, a framework that performs causal discovery and causally consistent imputation simultaneously. While existing work directly makes the assumption that all source variables in the causal graph are observed, we establish a verifiable criterion for this assumption under MCAR and MAR missingness, using the Algorithmic Markov Condition postulate. Building on this, LOGIC proceeds layer by layer, identifying sources, recovering downstream relations, and imputing missing values, while explicitly declaring unknowns when imputation is unsupported rather than forcefully completing the data. This preserves causal reasoning even in challenging missingness regimes. Experiments on synthetic and real-world datasets demonstrate that LOGIC achieves better performance than state-of-the-art baselines in both structure discovery and imputation accuracy.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13380", "url": null, "sourceid": 958, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=tNPrpkP1tb", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11073, "modified": "2026-03-29T20:42:56.540373-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=tNPrpkP1tb", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "176", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13381, "uid": "26751be1181460baf78db8d5eb7aad39", "name": "Spectral Thresholds in Correlated Spiked Models and Fundamental Limits of Partial Least Squares", "authors": [{"id": 19882, "fullname": "Pierre Mergny", "url": "http://virtual.aistats.org/api/miniconf/users/19882?format=json", "institution": "Ecole Normale Sup\u00e9rieure"}, {"id": 10115, "fullname": "Lenka Zdeborova", "url": "http://virtual.aistats.org/api/miniconf/users/10115?format=json", "institution": "EPFL"}], "abstract": "We provide a rigorous random matrix theory analysis of spiked cross-covariance models where the signals across two high-dimensional data channels are partially aligned. These models are motivated by multi-modal learning and form the standard generative setting underlying Partial Least Squares (PLS), a widely used yet theoretically underdeveloped method. We show that the leading singular values of the sample cross-covariance matrix undergo a Baik\u2013Ben Arous\u2013P\u00e9ch\u00e9 (BBP)-type phase transition, and we characterize the precise thresholds for the emergence of informative components. Our results yield the first sharp asymptotic description of the signal recovery capabilities of PLS in this setting, revealing a fundamental performance gap between PLS and the Bayes-optimal estimator. In particular, we identify the SNR and correlation regimes where PLS fails to recover any signal, despite detectability being possible in principle. These findings clarify the theoretical limits of PLS and provide guidance for the design of reliable multi-modal inference methods in high dimensions.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13381", "url": null, "sourceid": 1694, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=tAWZ4Pz5s9", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11074, "modified": "2026-03-29T20:42:56.582973-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=tAWZ4Pz5s9", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "169", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13387, "uid": "e7b24b112a44fdd9ee93bdf998c6ca0e", "name": "Variance Constrained Distribution Alignment in Few-shot Models", "authors": [{"id": 19820, "fullname": "Xiaohong Cai", "url": "http://virtual.aistats.org/api/miniconf/users/19820?format=json", "institution": "Beijing University of Posts and Telecommunications"}, {"id": 21983, "fullname": "Yi SUN", "url": "http://virtual.aistats.org/api/miniconf/users/21983?format=json", "institution": "Beijing University of Posts and Telecommunications Shenzhen Institude"}, {"id": 21984, "fullname": "Zhaowen Lin", "url": "http://virtual.aistats.org/api/miniconf/users/21984?format=json", "institution": "Beijing University of Posts and Telecommunications"}, {"id": 21985, "fullname": "Tianwei Cai", "url": "http://virtual.aistats.org/api/miniconf/users/21985?format=json", "institution": "Beijing University of Posts and Telecommunications"}], "abstract": "Learning generative models from the limited samples remains challenging due to unstable estimation of class conditional representations. Such instability often leads to intra-class distribution drift and degraded generalization under few sample regimes. To address these challenges, we propose a method that can model class level latent distributions for flexible and efficient few shot synthesis. Specifically, each input is represented by a learnable conditional latent distribution. Metric based statistical modeling effectively disentangles latent variables, contracts intra-class variance, and enlarges inter-class margins while enforcing cross task distributional alignment. Furthermore, we provide a variance based generalization analysis, showing that controlling class conditional variance tightens generalization bounds under few sample regimes. Experiments on the benchmark datasets demonstrate that our method surpasses prior works in visual quality and diversity, highlighting the benefit of statistical alignment for robust few shot generative modeling.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13387", "url": null, "sourceid": 360, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=s3ZDMBJGFV", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11080, "modified": "2026-03-29T20:42:56.851185-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=s3ZDMBJGFV", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "193", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13893, "uid": "c3c59e5f8b3e9753913f4d435b53c308", "name": "Sparse Linear Bandits with Blocking Constraints", "authors": [{"id": 22897, "fullname": "Adit Jain", "url": "http://virtual.aistats.org/api/miniconf/users/22897?format=json", "institution": "Cornell University"}, {"id": 23066, "fullname": "Soumyabrata Pal", "url": "http://virtual.aistats.org/api/miniconf/users/23066?format=json", "institution": "Adobe Systems"}, {"id": 23067, "fullname": "Sunav Choudhary", "url": "http://virtual.aistats.org/api/miniconf/users/23067?format=json", "institution": "Adobe Research"}, {"id": 23068, "fullname": "Ramasuri Narayanam", "url": "http://virtual.aistats.org/api/miniconf/users/23068?format=json", "institution": "International Business Machines"}, {"id": 23069, "fullname": "Harshita Chopra", "url": "http://virtual.aistats.org/api/miniconf/users/23069?format=json", "institution": "Department of Computer Science, University of Washington"}, {"id": 23070, "fullname": "Vikram Krishnamurthy", "url": "http://virtual.aistats.org/api/miniconf/users/23070?format=json", "institution": "Cornell University"}], "abstract": "We investigate the high-dimensional sparse linear bandits problem in a data-poor regime where the time horizon is much smaller than the ambient dimension and number of arms. We study the setting under the additional \\textit{blocking constraint} where each unique arm can be pulled only once. The blocking constraint is motivated by practical applications in personalized content recommendation and identification of datapoints to improve annotation efficiency for complex learning tasks. With mild assumptions on the arms, our proposed  online algorithm (\\texttt{BSLB}) achieves a regret guarantee of $\\widetilde{\\mathsf{O}}((1+\\beta_k)^2k^{\\frac{2}{3}} \\mathsf{T}^{\\frac{2}{3}})$ where the parameter vector has an (unknown) relative tail $\\beta_k$ - the ratio of $\\ell_1$ norm of the top-$k$ and remaining entries of the parameter vector. To this end, we show novel offline statistical guarantees of the lasso estimator for the linear model that is robust to the sparsity modeling assumption. Finally, we propose a meta-algorithm (\\texttt{C-BSLB})  based on corralling that does not need knowledge of optimal sparsity parameter $k$ at minimal cost to regret. Our experiments on multiple real-world datasets demonstrate the validity of our algorithms and theoretical framework.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13893", "url": null, "sourceid": 488, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=28pNNQtGf3", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11586, "modified": "2026-03-29T20:43:17.575646-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=28pNNQtGf3", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "170", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13389, "uid": "1b0114c51cc532ed34e1954b5b9e4b58", "name": "On the optimal regret of collaborative personalized linear bandits", "authors": [{"id": 17668, "fullname": "Bruce Huang", "url": "http://virtual.aistats.org/api/miniconf/users/17668?format=json", "institution": "University of California, Los Angeles"}, {"id": 12997, "fullname": "Ruida Zhou", "url": "http://virtual.aistats.org/api/miniconf/users/12997?format=json", "institution": "Texas A&amp;M University"}, {"id": 786, "fullname": "Lin Yang", "url": "http://virtual.aistats.org/api/miniconf/users/786?format=json", "institution": "UCLA"}, {"id": 17701, "fullname": "Suhas Diggavi", "url": "http://virtual.aistats.org/api/miniconf/users/17701?format=json", "institution": "University of California, Los Angeles"}], "abstract": "Stochastic linear bandits are a fundamental model for sequential decision making. Although well studied in the single-agent setting, many real-world scenarios involve multiple agents solving heterogeneous bandit problems, each with a different unknown parameter. This paper investigates the optimal regret achievable in collaborative personalized linear bandits.  We derive an information-theoretic lower bound showing how the number of agents, the number of rounds, and the degree of heterogeneity jointly affect regret. We propose a two-stage collaborative algorithm that achieves the optimal regret. We model heterogeneity via a hierarchical Bayesian framework and introduces a novel information-theoretic technique for bounding regret. Our results offer a complete characterization of when and how collaboration helps with a optimal regret bound $\\tilde{O}(d\\sqrt{mn})$, $\\tilde{O}(dm^{1-\\gamma}\\sqrt{n})$, $\\tilde{O}(dm\\sqrt{n})$ for the number of rounds $n$ in the range of $o \\left( \\frac{d}{m \\sigma^2} \\right)$, $\\Theta \\left( \\frac{d}{m^{2\\gamma} \\sigma^2} \\right)$ and $\\omega \\left( \\frac{d}{\\sigma^2}, \\right)$ respectively, where $\\sigma$ measures the level of heterogeneity, $m$ is the number of agents, and $\\gamma\\in[0, 1/2]$ is an absolute constant. In contrast,   without collaboration achieves a regret bound $O(dm\\sqrt{n})$ at best.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13389", "url": null, "sourceid": 1071, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=rqu1m6WgKB", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11082, "modified": "2026-03-29T20:42:56.922420-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=rqu1m6WgKB", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "118", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13900, "uid": "42998cf32d552343bc8e460416382dca", "name": "Integrating Feature Correlation in Differential Privacy with Applications in DP-ERM", "authors": [{"id": 5583, "fullname": "Tianyu Wang", "url": "http://virtual.aistats.org/api/miniconf/users/5583?format=json", "institution": "Columbia University"}, {"id": 23080, "fullname": "Luhao Zhang", "url": "http://virtual.aistats.org/api/miniconf/users/23080?format=json", "institution": "Johns Hopkins University"}, {"id": 18496, "fullname": "Rachel Cummings", "url": "http://virtual.aistats.org/api/miniconf/users/18496?format=json", "institution": "Columbia University"}], "abstract": "Standard differential privacy imposes uniform privacy constraints across all features, overlooking the inherent distinction between sensitive and insensitive features in practice. In this paper, we introduce a relaxed definition of differential privacy that accounts for such heterogeneity, allowing certain features to be treated as insensitive even when correlated with sensitive ones. We propose a correlation-aware framework, **CorrDP**, which relaxes privacy for insensitive features while accounting for their correlations with sensitive features, with the correlations quantified using total variation distance. We design algorithms for differentially private empirical risk minimization (DP-ERM) under the **CorrDP** framework, incorporating distance-dependent noise into gradients for improved theoretical utility guarantees. When the correlation distance is unknown, we estimate it from the dataset and show that it achieves a comparable privacy-utility guarantee. We perform experiments on synthetic and real-world datasets and show that **CorrDP**-based DP-ERM algorithms consistently outperform the standard DP framework.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13900", "url": null, "sourceid": 457, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=1YXnZiIRu2", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11593, "modified": "2026-03-29T20:43:17.838247-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=1YXnZiIRu2", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "91", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13904, "uid": "de03beffeed9da5f3639a621bcab5dd4", "name": "Impact of Positional Encoding: Clean and Adversarial Rademacher Complexity for Transformers under In-Context Regression", "authors": [{"id": 19775, "fullname": "Weiyi He", "url": "http://virtual.aistats.org/api/miniconf/users/19775?format=json", "institution": null}, {"id": 11006, "fullname": "Yue Xing", "url": "http://virtual.aistats.org/api/miniconf/users/11006?format=json", "institution": "Michigan State University"}], "abstract": "Positional encoding (PE) is a core architectural component of Transformers, yet its impact on the Transformer's generalization and robustness remains unclear. In this work, we provide the first generalization analysis for single-layer Transformer under in-context regression that explicitly accounts for a trainable PE module. Our result shows that PE systematically enlarges the generalization gap. Extending to the adversarial setting, we derive the adversarial Rademacher generalization bound. We find that the gap between models with and without PE is magnified under attack, demonstrating that PE amplifies the vulnerability of models. Our bounds are empirically validated by a simulation study. Together, this work establishes a new framework for understanding the clean and adversarial generalization in ICL with PE.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13904", "url": null, "sourceid": 1947, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=18OveXSw7d", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11597, "modified": "2026-03-29T20:43:17.986998-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=18OveXSw7d", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "83", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13905, "uid": "66e8ba8216a1e152d72653d99a4f03ab", "name": "$\\epsilon$-Identifiability of Causal Quantities", "authors": [{"id": 23086, "fullname": "Ang Li", "url": "http://virtual.aistats.org/api/miniconf/users/23086?format=json", "institution": "Computer Science Department, Florida State University"}, {"id": 23087, "fullname": "Scott Mueller", "url": "http://virtual.aistats.org/api/miniconf/users/23087?format=json", "institution": "Toyota Research Institute"}, {"id": 23088, "fullname": "Xin Shu", "url": "http://virtual.aistats.org/api/miniconf/users/23088?format=json", "institution": "Florida State University"}, {"id": 808, "fullname": "Judea Pearl", "url": "http://virtual.aistats.org/api/miniconf/users/808?format=json", "institution": "University of California, Los Angeles"}], "abstract": "Identifying the effects of causes and causes of effects is vital in virtually every scientific field. Often, however, the needed probabilities may not be fully identifiable from the available data sources. This paper shows how approximate identifiability is still possible for several probabilities of causation. We term this $\\epsilon\\text{-identifiability}$ and demonstrate its usefulness in cases where the behavior of certain subpopulations can be restricted within sufficiently narrow bounds. In particular, we show how unidentifiable causal effects and counterfactual probabilities can be $\\epsilon\\text{-identified}$ when such allowances are made. Often, these allowances are easily measured and reasonably assumed. Finally, $\\epsilon\\text{-identifiability}$ is applied to the unit selection problem.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13905", "url": null, "sourceid": 2265, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=0wiDik2c67", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11598, "modified": "2026-03-29T20:43:18.026801-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=0wiDik2c67", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "1", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13910, "uid": "705f2172834666788607efbfca35afb3", "name": "Robust Estimation of a Sparse Linear Model: Provable Guarantees with Non-convexity", "authors": [{"id": 23096, "fullname": "Deepak Maurya", "url": "http://virtual.aistats.org/api/miniconf/users/23096?format=json", "institution": ", Purdue University"}, {"id": 23097, "fullname": "Adarsh Barik", "url": "http://virtual.aistats.org/api/miniconf/users/23097?format=json", "institution": "Indian Institute of Technology, Delhi"}, {"id": 827, "fullname": "Jean Honorio", "url": "http://virtual.aistats.org/api/miniconf/users/827?format=json", "institution": "University of Melbourne"}], "abstract": "In this paper, we address the problem of sparse regression vector estimation in the presence of corrupted samples, with a particular focus on accurately identifying the support. Traditional methods, such as the Least Absolute Shrinkage and Selection Operator (LASSO), often fail in such scenarios, exhibiting inconsistency. To tackle this challenge, we propose a combinatorial, non-convex, and robust variant of LASSO framework, designed to enhance estimation accuracy under corruption. Our approach is supported by theoretical guarantees, which establish its reliability and robustness. Our method also handles corruption from heavy-tailed distributions, with only a few bounded moments. We validate our theoretical results through extensive experiments, comparing the performance of our method against the LASSO and its other robust variants. These comparisons highlight the efficacy of our framework, demonstrating its practical applicability in sparse regression tasks involving corrupted data.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13910", "url": null, "sourceid": 227, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=0bNcXCLrYA", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11603, "modified": "2026-03-29T20:43:18.235405-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=0bNcXCLrYA", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "159", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13383, "uid": "afe434653a898da20044041262b3ac74", "name": "A Proof of Learning Rate Transfer under $\\mu$P", "authors": [{"id": 21977, "fullname": "Soufiane Hayou", "url": "http://virtual.aistats.org/api/miniconf/users/21977?format=json", "institution": "Johns Hopkins University"}], "abstract": "We provide the first proof of learning rate transfer with a multi-layer perceptron (MLP) parametrized with $\\mu P$, a neural network parameterization designed to ``maximize'' feature learning in the infinite-width limit. We show that with $\\mu P$, the optimal learning rate converges to a non-zero constant as width goes to infinity. In contrast, we show that this doesn't hold with other parametrizations such as Standard Parameterization (SP) and Neural Tangent Parametrization (NTP). We provide extensive empirical results validating our theoretical findings.", "topic": null, "keywords": [], "decision": "Accept (Oral)", "session": "Oral Session 3: Optimization, Training Methods & Learning Dynamics", "eventtype": "Oral", "event_type": "Oral", "room_name": null, "virtualsite_url": "/virtual/2026/oral/13383", "url": null, "sourceid": -1444, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=2026-Oral--1444-d5d76247", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Oral%20Session%203:%20Optimization,%20Training%20Methods%20&%20Learning%20Dynamics?format=json", "parent_id": 11483, "eventmedia": [{"id": 11076, "modified": "2026-03-29T20:42:56.680740-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=2026-Oral--1444-d5d76247", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": null, "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13392, "uid": "69adc1e107f7f7d035d7baf04342e1ca", "name": "Scalable Spatiotemporal Inference with Biased Scan Attention Transformer Neural Processes", "authors": [{"id": 21991, "fullname": "Daniel Jenson", "url": "http://virtual.aistats.org/api/miniconf/users/21991?format=json", "institution": "Oxford University"}, {"id": 21992, "fullname": "Jhonathan Navott", "url": "http://virtual.aistats.org/api/miniconf/users/21992?format=json", "institution": "University of Oxford  / Imperial College London"}, {"id": 21993, "fullname": "Piotr Grynfelder", "url": "http://virtual.aistats.org/api/miniconf/users/21993?format=json", "institution": "University of Oxford"}, {"id": 21994, "fullname": "Mengyan Zhang", "url": "http://virtual.aistats.org/api/miniconf/users/21994?format=json", "institution": "University of Bristol"}, {"id": 21995, "fullname": "Makkunda Sharma", "url": "http://virtual.aistats.org/api/miniconf/users/21995?format=json", "institution": "Department of Computer Science, University of Oxford"}, {"id": 5135, "fullname": "Elizaveta Semenova", "url": "http://virtual.aistats.org/api/miniconf/users/5135?format=json", "institution": "Oxford University"}, {"id": 21996, "fullname": "Seth Flaxman", "url": "http://virtual.aistats.org/api/miniconf/users/21996?format=json", "institution": "University of Oxford"}], "abstract": "Neural Processes (NPs) are a rapidly evolving class of models designed to directly model the posterior predictive distribution of stochastic processes. While early architectures were developed primarily as a scalable alternative to Gaussian Processes (GPs), modern NPs tackle far more complex and data hungry applications spanning geology, epidemiology, climate, and robotics. These applications have placed increasing pressure on the scalability of these models, with many architectures compromising accuracy for scalability. In this paper, we demonstrate that this tradeoff is often unnecessary, particularly when modeling fully or partially translation invariant processes. We propose a versatile new architecture, the Biased Scan Attention Transformer Neural Process (BSA-TNP), which introduces Kernel Regression Blocks (KRBlocks), group-invariant attention biases, and memory-efficient Biased Scan Attention (BSA). BSA-TNP is able to: (1) match or exceed the accuracy of the best models while often training in a fraction of the time, (2) exhibit translation invariance, enabling learning at multiple resolutions simultaneously, (3) transparently model processes that evolve in both space and time, (4) support high dimensional fixed effects, and (5) scale gracefully -- running inference with over 1M test points with 100K context points in under a minute on a single 24GB GPU. All code is provided as part of the \\texttt{dl4bi} Python package.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13392", "url": null, "sourceid": 207, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=rRiACEi5Zp", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11085, "modified": "2026-03-29T20:42:57.055157-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=rRiACEi5Zp", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "165", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13721, "uid": "7fe1f8abaad094e0b5cb1b01d712f708", "name": "On Different Notions of Redundancy in Conditional-Independence-Based Discovery of Graphical Models", "authors": [{"id": 13103, "fullname": "Philipp M. Faller", "url": "http://virtual.aistats.org/api/miniconf/users/13103?format=json", "institution": "Karlsruhe Institute of Technology"}, {"id": 815, "fullname": "Dominik Janzing", "url": "http://virtual.aistats.org/api/miniconf/users/815?format=json", "institution": "Amazon"}], "abstract": "Conditional-independence-based discovery uses statistical tests to identify a graphical model that represents the independence structure of variables in a dataset. These test, however, can be unreliable and algorithms are sensitive to errors and violated assumptions.  Often there are tests that were not used in the construction of the graph. In this work, we show that these _redundant_ tests have the potential to _detect_ or sometimes _correct_ errors in the learned model. But we further show that not all tests contain this additional information and that such redundant tests have to be applied with care. Precisely, we argue that the conditional (in)dependence statements that  hold for every probability distribution are unlikely to detect and correct errors - in contrast to those that follow only from graphical assumptions.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13721", "url": null, "sourceid": 459, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=GAHWBv1DhM", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11414, "modified": "2026-03-29T20:43:10.213225-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=GAHWBv1DhM", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "121", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13631, "uid": "6ea2ef7311b482724a9b7b0bc0dd85c6", "name": "Filter, Augment, Forecast: Online Data Selection for Robust Time Series Forecasting", "authors": [{"id": 22464, "fullname": "Ege Taga", "url": "http://virtual.aistats.org/api/miniconf/users/22464?format=json", "institution": "University of Michigan - Ann Arbor"}, {"id": 22521, "fullname": "Halil Gozeten", "url": "http://virtual.aistats.org/api/miniconf/users/22521?format=json", "institution": "University of Michigan - Ann Arbor"}, {"id": 19816, "fullname": "Kutay Tire", "url": "http://virtual.aistats.org/api/miniconf/users/19816?format=json", "institution": "University of Texas at Austin"}, {"id": 22522, "fullname": "Rahul Dalvi", "url": "http://virtual.aistats.org/api/miniconf/users/22522?format=json", "institution": "University of Michigan - Ann Arbor"}, {"id": 22523, "fullname": "Reinhard Heckel", "url": "http://virtual.aistats.org/api/miniconf/users/22523?format=json", "institution": "Technische Universit\u00e4t M\u00fcnchen"}, {"id": 12995, "fullname": "Samet Oymak", "url": "http://virtual.aistats.org/api/miniconf/users/12995?format=json", "institution": "University of California, Riverside"}], "abstract": "Data curation pipelines play a central role in training deep learning architectures, with their impact in time series forecasting still relatively underexplored. In this work, we propose Filter, Augment, Forecast (FAF): an online data curation strategy based on (1) data selection to filter out low-quality (e.g., noisy) examples and (2) augmentation of the remaining high-quality data. We use reference model-based filtering inspired by the reducible holdout loss selection (RHO-LOSS) from the language modeling literature. We identify limitations of RHO-LOSS under domain shifts common in time series and introduce the adaptive RHO method (AdaRho), which improves performance by updating the reference model during training. Using random matrix theory, we further provide a statistical analysis that characterizes the role of the reference model, sample size, and noise statistics in data selection. FAF consistently improves forecasting accuracy across diverse architectures without modifying them, achieving state-of-the-art results. Specifically, applying FAF to eight state-of-the-art models yields a 6.55% mean reduction in MSE and a 3.79% mean reduction in MAE, averaged across nine benchmark datasets.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13631", "url": null, "sourceid": 480, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=PuVT8WarRl", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11324, "modified": "2026-03-29T20:43:06.481607-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=PuVT8WarRl", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "66", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13525, "uid": "5b69b9cb83065d403869739ae7f0995e", "name": "Differential Privacy in Kernelized Contextual Bandits via Random Projections", "authors": [{"id": 14353, "fullname": "Nikola Pavlovic", "url": "http://virtual.aistats.org/api/miniconf/users/14353?format=json", "institution": "Cornell University"}, {"id": 18105, "fullname": "Sudeep Salgia", "url": "http://virtual.aistats.org/api/miniconf/users/18105?format=json", "institution": "Carnegie Mellon University"}, {"id": 18098, "fullname": "Qing Zhao", "url": "http://virtual.aistats.org/api/miniconf/users/18098?format=json", "institution": "Cornell University"}], "abstract": "We consider the problem of contextual kernel bandits with stochastic contexts, where the underlying reward function belongs to a known Reproducing Kernel Hilbert Space. We study this problem under an additional constraint of Differential Privacy, where the agent needs to ensure that the sequence of query points is differentially private with respect to both the sequence of contexts and rewards. We propose a novel algorithm that achieves the state-of-the-art cumulative regret of $\\widetilde{\\mathcal{O}}(\\sqrt{\\gamma_TT}+\\frac{\\gamma_T}{\\varepsilon_{\\text{DP}}})$ and $\\widetilde{\\mathcal{O}}(\\sqrt{\\gamma_TT}+\\frac{\\gamma_T\\sqrt{T}}{\\varepsilon_{\\text{DP}}})$ over a time horizon of $T$ in the joint and local models of differential privacy, respectively, where $\\gamma_T$ is the effective dimension of the kernel and $\\varepsilon_{\\text{DP}} > 0$ is the privacy parameter. The key ingredient of the proposed algorithm is a novel private kernel-ridge regression estimator which is based on a combination of private covariance estimation and private random projections. It offers a significantly reduced sensitivity compared to its classical counterpart while maintaining a high prediction accuracy, allowing our algorithm to achieve the state-of-the-art performance guarantees.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13525", "url": null, "sourceid": 501, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=cj2jkRyYgj", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11218, "modified": "2026-03-29T20:43:02.218414-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=cj2jkRyYgj", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "46", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13747, "uid": "07563a3fe3bbe7e3ba84431ad9d055af", "name": "On Barycenter Computation: Analyzing Semi-Unbalanced Optimal Transport-based Method on Bures-Wasserstein manifold.", "authors": [{"id": 22757, "fullname": "Ngoc-Hai Nguyen", "url": "http://virtual.aistats.org/api/miniconf/users/22757?format=json", "institution": "Tufts University"}, {"id": 17686, "fullname": "Le Dung", "url": "http://virtual.aistats.org/api/miniconf/users/17686?format=json", "institution": "University of Texas at Austin"}, {"id": 22758, "fullname": "Hoang-Phi Nguyen", "url": "http://virtual.aistats.org/api/miniconf/users/22758?format=json", "institution": "VinAI Research"}, {"id": 22759, "fullname": "Tung Pham", "url": "http://virtual.aistats.org/api/miniconf/users/22759?format=json", "institution": "Qualcomm Inc, QualComm"}, {"id": 729, "fullname": "Nhat Ho", "url": "http://virtual.aistats.org/api/miniconf/users/729?format=json", "institution": "University of Texas at Austin"}], "abstract": "We explore a robust version of the barycenter problem among $n$ centered Gaussian probability measures, termed Semi-Unbalanced Optimal Transport (SUOT)-based Barycenter, wherein the barycenter remains fixed while the others are relaxed using Kullback-Leibler divergence. We develop optimization algorithms on Bures-Wasserstein manifold, named the Exact Geodesic Gradient Descent and Hybrid Gradient Descent algorithms. While the Exact Geodesic Gradient Descent method is based on computing the exact closed form of the first-order derivative of the objective function of the barycenter along a geodesic on the Bures manifold, the Hybrid Gradient Descent method utilizes optimizer components when solving the SUOT problem to replace contaminated measures before applying the Riemannian Gradient Descent. We establish the theoretical convergence guarantees for both methods and demonstrate that the Exact Geodesic Gradient Descent algorithm attains a dimension-free convergence rate. This is a novel theoretical result for Riemannian Gradient Descent applicable to an expanded class of averaging functions.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13747", "url": null, "sourceid": 521, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=EWl46BVj24", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11440, "modified": "2026-03-29T20:43:11.386919-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=EWl46BVj24", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "119", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13352, "uid": "8fecb20817b3847419bb3de39a609afe", "name": "Likelihood-Free Inference via Structured Score Matching", "authors": [{"id": 19417, "fullname": "Haoyu Jiang", "url": "http://virtual.aistats.org/api/miniconf/users/19417?format=json", "institution": "University of Illinois Urbana-Champaign"}, {"id": 21917, "fullname": "Yuexi Wang", "url": "http://virtual.aistats.org/api/miniconf/users/21917?format=json", "institution": "University of Illinois at Urbana-Champaign"}, {"id": 21918, "fullname": "Yun Yang", "url": "http://virtual.aistats.org/api/miniconf/users/21918?format=json", "institution": "University of Maryland, College Park"}], "abstract": "In many statistical problems, the data distribution is specified through a generative process for which the likelihood function is analytically intractable, yet inference on the associated model parameters remains of primary interest. We develop a likelihood-free inference framework that combines score matching with gradient-based optimization and bootstrap procedures to facilitate parameter estimation together with uncertainty quantification. The proposed methodology introduces tailored score-matching estimators for approximating likelihood score functions, and incorporates an architectural regularization scheme that embeds the statistical structure of log-likelihood scores to improve both accuracy and scalability. We provide theoretical guarantees and demonstrate the practical utility of the method through simulations and benchmark applications, where it performs favorably compared to existing approaches.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13352", "url": null, "sourceid": 675, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=x5QIOJx6W1", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11045, "modified": "2026-03-29T20:42:55.428066-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=x5QIOJx6W1", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "95", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13393, "uid": "f4be00279ee2e0a53eafdaa94a151e2c", "name": "Causal Partial Identification via Conditional Optimal Transport", "authors": [{"id": 21997, "fullname": "Sirui Lin", "url": "http://virtual.aistats.org/api/miniconf/users/21997?format=json", "institution": "Stanford University"}, {"id": 17922, "fullname": "Zijun Gao", "url": "http://virtual.aistats.org/api/miniconf/users/17922?format=json", "institution": "University of Southern California"}, {"id": 1323, "fullname": "Jose Blanchet", "url": "http://virtual.aistats.org/api/miniconf/users/1323?format=json", "institution": "Stanford University"}, {"id": 1324, "fullname": "Peter Glynn", "url": "http://virtual.aistats.org/api/miniconf/users/1324?format=json", "institution": "Stanford University"}], "abstract": "We study the estimation of causal estimand involving the joint distribution of treatment and control outcomes for a single unit. In typical causal inference settings, it is impossible to observe both outcomes simultaneously, which places our estimation within the domain of partial identification (PI). Pre-treatment covariates can substantially reduce estimation uncertainty by shrinking the partially identified set. Recently, it was shown that covariate-assisted PI sets can be characterized through conditional optimal transport (COT) problems. However, finite-sample estimation of COT poses significant challenges, primarily because the COT functional is discontinuous under the weak topology, rendering the direct plug-in estimator inconsistent. To circumvent this, existing literature relies on relaxations or indirect methods involving the estimation of non-parametric nuisance statistics. In this work, we demonstrate continuity of the COT problem under a stronger topology induced by the adapted Wasserstein distance. Leveraging this result, we propose a direct, consistent, non-parametric estimator for COT that avoids nuisance parameter estimation. We derive the convergence rate for our estimator and validate its effectiveness through comprehensive experiments, demonstrating its improved performance compared to existing techniques.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13393", "url": null, "sourceid": 528, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=rDufBj64yQ", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11086, "modified": "2026-03-29T20:42:57.084778-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=rDufBj64yQ", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "29", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13396, "uid": "a96b65a721e561e1e3de768ac819ffbb", "name": "Brenier Isotonic Regression", "authors": [{"id": 17677, "fullname": "Han Bao", "url": "http://virtual.aistats.org/api/miniconf/users/17677?format=json", "institution": "The Institute of Statistical Mathematics"}, {"id": 22002, "fullname": "Amirreza Eshraghi", "url": "http://virtual.aistats.org/api/miniconf/users/22002?format=json", "institution": "Illinois Institute of Technology"}, {"id": 22003, "fullname": "Yutong Wang", "url": "http://virtual.aistats.org/api/miniconf/users/22003?format=json", "institution": "Illinois Institute of Technology"}], "abstract": "Isotonic regression (IR) is shape-constrained regression to maintain a univariate fitting curve non-decreasing, which has numerous applications including single-index models and probability calibration. When it comes to multi-output regression, the classical IR is no longer applicable because the monotonicity is not readily extendable. We consider a novel multi-output regression problem where a regression function is \\emph{cyclically monotone}. Roughly speaking, a cyclically monotone function is the gradient of some convex potential. Whereas enforcing cyclic monotonicity is apparently challenging, we leverage the fact that Kantorovich's optimal transport (OT) always yields a cyclically monotone coupling as an optimal solution. This perspective naturally allows us to interpret a regression function and the convex potential as a link function in generalized linear models and Brenier's potential in OT, respectively, and hence we call this IR extension \\emph{Brenier isotonic regression}. We demonstrate experiments with probability calibration and generalized linear models. In particular, IR outperforms many famous baselines in probability calibration robustly.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13396", "url": null, "sourceid": 409, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=qQPhL4nHSZ", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11089, "modified": "2026-03-29T20:42:57.226138-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=qQPhL4nHSZ", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "26", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13781, "uid": "ca9c267dad0305d1a6308d2a0cf1c39c", "name": "Causal Additive Models with Unobserved Causal Paths and Backdoor Paths", "authors": [{"id": 19760, "fullname": "Thong Pham", "url": "http://virtual.aistats.org/api/miniconf/users/19760?format=json", "institution": "Shiga University"}, {"id": 22850, "fullname": "Takashi Nicholas Maeda", "url": "http://virtual.aistats.org/api/miniconf/users/22850?format=json", "institution": "Gakushuin University"}, {"id": 22851, "fullname": "Shohei Shimizu", "url": "http://virtual.aistats.org/api/miniconf/users/22851?format=json", "institution": "The University of Osaka"}], "abstract": "Causal additive models provide a tractable yet expressive framework for causal discovery in the presence of hidden variables. When unobserved backdoor or causal paths exist between two variables, their causal relationship is often unidentifiable under existing theories. We establish sufficient conditions under which causal directions can be identified in many such cases. These conditions rely on new characterizations of regression sets to determine independence among regression residuals and conditional independencies among observed variables. Building on these results, we introduce a search algorithm that incorporates these innovations and prove its soundness and completeness. Empirical evaluations demonstrate its competitive performance against state-of-the-art methods.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13781", "url": null, "sourceid": 679, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=C4QeriKzGf", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11474, "modified": "2026-03-29T20:43:12.792817-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=C4QeriKzGf", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "28", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13485, "uid": "10a5ab2db37feedfdeaab192ead4ac0e", "name": "Learning with Incomplete Context: Linear Contextual Bandits with Pretrained Imputation", "authors": [{"id": 14378, "fullname": "Hao Yan", "url": "http://virtual.aistats.org/api/miniconf/users/14378?format=json", "institution": "University of Wisconsin-Madison"}, {"id": 19421, "fullname": "Heyan Zhang", "url": "http://virtual.aistats.org/api/miniconf/users/19421?format=json", "institution": "University of Wisconsin - Madison"}, {"id": 22208, "fullname": "Yongyi Guo", "url": "http://virtual.aistats.org/api/miniconf/users/22208?format=json", "institution": "University of Wisconsin - Madison"}], "abstract": "The rise of large-scale pretrained models has made it feasible to generate predictive or synthetic features at low cost, raising the question of how to incorporate such surrogate predictions into downstream decision-making. We study this problem in the setting of online linear contextual bandits, where contexts may be complex, nonstationary, and only partially observed. In addition to bandit data, we assume access to an auxiliary historical dataset containing fully observed contexts--common in practice since such data are collected without adaptive interventions. We propose PULSE-UCB, an algorithm that leverages pretrained models trained on the auxiliary data to impute missing features during online decision-making. We establish regret guarantees that decompose into a standard bandit term plus an additional component reflecting pretrained model quality. In the i.i.d. context case with H\u00f6lder-smooth missing features, PULSE-UCB achieves near-optimal performance, supported by matching lower bounds. Our results quantify how uncertainty in predicted contexts affects decision quality and how much historical data is needed to improve downstream learning.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13485", "url": null, "sourceid": 691, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=gnSHM0Y1mU", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11178, "modified": "2026-03-29T20:43:00.741208-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=gnSHM0Y1mU", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "94", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13413, "uid": "f4552671f8909587cf485ea990207f3b", "name": "Thompson Sampling-like Algorithms for Stochastic Rising Bandits", "authors": [{"id": 22033, "fullname": "Marco Fiandri", "url": "http://virtual.aistats.org/api/miniconf/users/22033?format=json", "institution": "Polytechnic Institute of Milan"}, {"id": 118, "fullname": "Alberto Maria Metelli", "url": "http://virtual.aistats.org/api/miniconf/users/118?format=json", "institution": "Politecnico di Milano"}, {"id": 12318, "fullname": "Francesco Trov\u00f2", "url": "http://virtual.aistats.org/api/miniconf/users/12318?format=json", "institution": "Politecnico di Milano"}], "abstract": "Stochastic rising rested bandit (SRRB) is a setting where the arms' expected rewards increase as they are pulled. It models scenarios in which the performances of the options grow as an effect of an underlying learning process (e.g., online model selection). Even if the bandit literature provides specifically crafted algorithms based on upper-confidence bounds for such a setting, no study about Thompson sampling (TS)-like algorithms has been performed so far. The strong regularity of the expected rewards in the SRRB setting suggests that specific instances may be tackled effectively using adapted and sliding-window TS approaches. This work provides novel regret analyses for such algorithms in SRRBs, highlighting the challenges and providing new technical tools of independent interest. Our results allow us to identify under which assumptions TS-like algorithms succeed in achieving sublinear regret and which properties of the environment govern the complexity of the regret minimization problem when approached with TS. Furthermore, we provide a regret lower bound based on a complexity index we introduce. Finally, we conduct numerical simulations comparing TS-like algorithms with state-of-the-art approaches for SRRBs in synthetic and real-world settings.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13413", "url": null, "sourceid": 847, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=op4kXyewgG", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11106, "modified": "2026-03-29T20:42:57.893163-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=op4kXyewgG", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "160", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13723, "uid": "1f4477bad7af3616c1f933a02bfabe4e", "name": "Multi-Metric Adaptive Experimental Design Under a Fixed Budget with Validation", "authors": [{"id": 22720, "fullname": "Qining Zhang", "url": "http://virtual.aistats.org/api/miniconf/users/22720?format=json", "institution": "University of Michigan - Ann Arbor"}, {"id": 22721, "fullname": "Tanner Fiez", "url": "http://virtual.aistats.org/api/miniconf/users/22721?format=json", "institution": "Amazon"}, {"id": 22722, "fullname": "Yi Liu", "url": "http://virtual.aistats.org/api/miniconf/users/22722?format=json", "institution": "Amazon"}, {"id": 22723, "fullname": "Wenyang Liu", "url": "http://virtual.aistats.org/api/miniconf/users/22723?format=json", "institution": "Amazon"}], "abstract": "A/B tests in online experiments face statistical power challenges when testing multiple candidates simultaneously, while adaptive experimental designs (AED) alone fall short in inferring experiment statistics such as the average treatment effect, especially with many metrics (e.g., revenue, safety) and heterogeneous variances. This paper proposes a fixed-budget multi-metric AED framework with a two-phase structure: an adaptive exploration phase to identify the best treatment, and a validation phase with an A/B test to verify the treatment's quality and infer statistics. We propose SHRVar, which generalizes sequential halving (SH) with a novel relative-variance-based sampling and an elimination strategy built on reward $z$ values. It achieves a provable error probability that decreases exponentially, where the exponent $H_3$ generalizes the complexity measure for SH and SHVar with homogeneous and heterogeneous variances, respectively. Numerical experiments demonstrate its performance and robustness.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13723", "url": null, "sourceid": 927, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=G4nM0StcC2", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11416, "modified": "2026-03-29T20:43:10.280781-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=G4nM0StcC2", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "108", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13574, "uid": "df6d2338b2b8fce1ec2f6dda0a630eb0", "name": "Boosted GFlowNets: Improving Exploration via Sequential Learning", "authors": [{"id": 19475, "fullname": "Pedro Dall&#x27;Antonia", "url": "http://virtual.aistats.org/api/miniconf/users/19475?format=json", "institution": "FGV-EMAp"}, {"id": 22405, "fullname": "Tiago Silva", "url": "http://virtual.aistats.org/api/miniconf/users/22405?format=json", "institution": "Mohamed bin Zayed University of Artificial Intelligence"}, {"id": 220, "fullname": "Daniel Augusto de Souza", "url": "http://virtual.aistats.org/api/miniconf/users/220?format=json", "institution": "University College London"}, {"id": 424, "fullname": "C\u00e9sar Lincoln Mattos", "url": "http://virtual.aistats.org/api/miniconf/users/424?format=json", "institution": "Federal University of Cear\u00e1"}, {"id": 10049, "fullname": "Diego Mesquita", "url": "http://virtual.aistats.org/api/miniconf/users/10049?format=json", "institution": "Getulio Vargas Foundation (FGV EMAp)"}], "abstract": "Generative Flow Networks (GFlowNets) are powerful samplers for compositional objects that, by design, sample proportionally to a given non-negative reward. Nonetheless, in practice, they often struggle to explore the reward landscape evenly: trajectories toward easy-to-reach regions dominate training, while hard-to-reach modes receive vanishing or uninformative gradients, leading to poor coverage of high-reward areas. We address this imbalance with Boosted GFlowNets, a method that sequentially trains an ensemble of GFlowNets, each optimizing a residual reward that compensates for the mass already captured by previous models. This residual principle reactivates learning signals in underexplored regions and, under mild assumptions, ensures a monotone non-degradation property: adding boosters cannot worsen the learned distribution and typically improves it. Empirically, Boosted GFlowNets achieve substantially better exploration and sample diversity on multimodal synthetic benchmarks and peptide design tasks, while preserving the stability and simplicity of standard trajectory-balance training.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13574", "url": null, "sourceid": 987, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=XbH8QaJ6Mh", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11267, "modified": "2026-03-29T20:43:04.180451-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=XbH8QaJ6Mh", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "21", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13399, "uid": "cbef46321026d8404bc3216d4774c8a9", "name": "On the convergence and straightness of flow matching models", "authors": [{"id": 14734, "fullname": "Saptarshi Roy", "url": "http://virtual.aistats.org/api/miniconf/users/14734?format=json", "institution": "University of Texas at Austin"}, {"id": 14430, "fullname": "Vansh Bansal", "url": "http://virtual.aistats.org/api/miniconf/users/14430?format=json", "institution": "UT Austin"}, {"id": 22008, "fullname": "Purnamrita Sarkar", "url": "http://virtual.aistats.org/api/miniconf/users/22008?format=json", "institution": "University of Texas, Austin"}, {"id": 12618, "fullname": "Alessandro Rinaldo", "url": "http://virtual.aistats.org/api/miniconf/users/12618?format=json", "institution": "The University of Texas at Austin"}], "abstract": "Flow Matching has become a cornerstone of modern generative models like Stable Diffusion 3, largely due to the efficiency of its Rectified Flow (RF) variant. The success of RF hinges on iteratively learning straight trajectories, pushing generation towards fewer sampling steps. However, the theoretical link between path geometry and sampling efficiency has been under-explored. This paper fills this gap by introducing a novel \\textit{Piecewise Straightness} parameter, $\\gamma_{2,T}$. We establish the first Wasserstein convergence bound that explicitly links the discretization error of \\textit{any} general flow-model to $\\gamma_{2,T}$, proving that minimizing curvature is the key to achieving high-fidelity, one-step sampling.   Building on this theory, we establish the first theoretical framework to analyze the straightness of RF. We begin by offering intuitive geometric arguments for simple cases before identifying sufficient conditions under which a single rectification step (1-RF) yields a perfectly straight or even a Monge optimal coupling. While whether these sufficient conditions are met depends on the problem geometry, they enable the first concrete proofs in this area. Critically, fulfilling these conditions makes the subsequent flow (2-RF) perfectly straight ($\\gamma_{2,T}=0$). This eliminates the discretization error in our bound and makes flawless, single-step sampling possible.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13399", "url": null, "sourceid": 2075, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=qKKxzTucgy", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11092, "modified": "2026-03-29T20:42:57.335839-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=qKKxzTucgy", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "118", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13401, "uid": "c81e728d9d4c2f636f067f89cc14862c", "name": "Semi-Implicit Variational Inference via Kernelized Path Gradient Descent", "authors": [{"id": 22012, "fullname": "Tobias Pielok", "url": "http://virtual.aistats.org/api/miniconf/users/22012?format=json", "institution": "Department of Statistics, LMU Munich"}, {"id": 4577, "fullname": "Bernd Bischl", "url": "http://virtual.aistats.org/api/miniconf/users/4577?format=json", "institution": "LMU Munich"}, {"id": 14293, "fullname": "David R\u00fcgamer", "url": "http://virtual.aistats.org/api/miniconf/users/14293?format=json", "institution": "LMU Munich, MCML"}], "abstract": "Semi-implicit variational inference (SIVI) is a powerful framework for approximating complex posterior distributions, but training with the Kullback\u2013Leibler (KL) divergence can be challenging due to high variance and bias in high-dimensional settings. While current state-of-the-art score-based methods, particularly Kernel Semi-Implicit Variational Inference (K-SIVI), have been shown to also work in high dimensions, they can be \"blind'' to isolated components and mixing proportions, especially in multi-modal distributions. In this work, we propose a kernelized KL divergence estimator that stabilizes training through nonparametric smoothing, effectively addressing the \"blindness'' challenge. To further reduce the bias, we introduce an importance sampling correction. We provide a theoretical connection to the amortized version of the Stein variational gradient descent, which estimates the score gradient via Stein's identity, showing that both methods minimize the same objective, but our semi-implicit approach achieves lower gradient variance. In addition, our method's bias in function space is benign, leading to more stable and efficient optimization. Empirical results demonstrate that our method outperforms or matches state-of-the-art score matching methods in both performance and training efficiency.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13401", "url": null, "sourceid": 2, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=pzNqB3Jtq2", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11094, "modified": "2026-03-29T20:42:57.414633-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=pzNqB3Jtq2", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "167", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13885, "uid": "e6d8545daa42d5ced125a4bf747b3688", "name": "Power Transform Revisited: Numerically Stable, and Federated", "authors": [{"id": 23053, "fullname": "Xuefeng Xu", "url": "http://virtual.aistats.org/api/miniconf/users/23053?format=json", "institution": "University of Warwick"}, {"id": 23054, "fullname": "Graham Cormode", "url": "http://virtual.aistats.org/api/miniconf/users/23054?format=json", "institution": "University of Oxford"}], "abstract": "Power transforms are popular parametric methods for making data more Gaussian-like, and are widely used as preprocessing steps in statistical analysis and machine learning. However, we find that direct implementations of power transforms suffer from severe numerical instabilities, which can lead to incorrect results or even crashes. In this paper, we provide a comprehensive analysis of the sources of these instabilities and propose effective remedies. We further extend power transforms to the federated learning setting, addressing both numerical and distributional challenges that arise in this context. Experiments on real-world datasets demonstrate that our methods are both effective and robust, substantially improving stability compared to existing approaches.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13885", "url": "https://xuefeng-xu.github.io/powertf.html", "sourceid": 1247, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=3DxlMMknli", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11578, "modified": "2026-03-29T20:43:17.246939-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=3DxlMMknli", "resourcetype": "UriEventmedia"}, {"id": 11608, "file": "/media/PosterPDFs/AISTATS%202026/13885.png", "modified": "2026-03-30T00:38:08.810416-07:00", "display_section": 1, "type": "Poster", "name": "Poster", "visible": true, "sortkey": 0, "is_live_content": false, "detailed_kind": "", "generated_from": null, "resourcetype": "EventmediaImageFile"}, {"id": 11609, "file": "/media/PosterPDFs/AISTATS%202026/13885-thumb.png", "modified": "2026-03-30T00:38:08.918403-07:00", "display_section": 1, "type": "Poster", "name": "Poster", "visible": false, "sortkey": 0, "is_live_content": false, "detailed_kind": "thumb", "generated_from": null, "resourcetype": "EventmediaImageFile"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "128", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13883, "uid": "995665640dc319973d3173a74a03860c", "name": "Rank Lifting and Random Non-Linear Maps", "authors": [{"id": 23046, "fullname": "Andrea Drago", "url": "http://virtual.aistats.org/api/miniconf/users/23046?format=json", "institution": "University of Roma &quot;La Sapienza&quot;"}, {"id": 20598, "fullname": "Maria Sofia Bucarelli", "url": "http://virtual.aistats.org/api/miniconf/users/20598?format=json", "institution": "CNRS, I3S, INRIA"}, {"id": 23047, "fullname": "Francesco Caso", "url": "http://virtual.aistats.org/api/miniconf/users/23047?format=json", "institution": "University of Roma &quot;La Sapienza&quot;"}, {"id": 23048, "fullname": "Marius Michetti", "url": "http://virtual.aistats.org/api/miniconf/users/23048?format=json", "institution": "University of Roma &quot;La Sapienza&quot;"}, {"id": 23049, "fullname": "Federico Siciliano", "url": "http://virtual.aistats.org/api/miniconf/users/23049?format=json", "institution": "University of Roma &quot;La Sapienza&quot;"}, {"id": 18031, "fullname": "Fabrizio Silvestri", "url": "http://virtual.aistats.org/api/miniconf/users/18031?format=json", "institution": "Sapienza University of Rome"}, {"id": 23050, "fullname": "Luca Becchetti", "url": "http://virtual.aistats.org/api/miniconf/users/23050?format=json", "institution": "University of Roma &quot;La Sapienza&quot;"}], "abstract": "Deep neural networks exhibit improved training and generalization performance as the number of parameters grows well beyond the size of the training set, contradicting classical intuitions about overfitting. In order to gain a better understanding of this \u201cbenign overparameterization\u201d, we analyze the representational capacity of a random one-hidden-layer perceptron with Gaussian weights, no bias and threshold activations. More precisely, we investigate the following question: when does a hidden layer of dimension $n$ maps $k$ input vectors with pairwise angles at least $\\theta$, to a full-rank activation matrix, thus ensuring that a simple linear classifier can perfectly fit those inputs in feature space? This problem has an immediate impact on memorization capacity at initialization and we frame it as a question about hyperplane arrangements on the unit sphere, and we prove new isoperimetric-like inequalities.  This allows us to derive non-trivial lower bounds on the probability that a random embedding avoids the arrangement\u2019s zero-measure regions. Our results show that once the hidden dimension exceeds a threshold (depending on $\\theta$ and the input dimension), hidden representations are linearly independent with high probability. While the case we consider is challenging due to the sparsity of the solution space, this setting highlights crucial, underlying geometric problems and connections to related questions in spherical geometry and linear algebra.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13883", "url": null, "sourceid": 1279, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=3Tvd1qJ5UJ", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11576, "modified": "2026-03-29T20:43:17.181967-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=3Tvd1qJ5UJ", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "135", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13813, "uid": "6d3a1e06d6a06349436bc054313b648c", "name": "Learnability with Partial Labels and Adaptive Nearest Neighbors", "authors": [{"id": 22920, "fullname": "Nicolas A. Errandonea", "url": "http://virtual.aistats.org/api/miniconf/users/22920?format=json", "institution": "Basque Center for Applied Mathematics"}, {"id": 13578, "fullname": "Santiago Mazuelas", "url": "http://virtual.aistats.org/api/miniconf/users/13578?format=json", "institution": "Basque Center for Applied Mathematics"}, {"id": 13195, "fullname": "Jose A Lozano", "url": "http://virtual.aistats.org/api/miniconf/users/13195?format=json", "institution": "UPV/EHU"}, {"id": 4422, "fullname": "Sanjoy Dasgupta", "url": "http://virtual.aistats.org/api/miniconf/users/4422?format=json", "institution": "UCSD"}], "abstract": "Prior work on partial labels learning (PLL) has shown that  learning is possible  even when each instance is associated with a bag of labels, rather than a single  accurate but costly label. However, the necessary conditions for learning with partial labels remain unclear, and existing PLL methods are effective only in specific scenarios. In this work, we mathematically characterize the scenarios in which PLL is feasible. In addition, we present PL A-$k$NN, an adaptive nearest-neighbors algorithm for PLL that is effective in general scenarios and enjoys strong performance guarantees. Experimental results corroborate that PL A-$k$NN  can outperform state-of-the-art methods in general PLL scenarios", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13813", "url": null, "sourceid": 1487, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=9Xf9ZUEF3F", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11506, "modified": "2026-03-29T20:43:14.077569-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=9Xf9ZUEF3F", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "89", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13406, "uid": "cb2c2041d9763d84d7d655e81178f444", "name": "Amortized In-Context Mixed Effect Transformer Models: A Zero-Shot Approach for Pharmacokinetics", "authors": [{"id": 19853, "fullname": "C\u00e9sar Ali Ojeda Marin", "url": "http://virtual.aistats.org/api/miniconf/users/19853?format=json", "institution": "University of Potsdam"}, {"id": 22021, "fullname": "Niklas Hartung", "url": "http://virtual.aistats.org/api/miniconf/users/22021?format=json", "institution": "Universit\u00e4t Potsdam"}, {"id": 22022, "fullname": "Wilhelm Huisinga", "url": "http://virtual.aistats.org/api/miniconf/users/22022?format=json", "institution": "Universit\u00e4t Potsdam"}, {"id": 22023, "fullname": "Ramses Sanchez", "url": "http://virtual.aistats.org/api/miniconf/users/22023?format=json", "institution": "Rheinische Friedrich-Wilhelms Universit\u00e4t Bonn"}], "abstract": "Accurate dose\u2013response forecasting under sparse sampling is central to precision pharmacotherapy.  We present the Amortized In-Context Mixed-Effect Transformer (AICMET) model, a transformer\u2011based latent\u2011variable framework that unifies mechanistic compartmental priors with amortized in\u2011context Bayesian inference.  AICMET is pre\u2011trained on hundreds of thousands of synthetic pharmacokinetic trajectories with  Ornstein-Uhlenbeck priors over the parameters of compartment models, endowing the model with strong inductive biases and enabling zero\u2011shot adaptation to new compounds. At inference time, the decoder conditions on the collective context of previously profiled trial participants, generating calibrated posterior predictions for newly enrolled patients after a few early drug concentration measurements.  This capability collapses traditional model\u2011development cycles from weeks to hours while preserving some degree of expert modelling.  Experiments across public datasets show that AICMET attains state\u2011of\u2011the\u2011art predictive accuracy and faithfully quantifies inter\u2011patient variability\u2014outperforming both nonlinear mixed\u2011effects baselines and recent neural ODE variants.  Our results highlight the feasibility of transformer\u2011based, population\u2011aware neural architectures as offering a new alternative for bespoke pharmacokinetic modeling pipelines, charting a path toward truly population-aware personalized dosing regimens.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13406", "url": null, "sourceid": 2414, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=pdamyf7o8W", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11099, "modified": "2026-03-29T20:42:57.620620-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=pdamyf7o8W", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "20", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13411, "uid": "0233f3bb964cf325a30f8b1c2ed2da93", "name": "Bandits in Flux: Dynamic Regret under Adversarial Constraints", "authors": [{"id": 22030, "fullname": "Tareq Si Salem", "url": "http://virtual.aistats.org/api/miniconf/users/22030?format=json", "institution": "Huawei"}], "abstract": "We investigate the challenging problem of adversarial multi-armed bandits operating under time-varying constraints, a scenario motivated by numerous real-world applications. To address this complex setting, we propose a novel primal-dual algorithm that extends online mirror descent through the incorporation of suitable gradient estimators and effective constraint handling. We provide theoretical guarantees establishing sublinear dynamic regret and sublinear constraint violation for our proposed policy. Our algorithm achieves state-of-the-art performance in terms of both regret and constraint violation. Empirical evaluations demonstrate the superiority of our approach.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13411", "url": null, "sourceid": 2262, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=owSVNdQ3Hi", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11104, "modified": "2026-03-29T20:42:57.815412-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=owSVNdQ3Hi", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "28", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13773, "uid": "d63fbf8c3173730f82b150c5ef38b8ff", "name": "Recovery Guarantees for Continual Learning of Dependent Tasks: Memory, Data-Dependent Regularization, and Data-Dependent Weights", "authors": [{"id": 22828, "fullname": "Liangzu Peng", "url": "http://virtual.aistats.org/api/miniconf/users/22828?format=json", "institution": "University of Pennsylvania"}, {"id": 14362, "fullname": "Uday Kiran Reddy Tadipatri", "url": "http://virtual.aistats.org/api/miniconf/users/14362?format=json", "institution": "University of Pennsylvania"}, {"id": 17963, "fullname": "Ziqing Xu", "url": "http://virtual.aistats.org/api/miniconf/users/17963?format=json", "institution": "The Wharton School, University of Pennsylvania"}, {"id": 22829, "fullname": "Eric Eaton", "url": "http://virtual.aistats.org/api/miniconf/users/22829?format=json", "institution": "University of Pennsylvania"}, {"id": 12922, "fullname": "Rene Vidal", "url": "http://virtual.aistats.org/api/miniconf/users/12922?format=json", "institution": "University of Pennsylvania"}], "abstract": "Continual learning (CL) is concerned with learning multiple tasks sequentially without forgetting previously learned tasks. Despite substantial empirical advances over recent years, the theoretical development of CL remains in its infancy. At the heart of developing CL theory lies the challenge that the data distribution varies across tasks, and we argue that properly addressing this challenge requires understanding this variation---dependency among tasks. To explicitly model task dependency, we consider nonlinear regression tasks and propose the assumption that these tasks are dependent in such a way that the data of the current task is a nonlinear transformation of previous data. With this model and under natural assumptions, we prove statistical recovery guarantees (more specifically, bounds on estimation errors) for several CL paradigms in practical use, including experience replay with data-independent regularization and data-independent weights that balance the losses of tasks, replay with data-dependent weights, and continual learning with data-dependent regularization (e.g., knowledge distillation). To the best of our knowledge, our bounds are novel in several aspects, and they are informative in cases where prior work gives vacuous bounds.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13773", "url": null, "sourceid": 1534, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=CZ0GoxYw4k", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11466, "modified": "2026-03-29T20:43:12.464685-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=CZ0GoxYw4k", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "151", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13642, "uid": "88a199611ac2b85bd3f76e8ee7e55650", "name": "It\u2019s All In The (Exponential) Family: An Equivalence Between Maximum Likelihood Estimation and Control Variates For Sketching Algorithms", "authors": [{"id": 19779, "fullname": "Keegan Kang", "url": "http://virtual.aistats.org/api/miniconf/users/19779?format=json", "institution": "Bucknell University"}, {"id": 22542, "fullname": "Kerong Wang", "url": "http://virtual.aistats.org/api/miniconf/users/22542?format=json", "institution": "University of California, Santa Barbara"}, {"id": 22543, "fullname": "Ding Zhang", "url": "http://virtual.aistats.org/api/miniconf/users/22543?format=json", "institution": "University of Virginia, Charlottesville"}, {"id": 22544, "fullname": "Rameshwar Pratap", "url": "http://virtual.aistats.org/api/miniconf/users/22544?format=json", "institution": "Indian Institute of Technology, Hyderabad"}, {"id": 22545, "fullname": "Bhisham Verma", "url": "http://virtual.aistats.org/api/miniconf/users/22545?format=json", "institution": "Wake Forest University"}, {"id": 22546, "fullname": "Benedict Wong", "url": "http://virtual.aistats.org/api/miniconf/users/22546?format=json", "institution": "A*STAR"}], "abstract": "Maximum likelihood estimators (MLE) and control variate estimators (CVE) have been used in conjunction with known information across sketching algorithms and applications in machine learning. We prove that under certain conditions in an exponential family, an optimal CVE will achieve the same asymptotic variance as the MLE, giving an Expectation-Maximization (EM) algorithm for the MLE. Experiments show the EM algorithm is faster and numerically stable compared to other root finding algorithms for the MLE for the bivariate Normal distribution, and we expect this to hold across distributions satisfying these conditions. We show how the EM algorithm leads to reproducibility for algorithms using MLE / CVE, and demonstrate how the EM algorithm leads to finding the MLE when the CV weights are known.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13642", "url": null, "sourceid": 1581, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=OXTdMSkLxv", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11335, "modified": "2026-03-29T20:43:06.880425-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=OXTdMSkLxv", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "88", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13405, "uid": "96de2547f44254c97f5f4f1f402711c1", "name": "Provable Affine Identifiability of Nonlinear CCA under Latent Distributional Priors", "authors": [{"id": 22019, "fullname": "Zhiwei Han", "url": "http://virtual.aistats.org/api/miniconf/users/22019?format=json", "institution": "Technical University of Munich"}, {"id": 22020, "fullname": "Stefan Matthes", "url": "http://virtual.aistats.org/api/miniconf/users/22020?format=json", "institution": "Technical University of Munich"}, {"id": 18638, "fullname": "Hao Shen", "url": "http://virtual.aistats.org/api/miniconf/users/18638?format=json", "institution": "Fortiss GmbH"}], "abstract": "In this work, we establish the sufficient conditions under which nonlinear Canonical Correlation Analysis (CCA) recovers ground-truth latent factors up to an affine transformation. By transporting the analysis from the observation space to the source space, we extend classical statistical results on orthogonal polynomial expansions of bivariate distributions to representation learning, proving affine identifiability under specific distributional priors. We formally demonstrate that whitening is strictly necessary to ensure the boundedness and well-conditioning of the learned mappings. Furthermore, we bridge the gap between theory and practice by proving that ridge-regularized empirical CCA converges to its population counterpart in the finite-sample regime. Finally, our findings provide a rigorous theoretical foundation explaining the empirical success of recent correlation-based non-contrastive learning methods. Experiments on synthetic and rendered image datasets, alongside systematic ablations, validate the predicted recovery behavior and illustrate the failure modes that arise when the assumptions are violated.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13405", "url": null, "sourceid": 1649, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=plbSNDbYkb", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11098, "modified": "2026-03-29T20:42:57.585720-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=plbSNDbYkb", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "142", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13415, "uid": "5f6371c9126149517d9ba475def53139", "name": "UniPROT: Uniform Prototype Selection via Partial Optimal Transport with Submodular Guarantees", "authors": [{"id": 22036, "fullname": "Prateek Chanda", "url": "http://virtual.aistats.org/api/miniconf/users/22036?format=json", "institution": "Indian Institute of Technology, Bombay"}, {"id": 22037, "fullname": "Prayas Agrawal", "url": "http://virtual.aistats.org/api/miniconf/users/22037?format=json", "institution": "Indian Institute of Technology Bombay, Indian Institute of Technology, Bombay"}, {"id": 22038, "fullname": "Karthik Gurumoorthy", "url": "http://virtual.aistats.org/api/miniconf/users/22038?format=json", "institution": "Walmart Global Tech"}, {"id": 22039, "fullname": "Ganesh Ramakrishnan", "url": "http://virtual.aistats.org/api/miniconf/users/22039?format=json", "institution": "Indian Institute of Technology Bombay, Indian Institute of Technology Bombay"}, {"id": 9537, "fullname": "Bamdev Mishra", "url": "http://virtual.aistats.org/api/miniconf/users/9537?format=json", "institution": "Microsoft"}, {"id": 9538, "fullname": "Pratik Jawanpuria", "url": "http://virtual.aistats.org/api/miniconf/users/9538?format=json", "institution": "Microsoft"}], "abstract": "Selecting prototypical examples from a source distribution to represent a target data distribution is a fundamental problem in machine learning. Existing subset selection methods often rely on implicit importance scores, which can be skewed towards majority classes and lead to low-quality prototypes for minority classes. We introduce UniPROT, a novel subset selection framework that minimizes the optimal transport (OT) distance between a uniformly weighted prototypical distribution and the target distribution. While intuitive, this formulation leads to a cardinality constrained super-additive maximization problem that is challenging to approximate efficiently. To address this, we propose a principled reformulation of the OT marginal constraints, yielding a partial optimal transport-based submodular objective. We prove that this reformulation is tight and enables a greedy algorithm with a ((1 - \\frac{1}{e})) approximation guarantee relative to the original sub-additive maximization problem. Empirically, we showcase that enforcing uniform prototype weights in UniPROT consistently improves minority-class representation in imbalanced classification benchmarks without compromising majority-class accuracy. In both finetuning and pretraining regimes for large language models under domain imbalance, UniPROT enforces uniform source contributions, yielding robust performance gains. Our results establish UniPROT as a scalable, theoretically grounded solution for uniform-weighted prototype selection.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13415", "url": null, "sourceid": 2293, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=oesdMIku1h", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11108, "modified": "2026-03-29T20:42:57.997266-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=oesdMIku1h", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "186", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13418, "uid": "6c4b761a28b734fe93831e3fb400ce87", "name": "A Scalable Lift-and-Project Differentiable Approach For the Maximum Cut Problem", "authors": [{"id": 22041, "fullname": "Ismail Alkhouri", "url": "http://virtual.aistats.org/api/miniconf/users/22041?format=json", "institution": "Los Alamos National Laboratory"}, {"id": 22042, "fullname": "Mian Wu", "url": "http://virtual.aistats.org/api/miniconf/users/22042?format=json", "institution": "Ohio State University, Columbus"}, {"id": 22043, "fullname": "CUNXI YU", "url": "http://virtual.aistats.org/api/miniconf/users/22043?format=json", "institution": "University of Maryland, College Park"}, {"id": 22044, "fullname": "Jia Liu", "url": "http://virtual.aistats.org/api/miniconf/users/22044?format=json", "institution": "The Ohio State University"}, {"id": 22045, "fullname": "Rongrong Wang", "url": "http://virtual.aistats.org/api/miniconf/users/22045?format=json", "institution": "Michigan State University"}, {"id": 22046, "fullname": "Alvaro Velasquez", "url": "http://virtual.aistats.org/api/miniconf/users/22046?format=json", "institution": "Defense Advanced Research Projects Agency"}], "abstract": "We propose a scalable framework for solving the Maximum Cut (MaxCut) problem in large graphs using projected gradient ascent on quadratic objectives. Our approach is differentiable and leverages GPUs for gradient-based optimization. It is not a machine learning method and does not require training data. Starting from a continuous relaxation of the classical quadratic binary formulation, we present a parallelized strategy that explores multiple initialization vectors in batch. We analyze the relaxed objective, showing it is convex and has fixed-points corresponding to local optima\u2014particularly at boundary points\u2014highlighting a key challenge in non-convex optimization. To improve exploration, we introduce a lifted quadratic formulation that over-parameterizes the solution space. We also provide a theoretical characterization of these lifted fixed-points. Finally, we propose DECO, a dimension-alternating algorithm that switches between the unlifted and lifted formulations, combined with importance-based degree initialization and a population-based evolutionary hyper-parameter search. Experiments on diverse graph families show that our methods attain comparable or superior performance relative to recent neural networks and GPU-accelerated sampling approaches.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13418", "url": null, "sourceid": 157, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=ob0TVfqMIg", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11111, "modified": "2026-03-29T20:42:58.125576-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=ob0TVfqMIg", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "10", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13470, "uid": "17ed8abedc255908be746d245e50263a", "name": "Adaptive Memory Momentum via a Model-Based Framework for Deep Learning Optimization", "authors": [{"id": 20624, "fullname": "Kristi Topollai", "url": "http://virtual.aistats.org/api/miniconf/users/20624?format=json", "institution": "New York University"}, {"id": 4387, "fullname": "Anna Choromanska", "url": "http://virtual.aistats.org/api/miniconf/users/4387?format=json", "institution": "NYU"}], "abstract": "The vast majority of modern deep learning models are trained with momentum-based first-order optimizers. The momentum term governs the optimizer's memory by determining how much each past gradient contributes to the current convergence direction. Fundamental momentum methods, such as Nesterov Accelerated Gradient and the Heavy Ball method, as well as more recent optimizers such as AdamW and Lion, all rely on the momentum coefficient that is customarily set to $\\beta = 0.9$ and kept constant during model training, a strategy widely used by practitioners, yet suboptimal. In this paper, we introduce an adaptive memory mechanism that replaces constant momentum with a dynamic momentum coefficient that is adjusted online during optimization. We derive our method by approximating the objective function using two planes: one derived from the gradient at the current iterate and the other obtained from the accumulated memory of the past gradients. To the best of our knowledge, such a proximal framework was never used for momentum-based optimization. Our proposed approach is novel, extremely simple to use, and does not rely on extra assumptions or hyperparameter tuning. We implement adaptive memory variants of both SGD and AdamW across a wide range of learning tasks, from simple convex problems to large-scale deep learning scenarios, demonstrating that our approach can outperform standard SGD and Adam with hand-tuned momentum coefficients. Finally, our work opens doors for new ways of inducing adaptivity in optimization.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13470", "url": null, "sourceid": 1762, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=iV81utmf9x", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11163, "modified": "2026-03-29T20:43:00.169599-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=iV81utmf9x", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "16", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13646, "uid": "e5a4d6bf330f23a8707bb0d6001dfbe8", "name": "ZipMoE: A Theoretically-Grounded Mixture of Experts Approach forParameter-Efficient Deep Learning", "authors": [{"id": 155, "fullname": "Lin Chen", "url": "http://virtual.aistats.org/api/miniconf/users/155?format=json", "institution": "Google"}, {"id": 22552, "fullname": "Kyriakos Axiotis", "url": "http://virtual.aistats.org/api/miniconf/users/22552?format=json", "institution": "Google"}, {"id": 22553, "fullname": "Gang Fu", "url": "http://virtual.aistats.org/api/miniconf/users/22553?format=json", "institution": "Google Research"}, {"id": 22554, "fullname": "Kaiyuan Wang", "url": "http://virtual.aistats.org/api/miniconf/users/22554?format=json", "institution": "Google"}, {"id": 12526, "fullname": "MohammadHossein Bateni", "url": "http://virtual.aistats.org/api/miniconf/users/12526?format=json", "institution": "Google Research"}, {"id": 22555, "fullname": "Vahab Mirrokni", "url": "http://virtual.aistats.org/api/miniconf/users/22555?format=json", "institution": "Google Research"}], "abstract": "The growing size of large language models (LLMs) presents significant challenges to their efficient training and deployment. To address this, we introduce ZipMoE, a novel family of parameter-efficient building blocks inspired by the Mixture of Experts (MoE) paradigm. ZipMoE provides a modular and efficient alternative to traditional fully connected layers. We theoretically analyze the expressiveness of ZipMoE, demonstrating its advantage over low-rank factorization in terms of representational capacity and test error in a least squares regression setting. Empirical results, including comparisons with low-rank, Monarch, and Kronecker methods, confirm that ZipMoE achieves superior model quality under equivalent parameter or FLOP budgets, highlighting its effectiveness as a parameter-efficient building block for LLMs.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13646", "url": null, "sourceid": 1770, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=NtcXbO1Qza", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11339, "modified": "2026-03-29T20:43:07.063204-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=NtcXbO1Qza", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "191", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13888, "uid": "b865367fc4c0845c0682bd466e6ebf4c", "name": "An Illusion of Unlearning? Assessing Machine Unlearning Through Internal Representation", "authors": [{"id": 19634, "fullname": "Yichen Gao", "url": "http://virtual.aistats.org/api/miniconf/users/19634?format=json", "institution": "OSU"}, {"id": 23058, "fullname": "Altay Unal", "url": "http://virtual.aistats.org/api/miniconf/users/23058?format=json", "institution": "New Jersey Institute of Technology"}, {"id": 23059, "fullname": "Akshay Rangamani", "url": "http://virtual.aistats.org/api/miniconf/users/23059?format=json", "institution": "New Jersey Institute of Technology"}, {"id": 1206, "fullname": "Zhihui Zhu", "url": "http://virtual.aistats.org/api/miniconf/users/1206?format=json", "institution": "University of Denver"}], "abstract": "While numerous machine unlearning (MU) methods have recently been developed with promising results in erasing the influence of forgotten data, classes, or concepts, they are also highly vulnerable\u2014for example, simple fine-tuning can inadvertently reintroduce erased concepts. In this paper, we address this contradiction by examining the {\\it internal feature representations} of unlearned models, in contrast to prior work that focuses primarily on output-level behavior. Our analysis shows that many state-of-the-art MU methods appear successful mainly due to a misalignment between last-layer features and the classifier\u2014a phenomenon we call {\\it feature\u2013classifier misalignment}. In fact, hidden features remain highly discriminative, and simple linear probing can recover near-original accuracy. Assuming neural collapse in the original model, we further demonstrate that adjusting only the classifier can achieve negligible forget accuracy while preserving retain accuracy, and we corroborate this with experiments using classifier-only fine-tuning. Motivated by these findings, we propose MU methods based on a class-mean features (CMF) classifier, which explicitly enforces alignment between features and classifiers. Experiments on standard benchmarks show that CMF-based unlearning reduces forgotten information in representations while maintaining high retain accuracy, highlighting the need for faithful representation-level evaluation of MU.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13888", "url": null, "sourceid": 1783, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=2kwnpGKxUo", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11581, "modified": "2026-03-29T20:43:17.354446-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=2kwnpGKxUo", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "23", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13328, "uid": "4496bf24afe7fab6f046bf4923da8de6", "name": "Inverse-Free Sparse Variational Gaussian Processes", "authors": [{"id": 21874, "fullname": "Stefano Cortinovis", "url": "http://virtual.aistats.org/api/miniconf/users/21874?format=json", "institution": "University of Oxford"}, {"id": 21875, "fullname": "Laurence Aitchison", "url": "http://virtual.aistats.org/api/miniconf/users/21875?format=json", "institution": "University of Bristol"}, {"id": 21876, "fullname": "Stefanos Eleftheriadis", "url": "http://virtual.aistats.org/api/miniconf/users/21876?format=json", "institution": "Unaffiliated"}, {"id": 21877, "fullname": "Mark van der Wilk", "url": "http://virtual.aistats.org/api/miniconf/users/21877?format=json", "institution": "University of Oxford"}], "abstract": "Gaussian processes (GPs) offer appealing properties but are costly to train at scale. Sparse variational GP (SVGP) approximations reduce cost yet still rely on Cholesky decompositions of kernel matrices, ill-suited to low-precision, massively parallel hardware. While one can construct valid variational bounds that rely only on matrix multiplications (matmuls) via an auxiliary matrix parameter, optimising them with off-the-shelf first-order methods is challenging. We make the inverse-free approach practical by proposing a better-conditioned bound and deriving a matmul-only natural-gradient update for the auxiliary parameter, markedly improving stability and convergence. We further provide simple heuristics, such as step-size schedules and stopping criteria, that make the overall optimisation routine fit seamlessly into existing workflows. Across regression and classification benchmarks, we demonstrate that our method 1) serves as a drop-in replacement in SVGP-based models (e.g., deep GPs), 2) recovers similar performance to traditional methods, and 3) can be faster than baselines when well tuned.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13328", "url": null, "sourceid": 1828, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=zQZ6CSVoQj", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11021, "modified": "2026-03-29T20:42:54.316943-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=zQZ6CSVoQj", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "86", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13431, "uid": "fc4ddc15f9f4b4b06ef7844d6bb53abf", "name": "Replicable Machine Learning: Theory and Algorithms for Stochastic Convex and Non-Convex Optimization", "authors": [{"id": 656, "fullname": "Raman Arora", "url": "http://virtual.aistats.org/api/miniconf/users/656?format=json", "institution": "Johns Hopkins University"}, {"id": 22085, "fullname": "Kaibo Zhang", "url": "http://virtual.aistats.org/api/miniconf/users/22085?format=json", "institution": "Department of Computer Science, Whiting School of Engineering"}], "abstract": "Replicable algorithms produce identical outputs with high probability when run on independent samples from the same distribution. We study replicable stochastic optimization providing algorithms with near-optimal sample complexity across convex and non-convex settings. For general Lipschitz losses, the exponential mechanism with correlated sampling achieves optimal $O(1/\\sqrt{n})$ excess risk and $\\rho$-replicability, but with exponential runtime. For strongly convex losses, empirical risk minimization (ERM) with randomized rounding achieves $\\tilde{O}(\\sqrt{d}/(\\rho\\sqrt{n}))$ excess risk in polynomial time. For general convex losses, regularized ERM yields $\\tilde{O}(n^{-1/4})$ rates. We extend our techniques to neural networks in the NTK regime. Our work reveals a fundamental computational statistical tradeoff. Optimal replicability requires exponential time, while efficient algorithms incur modest statistical penalties.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13431", "url": null, "sourceid": 1902, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=n3PY5syAha", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11124, "modified": "2026-03-29T20:42:58.553079-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=n3PY5syAha", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "156", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13420, "uid": "3bf55bbad370a8fcad1d09b005e278c2", "name": "ReTrack: Data Unlearning in Diffusion Models through Redirecting the Denoising Trajectory", "authors": [{"id": 19831, "fullname": "Qitan Shi", "url": "http://virtual.aistats.org/api/miniconf/users/19831?format=json", "institution": "Tsinghua University"}, {"id": 22048, "fullname": "Cheng Jin", "url": "http://virtual.aistats.org/api/miniconf/users/22048?format=json", "institution": "Tsinghua University"}, {"id": 22049, "fullname": "Jiawei Zhang", "url": "http://virtual.aistats.org/api/miniconf/users/22049?format=json", "institution": "Tsinghua University"}, {"id": 9338, "fullname": "Yuantao Gu", "url": "http://virtual.aistats.org/api/miniconf/users/9338?format=json", "institution": "Tsinghua University"}], "abstract": "Diffusion models excel at generating high-quality, diverse images but also suffer from undesirable training data memorization, raising critical privacy and safety concerns. Data unlearning has emerged to mitigate this issue by removing the influence of specific data through fine-tuning rather than retraining from scratch. We propose ReTrack, a fast and effective data unlearning method for diffusion models. ReTrack employs importance sampling to construct a more efficient unbiased fine-tuning loss. This loss is further approximated by retaining only the dominant terms, thereby reducing computational cost. This yields an interpretable objective that redirects denoising trajectories toward the $k$-nearest neighbors, enabling efficient unlearning while preserving generative quality. Experiments on MNIST T-Shirt, CelebA-HQ, CIFAR-10, and Stable Diffusion show that ReTrack achieves state-of-the-art performance, striking the best trade-off between unlearning strength and generation quality preservation.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13420", "url": null, "sourceid": 1255, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=oTUZfa0iPv", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11113, "modified": "2026-03-29T20:42:58.188213-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=oTUZfa0iPv", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "143", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13800, "uid": "0950ca92a4dcf426067cfd2246bb5ff3", "name": "A Unifying Framework for Unsupervised Concept Extraction", "authors": [{"id": 18354, "fullname": "Chandler Squires", "url": "http://virtual.aistats.org/api/miniconf/users/18354?format=json", "institution": "CMU, Carnegie Mellon University"}], "abstract": "Techniques for *concept extraction*, such as sparse autoencoders and transcoders, aim to extract high-level symbolic concepts from low-level nonsymbolic representations. When these extracted concepts are used for downstream tasks such as model steering and unlearning, it is essential to understand their guarantees, or lack thereof. In this work, we present a unified theoretical framework for unsupervised concept extraction, in which we frame the task of concept extraction as identifying a generative model. We present a general meta-theorem for identifiability, which reduces the problem of establishing identifiability guarantees to the problem of characterizing the intersection of two sets. As we demonstrate on a range of widely-used approaches, this meta-theorem substantially simplifies the task of proving such guarantees, thus paving the way for the development of new, principled approaches for concept extraction.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13800", "url": null, "sourceid": 1925, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=Am9YJMkNQi", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11493, "modified": "2026-03-29T20:43:13.546695-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=Am9YJMkNQi", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "8", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13660, "uid": "e5b294b70c9647dcf804d7baa1903918", "name": "Nearly Optimal Best Arm Identification for Semiparametric Bandits", "authors": [{"id": 19746, "fullname": "Seok-Jin Kim", "url": "http://virtual.aistats.org/api/miniconf/users/19746?format=json", "institution": "Columbia University"}], "abstract": "We study the fixed confidence Best Arm Identification (BAI) problem in semiparametric bandits, where rewards follow a linear model with an unknown additive baseline shift. While BAI is well understood for linear bandits, optimality in this semiparametric setting has remained open. Firstly, we establish a sample complexity lower bound for semiparametric bandit BAI problem.  Next, we propose an efficient BAI algorithm and proved it's sample complexity matches lower bound up to logarithmic factors. We also extend our results to transductive BAI problem and obtained nearly optimal results.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13660", "url": null, "sourceid": 1927, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=Mn1Vx6RZ1e", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11353, "modified": "2026-03-29T20:43:07.667606-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=Mn1Vx6RZ1e", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "109", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13650, "uid": "52d2752b150f9c35ccb6869cbf074e48", "name": "On the Finite-Sample Bias of Minimizing Expected Wasserstein Loss Between Empirical Distributions", "authors": [{"id": 19918, "fullname": "Cheongjae Jang", "url": "http://virtual.aistats.org/api/miniconf/users/19918?format=json", "institution": "Hanyang University"}, {"id": 22560, "fullname": "Yung-Kyun Noh", "url": "http://virtual.aistats.org/api/miniconf/users/22560?format=json", "institution": "Hanyang University"}], "abstract": "We show that minimizing the expected Wasserstein loss between empirical distributions can lead to biased parameter estimates in the finite-sample regime. Remarkably, such bias arises even in well-specified settings where both empirical distributions are drawn from the same parametric family: unlike maximum likelihood estimation\u2014understood here as maximizing the expected log-likelihood\u2014optimizing one parameter while fixing another fails to recover the true fixed value. We derive closed-form expressions for the expected Wasserstein loss in one dimension and, focusing on location\u2013scale models, provide an analytic characterization of the bias. This analysis reveals that finite-sample bias occurs whenever the expected loss varies along the diagonal subspace where parameter values coincide, and we propose a simple correction scheme that removes this effect. We extend our analysis to misspecified models and the Sinkhorn divergence, demonstrating that finite-sample bias persists in more practical settings. Experiments on synthetic and real data confirm that stochastic optimization of Wasserstein-based objectives converges to biased solutions, and validate the effectiveness of the proposed correction scheme.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13650", "url": null, "sourceid": 1932, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=NkKkUL9380", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11343, "modified": "2026-03-29T20:43:07.261506-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=NkKkUL9380", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "119", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13464, "uid": "4a3e00961a08879c34f91ca0070ea2f5", "name": "A New Perspective on Least-Norm Interpolation Under Gaussian Covariates", "authors": [{"id": 18034, "fullname": "Gil Kur", "url": "http://virtual.aistats.org/api/miniconf/users/18034?format=json", "institution": "Department of Computer Science, ETHZ - ETH Zurich"}, {"id": 22155, "fullname": "Zong Shang", "url": "http://virtual.aistats.org/api/miniconf/users/22155?format=json", "institution": "Ecole Nationale de la Statistique et de l&#x27;Administration Economique"}, {"id": 22156, "fullname": "Paul Simanjuntak", "url": "http://virtual.aistats.org/api/miniconf/users/22156?format=json", "institution": "Texas A&amp;M University - College Station"}, {"id": 22157, "fullname": "Guillaume Lecu\u00e9", "url": "http://virtual.aistats.org/api/miniconf/users/22157?format=json", "institution": "ESSEC, business school"}, {"id": 22158, "fullname": "Reese Pathak", "url": "http://virtual.aistats.org/api/miniconf/users/22158?format=json", "institution": "University of California Berkeley"}], "abstract": "Least-Norm Interpolators (LNI) in overparameterized linear models have gained attention as a tractable framework for studying interpolation phenomena that resemble empirical observations in neural networks. Most prior work on these interpolators exploits closed-form solutions when available or heavily relies on Gaussian comparison results, such as the convex Gaussian Min-Max Theorem (CGMT). In this paper, we introduce a new perspective on LNI under Gaussian covariates by leveraging tools from high-dimensional geometry. First, we obtain a new variational formula for the bias of any LNI under isotropic Gaussian covariates when the norm is in Milman's $M$-position. Next, we prove the sharp rates on $\\ell_1$-LNI that were obtained by Wang et al. 22', using techniques from Gaussian polytopes and super-concentration. Crucially, our approach does not rely on CGMT.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13464", "url": null, "sourceid": 1989, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=jpsrJrRzBS", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11157, "modified": "2026-03-29T20:42:59.945073-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=jpsrJrRzBS", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "5", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13424, "uid": "788d986905533aba051261497ecffcbb", "name": "DISPO: Enhancing Training Efficiency and Stability in Reinforcement Learning for Large Language Model Mathematical Reasoning", "authors": [{"id": 19386, "fullname": "Batuhan Karaman", "url": "http://virtual.aistats.org/api/miniconf/users/19386?format=json", "institution": "Cornell University"}, {"id": 22061, "fullname": "Aditya Rawal", "url": "http://virtual.aistats.org/api/miniconf/users/22061?format=json", "institution": "Amazon"}, {"id": 1180, "fullname": "Mohammad Ghavamzadeh", "url": "http://virtual.aistats.org/api/miniconf/users/1180?format=json", "institution": "Google Research"}, {"id": 22062, "fullname": "Suhaila Shakiah", "url": "http://virtual.aistats.org/api/miniconf/users/22062?format=json", "institution": "Amazon"}, {"id": 22063, "fullname": "Arijit Biswas", "url": "http://virtual.aistats.org/api/miniconf/users/22063?format=json", "institution": "Amazon"}, {"id": 12997, "fullname": "Ruida Zhou", "url": "http://virtual.aistats.org/api/miniconf/users/12997?format=json", "institution": "Texas A&amp;M University"}], "abstract": "Reinforcement learning with verifiable rewards has emerged as a promising paradigm for enhancing the reasoning capabilities of large language models particularly in mathematics. Current approaches in this domain present a clear trade-off: PPO-style methods (e.g., GRPO/DAPO) offer training stability but exhibit slow learning trajectories due to their trust-region constraints on policy updates, while REINFORCE-style approaches (e.g., CISPO) demonstrate improved learning efficiency but suffer from performance instability as they clip importance sampling weights while still permitting non-zero gradients outside the trust-region. To address these limitations, we introduce DISPO, a simple yet effective REINFORCE-style algorithm that decouples the up-clipping and down-clipping of importance sampling weights for correct and incorrect responses, yielding four controllable policy update regimes. Through targeted ablations, we uncover how each regime impacts training: for correct responses, weights $>1$ increase the average token entropy (i.e., exploration) while weights $<1$ decrease it (i.e., distillation) - both beneficial but causing gradual performance degradation when excessive. For incorrect responses, overly restrictive clipping triggers sudden performance collapse through repetitive outputs (when weights $>1$) or vanishing response lengths (when weights $<1$). By separately tuning these four clipping parameters, DISPO maintains the exploration-distillation balance while preventing catastrophic failures, achieving 61.04\\% on AIME'24 (vs.\\ 55.42\\% CISPO and 50.21\\% DAPO) with similar gains across various benchmarks and models.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13424", "url": null, "sourceid": 717, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=ntslllFNE7", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11117, "modified": "2026-03-29T20:42:58.299205-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=ntslllFNE7", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "50", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13458, "uid": "d1a21da7bca4abff8b0b61b87597de73", "name": "CoreSPECT: Enhancing Clustering Algorithms via an Interplay of Density and Geometry", "authors": [{"id": 19851, "fullname": "Chandra Sekhar Mukherjee", "url": "http://virtual.aistats.org/api/miniconf/users/19851?format=json", "institution": "University of Southern California"}, {"id": 22144, "fullname": "Joonyoung Bae", "url": "http://virtual.aistats.org/api/miniconf/users/22144?format=json", "institution": "University of Southern California"}, {"id": 22145, "fullname": "Jiapeng Zhang", "url": "http://virtual.aistats.org/api/miniconf/users/22145?format=json", "institution": "University of Southern California"}], "abstract": "In this paper, we provide a novel perspective on the underlying structure of real-world data with ground-truth clustering via characterization of an abundantly observed yet often overlooked *density\u2013geometry* correlation.  We leverage this correlation to design CoreSPECT (Core Space Projection based Enhancement of Clustering Techniques), a general framework that improves the performance of generic clustering algorithms. Our framework boosts the performance of clustering algorithms by applying them to strategically selected regions, then extending the partial partition to a complete partition for the dataset using a novel neighborhood graph based multi-layer propagation procedure.   We provide initial theoretical support of the functionality of our framework under the assumption of our model, and then provide large-scale real-world experiments on 20 datasets that include standard image datasets as well as genomics datasets.   We observe two notable improvements. First, CoreSPECT improves the NMI of K-Means by 20 % on average, making it competitive (and in some cases surpassing) the state-of-the art manifold-based clustering algorithms, while being orders of magnitude faster.    Secondly, our framework boosts the NMI of HDBSCAN by more than 100 % on average, making it competitive to the state-of-the-art in several cases *without requiring the true number of clusters and hyper-parameter tuning*. The overall ARI improvements are higher.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13458", "url": null, "sourceid": 2174, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=kS3IE6cexb", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11151, "modified": "2026-03-29T20:42:59.709307-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=kS3IE6cexb", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "46", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13808, "uid": "211c1e0b83b9c69fa9c4bdede203c1e3", "name": "Deep Feedback Models", "authors": [{"id": 22909, "fullname": "David Calhas", "url": "http://virtual.aistats.org/api/miniconf/users/22909?format=json", "institution": "INESC-ID VAT PT 504547593"}, {"id": 22910, "fullname": "Arlindo Oliveira", "url": "http://virtual.aistats.org/api/miniconf/users/22910?format=json", "institution": "Instituto Superior T\u00e9cnico"}], "abstract": "Deep Feedback Models (DFMs) are a new class of stateful neural networks that combine bottom up input with high level representations over time. This feedback mechanism introduces dynamics into otherwise static architectures, enabling DFMs to iteratively refine their internal state and mimic aspects of biological decision making. We model this process as a differential equation solved through a recurrent neural network, stabilized via exponential decay to ensure convergence. To evaluate their effectiveness, we measure DFMs under two key conditions: robustness to noise and generalization with limited data. In both object recognition and segmentation tasks, DFMs consistently outperform their feedforward counterparts, particularly in low data or high noise regimes. In addition, DFMs translate to medical imaging settings, while being robust against various types of noise corruption. These findings highlight the importance of feedback in achieving stable, robust, and generalizable learning. Code is available at github.com/DCalhas/deep_feedback_models.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13808", "url": null, "sourceid": 2175, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=A7Vk9WWlU5", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11501, "modified": "2026-03-29T20:43:13.879269-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=A7Vk9WWlU5", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "48", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13425, "uid": "208e43f0e45c4c78cafadb83d2888cb6", "name": "Optimistic Actor-Critic with Parametric Policies for Linear Markov Decision Processes", "authors": [{"id": 22064, "fullname": "Max Lin", "url": "http://virtual.aistats.org/api/miniconf/users/22064?format=json", "institution": "Simon Fraser University"}, {"id": 14707, "fullname": "Reza Asad", "url": "http://virtual.aistats.org/api/miniconf/users/14707?format=json", "institution": "Simon Fraser University"}, {"id": 22065, "fullname": "Kevin Tan", "url": "http://virtual.aistats.org/api/miniconf/users/22065?format=json", "institution": "Wharton Statistics Department, The Wharton School"}, {"id": 22066, "fullname": "Haque Ishfaq", "url": "http://virtual.aistats.org/api/miniconf/users/22066?format=json", "institution": "McGill University, McGill University"}, {"id": 13560, "fullname": "Csaba Szepesvari", "url": "http://virtual.aistats.org/api/miniconf/users/13560?format=json", "institution": "University of Alberta"}, {"id": 698, "fullname": "Sharan Vaswani", "url": "http://virtual.aistats.org/api/miniconf/users/698?format=json", "institution": "Simon Fraser University"}], "abstract": "Although actor-critic methods have been successful in practice, their theoretical analyses have several limitations. Specifically, existing theoretical work either sidesteps the exploration problem by making strong assumptions or analyzes impractical methods with complicated algorithmic modifications. Moreover, the actor-critic methods analyzed for linear MDPs often employ natural policy gradient (NPG) and construct \"implicit'\" policies without explicit parameterization. Such policies are computationally expensive to sample from, making the environment interactions inefficient. To that end, we focus on the finite-horizon linear MDPs and propose an optimistic actor-critic framework that uses parametric log-linear policies. In particular, we introduce a tractable \\textit{logit-matching} regression objective for the actor. For the critic, we use approximate Thompson sampling via Langevin Monte Carlo to obtain optimistic value estimates. We prove that the resulting algorithm achieves $\\widetilde{\\mathcal{O}}(\\epsilon^{-4})$ and $\\widetilde{\\mathcal{O}}(\\epsilon^{-2})$ sample complexity in the on-policy and off-policy setting, respectively. Our results match prior theoretical works in achieving the state-of-the-art sample complexity, while our algorithm is more aligned with practice.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13425", "url": null, "sourceid": 1143, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=npMh82yS0S", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11118, "modified": "2026-03-29T20:42:58.327285-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=npMh82yS0S", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "123", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13427, "uid": "5ea1649a31336092c05438df996a3e59", "name": "Beyond Binary Out of Distribution Detection: Characterizing Distributional Shifts with Multi-Statistic Diffusion Trajectories", "authors": [{"id": 22069, "fullname": "Achref Jaziri", "url": "http://virtual.aistats.org/api/miniconf/users/22069?format=json", "institution": "Johann Wolfgang Goethe Universit\u00e4t Frankfurt am Main"}, {"id": 22070, "fullname": "Martin Rogmann", "url": "http://virtual.aistats.org/api/miniconf/users/22070?format=json", "institution": "Johann Wolfgang Goethe Universit\u00e4t Frankfurt am Main"}, {"id": 22071, "fullname": "Martin Mundt", "url": "http://virtual.aistats.org/api/miniconf/users/22071?format=json", "institution": "TU Darmstadt &amp; hessian.AI"}, {"id": 22072, "fullname": "Visvanathan Ramesh", "url": "http://virtual.aistats.org/api/miniconf/users/22072?format=json", "institution": "Goethe University"}], "abstract": "Detecting out-of-distribution (OOD) data is critical for machine learning, be it for safety reasons or to enable open-ended learning. However, beyond mere detection, choosing an appropriate course of action typically hinges on the type of OOD data encountered. Unfortunately, the latter is generally not distinguished in practice, as modern OOD detection methods collapse distributional shifts into single scalar outlier scores. This work argues that scalar-based methods are thus insufficient for OOD data to be properly contextualized and prospectively exploited, a limitation we overcome with the introduction of DISC: Diffusion-based Statistical Characterization. DISC leverages the iterative denoising process of diffusion models to extract a rich, multi-dimensional feature vector that captures statistical discrepancies across multiple noise levels. Extensive experiments on image and tabular benchmarks show that DISC matches or surpasses state-of-the-art detectors for OOD detection and, crucially, also classifies OOD type, a capability largely absent from prior work. As such, our work enables a shift from simple binary OOD detection to a more granular detection.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13427", "url": null, "sourceid": 537, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=nPooplHfR4", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11120, "modified": "2026-03-29T20:42:58.383533-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=nPooplHfR4", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "22", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13421, "uid": "934b535800b1cba8f96a5d72f72f1611", "name": "A Polynomial-Time Approximation for Pairwise Fair $k$-Median Clustering", "authors": [{"id": 19767, "fullname": "Sayan Bandyapadhyay", "url": "http://virtual.aistats.org/api/miniconf/users/19767?format=json", "institution": "Port"}, {"id": 22050, "fullname": "Eden Chlamt\u00e1\u010d", "url": "http://virtual.aistats.org/api/miniconf/users/22050?format=json", "institution": ", Ben Gurion University of the Negev"}, {"id": 22051, "fullname": "Zachary Friggstad", "url": "http://virtual.aistats.org/api/miniconf/users/22051?format=json", "institution": "University of Alberta"}, {"id": 22052, "fullname": "Mahya Jamshidian", "url": "http://virtual.aistats.org/api/miniconf/users/22052?format=json", "institution": "University of Alberta"}, {"id": 22053, "fullname": "Yury Makarychev", "url": "http://virtual.aistats.org/api/miniconf/users/22053?format=json", "institution": "University of Chicago"}, {"id": 4448, "fullname": "Ali Vakilian", "url": "http://virtual.aistats.org/api/miniconf/users/4448?format=json", "institution": "Virginia Tech"}], "abstract": "In this work, we study pairwise fair $k$-Median with $\\ell \\ge 2$ groups, where for every cluster $C$ and every group $i \\in [\\ell]$, the number of points in $C$ from group $i$  must be at most $t$ times the number of points in $C$ from any other group $j \\in [\\ell]$, for a given integer $t$. Only bi-criteria approximation and exponential-time algorithms follow for this problem from the prior work on fair clustering problems when $\\ell > 2$.   We present the first polynomial-time $O(k^2\\cdot \\ell \\cdot t)$-approximation for this problem that does not violate the fairness constraints. We also implemented our algorithm on a variety of datasets to test the \"price of fairness\" achieved by our approach in real data, which turned out to be significantly smaller than the theoretical guarantee.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13421", "url": null, "sourceid": 2222, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=oT9q9wfnAh", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11114, "modified": "2026-03-29T20:42:58.215424-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=oT9q9wfnAh", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "6", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13865, "uid": "cc7e2b878868cbae992d1fb743995d8f", "name": "Generalization Bounds under Heavy-Tailed Losses", "authors": [{"id": 5701, "fullname": "Gholamali Aminian", "url": "http://virtual.aistats.org/api/miniconf/users/5701?format=json", "institution": "The Alan Turing Institute"}], "abstract": "The generalization error of a supervised statistical learning algorithm, defined as the difference between the population risk and the empirical risk, quantifies its  ability to predict performance on previously unseen data. In this work, we analyze the generalization error under the heavy-tailed assumption on the loss function with respect to the data-generating distribution. Specifically, we derive uniform, information-theoretic, and PAC-Bayesian bounds on the generalization error under the assumption that the $(1+\\epsilon)$-th moment of the loss function is bounded for $\\epsilon\\in(0,1]$. The generalization error is shown to have a convergence rate of $O(n^{-\\epsilon/(1+\\epsilon)})$ where $n$ is the number of training samples. Furthermore, we apply our results to study the generalization error of the Gibbs posterior and noisy iterative learning algorithms under the heavy-tailed assumption.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13865", "url": null, "sourceid": 2278, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=5HbgyFDDDt", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11558, "modified": "2026-03-29T20:43:16.461293-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=5HbgyFDDDt", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "76", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13911, "uid": "7e3b7a5bafcb0fa8e8dfe3ea6aca9186", "name": "From Decision to Acquisition: Loss-Driven Bayesian Active Learning", "authors": [{"id": 19595, "fullname": "Zhuoyue Huang", "url": "http://virtual.aistats.org/api/miniconf/users/19595?format=json", "institution": "University of Oxford"}, {"id": 10274, "fullname": "Freddie Bickford Smith", "url": "http://virtual.aistats.org/api/miniconf/users/10274?format=json", "institution": "University of Oxford"}, {"id": 1406, "fullname": "Tom Rainforth", "url": "http://virtual.aistats.org/api/miniconf/users/1406?format=json", "institution": "University of Oxford"}], "abstract": "Bayesian active learning has become practically synonymous with maximising expected information gain (EIG) during data acquisition. We highlight that this standard practice makes an implicit assumption about the underlying decision problem of interest, with this assumption not reflecting our goals in many cases. Generalising EIG maximisation, we propose an explicitly loss-driven approach to Bayesian active learning, with which we can target reduced loss in a much broader class of decision problems. We also identify a large family of losses for which we can derive efficient estimators of our principled data-acquisition objective. In practical classification and regression demonstrations, we find our approach can produce a notable performance boost over existing techniques.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13911", "url": null, "sourceid": 2334, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=0B1RuEcUph", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11604, "modified": "2026-03-29T20:43:18.276546-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=0B1RuEcUph", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "72", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13486, "uid": "8208974663db80265e9bfe7b222dcb18", "name": "Fundamental limits for weighted empirical approximations of exponentially tilted distributions", "authors": [{"id": 19693, "fullname": "Sarvesh Iyer", "url": "http://virtual.aistats.org/api/miniconf/users/19693?format=json", "institution": "Ashoka University"}, {"id": 22209, "fullname": "Himadri Mandal", "url": "http://virtual.aistats.org/api/miniconf/users/22209?format=json", "institution": "Indian Statistical Institute, Kolkata"}, {"id": 22210, "fullname": "Dhruman Gupta", "url": "http://virtual.aistats.org/api/miniconf/users/22210?format=json", "institution": "Ashoka University"}, {"id": 22211, "fullname": "Rushil Gupta", "url": "http://virtual.aistats.org/api/miniconf/users/22211?format=json", "institution": "Ashoka University"}, {"id": 22212, "fullname": "Agniv Bandyopadhyay", "url": "http://virtual.aistats.org/api/miniconf/users/22212?format=json", "institution": "Tata Institute of Fundamental Research"}, {"id": 22213, "fullname": "Achal Bassamboo", "url": "http://virtual.aistats.org/api/miniconf/users/22213?format=json", "institution": "Northwestern University"}, {"id": 22214, "fullname": "Sandeep Juneja", "url": "http://virtual.aistats.org/api/miniconf/users/22214?format=json", "institution": "Ashoka University"}, {"id": 22215, "fullname": "Varun Gupta", "url": "http://virtual.aistats.org/api/miniconf/users/22215?format=json", "institution": "University of Utah"}], "abstract": "Generating samples from exponentially tilting a given distribution of random vectors when samples from the given distribution are available finds applications in fields such as finance and climate science and in the broad area of rare event simulation. In this article, we discuss the asymptotic efficiency of an estimator obtained by exponentially tilting the empirical distribution. We provide a sharp characterization of how much one can accurately tilt distributions given a certain number of samples. Our findings reveal a surprising dichotomy: While twisting unbounded distributions is a fundamentally hard task, for bounded distributions, one can accurately tilt by a large amount using much fewer samples.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13486", "url": null, "sourceid": 2404, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=gmmtcjRs0O", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11179, "modified": "2026-03-29T20:43:00.778857-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=gmmtcjRs0O", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "75", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13529, "uid": "e3ca0449fa2ea7701a7ac53fb719c51a", "name": "TESLA: Taylor Expansion of Sinusoidal Learnable Activations", "authors": [{"id": 19576, "fullname": "Daehwa Ko", "url": "http://virtual.aistats.org/api/miniconf/users/19576?format=json", "institution": "Korea Aerospace University"}, {"id": 22296, "fullname": "Jay Hoon Jung", "url": "http://virtual.aistats.org/api/miniconf/users/22296?format=json", "institution": "Korea Aerospace University"}, {"id": 22297, "fullname": "SeungHyun Ham", "url": "http://virtual.aistats.org/api/miniconf/users/22297?format=json", "institution": "Korea Aerospace University"}, {"id": 22298, "fullname": "JaeHyeon Kim", "url": "http://virtual.aistats.org/api/miniconf/users/22298?format=json", "institution": "Korea Aerospace University"}], "abstract": "The parity problem\u2014deciding whether the number of ones in a binary vector is odd or even\u2014remains challenging for standard neural networks due to linear inseparability and the need for global interactions. We propose TESLA, an activation defined as a learnable combination of sine and cosine terms whose Taylor coefficients are trained directly, enabling explicit control over polynomial degree and selective amplification of high-order components. Theoretically, we show that constraining TESLA\u2019s coefficients yields Lipschitz/Rademacher complexity bounds and shapes the training dynamics to emphasize higher-frequency structure. Empirically, on parity with input length $n=32$, TESLA achieve perfect generalization using ~100K training sample ($\\approx 0.002\\%$ of the $2^{32}$ input space). Notably, TESLA maintains strong generalization under heavy corruption, retaining high accuracy with up to 30\\% label noise in the parity signals. Beyond synthetic structure, TESLA delivers comparable performance on ImageNet-100, indicating that activation-level degree control transfers to more general vision workloads. These findings suggest that TESLA is an effective mechanism to improve expressivity and sample efficiency on tasks requiring global structure.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13529", "url": null, "sourceid": 2415, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=cMQ9GE4msQ", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11222, "modified": "2026-03-29T20:43:02.395281-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=cMQ9GE4msQ", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "173", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13435, "uid": "7eb7eabbe9bd03c2fc99881d04da9cbd", "name": "Adaptive Diffusion Guidance via Stochastic Optimal Control", "authors": [{"id": 22095, "fullname": "Iskander Azangulov", "url": "http://virtual.aistats.org/api/miniconf/users/22095?format=json", "institution": "Oxofrd, University of Oxford"}, {"id": 22096, "fullname": "Peter Potaptchik", "url": "http://virtual.aistats.org/api/miniconf/users/22096?format=json", "institution": "University of Oxford"}, {"id": 22097, "fullname": "Qinyu Li", "url": "http://virtual.aistats.org/api/miniconf/users/22097?format=json", "institution": "University of Oxford"}, {"id": 22098, "fullname": "Eddie Aamari", "url": "http://virtual.aistats.org/api/miniconf/users/22098?format=json", "institution": "CNRS - \u00c9cole Normale Sup\u00e9rieure PSL"}, {"id": 704, "fullname": "George Deligiannidis", "url": "http://virtual.aistats.org/api/miniconf/users/704?format=json", "institution": "Oxford"}, {"id": 22099, "fullname": "Judith Rousseau", "url": "http://virtual.aistats.org/api/miniconf/users/22099?format=json", "institution": "Universit\u00e9 Paris-Dauphine"}], "abstract": "Classifier-Free Guidance (CFG) is a cornerstone of modern diffusion models, playing a pivotal role in conditional generation and enhancing the quality of unconditional samples. However, current approaches to CFG scheduling\u2014determining the appropriate guidance weight\u2014are largely heuristic and lack a solid theoretical foundation. This work addresses these limitations on two fronts. First, we provide a theoretical formalization that precisely characterizes the relationship between guidance strength and classifier confidence. Second, building on this insight, we introduce a stochastic optimal control framework that casts CFG scheduling as an adaptive optimization problem. In this formulation, guidance strength is not fixed but dynamically selected based on time, the current sample, and the conditioning class, either independently or in combination. By solving the resulting control problem, we establish a principled foundation for more effective guidance in diffusion models.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13435", "url": null, "sourceid": 1769, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=mef54yiI1L", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11128, "modified": "2026-03-29T20:42:58.804022-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=mef54yiI1L", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "15", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13436, "uid": "e3796ae838835da0b6f6ea37bcf8bcb7", "name": "An Indicator of Membership Inference Security in Post-Training Quantized Models", "authors": [{"id": 22100, "fullname": "Eric AUBINAIS", "url": "http://virtual.aistats.org/api/miniconf/users/22100?format=json", "institution": "Universit\u00e9 Paris-Saclay"}, {"id": 22101, "fullname": "Philippe Formont", "url": "http://virtual.aistats.org/api/miniconf/users/22101?format=json", "institution": "\u00c9cole de technologie sup\u00e9rieure, Universit\u00e9 du Qu\u00e9bec"}, {"id": 19338, "fullname": "Pablo Piantanida", "url": "http://virtual.aistats.org/api/miniconf/users/19338?format=json", "institution": "MILA &amp; ILLS, CNRS Universit\u00e9 Paris-Saclay"}, {"id": 22102, "fullname": "Elisabeth Gassiat", "url": "http://virtual.aistats.org/api/miniconf/users/22102?format=json", "institution": "Universit\u00e9 Paris-Saclay"}], "abstract": "Quantizing machine learning models has demonstrated its effectiveness in lowering memory and inference costs while maintaining performance levels comparable to those of the original models. In this work, we investigate the impact of quantization procedures on privacy in data-driven models, focusing on their vulnerability to membership inference attacks. Membership Inference Security (MIS) has recently been proposed to characterize the privacy of machine learning models against the most powerful (and possibly unknown) attacks. However, quantifying MIS appears to be computationally very difficult. In this paper, we propose a new MIS indicator for post-training quantization procedures of machine learning models that minimize an empirical loss. This new indicator is a byproduct of a theoretical asymptotic analysis of the MIS in this context. We also present a methodology for empirically estimating our MIS indicator. Using synthetic datasets and real-world data (in the context of drug discovery), we demonstrate the effectiveness of our approach in assessing and ranking the MIS of different quantizers.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13436", "url": null, "sourceid": 281, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=me03yxfOH7", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11129, "modified": "2026-03-29T20:42:58.840432-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=me03yxfOH7", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "17", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13437, "uid": "b440509a0106086a67bc2ea9df0a1dab", "name": "Differentially Private and Federated Structure Learning in Bayesian Networks", "authors": [{"id": 22103, "fullname": "Ghita Fassy El Fehri", "url": "http://virtual.aistats.org/api/miniconf/users/22103?format=json", "institution": "INRIA"}, {"id": 535, "fullname": "Aur\u00e9lien Bellet", "url": "http://virtual.aistats.org/api/miniconf/users/535?format=json", "institution": "INRIA"}, {"id": 22104, "fullname": "Philippe Bastien", "url": "http://virtual.aistats.org/api/miniconf/users/22104?format=json", "institution": "L&#x27;Or\u00e9al"}], "abstract": "Learning the structure of a Bayesian network from decentralized data poses two major challenges: (i) ensuring rigorous privacy guarantees for participants, and (ii) avoiding communication costs that scale poorly with dimensionality. In this work, we introduce Fed-BNSL, a novel federated method for learning linear Gaussian Bayesian network structures that addresses both challenges. By combining differential privacy with greedy updates that target only a few relevant edges per participant, Fed-BNSL efficiently uses the privacy budget while keeping communication costs low. Our careful algorithmic design preserves model identifiability and enables accurate structure estimation. Experiments on synthetic and real datasets demonstrate that Fed-BNSL achieves utility close to non-private baselines while offering substantially stronger privacy and communication efficiency.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13437", "url": null, "sourceid": 2188, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=ma9gVnospd", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11130, "modified": "2026-03-29T20:42:58.874918-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=ma9gVnospd", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "52", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13438, "uid": "7810ccd41bf26faaa2c4e1f20db70a71", "name": "GL-LowPopArt: A Nearly Instance-Wise Minimax Estimator for Generalized Low-Rank Trace Regression", "authors": [{"id": 1869, "fullname": "Junghyun Lee", "url": "http://virtual.aistats.org/api/miniconf/users/1869?format=json", "institution": "KAIST"}, {"id": 5305, "fullname": "Kyoungseok Jang", "url": "http://virtual.aistats.org/api/miniconf/users/5305?format=json", "institution": "Chung-Ang University"}, {"id": 22105, "fullname": "Kwang-Sung Jun", "url": "http://virtual.aistats.org/api/miniconf/users/22105?format=json", "institution": "Pohang University of Science and Technology"}, {"id": 22106, "fullname": "Milan Vojnovic", "url": "http://virtual.aistats.org/api/miniconf/users/22106?format=json", "institution": "London School of Economics"}, {"id": 9815, "fullname": "Se-Young Yun", "url": "http://virtual.aistats.org/api/miniconf/users/9815?format=json", "institution": "KAIST"}], "abstract": "We present **GL-LowPopArt**, a novel Catoni-style estimator for generalized low-rank trace regression. Building on *LowPopArt* (Jang et al., 2024), it employs a two-stage approach: nuclear norm regularization followed by matrix Catoni estimation. We establish state-of-the-art estimation error bounds, surpassing existing guarantees (Fan et al., 2019; Kang et al., 2022), and reveal a novel experimental design objective, **GL(\u03c0)**. The key technical challenge is controlling bias from the nonlinear inverse link function, which we address with our two-stage approach. We prove a *local minimax lower bound*, showing that **GL-LowPopArt** enjoys instance-wise optimality up to the condition number of the ground-truth Hessian. Our method immediately achieves an improved Frobenius error guarantee for generalized linear matrix completion. We also introduce a new problem setting, **bilinear dueling bandits**, a contextualized version of dueling bandits with a general preference model. Using an explore-then-commit strategy with **GL-LowPopArt**, we demonstrate an improved Borda regret bound over na\u00efve vectorization (Wu et al., 2024).", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13438", "url": null, "sourceid": 1272, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=mMxiIP3Gxu", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11131, "modified": "2026-03-29T20:42:58.908609-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=mMxiIP3Gxu", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "68", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13443, "uid": "fa83a11a198d5a7f0bf77a1987bcd006", "name": "Variational inference via radial transport", "authors": [{"id": 20638, "fullname": "Luca Ghafourpour", "url": "http://virtual.aistats.org/api/miniconf/users/20638?format=json", "institution": "University of Cambridge"}, {"id": 22116, "fullname": "Sinho Chewi", "url": "http://virtual.aistats.org/api/miniconf/users/22116?format=json", "institution": "Yale University"}, {"id": 22117, "fullname": "Alessio Figalli", "url": "http://virtual.aistats.org/api/miniconf/users/22117?format=json", "institution": "ETHZ - ETH Zurich"}, {"id": 22118, "fullname": "Aram-Alexandre Pooladian", "url": "http://virtual.aistats.org/api/miniconf/users/22118?format=json", "institution": "Yale University"}], "abstract": "In variational inference (VI), the practitioner approximates a high-dimensional distribution $\\pi$ with a simple surrogate one, often a (product) Gaussian distribution. However, in many cases of practical interest, Gaussian distributions might not capture the correct radial profile of $\\pi$, resulting in poor coverage. In this work, we approach the VI problem from the perspective of optimizing over these radial profiles. Our algorithm $\\texttt{radVI}$ is a cheap, effective add-on to many existing VI schemes, such as Gaussian (mean-field) VI and Laplace approximation. We provide theoretical convergence guarantees for our algorithm, owing to recent developments in optimization over the Wasserstein space\u2014the space of probability distributions endowed with the Wasserstein distance\u2014and new regularity properties of radial transport maps in the style of Caffarelli (2000).", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13443", "url": null, "sourceid": 840, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=lomAtGpQ6n", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11136, "modified": "2026-03-29T20:42:59.124220-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=lomAtGpQ6n", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "179", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13447, "uid": "b056eb1587586b71e2da9acfe4fbd19e", "name": "Sharp risk bounds for early-stopping in Gaussian linear regression", "authors": [{"id": 18378, "fullname": "Tobias Wegel", "url": "http://virtual.aistats.org/api/miniconf/users/18378?format=json", "institution": "ETHZ - ETH Zurich"}, {"id": 18034, "fullname": "Gil Kur", "url": "http://virtual.aistats.org/api/miniconf/users/18034?format=json", "institution": "Department of Computer Science, ETHZ - ETH Zurich"}, {"id": 593, "fullname": "Patrick Rebeschini", "url": "http://virtual.aistats.org/api/miniconf/users/593?format=json", "institution": "University of Oxford"}], "abstract": "We study early-stopped mirror descent (ESMD) for high-dimensional Gaussian linear regression over arbitrary convex bodies and design matrices, where the task is to minimize the in-sample mean squared error. Our main result shows that some of the sharpest risk bounds for the least squares estimator (LSE), based on the local Gaussian width, extend to ESMD. We derive sufficient conditions on the potential, expressed via the Minkowski functional, under which our result holds. These conditions allow us to construct new potentials and analyze existing ones. Our results then yield general sufficient conditions for minimax optimality of ESMD, provide a systematic comparison with the LSE, and establish the tightest known risk bound in the $\\ell_1$-constrained setting.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13447", "url": null, "sourceid": 749, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=lbOy4AHvWF", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11140, "modified": "2026-03-29T20:42:59.288302-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=lbOy4AHvWF", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "149", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13448, "uid": "37a749d808e46495a8da1e5352d03cae", "name": "Dashed Line Defense: Plug-And-Play Defense Against Adaptive Score-Based Query Attacks", "authors": [{"id": 19814, "fullname": "Yanzhang Fu", "url": "http://virtual.aistats.org/api/miniconf/users/19814?format=json", "institution": "Harbin Institute of Technology"}, {"id": 22127, "fullname": "Jizhou Luo", "url": "http://virtual.aistats.org/api/miniconf/users/22127?format=json", "institution": "Harbin Institute of Technology"}, {"id": 22126, "fullname": "Zizheng Guo", "url": "http://virtual.aistats.org/api/miniconf/users/22126?format=json", "institution": "Harbin Institute of Technology"}], "abstract": "Score-based query attacks pose a serious threat to deep learning models by crafting adversarial examples (AEs) using only black-box access to model output scores, iteratively optimizing inputs based on observed loss values. While recent runtime defenses attempt to disrupt this process via output perturbation, most either require access to model parameters or fail when attackers adapt their tactics. In this paper, we first reveal that even the state-of-the-art plug-and-play defense can be bypassed by adaptive attacks, exposing a critical limitation of existing runtime defenses. We then propose Dashed Line Defense (DLD), a plug-and-play post-processing method specifically designed to withstand adaptive query strategies. By introducing ambiguity in how the observed loss reflects the true adversarial strength of candidate examples, DLD prevents attackers from reliably analyzing and adapting their queries, effectively disrupting the AE generation process. We provide theoretical guarantees of DLD\u2019s defense capability and validate its effectiveness through experiments on ImageNet, demonstrating that DLD consistently outperforms prior defenses\u2014even under worst-case adaptive attacks\u2014while preserving the model\u2019s predicted labels.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13448", "url": null, "sourceid": 152, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=lYqi9b9i6o", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11141, "modified": "2026-03-29T20:42:59.325207-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=lYqi9b9i6o", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "42", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13451, "uid": "f47330643ae134ca204bf6b2481fec47", "name": "ECAI: Efficient Convolution Activation Inversion for Constant-Memory Convolutional Neural Networks Training", "authors": [{"id": 19782, "fullname": "Changhyeon Lee", "url": "http://virtual.aistats.org/api/miniconf/users/19782?format=json", "institution": "UNIST"}, {"id": 22132, "fullname": "Seulki Lee", "url": "http://virtual.aistats.org/api/miniconf/users/22132?format=json", "institution": "KAIST (Korea Advanced Institute of Science &amp; Technology)"}], "abstract": "We propose a novel approach that achieves constant activation memory usage during the training of convolutional neural networks (CNNs), addressing a key memory bottleneck in the backward pass. By reconstructing activations required for gradient matrix calculation through the proposed efficient convolution activation inversion (ECAI) rather than storing them in memory during forward pass, it becomes possible to maintain constant activation memory usage across convolution layers. We formulate the activation inversion problem as a set of $n$ systems of linear equations derived from forward convolution operations, and solve them with an accelerated method that achieves $\\mathcal{O}(n^2)$ complexity. The proposed approach enables memory-constrained mobile, edge, and embedded devices to perform CNN training without a growth of activation memory over the model capacity while also enhancing training memory efficiency for large-sized images on commercial GPUs. The experimental results demonstrate that the proposed approach maintains constant activation memory by reusing a fixed memory space, improving memory efficiency without degradation in model accuracy. The memory savings achieved by the proposed method increase when using more convolution layers, potentially achieving near-zero activation (e.g., $30\\times$ or more activation memory reduction in specific setups). The code implementation is available at an anonymous GitHub.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13451", "url": null, "sourceid": 1182, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=l6VKhcivei", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11144, "modified": "2026-03-29T20:42:59.461589-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=l6VKhcivei", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "42", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13456, "uid": "934815ad542a4a7c5e8a2dfa04fea9f5", "name": "LatticeVision: Image to Image Networks for Modeling Non-Stationary Spatial Data", "authors": [{"id": 19863, "fullname": "Antony Sikorski", "url": "http://virtual.aistats.org/api/miniconf/users/19863?format=json", "institution": "Colorado School of Mines"}, {"id": 22141, "fullname": "Michael Ivanitskiy", "url": "http://virtual.aistats.org/api/miniconf/users/22141?format=json", "institution": "Colorado School of Mines"}, {"id": 22142, "fullname": "Nathan Lenssen", "url": "http://virtual.aistats.org/api/miniconf/users/22142?format=json", "institution": "Colorado School of Mines &amp; NSF National Center for Atmospheric Research"}, {"id": 20604, "fullname": "Douglas Nychka", "url": "http://virtual.aistats.org/api/miniconf/users/20604?format=json", "institution": "Colorado School of Mines"}, {"id": 22143, "fullname": "Daniel McKenzie", "url": "http://virtual.aistats.org/api/miniconf/users/22143?format=json", "institution": "Colorado School of Mines"}], "abstract": "In many applications, we wish to fit a parametric statistical model to a small ensemble of spatially distributed random variables ('fields'). However, parameter inference using maximum likelihood estimation (MLE) is computationally prohibitive, especially for large, non-stationary fields. Thus, many recent works train neural networks to estimate parameters given spatial fields as input, sidestepping MLE completely. In this work we focus on a popular class of parametric, spatially autoregressive (SAR) models. We make a simple yet impactful observation; because the SAR parameters can be arranged on a regular grid, both inputs (spatial fields) and outputs (model parameters) can be viewed as images. Using this insight, we demonstrate that image-to-image (I2I) networks enable faster and more accurate parameter estimation for a class of non-stationary SAR models with unprecedented complexity.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13456", "url": null, "sourceid": 994, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=km8KA33TUu", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11149, "modified": "2026-03-29T20:42:59.639922-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=km8KA33TUu", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "88", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13457, "uid": "e5841df2166dd424a57127423d276bbe", "name": "SPIRE: Conditional Personalization for Federated Diffusion Generative Models", "authors": [{"id": 17682, "fullname": "Kaan Ozkara", "url": "http://virtual.aistats.org/api/miniconf/users/17682?format=json", "institution": "University of California, Los Angeles"}, {"id": 12997, "fullname": "Ruida Zhou", "url": "http://virtual.aistats.org/api/miniconf/users/12997?format=json", "institution": "Texas A&amp;M University"}, {"id": 17701, "fullname": "Suhas Diggavi", "url": "http://virtual.aistats.org/api/miniconf/users/17701?format=json", "institution": "University of California, Los Angeles"}], "abstract": "Two defining characteristics of federated learning (FL) client data are distributional heterogeneity and small local sample sizes. These properties necessitate data efficient, and client specific adaptation rather than a one-size-fits-all model. Recent advances in diffusion models have revolutionized generative AI. However, their scale is too large for straightforward fine-tuning; making personalization difficult. To enable personalized diffusion generative models, we propose Shared\u2011backbone Personal Identity Representation Embeddings (SPIRE), a framework that casts per\u2011client diffusion based generation as conditional generation in FL. SPIRE factorizes the network into (i) a high\u2011capacity global backbone that learns a population\u2011level score function and (ii) lightweight, learnable client embeddings that encode local data statistics. This separation enables parameter\u2011efficient fine\u2011tuning that touches $<0.01\\%$ of weights. We provide the first theoretical bridge between conditional diffusion training and maximum\u2011likelihood estimation in Gaussian\u2011mixture models. For a two\u2011component mixture we prove that gradient descent on the DDPM with respect to mixing weights loss recovers the optimal mixing weights and enjoys dimension\u2011free error bounds. Our analysis also hints at how client embeddings act as biases that steer a shared score network toward personalized distributions. Empirically, SPIRE matches or surpasses strong baselines during collaborative pre\u2011training, and vastly outperforms them when adapting to unseen/new clients\u2014reducing Kernel Inception Distance while updating only hundreds of parameters. SPIRE further mitigates catastrophic forgetting and remains robust across fine-tuning learning\u2011rate and epoch choices.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13457", "url": null, "sourceid": 700, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=kgFbRCzqM9", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11150, "modified": "2026-03-29T20:42:59.675344-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=kgFbRCzqM9", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "174", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13459, "uid": "fba9d88164f3e2d9109ee770223212a0", "name": "On the Misinformation in a Statistical Experiment", "authors": [{"id": 19755, "fullname": "Jake Callahan", "url": "http://virtual.aistats.org/api/miniconf/users/19755?format=json", "institution": "The University of Arizona"}, {"id": 22146, "fullname": "Tommie Catanach", "url": "http://virtual.aistats.org/api/miniconf/users/22146?format=json", "institution": "Sandia National Laboratories"}], "abstract": "The principle that more informative experiments are always better is a cornerstone of Bayesian experimental design. This principle assumes the practitioner's model and inference are correct. In practice, both the data-generating model and the inferential approximation are inevitably misspecified, and we show that under these conditions the classical framework for comparing experiments breaks down. Designs ranked as most informative can become actively harmful, amplifying bias to produce confident but incorrect inferences. We demonstrate that the commonly-accepted axioms of experimental utility, such as Blackwell monotonicity, fail under misspecification, and that information measures proposed to handle it, like the Expected Generalized Information Gain (EGIG), do not obey these axioms.  To resolve this, we propose a generalized axiomatic framework for robust Bayesian experimental design. We prove that EGIG satisfies our axioms as a criterion that penalizes inferential error, providing a principled foundation for its use in Bayesian experimental design. As a complementary approach, we derive a new measure that instead penalizes model error. Finally, we demonstrate our framework's utility across common modes of misspecification, showing it provides a reliable guide for experimental design where classical methods fail.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13459", "url": null, "sourceid": 1002, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=kO66LSLmB8", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11152, "modified": "2026-03-29T20:42:59.741051-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=kO66LSLmB8", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "117", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13460, "uid": "fe73f687e5bc5280214e0486b273a5f9", "name": "Optimal rates for density and mode estimation with expand-and-sparsify representations", "authors": [{"id": 22147, "fullname": "Kaushik Sinha", "url": "http://virtual.aistats.org/api/miniconf/users/22147?format=json", "institution": "Wichita State University"}, {"id": 20609, "fullname": "Christopher Tosh", "url": "http://virtual.aistats.org/api/miniconf/users/20609?format=json", "institution": "Memorial Sloan Kettering Cancer Center"}], "abstract": "Expand-and-sparsify representations are a class of theoretical models that capture sparse  representation phenomena observed in the sensory systems of many animals. At a high level, these representations map an input $x \\in \\mathbb{R}^d$ to a much higher dimension $m \\gg d$ via random linear projections before zeroing out all but the $k \\ll m$ largest entries. The result is a $k$-sparse vector in $\\\\{0,1\\\\}^m$. We study the suitability of this representation for two fundamental statistical problems: density estimation and mode estimation. For density estimation, we show that a simple linear function of the expand-and-sparsify representation produces an estimator with  minimax-optimal $\\ell_{\\infty}$ convergence rates. In mode estimation, we provide simple algorithms on top of our density estimator that recover single or multiple modes at optimal rates up to logarithmic factors under mild conditions.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13460", "url": null, "sourceid": 330, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=kHuqnMOJwD", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11153, "modified": "2026-03-29T20:42:59.776671-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=kHuqnMOJwD", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "131", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13462, "uid": "0188e8b8b014829e2fa0f430f0a95961", "name": "Optimal Query Allocation in Extractive QA with LLMs: A Learning-to-Defer Framework with Theoretical Guarantees", "authors": [{"id": 22148, "fullname": "Yannis Montreuil", "url": "http://virtual.aistats.org/api/miniconf/users/22148?format=json", "institution": "National University of Singapore"}, {"id": 22149, "fullname": "Yeo Heng", "url": "http://virtual.aistats.org/api/miniconf/users/22149?format=json", "institution": "National University of Singapore"}, {"id": 22151, "fullname": "Axel Carlier", "url": "http://virtual.aistats.org/api/miniconf/users/22151?format=json", "institution": "Institut Sup\u00e9rieur de l&#x27;A\u00e9ronautique et de l&#x27;Espace"}, {"id": 22150, "fullname": "Lai Xing Ng", "url": "http://virtual.aistats.org/api/miniconf/users/22150?format=json", "institution": "Institute for Infocomm Research (I2R), A*STAR"}, {"id": 22152, "fullname": "Wei Ooi", "url": "http://virtual.aistats.org/api/miniconf/users/22152?format=json", "institution": "National University of Singapore"}], "abstract": "Large Language Models (LLMs) excel at generative language tasks but remain unreliable for structured prediction, particularly in extractive question answering (EQA), where success depends on precise span selection. These challenges are amplified in resource-constrained environments, such as mobile or embedded systems, where deploying high-capacity models is often infeasible. We propose a Learning-to-Defer framework that routes EQA queries across a pool of models with varying capabilities and costs to balance accuracy and efficiency. Our approach is grounded in statistical decision theory: we define a differentiable surrogate loss whose minimizer provably converges to the Bayes-optimal allocation policy. Experiments on SQuADv1, SQuADv2, and TriviaQA show that our method consistently improves the accuracy-efficiency trade-off relative to static baselines and prior routing heuristics. Overall, our framework provides a principled and scalable solution for EQA in both high-performance and on-device deployment settings.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13462", "url": null, "sourceid": 1195, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=kEVupwepTq", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11155, "modified": "2026-03-29T20:42:59.853515-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=kEVupwepTq", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "122", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13477, "uid": "07e1cd7dca89a1678042477183b7ac3f", "name": "Variance Reduction Methods Do Not Need to Compute Full Gradients: Improved Efficiency through Shuffling", "authors": [{"id": 22190, "fullname": "Daniil Medyakov", "url": "http://virtual.aistats.org/api/miniconf/users/22190?format=json", "institution": "Moscow Independent Research Institute of Artificial Intelligence"}, {"id": 22191, "fullname": "Gleb Molodtsov", "url": "http://virtual.aistats.org/api/miniconf/users/22191?format=json", "institution": "Moscow Independent Research Institute of Artificial Intelligence"}, {"id": 22192, "fullname": "Savelii Chezhegov", "url": "http://virtual.aistats.org/api/miniconf/users/22192?format=json", "institution": "Moscow Independent Research Institute of Artificial Intelligence"}, {"id": 22193, "fullname": "Alexey Rebrikov", "url": "http://virtual.aistats.org/api/miniconf/users/22193?format=json", "institution": "Independent Author"}, {"id": 9269, "fullname": "Aleksandr Beznosikov", "url": "http://virtual.aistats.org/api/miniconf/users/9269?format=json", "institution": "Moscow Institute of Physics and Technology (National Research University)"}], "abstract": "Stochastic optimization algorithms are widely used for machine learning with large-scale data. However, their convergence often suffers from non-vanishing variance. Variance Reduction (VR) methods, such as SVRG and SARAH, address this issue but introduce a bottleneck by requiring periodic full gradient computations. In this paper, we explore popular VR techniques and propose an approach that eliminates the necessity for expensive full gradient calculations. To avoid these computations and make our approach memory-efficient, we employ two key techniques: the shuffling heuristic and the concept of SAG/SAGA methods. For non-convex objectives, our convergence rates match those of standard shuffling methods, while under strong convexity, they demonstrate an improvement. We empirically validate the efficiency of our approach and demonstrate its scalability on large-scale machine learning tasks including image classification problem on CIFAR-10 and CIFAR-100 datasets.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13477", "url": null, "sourceid": 119, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=hIx68tWj9w", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11170, "modified": "2026-03-29T20:43:00.429308-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=hIx68tWj9w", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "194", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13479, "uid": "fed33392d3a48aa149a87a38b875ba4a", "name": "Undersmoothing Black-Box Models for Functional Estimation", "authors": [{"id": 19479, "fullname": "Yue Yu", "url": "http://virtual.aistats.org/api/miniconf/users/19479?format=json", "institution": "University of Michigan, Department of Statistics"}, {"id": 22195, "fullname": "Debarghya Mukherjee", "url": "http://virtual.aistats.org/api/miniconf/users/22195?format=json", "institution": "Boston University, Boston University"}, {"id": 17747, "fullname": "Moulinath Banerjee", "url": "http://virtual.aistats.org/api/miniconf/users/17747?format=json", "institution": "University of Michigan - Ann Arbor"}, {"id": 22196, "fullname": "Yaacov Ritov", "url": "http://virtual.aistats.org/api/miniconf/users/22196?format=json", "institution": "University of Michigan - Ann Arbor"}], "abstract": "We study functional estimation using black-box models through a model-agnostic undersmoothing framework. The proposed procedure \\texttt{Rep} operates by augmenting the original dataset through replicating a proportion of samples multiple times, and subsequently applying the black-box algorithm to the augmented dataset. This construction automatically induces undersmoothing and removes the need for manual hyperparameter tuning. We provide several empirical demonstrations (including neural network based learners) showing that compared to the plug-in estimator, the proposed algorithm \\texttt{Rep} improves the estimation accuracy of functional estimation  \\textit{without} requiring explicit expressions for the associated influence functions. Furthermore, we develop a theoretical analysis in two representative settings, the Nadaraya\u2013Watson estimator and the random feature model, establishing that replication provides explicit prescriptions for the replication proportion and number of copies, and yields  optimal convergence rates for functional estimation. In the classical nonparametric regression setting, we extend \\texttt{Rep} with a Lepski-style method that adapts to unknown structural features of the regression function.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13479", "url": null, "sourceid": 1004, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=hE0fklpjVk", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11172, "modified": "2026-03-29T20:43:00.504127-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=hE0fklpjVk", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "174", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13480, "uid": "ef4e3b775c934dada217712d76f3d51f", "name": "Time Series Forecasting with Hahn Kolmogorov-Arnold Networks", "authors": [{"id": 22197, "fullname": "Zahidul Hasan", "url": "http://virtual.aistats.org/api/miniconf/users/22197?format=json", "institution": "Concordia University"}, {"id": 19849, "fullname": "Abdessamad Ben Hamza", "url": "http://virtual.aistats.org/api/miniconf/users/19849?format=json", "institution": "Concordia University"}, {"id": 22198, "fullname": "Nizar Bouguila", "url": "http://virtual.aistats.org/api/miniconf/users/22198?format=json", "institution": "Concordia University"}], "abstract": "Recent Transformer- and MLP-based models have demonstrated strong performance in long-term time series forecasting, yet Transformers remain limited by their quadratic complexity and permutation-equivariant attention, while MLPs exhibit spectral bias. We propose HaKAN, a versatile model based on Kolmogorov-Arnold Networks (KANs), leveraging Hahn polynomial-based learnable activation functions and providing a lightweight and interpretable alternative for multivariate time series forecasting. Our model integrates channel independence, patching, a stack of Hahn-KAN blocks with residual connections, and a bottleneck structure comprised of two fully connected layers. The Hahn-KAN block consists of inter- and intra-patch KAN layers to effectively capture both global and local temporal patterns. Extensive experiments on various forecasting benchmarks demonstrate that our model consistently outperforms recent state-of-the-art methods, with ablation studies validating the effectiveness of its core components.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13480", "url": null, "sourceid": 955, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=hA1JZ2Sbz5", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11173, "modified": "2026-03-29T20:43:00.536066-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=hA1JZ2Sbz5", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "165", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13482, "uid": "3f53d7190148675e3cd472fc826828c5", "name": "Network Inversion for Extreme-Case Training-Like Data Reconstruction", "authors": [{"id": 22199, "fullname": "Pirzada Suhail", "url": "http://virtual.aistats.org/api/miniconf/users/22199?format=json", "institution": "Indian Institute of Technology Bombay"}, {"id": 22200, "fullname": "Sunny Gupta", "url": "http://virtual.aistats.org/api/miniconf/users/22200?format=json", "institution": "IIT Bombay"}, {"id": 22201, "fullname": "Amit Sethi", "url": "http://virtual.aistats.org/api/miniconf/users/22201?format=json", "institution": "Indian Institute of Technology, Bombay"}], "abstract": "Machine learning models are often trained on proprietary or private datasets that cannot be openly shared. However, the trained model weights are frequently distributed under the assumption that sharing model parameters does not compromise the confidentiality or privacy of the training data. In this work, we challenge this assumption by presenting \\textbf{Training-Like Data Reconstruction (TLDR)}, as a general-purpose and architecture-agnostic framework for reconstructing training data from a fully trained classifier. Our approach leverages network inversion techniques to recover data that closely resembles the original training samples by exploiting key properties of the classifier with respect to the training data, without requiring access to training dynamics, gradients, pre-trained models, auxiliary datasets, or unobvious priors. Operating in this extreme setting, we demonstrate successful reconstruction of samples with high similarity to the original training data from diverse classifier architectures highlighting critical privacy concerns associated with sharing model parameters. While prior work in this extreme setting has been limited to binary MLP classifiers trained on small datasets, our framework extends to multi-class classification tasks for models based on diverse architectures trained on significantly larger and more complex datasets. Furthermore, we provide quantitative evaluation using the Structural Similarity Index Measure (SSIM) to compare the reconstructed samples with the training samples.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13482", "url": null, "sourceid": 2355, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=h1OTchW77a", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11175, "modified": "2026-03-29T20:43:00.624601-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=h1OTchW77a", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "110", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13487, "uid": "c6335734dbc0b1ded766421cfc611750", "name": "Adaptive A/B Testing under Nonstationary Dynamics using State-Space Models", "authors": [{"id": 13454, "fullname": "Junzhe Shao", "url": "http://virtual.aistats.org/api/miniconf/users/13454?format=json", "institution": "UC Berkeley"}, {"id": 22216, "fullname": "Waverly Wei", "url": "http://virtual.aistats.org/api/miniconf/users/22216?format=json", "institution": "University of Southern California"}, {"id": 12869, "fullname": "Jingshen Wang", "url": "http://virtual.aistats.org/api/miniconf/users/12869?format=json", "institution": "UC Berkeley"}], "abstract": "A/B testing is central to evaluating how modifications to products, services, and user experiences impact user outcomes. Yet in practice, experiments rarely occur in stationary environments: seasonality, feature launches, and dynamically evolved user demographics make the underlying treatment effects shift over time. Conventional fixed-allocation designs fail to adapt to this nonstationarity, relying on static treatment allocations that potentially compromise estimation efficiency and lead to inefficient use of experimental resources. Response-adaptive randomization (RAR) design provides a natural alternative, adaptively allocating participants over time based on accrued information. However, deploying RAR designs in nonstationary environments raises fundamental challenges: the underlying treatment effects drift over time, noise levels could vary, and continuous monitoring is required to maintain valid statistical inference. In this work, we propose a methodology framework that addresses these challenges. On the one hand, we model period-level treatment arm means as autoregressive state-space processes and develop a Kalman filter estimator that exploits temporal dependence. On the other hand, we propose an RAR design that accommodates nonstationarity by incorporating state uncertainty via predicted Kalman variances. Our theoretical analysis establishes asymptotic normality of the treatment effect estimator, establishes asymptotic normality, compares relative efficiency, and enables the construction of anytime-valid confidence sequences for continuous monitoring. Simulation studies demonstrate that our method is significantly more efficient than a benchmark time-averaging estimator and fixed allocation strategy, particularly under treatment effect drift and variance imbalance.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13487", "url": null, "sourceid": 2106, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=gjIn18VuA7", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11180, "modified": "2026-03-29T20:43:00.817045-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=gjIn18VuA7", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "13", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13488, "uid": "a1519de5b5d44b31a01de013b9b51a80", "name": "Gradient-flow SDEs have unique transient population dynamics", "authors": [{"id": 19351, "fullname": "Vincent Guan", "url": "http://virtual.aistats.org/api/miniconf/users/19351?format=json", "institution": "University of British Columbia"}, {"id": 5528, "fullname": "Joseph Janssen", "url": "http://virtual.aistats.org/api/miniconf/users/5528?format=json", "institution": "University of British Columbia"}, {"id": 22217, "fullname": "Nicolas Lanzetti", "url": "http://virtual.aistats.org/api/miniconf/users/22217?format=json", "institution": "Deparment of Computing + Mathematical Sciences, California Institute of Technology"}, {"id": 22218, "fullname": "Antonio Terpin", "url": "http://virtual.aistats.org/api/miniconf/users/22218?format=json", "institution": "ETHZ - ETH Zurich"}, {"id": 22219, "fullname": "Geoffrey Schiebinger", "url": "http://virtual.aistats.org/api/miniconf/users/22219?format=json", "institution": "University of British Columbia"}, {"id": 9594, "fullname": "Elina Robeva", "url": "http://virtual.aistats.org/api/miniconf/users/9594?format=json", "institution": "University of British Columbia"}], "abstract": "Identifying the drift and diffusion of an SDE from its population dynamics is a notoriously challenging task. Researchers in machine learning and single cell biology have only been able to prove a partial identifiability result: for potential-driven SDEs, the gradient-flow drift can be identified from temporal marginals if the Brownian diffusivity is already known. Existing methods therefore assume that the diffusivity is known a priori, despite it being unknown in practice. We dispel the need for this assumption by providing a complete characterization of identifiability: the gradient-flow drift and Brownian diffusivity are jointly identifiable from temporal marginals if and only if the process is observed outside of equilibrium. Given this fundamental result, we propose nn-APPEX, the first Schr\u00f6dinger Bridge\u2013based inference method that can simultaneously learn the drift and diffusion of gradient-flow SDEs solely from observed marginals. Extensive numerical experiments show that nn-APPEX's ability to adjust its diffusion estimate enables accurate inference, while previous Schr\u00f6dinger Bridge methods obtain biased drift estimates due to their assumed, and likely incorrect, diffusion.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13488", "url": null, "sourceid": 1072, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=gfF0rwtRMp", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11181, "modified": "2026-03-29T20:43:00.848814-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=gfF0rwtRMp", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "70", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13490, "uid": "4588e674d3f0faf985047d4c3f13ed0d", "name": "From Hawkes Processes to Attention: Time-Modulated Mechanisms for Event Sequences", "authors": [{"id": 22223, "fullname": "Xinzi Tan", "url": "http://virtual.aistats.org/api/miniconf/users/22223?format=json", "institution": "National University of Singapore"}, {"id": 22224, "fullname": "Kejian Zhang", "url": "http://virtual.aistats.org/api/miniconf/users/22224?format=json", "institution": "National University of Singapore"}, {"id": 22225, "fullname": "Junhan Yu", "url": "http://virtual.aistats.org/api/miniconf/users/22225?format=json", "institution": "National University of Singaore, National University of Singapore"}, {"id": 22226, "fullname": "Doudou Zhou", "url": "http://virtual.aistats.org/api/miniconf/users/22226?format=json", "institution": "National University of Singapore"}], "abstract": "Marked Temporal Point Processes (MTPPs) arise naturally in medical, social, commercial, and financial domains. However, existing Transformer-based methods mostly inject temporal information only via positional encodings, relying on shared or parametric decay structures, which limits their ability to capture heterogeneous and type-specific temporal effects. Inspired by this observation, we derive a novel attention operator called Hawkes Attention from the multivariate Hawkes process theory for MTPP, using learnable per-type neural kernels to modulate query, key and value projections, thereby replacing the corresponding parts in the traditional attention. Benefited from the design, Hawkes Attention unifies event timing and content interaction, learning both the time-relevant behavior and type-specific excitation patterns from the data. The experimental results show that our method achieves better performance compared to the baselines. In addition to the general MTPP, our attention mechanism can also be easily applied to specific temporal structures, such as time series forecasting.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13490", "url": null, "sourceid": 1144, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=gOOxbzB9Y5", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11183, "modified": "2026-03-29T20:43:00.917706-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=gOOxbzB9Y5", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "64", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13491, "uid": "077e29b11be80ab57e1a2ecabb7da330", "name": "Towards Motion-aware Referring Image Segmentation", "authors": [{"id": 22227, "fullname": "Chaeyun Kim", "url": "http://virtual.aistats.org/api/miniconf/users/22227?format=json", "institution": "Datumo Inc."}, {"id": 19844, "fullname": "Seunghoon Yi", "url": "http://virtual.aistats.org/api/miniconf/users/19844?format=json", "institution": "Seoul National University"}, {"id": 22228, "fullname": "Yejin Kim", "url": "http://virtual.aistats.org/api/miniconf/users/22228?format=json", "institution": "Seoul National University"}, {"id": 22229, "fullname": "Yohan Jo", "url": "http://virtual.aistats.org/api/miniconf/users/22229?format=json", "institution": "Seoul National University"}, {"id": 11069, "fullname": "Joonseok Lee", "url": "http://virtual.aistats.org/api/miniconf/users/11069?format=json", "institution": "Seoul National Univ."}], "abstract": "Referring Image Segmentation (RIS) requires identifying objects from images based on textual descriptions. We observe that existing methods significantly underperform on motion-related queries compared to appearance-based ones, and propose to address this from both data and algorithmic perspectives. First, we introduce an efficient data augmentation scheme that extracts motion-centric phrases from original captions, exposing models to more motion expressions without additional annotations. Second, since the same object can be described differently depending on the context, we propose Multimodal Radial Contrastive Learning (MRaCL), performed on fused image-text embeddings rather than unimodal representations. For comprehensive evaluation, we introduce a new test split focusing on motion-centric queries, and introduce a new benchmark called M-Bench, where objects are distinguished primarily by actions. Extensive experiments show our method substantially improves performance on motion-centric queries across multiple RIS models, maintaining competitive results on appearance-based descriptions.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13491", "url": null, "sourceid": 249, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=gI9IV3cmnC", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11184, "modified": "2026-03-29T20:43:00.949961-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=gI9IV3cmnC", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "187", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13493, "uid": "b7ee6f5f9aa5cd17ca1aea43ce848496", "name": "Hypergraph Neural Networks Accelerate MUS Enumeration", "authors": [{"id": 22231, "fullname": "Hiroya Ijima", "url": "http://virtual.aistats.org/api/miniconf/users/22231?format=json", "institution": "Hitachi, Ltd."}, {"id": 22232, "fullname": "Koichiro Yawata", "url": "http://virtual.aistats.org/api/miniconf/users/22232?format=json", "institution": "Hitachi, Ltd."}], "abstract": "Enumerating Minimal Unsatisfiable Subsets (MUSes) is a fundamental task in constraint satisfaction problems (CSPs). Its major challenge is the exponential growth of the search space, which becomes particularly severe when satisfiability checks are expensive. Recent machine learning approaches reduce this cost for Boolean satisfiability problems but rely on explicit variable-constraint relationships, limiting their application domains. This paper proposes a domain-agnostic method to accelerate MUS enumeration using Hypergraph Neural Networks (HGNNs). The proposed method incrementally builds a hypergraph with constraints as vertices and MUSes enumerated until the current step as hyperedges, and employs an HGNN-based agent trained via reinforcement learning to minimize the number of satisfiability checks required to obtain an MUS. Experimental results demonstrate the effectiveness of our approach in accelerating MUS enumeration, showing that our method can enumerate more MUSes within the same satisfiability check budget compared to conventional methods.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13493", "url": null, "sourceid": 771, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=gCHIAqnoip", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11186, "modified": "2026-03-29T20:43:01.009710-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=gCHIAqnoip", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "75", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13497, "uid": "147702db07145348245dc5a2f2fe5683", "name": "Momentum SVGD-EM for Accelerated Maximum Marginal Likelihood Estimation", "authors": [{"id": 19910, "fullname": "Adam Rozzio", "url": "http://virtual.aistats.org/api/miniconf/users/19910?format=json", "institution": "ENS Paris Saclay"}, {"id": 22243, "fullname": "Rafael Athanasiades", "url": "http://virtual.aistats.org/api/miniconf/users/22243?format=json", "institution": "Imperial College London"}, {"id": 22154, "fullname": "O. Deniz Akyildiz", "url": "http://virtual.aistats.org/api/miniconf/users/22154?format=json", "institution": "Imperial College London"}], "abstract": "Maximum marginal likelihood estimation (MMLE) can be recast as the optimization of the so-called free energy. The celebrated expectation-maximisation (EM) procedure can then be interpreted as a coordinate-descent scheme in the space of parameters and measures. Recently, a significant body of work has adopted this perspective, leading to interacting particle algorithms for MMLE. In this paper, we propose an accelerated version of one such procedure, based on Stein variational gradient descent (SVGD), by introducing Nesterov momentum in both the parameter updates and in the space of probability measures. The resulting method, termed Momentum SVGD-EM, consistently accelerates convergence in terms of required iterations across various tasks of increasing difficulty, demonstrating effectiveness in both low- and high-dimensional settings.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13497", "url": null, "sourceid": 1202, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=g6lWuOA8KR", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11190, "modified": "2026-03-29T20:43:01.193420-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=g6lWuOA8KR", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "106", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13506, "uid": "30c8e1ca872524fbf7ea5c519ca397ee", "name": "Learning Physical Operators using Neural Operators", "authors": [{"id": 22264, "fullname": "Vignesh Gopakumar", "url": "http://virtual.aistats.org/api/miniconf/users/22264?format=json", "institution": "University College London, UK Atomic Energy Authority"}, {"id": 22265, "fullname": "Ander Gray", "url": "http://virtual.aistats.org/api/miniconf/users/22265?format=json", "institution": "UK Atomic Energy Authority"}, {"id": 9195, "fullname": "Daniel Giles", "url": "http://virtual.aistats.org/api/miniconf/users/9195?format=json", "institution": "University College London"}, {"id": 22266, "fullname": "Lorenzo Zanisi", "url": "http://virtual.aistats.org/api/miniconf/users/22266?format=json", "institution": "UK atomic energy authority"}, {"id": 22267, "fullname": "Matt Kusner", "url": "http://virtual.aistats.org/api/miniconf/users/22267?format=json", "institution": "Mila - Quebec Artificial Intelligence Institute"}, {"id": 22268, "fullname": "Timo Betcke", "url": "http://virtual.aistats.org/api/miniconf/users/22268?format=json", "institution": "University College London, University of London"}, {"id": 22269, "fullname": "Stanislas Pamela", "url": "http://virtual.aistats.org/api/miniconf/users/22269?format=json", "institution": "UKAEA"}, {"id": 23263, "fullname": "Marc Deisenroth", "url": "http://virtual.aistats.org/api/miniconf/users/23263?format=json", "institution": "Google DeepMind"}], "abstract": "Neural operators have emerged as promising surrogate models for solving partial differential equations (PDEs), but struggle to generalise beyond training distributions and are often constrained to a fixed temporal discretisation. This work introduces a physics-informed training framework that addresses these limitations by decomposing PDEs using operator splitting methods, training separate neural operators to learn individual non-linear physical operators while approximating linear operators with fixed finite-difference convolutions. This modular mixture-of-experts architecture enables generalisation to novel physical regimes by explicitly encoding the underlying operator structure. We formulate the modelling task as a neural ordinary differential equation (ODE) where these learned operators constitute the right-hand side, enabling continuous-in-time predictions through standard ODE solvers and implicitly enforcing PDE constraints. Demonstrated on incompressible and compressible Navier--Stokes equations, our approach achieves superior performance when generalising to unseen physics while remaining parameter-efficient, enables temporal extrapolation beyond training horizons, and provides interpretable components whose behaviour can be verified against known physics.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13506", "url": null, "sourceid": 1419, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=elPH36vmoZ", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11199, "modified": "2026-03-29T20:43:01.543856-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=elPH36vmoZ", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "92", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13510, "uid": "33e8075e9970de0cfea955afd4644bb2", "name": "Topological Alignment of Shared Vision-Language Embedding Space", "authors": [{"id": 22275, "fullname": "Junwon You", "url": "http://virtual.aistats.org/api/miniconf/users/22275?format=json", "institution": "Korea Advanced Institute of Science and Technology"}, {"id": 22276, "fullname": "Kang Dasol", "url": "http://virtual.aistats.org/api/miniconf/users/22276?format=json", "institution": "Google"}, {"id": 22277, "fullname": "Jae-Hun Jung", "url": "http://virtual.aistats.org/api/miniconf/users/22277?format=json", "institution": "POSTECH (Pohang University of Science and Technology)"}], "abstract": "Contrastive Vision-Language Models (VLMs) have demonstrated strong zero-shot capabilities.  However, their cross-modal alignment remains biased toward English due to limited multilingual multimodal data.  Recent multilingual extensions have alleviated this gap but enforce instance-level alignment while neglecting the global geometry of the shared embedding space.  We address this problem by introducing **ToMCLIP** (**To**pological Alignment for **M**ultilingual **CLIP**), a topology-aware framework aligning embedding spaces with topology-preserving constraints.  The proposed method applies persistent homology to define a topological alignment loss and approximates persistence diagram with theoretical error bounds using graph sparsification strategy.  This work validates the proposed approach, showing enhanced structural coherence of multilingual representations, higher zero-shot accuracy on the CIFAR-100, and stronger multilingual retrieval performance on the xFlickr\\&CO.  Beyond VLMs, the proposed approach provides a general method for incorporating topological alignment into representation learning.  Code is available at https://github.com/junwon0/ToMCLIP.git.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13510", "url": null, "sourceid": 535, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=ecd8cgWZr6", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11203, "modified": "2026-03-29T20:43:01.686384-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=ecd8cgWZr6", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "186", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13512, "uid": "8e82ab7243b7c66d768f1b8ce1c967eb", "name": "AlphaFold's Bayesian Roots in Probability Kinematics", "authors": [{"id": 20599, "fullname": "Thomas Hamelryck", "url": "http://virtual.aistats.org/api/miniconf/users/20599?format=json", "institution": "University of Copenhagen"}, {"id": 22279, "fullname": "Kanti Mardia", "url": "http://virtual.aistats.org/api/miniconf/users/22279?format=json", "institution": "University of Leeds"}], "abstract": "The seminal breakthrough of AlphaFold in protein structure prediction relied on a learned potential energy function parameterized by deep models, in contrast to its successors AlphaFold2 and AlphaFold3, which lack an explicit probabilistic interpretation. While AlphaFold\u2019s potential was originally justified by heuristic analogy to physical potentials of mean force, we show that it can instead be understood as a principled instance of probability kinematics (PK), also known as Jeffrey conditioning, a generalization of Bayesian updating. This reinterpretation reveals that AlphaFold is a generalized Bayesian model that explicitly defines a posterior distribution over structures, providing a deeper explanation of its success and a foundation for future model design. To demonstrate this framework with precision, we introduce a tractable synthetic model in which an angular random walk prior is updated with distance-based evidence via PK, directly mirroring AlphaFold\u2019s mechanism. This setting allows us to explore the probabilistic foundations of AlphaFold in a clear and interpretable way. Our work connects a landmark in protein structure prediction to a broader class of compositional deep generative models and points to new opportunities for principled probabilistic approaches.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13512", "url": null, "sourceid": 830, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=eX9yYimUDx", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11205, "modified": "2026-03-29T20:43:01.772224-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=eX9yYimUDx", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "12", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13513, "uid": "9fd81843ad7f202f26c1a174c7357585", "name": "Learning in Continuous State-Space MDPs for Network Inventory Management", "authors": [{"id": 12327, "fullname": "Hansheng Jiang", "url": "http://virtual.aistats.org/api/miniconf/users/12327?format=json", "institution": "University of Toronto"}, {"id": 22280, "fullname": "Shunan Jiang", "url": "http://virtual.aistats.org/api/miniconf/users/22280?format=json", "institution": "University of California, Berkeley"}, {"id": 22281, "fullname": "Zuo-Jun Shen", "url": "http://virtual.aistats.org/api/miniconf/users/22281?format=json", "institution": "Northwestern University"}], "abstract": "We consider online learning in infinite-horizon, average-cost Markov Decision Processes (MDPs) with multi-dimensional, continuous state spaces and censored feedback. Our model setting, motivated by network inventory management applications such as vehicle sharing, is characterized by complex, correlated state transitions and the absence of value function convexity, rendering standard analytical techniques for both MDPs and inventory control inapplicable. Our primary contribution is an integrated framework establishing and leveraging the Lipschitz property of the long-run average cost function. This insight allows us to analyze the problem through the lens of Lipschitz bandits, for which we design a provably efficient online learning algorithm that learns a near-optimal policy from censored demand data. We derive a high-probability regret bound of $O(T^{\\frac{n}{n+1}} (\\log T)^{\\frac{1}{n+1}})$, where $n$ is the network size through customized concentration inequalities for cumulative costs in MDPs with state-dependent transitions. Furthermore, we devise a matching lower bound for this learning problem, which captures the inherent dimensionality challenge.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13513", "url": null, "sourceid": 297, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=e4ATBG2Oh5", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11206, "modified": "2026-03-29T20:43:01.804999-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=e4ATBG2Oh5", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "93", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13519, "uid": "64223ccf70bbb65a3a4aceac37e21016", "name": "Incentivizing Truthful Submissions in a Data Marketplace for Mean Estimation", "authors": [{"id": 19467, "fullname": "Keran Chen", "url": "http://virtual.aistats.org/api/miniconf/users/19467?format=json", "institution": "University of Wisconsin Madison"}, {"id": 22287, "fullname": "Alex Clinton", "url": "http://virtual.aistats.org/api/miniconf/users/22287?format=json", "institution": "University of Wisconsin - Madison"}, {"id": 9681, "fullname": "Kirthevasan Kandasamy", "url": "http://virtual.aistats.org/api/miniconf/users/9681?format=json", "institution": "UC Berkeley"}], "abstract": "We study a data marketplace where a broker intermediates between buyers, who seek to estimate the mean $\\mu$ of an unknown normal distribution $N(\\mu, \\sigma^2)$, and contributors, who can collect data from this distribution at a cost.  The broker delegates data collection work to contributors, aggregates reported datasets, sells it to buyers, and redistributes revenue as payments to contributors. We aim to maximize welfare or profit under key constraints: individual rationality for buyers and contributors,  incentive compatibility (contributors are incentivized to comply with data collection instructions and truthfully report the collected data), and budget balance (total contributor payments equals total revenue). We first compute welfare/profit-optimal prices under truthful reporting; however, to incentivize data collection and truthful data reporting, we adjust them based on discrepancies in contributors' reported data. This yields a Nash equilibrium (NE) where the two lowest-cost contributors collect all data. We complement this with two hardness results: $\\mathcal{(i)}$ no nontrivial dominant-strategy incentive-compatible mechanism exists in this problem, and $\\mathcal{(ii)}$ no mechanism outperforms ours in a NE.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13519", "url": null, "sourceid": 944, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=dDryfNkrL4", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11212, "modified": "2026-03-29T20:43:01.989002-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=dDryfNkrL4", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "80", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13521, "uid": "300891a62162b960cf02ce3827bb363c", "name": "Empirically Calibrated Conditional Independence Tests", "authors": [{"id": 19852, "fullname": "Milleno Pan", "url": "http://virtual.aistats.org/api/miniconf/users/19852?format=json", "institution": "Memorial Sloan Kettering Cancer Center"}, {"id": 22289, "fullname": "Antoine de Mathelin", "url": "http://virtual.aistats.org/api/miniconf/users/22289?format=json", "institution": "CMLA - ENS Paris Saclay, ENS Paris-Saclay"}, {"id": 22290, "fullname": "Wesley Tansey", "url": "http://virtual.aistats.org/api/miniconf/users/22290?format=json", "institution": "Memorial Sloan Kettering Cancer Center"}], "abstract": "Conditional independence tests (CIT) are widely used for causal discovery and feature selection. Even with false discovery rate (FDR) control procedures, they often fail to provide frequentist guarantees in practice. We highlight two common failure modes: (i) in small samples, asymptotic guarantees for many CITs can be inaccurate and even correctly specified models fail to estimate the noise levels and control the error, and (ii) when sample sizes are large but models are misspecified, unaccounted dependencies skew the test's behavior and fail to return uniform p-values under the null. We propose Empirically Calibrated Conditional Independence Tests (ECCIT), a method that measures and corrects for miscalibration. For a chosen base CIT (e.g., GCM, HRT), ECCIT optimizes an adversary that selects features and response functions to maximize a miscalibration metric. ECCIT then fits a monotone calibration map that adjusts the base-test p-values in proportion to the observed miscalibration. Across empirical benchmarks on synthetic and real data, ECCIT achieves valid FDR with higher power than existing calibration strategies while remaining test agnostic. Code is available at https://github.com/tansey-lab/ECCIT.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13521", "url": null, "sourceid": 2367, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=d9wlMAJkSD", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11214, "modified": "2026-03-29T20:43:02.064453-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=d9wlMAJkSD", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "65", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13536, "uid": "a597e50502f5ff68e3e25b9114205d4a", "name": "SFBD Flow: A Continuous-Optimization Framework for Training Diffusion Models with Noisy Samples", "authors": [{"id": 14410, "fullname": "Haoye Lu", "url": "http://virtual.aistats.org/api/miniconf/users/14410?format=json", "institution": "University of Waterloo; Vector Institute"}, {"id": 22307, "fullname": "Darren Lo", "url": "http://virtual.aistats.org/api/miniconf/users/22307?format=json", "institution": "University of Waterloo"}, {"id": 12766, "fullname": "Yaoliang Yu", "url": "http://virtual.aistats.org/api/miniconf/users/12766?format=json", "institution": "University of Waterloo"}], "abstract": "Diffusion models achieve strong generative performance but often rely on large datasets that may include sensitive content. This challenge is compounded by the models\u2019 tendency to memorize training data, raising privacy concerns. SFBD (Lu et al., 2025) addresses this by training on corrupted data and using limited clean samples to capture local structure and improve convergence. However, its iterative denoising and fine-tuning loop requires manual coordination, making it burdensome to implement. We reinterpret SFBD as an alternating projection algorithm and introduce a continuous variant, SFBD flow, that removes the need for alternating steps. We further show its connection to consistency constraint-based methods, and demonstrate that its practical instantiation, Online SFBD, consistently outperforms strong baselines across benchmarks.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13536", "url": null, "sourceid": 194, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=bm9E45jCdu", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11229, "modified": "2026-03-29T20:43:02.623996-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=bm9E45jCdu", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "169", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13537, "uid": "2cad8fa47bbef282badbb8de5374b894", "name": "Learning Linear Regression with Low-Rank Tasks In-Context", "authors": [{"id": 22308, "fullname": "Kaito Takanami", "url": "http://virtual.aistats.org/api/miniconf/users/22308?format=json", "institution": "The University of Tokyo"}, {"id": 9281, "fullname": "Takashi Takahashi", "url": "http://virtual.aistats.org/api/miniconf/users/9281?format=json", "institution": "The University of Tokyo"}, {"id": 9282, "fullname": "Yoshiyuki Kabashima", "url": "http://virtual.aistats.org/api/miniconf/users/9282?format=json", "institution": "The University of Tokyo"}], "abstract": "In-context learning (ICL) is a key building block of modern large language models, yet its theoretical mechanisms remain poorly understood. It is particularly mysterious how ICL operates in real-world applications where tasks have a structure. In this work, we address this problem by analyzing a linear attention model trained on low-rank regression tasks. Within this setting, we precisely characterize the distribution of predictions and the generalization error in the high-dimensional limit. Moreover, we find that statistical fluctuations in finite pre-training data induce an implicit regularization. Finally, we identify a sharp phase transition of the generalization error governed by task structure. These results provide a framework for understanding how transformers learn to learn the task structure.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13537", "url": null, "sourceid": 2100, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=bkhqasdf2u", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11230, "modified": "2026-03-29T20:43:02.654801-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=bkhqasdf2u", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "91", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13538, "uid": "cf004fdc76fa1a4f25f62e0eb5261ca3", "name": "Neural Additive Experts: Context-Gated Experts for Controllable Model Additivity", "authors": [{"id": 22309, "fullname": "Guangzhi Xiong", "url": "http://virtual.aistats.org/api/miniconf/users/22309?format=json", "institution": "University of Virginia, Charlottesville"}, {"id": 22310, "fullname": "Sanchit Sinha", "url": "http://virtual.aistats.org/api/miniconf/users/22310?format=json", "institution": "University of Virginia, Charlottesville"}, {"id": 22311, "fullname": "Aidong Zhang", "url": "http://virtual.aistats.org/api/miniconf/users/22311?format=json", "institution": "University of Virginia"}], "abstract": "The trade-off between interpretability and accuracy remains a core challenge in machine learning. Standard Generalized Additive Models (GAMs) offer clear feature attributions but are often constrained by their strictly additive nature, which can limit predictive performance. Introducing feature interactions can boost accuracy yet may obscure individual feature contributions. To address these issues, we propose Neural Additive Experts (NAEs), a novel framework that seamlessly balances interpretability and accuracy. NAEs employ a mixture of experts framework, learning multiple specialized networks per feature, while a dynamic gating mechanism integrates information across features, thereby relaxing rigid additive constraints. Furthermore, we propose targeted regularization techniques to mitigate variance among expert predictions, facilitating a smooth transition from an exclusively additive model to one that captures intricate feature interactions while maintaining clarity in feature attributions. Our theoretical analysis and experiments on synthetic data illustrate the model's flexibility, and extensive evaluations on real-world datasets confirm that NAEs achieve an optimal balance between predictive accuracy and transparent, feature-level explanations. The code is available at https://github.com/Teddy-XiongGZ/NAE.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13538", "url": null, "sourceid": 368, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=bdqw4bSP4l", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11231, "modified": "2026-03-29T20:43:02.684188-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=bdqw4bSP4l", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "114", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13542, "uid": "c0c7c76d30bd3dcaefc96f40275bdc0a", "name": "On the Weight Density of L2-Regularized Linear Classification and Regression", "authors": [{"id": 19274, "fullname": "He Zhe Lin", "url": "http://virtual.aistats.org/api/miniconf/users/19274?format=json", "institution": "National Taiwan University"}, {"id": 22321, "fullname": "Zhi-Bao Lu", "url": "http://virtual.aistats.org/api/miniconf/users/22321?format=json", "institution": "National Taiwan University"}, {"id": 19898, "fullname": "Sheng-Wei Chen", "url": "http://virtual.aistats.org/api/miniconf/users/19898?format=json", "institution": "National Taiwan University"}, {"id": 22322, "fullname": "Cheng-Hung Liu", "url": "http://virtual.aistats.org/api/miniconf/users/22322?format=json", "institution": "Department of computer science and informational engineering, National Taiwan University"}, {"id": 22323, "fullname": "Chih-Jen Lin", "url": "http://virtual.aistats.org/api/miniconf/users/22323?format=json", "institution": "National Taiwan Univ / MBZUAI"}], "abstract": "For traditional linear models with the widely used $L_2$-regularizer, it is often assumed that the resulting models are dense. As a result, little attention has been paid to when the optimal solution for an $L_2$-regularized problem can actually be sparse. In this work, we rigorously prove that for $L_2$-regularized support vector classification/regression, the theoretical optimum can indeed be sparse when the data have sparse feature values. Surprisingly, we observe that some optimization methods fail to preserve this sparsity and instead produce fully dense numerical solutions, leading to unnecessary storage overhead. We explain this phenomenon through detailed analysis. In particular, we novelly show that certain projected gradient methods for solving the dual problem naturally yields sparser numerical solutions compared to other optimization algorithms. By applying suitable algorithms that preserve numerical sparsity, the storage can be reduced by up to 50%, which is highly advantageous for large-scale industrial applications.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13542", "url": null, "sourceid": 50, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=bFDTISBkLu", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11235, "modified": "2026-03-29T20:43:02.833768-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=bFDTISBkLu", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "128", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13547, "uid": "e7e23670481ac78b3c4122a99ba60573", "name": "RamPINN: Recovering Raman Spectra From Coherent Anti-Stokes Spectra Using Embedded Physics", "authors": [{"id": 19907, "fullname": "Sai Karthikeya Vemuri", "url": "http://virtual.aistats.org/api/miniconf/users/19907?format=json", "institution": "Friedrich Schiller University Jena"}, {"id": 22337, "fullname": "Adithya Ashok Chalain Valapil", "url": "http://virtual.aistats.org/api/miniconf/users/22337?format=json", "institution": "Friedrich-Schiller Universit\u00e4t Jena"}, {"id": 22338, "fullname": "Tim B\u00fcchner", "url": "http://virtual.aistats.org/api/miniconf/users/22338?format=json", "institution": "Friedrich-Schiller Universit\u00e4t Jena"}, {"id": 22339, "fullname": "Joachim Denzler", "url": "http://virtual.aistats.org/api/miniconf/users/22339?format=json", "institution": "Friedrich-Schiller-University Jena"}], "abstract": "Transferring the recent advancements in deep learning into scientific disciplines is hindered by the lack of the required large-scale datasets for training. We argue that in these knowledge-rich domains, the established body of scientific theory provides reliable inductive biases in the form of governing physical laws. We address the ill-posed inverse problem of recovering Raman spectra from noisy Coherent Anti-Stokes Raman Scattering (CARS) measurements, as the true Raman signal here is suppressed by a dominating non-resonant background. We propose RamPINN, a model that learns to recover Raman spectra from given CARS spectra. Our core methodological contribution is a physics-informed neural network that utilizes a dual-decoder architecture to disentangle resonant and non-resonant signals.  This is done by enforcing the Kramers-Kronig causality relations via a differentiable Hilbert transform loss on the resonant and a smoothness prior on the non-resonant part of the signal.  Trained entirely on synthetic data, RamPINN demonstrates strong zero-shot generalization to real-world experimental data, explicitly closing this gap and significantly outperforming existing baselines.  Furthermore, we show that training with these physics-based losses alone, without access to any ground-truth Raman spectra, still yields competitive results.  This work highlights a broader concept: formal scientific rules can act as a potent inductive bias, enabling robust, self-supervised learning in data-limited scientific domains.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13547", "url": null, "sourceid": 2104, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=amJjICHBEd", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11240, "modified": "2026-03-29T20:43:03.015539-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=amJjICHBEd", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "148", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13548, "uid": "2715518c875999308842e3455eda2fe3", "name": "CTRLS: Chain-of-Thought Reasoning via Latent State-Transition", "authors": [{"id": 22340, "fullname": "Junda Wu", "url": "http://virtual.aistats.org/api/miniconf/users/22340?format=json", "institution": "University of California, San Diego"}, {"id": 22341, "fullname": "Yuxin Xiong", "url": "http://virtual.aistats.org/api/miniconf/users/22341?format=json", "institution": "University of California, San Diego"}, {"id": 22342, "fullname": "Xintong Li", "url": "http://virtual.aistats.org/api/miniconf/users/22342?format=json", "institution": "University of California, San Diego"}, {"id": 22343, "fullname": "Sheldon Yu", "url": "http://virtual.aistats.org/api/miniconf/users/22343?format=json", "institution": "University of California, San Diego"}, {"id": 9212, "fullname": "Zhengmian Hu", "url": "http://virtual.aistats.org/api/miniconf/users/9212?format=json", "institution": "University of Pittsburgh"}, {"id": 18656, "fullname": "Tong Yu", "url": "http://virtual.aistats.org/api/miniconf/users/18656?format=json", "institution": "Adobe Research"}, {"id": 22344, "fullname": "Rui Wang", "url": "http://virtual.aistats.org/api/miniconf/users/22344?format=json", "institution": "Adobe Systems"}, {"id": 22345, "fullname": "Xiang Chen", "url": "http://virtual.aistats.org/api/miniconf/users/22345?format=json", "institution": "Adobe Systems"}, {"id": 22346, "fullname": "Jingbo Shang", "url": "http://virtual.aistats.org/api/miniconf/users/22346?format=json", "institution": "University of California, San Diego"}, {"id": 22347, "fullname": "Julian McAuley", "url": "http://virtual.aistats.org/api/miniconf/users/22347?format=json", "institution": "University of California, San Diego, University of California, San Diego"}], "abstract": "Chain-of-thought (CoT) reasoning enables large language models (LLMs) to break down complex problems into interpretable intermediate steps, significantly enhancing model transparency and performance in reasoning tasks. However, conventional CoT methods rely on heuristic sampling without structured modeling of reasoning transitions, constraining their ability to systematically explore and discover diverse and effective reasoning trajectories. In this work, we introduce CTRLS, a framework that formulates CoT reasoning as a Markov decision process (MDP) with latent state transitions, enabling principled and state-aware exploration via distributional reinforcement learning. By modelling reasoning actions as explicit probability distributions in latent space, our approach explicitly models epistemic uncertainty, facilitating robust exploration of the reasoning space. As part of our framework, we introduce an on-policy reinforcement learning strategy incorporating epsilon-greedy exploration and entropy-based regularization to iteratively refine latent state transitions without requiring additional fine-tuning of the underlying LLM. Theoretical analyses provide evidence lower bounds (ELBO), theoretically grounding our transition-aware modeling of latent reasoning dynamics. Further experiments demonstrate improvements in reasoning accuracy, diversity, and exploration efficiency across benchmark reasoning tasks.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13548", "url": null, "sourceid": 1219, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=afld4XGbQe", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11241, "modified": "2026-03-29T20:43:03.056059-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=afld4XGbQe", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "35", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13553, "uid": "598b3e71ec378bd83e0a727608b5db01", "name": "Linear Convergence of the Frank-Wolfe Algorithm over Product Polytopes", "authors": [{"id": 22354, "fullname": "Gabriele Iommazzo", "url": "http://virtual.aistats.org/api/miniconf/users/22354?format=json", "institution": "Zuse Institute Berlin"}, {"id": 22355, "fullname": "David Mart\u00ednez-Rubio", "url": "http://virtual.aistats.org/api/miniconf/users/22355?format=json", "institution": "IMDEA Software Institute"}, {"id": 22356, "fullname": "Francisco Criado", "url": "http://virtual.aistats.org/api/miniconf/users/22356?format=json", "institution": "CUNEF Universidad"}, {"id": 17828, "fullname": "Elias Wirth", "url": "http://virtual.aistats.org/api/miniconf/users/17828?format=json", "institution": "TU Berlin"}, {"id": 295, "fullname": "Sebastian Pokutta", "url": "http://virtual.aistats.org/api/miniconf/users/295?format=json", "institution": "ZIB"}], "abstract": "We study the linear convergence of Frank-Wolfe algorithms over product polytopes. We analyze two condition numbers for the product polytope, namely the pyramidal width and the vertex-facet distance, based on the condition numbers of individual polytope components.  As a result, for convex objectives that are $\\mu$-Polyak-\u0141ojasiewicz, we show linear convergence rates quantified in terms of the resulting condition numbers. We apply our results to the problem of approximately finding a feasible point in a polytope intersection in high-dimensions, and demonstrate the practical efficiency of our algorithms through empirical results.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13553", "url": null, "sourceid": 476, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=a6GOkAeSRD", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11246, "modified": "2026-03-29T20:43:03.249833-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=a6GOkAeSRD", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "96", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13558, "uid": "ebd6d2f5d60ff9afaeda1a81fc53e2d0", "name": "Identification and Estimation of \"Probabilities of Causation\" in the Presence of Confounding and Selection Bias", "authors": [{"id": 11066, "fullname": "Ryusei Shingaki", "url": "http://virtual.aistats.org/api/miniconf/users/11066?format=json", "institution": "Yokohama National University"}, {"id": 22363, "fullname": "Haruka Yoshida", "url": "http://virtual.aistats.org/api/miniconf/users/22363?format=json", "institution": "Yokohama National University"}, {"id": 12426, "fullname": "Manabu Kuroki", "url": "http://virtual.aistats.org/api/miniconf/users/12426?format=json", "institution": "Yokohama National University"}], "abstract": "Probabilities of causation are valuable concepts for explainable artificial intelligence (XAI) and personalized decision-making. Pearl (2009) defined the probabilities of causation from the viewpoint of \"necessity\", \"sufficiency\", and \"necessity and sufficiency\" in the context of structural causal models. In addition, Tian and Pearl (2000) and Kuroki and Cai (2011) provided the identification conditions of the probabilities of causation under the monotonicity assumption. However, these identification conditions are described based on \"the joint probabilities of observed random variables\" and/or \"causal risks\" without selection biases. Thus, they are not applicable to studies in the presence of confounding and selection biases. To address this problem, this paper provides novel identification conditions for the probabilities of causation by using (i) two proxy covariates and (ii) an instrumental variable and a proxy covariate. When the probabilities of causation can be evaluated through the proposed identification conditions, new plug-in estimators for these probabilities are presented. Finally, we illustrate the application of our results on a real-world dataset.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13558", "url": null, "sourceid": 1508, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=ZNNhIwXU4m", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11251, "modified": "2026-03-29T20:43:03.498608-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=ZNNhIwXU4m", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "77", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13559, "uid": "f516dfb84b9051ed85b89cdc3a8ab7f5", "name": "Generalization Bounds for Spectral GNNs via Fourier Domain Analysis", "authors": [{"id": 22364, "fullname": "Vahan Martirosyan", "url": "http://virtual.aistats.org/api/miniconf/users/22364?format=json", "institution": "Centrale Sup\u00e9lec"}, {"id": 22365, "fullname": "Daniele Malitesta", "url": "http://virtual.aistats.org/api/miniconf/users/22365?format=json", "institution": "Centrale Sup\u00e9lec"}, {"id": 22366, "fullname": "Hugues Talbot", "url": "http://virtual.aistats.org/api/miniconf/users/22366?format=json", "institution": "CentraleSupelec"}, {"id": 22367, "fullname": "Jhony Giraldo", "url": "http://virtual.aistats.org/api/miniconf/users/22367?format=json", "institution": "T\u00e9l\u00e9com Paris, Institut Polytechnique de Paris"}, {"id": 22368, "fullname": "Fragkiskos Malliaros", "url": "http://virtual.aistats.org/api/miniconf/users/22368?format=json", "institution": "CentraleSup\u00e9lec, Inria, Paris-Saclay University"}], "abstract": "Spectral graph neural networks learn graph filters, but their behavior with increasing depth and polynomial order is not well understood.  We analyze these models in the graph Fourier domain, where each layer becomes an element-wise frequency update, separating the fixed spectrum from trainable parameters and making depth and order explicit.  In this setting, we show that Gaussian complexity is invariant under the Graph Fourier Transform, which allows us to derive data-dependent, depth, and order-aware generalization bounds together with stability estimates. In the linear case, our bounds are tighter, and on real graphs, the data-dependent term correlates with the generalization gap across polynomial bases, highlighting practical choices that avoid frequency amplification across layers.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13559", "url": null, "sourceid": 2309, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=YyqkBu96tG", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11252, "modified": "2026-03-29T20:43:03.536297-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=YyqkBu96tG", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "77", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13562, "uid": "f3144cefe89a60d6a1afaf7859c5076b", "name": "Value Gradient Sampler: Learning Invariant Value Functions for Equivariant Diffusion Sampling", "authors": [{"id": 22373, "fullname": "Himchan Hwang", "url": "http://virtual.aistats.org/api/miniconf/users/22373?format=json", "institution": "Seoul National University"}, {"id": 22374, "fullname": "Hyeokju Jeong", "url": "http://virtual.aistats.org/api/miniconf/users/22374?format=json", "institution": "Seoul National University"}, {"id": 22375, "fullname": "Dong Shin", "url": "http://virtual.aistats.org/api/miniconf/users/22375?format=json", "institution": "Korea Advanced Institute of Science &amp; Technology"}, {"id": 22376, "fullname": "Che-Sang Park", "url": "http://virtual.aistats.org/api/miniconf/users/22376?format=json", "institution": "Seoul National University"}, {"id": 22377, "fullname": "Sehee Kweon", "url": "http://virtual.aistats.org/api/miniconf/users/22377?format=json", "institution": "SAIGE"}, {"id": 22378, "fullname": "Sangwoong Yoon", "url": "http://virtual.aistats.org/api/miniconf/users/22378?format=json", "institution": "UNIST"}, {"id": 22379, "fullname": "Frank Park", "url": "http://virtual.aistats.org/api/miniconf/users/22379?format=json", "institution": "Seoul National University"}], "abstract": "We propose the Value Gradient Sampler (VGS), a diffusion sampler parameterized by value functions. VGS generates samples from an unnormalized target density (i.e., energy) by evolving randomly initialized particles along the gradient of the value function. In many sampling problems where the target density exhibits equivariant symmetries, we show that value functions enable a novel approach to leveraging invariant neural networks for sampling, as an invariant value function induces an equivariant gradient flow. The value functions are trained via temporal-difference learning, which supports off-policy training and other established reinforcement learning (RL) techniques. By combining efficient invariant neural networks with advanced RL methods, VGS achieves strong performance in high-dimensional particle systems, including Lennard-Jones systems with up to 55 particles.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13562", "url": null, "sourceid": 1719, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=YmV7Sc80YK", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11255, "modified": "2026-03-29T20:43:03.653954-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=YmV7Sc80YK", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "187", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13564, "uid": "ce5140df15d046a66883807d18d0264b", "name": "Provable Accelerated Bayesian Optimization with Knowledge Transfer", "authors": [{"id": 22382, "fullname": "Haitao Lin", "url": "http://virtual.aistats.org/api/miniconf/users/22382?format=json", "institution": "University of Chicago"}, {"id": 22383, "fullname": "Boxin Zhao", "url": "http://virtual.aistats.org/api/miniconf/users/22383?format=json", "institution": "University of Chicago"}, {"id": 18116, "fullname": "Mladen Kolar", "url": "http://virtual.aistats.org/api/miniconf/users/18116?format=json", "institution": "University of Southern California"}, {"id": 18477, "fullname": "Chong Liu", "url": "http://virtual.aistats.org/api/miniconf/users/18477?format=json", "institution": "University at Albany, State University of New York"}], "abstract": "We study how to accelerate Bayesian optimization (BO) on a target task by transferring historical knowledge from related source tasks. Existing work on BO with knowledge transfer either lacks theoretical guarantees or achieves the same regret as BO in the non-transfer setting, $\\tilde{\\mathcal{O}}(\\sqrt{T \\gamma_f})$, where $T$ is the number of evaluations of the target function and $\\gamma_f$ denotes its information gain. In this paper, we propose the DeltaBO algorithm, which builds a novel uncertainty-quantification approach on the difference function $\\delta$ between the source and target functions, which are allowed to belong to different Reproducing Kernel Hilbert Spaces (RKHSs). Under mild assumptions, we prove that the regret of DeltaBO is of order $\\tilde{\\mathcal{O}}(\\sqrt{T(T/N+\\gamma_\\delta)})$, where $N$ denotes the number of evaluations from source tasks and typically $N \\gg T$. In many applications, source and target tasks are similar, which implies that $\\gamma_\\delta$ can be much smaller than $\\gamma_f$. Empirical studies on both real-world hyperparameter-tuning tasks and synthetic functions show that DeltaBO outperforms other baseline methods and also verify our theoretical claims. Our code is available on GitHub.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13564", "url": null, "sourceid": 1023, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=YaYwkfkLCa", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11257, "modified": "2026-03-29T20:43:03.719139-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=YaYwkfkLCa", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "131", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13566, "uid": "d1f491a404d6854880943e5c3cd9ca25", "name": "Rethinking Cross-Modal Fine-Tuning: Optimizing the Interaction Between Feature Alignment and Target Fitting", "authors": [{"id": 22385, "fullname": "Trong Khiem Tran", "url": "http://virtual.aistats.org/api/miniconf/users/22385?format=json", "institution": "Hanoi University of Science and Technology"}, {"id": 22386, "fullname": "Manh Dao", "url": "http://virtual.aistats.org/api/miniconf/users/22386?format=json", "institution": "National University of Singapore"}, {"id": 22387, "fullname": "Phi Le Nguyen", "url": "http://virtual.aistats.org/api/miniconf/users/22387?format=json", "institution": "Hanoi University of Science and Technology"}, {"id": 22388, "fullname": "Thao Nguyen Truong", "url": "http://virtual.aistats.org/api/miniconf/users/22388?format=json", "institution": "AIST, National Institute of Advanced Industrial Science and Technology"}, {"id": 3617, "fullname": "Nghia Hoang", "url": "http://virtual.aistats.org/api/miniconf/users/3617?format=json", "institution": "Independent Researcher"}], "abstract": "Adapting pre-trained models to unseen feature modalities has become increasingly important due to the growing need for cross-disciplinary knowledge integration. A key challenge here is how to align the representation of new modalities with the most relevant parts of the pre-trained model's representation space to enable accurate knowledge transfer. This requires combining feature alignment with target fine-tuning, but uncalibrated combinations can exacerbate misalignment between the source and target feature-label structures and reduce target generalization. Existing work however lacks a theoretical understanding of this critical interaction between feature alignment and target fitting. To bridge this gap, we develop a principled framework that establishes a provable generalization bound on the target error, which explains the interaction between feature alignment and target fitting through a novel concept of feature-label distortion. This bound offers actionable insights into how this interaction should be optimized for practical algorithm design. The resulting approach achieves significantly improved performance over state-of-the-art methods across a wide range of benchmark datasets.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13566", "url": null, "sourceid": 129, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=YXPoM9GI12", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11259, "modified": "2026-03-29T20:43:03.801984-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=YXPoM9GI12", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "154", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13573, "uid": "b7bb35b9c6ca2aee2df08cf09d7016c2", "name": "Structured Difference-of-Q via Orthogonal Learning", "authors": [{"id": 22403, "fullname": "Defu Cao", "url": "http://virtual.aistats.org/api/miniconf/users/22403?format=json", "institution": "University of Southern California"}, {"id": 22404, "fullname": "Angela Zhou", "url": "http://virtual.aistats.org/api/miniconf/users/22404?format=json", "institution": "University of Southern California"}], "abstract": "Offline reinforcement learning is important in many settings with available observational data but the inability to deploy new policies online due to safety, cost, and other concerns. Many recent advances in causal inference and machine learning target estimation of ``causal contrast\" functions such as CATE, which can adapt to potentially smoother structure. We develop a dynamic generalization of the R-learner (Nie et al 2021, Lewis and Syrgkanis 2021) for estimating and optimizing the difference of $Q^\\pi$-functions, $Q^\\pi(s,a)-Q^\\pi(s,a_0)$, for potential discrete-valued actions $a,a_0$, which can be used to optimize multiple-valued actions without loss of generality. We leverage orthogonal estimation to improve convergence rates, even if $Q$ and behavior policy (so-called nuisance functions) converge at slower rates and prove consistency of policy optimization under a margin condition. The method can leverage black-box  estimators of the $Q$-function and behavior policy to target estimation of a more structured $Q$-function contrast, and comprises of simple squared-loss minimization.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13573", "url": null, "sourceid": 631, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=XgFOXbm9d7", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11266, "modified": "2026-03-29T20:43:04.154906-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=XgFOXbm9d7", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "179", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13582, "uid": "eaa32c96f620053cf442ad32258076b9", "name": "Zeroth-Order Stochastic Compositional Gradient Descent: Towards Black-Box Sparse AUC Maximization", "authors": [{"id": 19889, "fullname": "WENKANG WANG", "url": "http://virtual.aistats.org/api/miniconf/users/19889?format=json", "institution": "Jilin University"}, {"id": 22419, "fullname": "Dongxu Liu", "url": "http://virtual.aistats.org/api/miniconf/users/22419?format=json", "institution": "Jilin University"}, {"id": 9956, "fullname": "Bin Gu", "url": "http://virtual.aistats.org/api/miniconf/users/9956?format=json", "institution": "MBZUAI"}], "abstract": "The area under the ROC curve (AUC) is a key metric for classification tasks, valued for its robustness to class imbalance. Sparse models trained with $\\ell_0$ constraints further enhance interpretability and generalization. Building on prior work that reformulates nonlinear AUC maximization as a pointwise compositional optimization problem, we revisit this formulation as the basis for addressing the black-box setting, where only function evaluations are available. A central challenge arises from integrating zeroth-order gradient estimation with hard-thresholding operators in the compositional framework, which has remained unresolved. To overcome this difficulty, we propose the Zeroth-Order Stochastic Compositional Hard-Thresholding (ZO-SCHT) algorithm, which, to the best of our knowledge, is the first method for black-box sparse AUC maximization. We establish that ZO-SCHT achieves linear convergence up to a tolerance bound under a fixed step size. Extensive experiments on both black-box sparse AUC maximization and black-box adversarial attack tasks demonstrate the effectiveness and versatility of our approach.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13582", "url": null, "sourceid": 1436, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=WtT8aqsoL9", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11275, "modified": "2026-03-29T20:43:04.422482-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=WtT8aqsoL9", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "183", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13583, "uid": "03255088ed63354a54e0e5ed957e9008", "name": "Adversary-Free Counterfactual Prediction via Information-Regularized Representations", "authors": [{"id": 22420, "fullname": "Shiqin Tang", "url": "http://virtual.aistats.org/api/miniconf/users/22420?format=json", "institution": "Centre for Artificial Intelligence and Robotics Hong Kong Institute of Science &amp; Innovation, Chinese Academy of Sciences"}, {"id": 22421, "fullname": "Rong Feng", "url": "http://virtual.aistats.org/api/miniconf/users/22421?format=json", "institution": "City University of Hong Kong"}, {"id": 22422, "fullname": "Shuxin Zhuang", "url": "http://virtual.aistats.org/api/miniconf/users/22422?format=json", "institution": "City University of Hong Kong"}, {"id": 23267, "fullname": "Hongzong LI", "url": "http://virtual.aistats.org/api/miniconf/users/23267?format=json", "institution": "Hong Kong University of Science and Technology"}, {"id": 22424, "fullname": "Youzhi Zhang", "url": "http://virtual.aistats.org/api/miniconf/users/22424?format=json", "institution": "Centre for Artificial Intelligence and Robotics, Hong Kong Institute of Science &amp; Innovation, Chinese Academy of Sciences"}], "abstract": "We study counterfactual prediction under assignment bias and propose a mathematically grounded, information-theoretic approach that removes treatment\u2013covariate dependence without adversarial training. Starting from a bound that links the counterfactual\u2013factual risk gap to mutual information, we learn a stochastic representation $Z$ that is predictive of outcomes while minimizing $I(Z;T)$. We derive a tractable variational objective that upper-bounds the information term and couples it with a supervised decoder, yielding a stable, provably motivated training criterion. The framework extends naturally to dynamic settings by applying the information penalty to sequential representations at each decision time. We evaluate the method on controlled numerical simulations and a real-world clinical dataset, comparing against recent state-of-the-art balancing, reweighting, and adversarial baselines. Across metrics of likelihood, counterfactual error, and policy evaluation, our approach performs favorably while avoiding the training instabilities and tuning burden of adversarial schemes.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13583", "url": null, "sourceid": 2327, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=WsAzOdVg7X", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11276, "modified": "2026-03-29T20:43:04.459343-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=WsAzOdVg7X", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "18", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13584, "uid": "2612aa892d962d6f8056b195ca6e550d", "name": "CADENT: Gated Hybrid Distillation for Sample-Efficient Transfer in Reinforcement Learning", "authors": [{"id": 22425, "fullname": "Mahyar Alinejad", "url": "http://virtual.aistats.org/api/miniconf/users/22425?format=json", "institution": "University of Central Florida"}, {"id": 22426, "fullname": "Yue Wang", "url": "http://virtual.aistats.org/api/miniconf/users/22426?format=json", "institution": "University of Central Florida"}, {"id": 20591, "fullname": "George Atia", "url": "http://virtual.aistats.org/api/miniconf/users/20591?format=json", "institution": "University of Central Florida"}], "abstract": "Transfer learning promises to reduce the high sample complexity of deep reinforcement learning (RL), yet existing methods struggle with domain shift between source and target environments. Policy distillation provides powerful tactical guidance but fails to transfer long-term strategic knowledge, while automaton-based methods capture task structure but lack fine-grained action guidance. We introduce Context-Aware Distillation with Experience-gated Transfer (CADENT), a framework that unifies strategic automaton-based knowledge with tactical policy-level knowledge into a coherent guidance signal. CADENT's key innovation is an experience-gated trust mechanism that dynamically weighs teacher guidance against the student's own experience at the state-action level, enabling graceful adaptation to target domain specifics. Across challenging environments, from sparse-reward grid worlds to continuous control tasks, CADENT achieves 40-60\\% better sample efficiency than baselines while maintaining superior asymptotic performance, establishing a robust approach for adaptive knowledge transfer in RL.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13584", "url": null, "sourceid": 1758, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=WkCSlfSa0B", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11277, "modified": "2026-03-29T20:43:04.490676-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=WkCSlfSa0B", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "36", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13588, "uid": "96671501524948bc3937b4b30d0e57b9", "name": "Bayesian Fourier Features for Reduced Rank Gaussian Processes", "authors": [{"id": 19585, "fullname": "Cristian A. Galvis-Florez", "url": "http://virtual.aistats.org/api/miniconf/users/19585?format=json", "institution": "Aalto University"}, {"id": 22443, "fullname": "George Whittle", "url": "http://virtual.aistats.org/api/miniconf/users/22443?format=json", "institution": "University of Oxford"}, {"id": 4535, "fullname": "Michael A. Osborne", "url": "http://virtual.aistats.org/api/miniconf/users/4535?format=json", "institution": "University of Oxford"}, {"id": 18340, "fullname": "Simo S\u00e4rkk\u00e4", "url": "http://virtual.aistats.org/api/miniconf/users/18340?format=json", "institution": "Aalto University"}], "abstract": "Gaussian processes are probabilistic models used in machine learning and the physical sciences, although they are limited by cubic complexity in the number of training observations. To mitigate this problem, various low-rank kernel approximation methods, including Fourier feature methods, Hilbert space methods, and inducing point methods, have been developed.     In this paper, we propose a novel Fourier feature approach leveraging Bayesian quadrature methods to construct reduced-rank approximations of the Gaussian process kernel. The new Bayesian Fourier feature framework also unifies many previously proposed low-rank methods, as they can be seen as instances of Bayesian quadrature-based approximations of Gaussian process kernels. Due to its probabilistic nature, the unified framework also enables the quantification of uncertainty in the approximation. Furthermore, the framework allows for the design of entirely new low-rank kernel approximations.    We compare the performance of the proposed methods with other approaches across different kernel length scales. Our experimental results demonstrate that it outperforms other popular low-rank kernel approximation methods across a wide range of length scales.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13588", "url": null, "sourceid": 2279, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=WVEVvwrUxx", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11281, "modified": "2026-03-29T20:43:04.640993-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=WVEVvwrUxx", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "30", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13589, "uid": "0ce2ffd21fc958d9ef0ee9ba5336e357", "name": "Computationally Lightweight Classifiers with Frequentist Bounds on Prediction Errors", "authors": [{"id": 19856, "fullname": "Shreeram Murali", "url": "http://virtual.aistats.org/api/miniconf/users/19856?format=json", "institution": "Aalto University"}, {"id": 22444, "fullname": "Cristian Rojas", "url": "http://virtual.aistats.org/api/miniconf/users/22444?format=json", "institution": "KTH Royal Institute of Technology"}, {"id": 18123, "fullname": "Dominik Baumann", "url": "http://virtual.aistats.org/api/miniconf/users/18123?format=json", "institution": "Aalto University"}], "abstract": "While both classical and neural network classifiers can achieve high accuracy, they fall short on offering uncertainty bounds on their predictions, making them unfit for safety-critical applications. Existing kernel-based classifiers that provide such bounds scale with $\\mathcal O (n^{\\sim3})$ in time, making them computationally intractable for large datasets. To address this, we propose a novel, computationally efficient classification algorithm based on the Nadaraya-Watson estimator, for whose estimates we derive frequentist uncertainty intervals. We evaluate our classifier on synthetically generated data and on electrocardiographic heartbeat signals from the MIT-BIH Arrhythmia database. We show that the method achieves competitive accuracy >96% at $\\mathcal O(n)$ operations, while providing actionable uncertainty bounds. These bounds can, e.g., aid in flagging low-confidence predictions, making them suitable for real-time settings with resource constraints, such as diagnostic monitoring or implantable devices.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13589", "url": null, "sourceid": 1265, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=WS0lO2axHh", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11282, "modified": "2026-03-29T20:43:04.685925-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=WS0lO2axHh", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "28", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13590, "uid": "ec8ce6abb3e952a85b8551ba726a1227", "name": "Enhancing LLM Safety through a Theoretical Minimax Game Lens", "authors": [{"id": 17959, "fullname": "Yihe Deng", "url": "http://virtual.aistats.org/api/miniconf/users/17959?format=json", "institution": "University of California, Los Angeles"}, {"id": 22445, "fullname": "Yu Yang", "url": "http://virtual.aistats.org/api/miniconf/users/22445?format=json", "institution": "OpenAI"}, {"id": 22446, "fullname": "Junkai Zhang", "url": "http://virtual.aistats.org/api/miniconf/users/22446?format=json", "institution": "Google"}, {"id": 22447, "fullname": "Wei Wang", "url": "http://virtual.aistats.org/api/miniconf/users/22447?format=json", "institution": "University of California, Los Angeles"}, {"id": 1263, "fullname": "Bo Li", "url": "http://virtual.aistats.org/api/miniconf/users/1263?format=json", "institution": "UIUC"}], "abstract": "The rapid advancement of large language models (LLMs) necessitates effective mechanisms to ensure their responsible deployment by accurately distinguishing unsafe content from benign content. While substantial safety datasets are available in English, multilingual safety modeling remains underexplored due to limited open-source safety datasets in other languages. Even within English datasets, safe yet sensitive corner-case content is scarce, leading to shortcut learning by models and non-trivial false-positive rates. To mitigate these issues, we introduce a novel minimax reinforcement learning (RL) framework wherein a data generator and a classifier model co-evolve, facilitating the production of high-quality synthetic multilingual safety data. We theoretically formalize this interaction as a minimax game and rigorously demonstrate convergence to a Nash equilibrium. Empirical evaluations confirm that our synthetic data generation method significantly enhances the classifier model performance, enabling a substantially smaller model to surpass the state-of-the-art by nearly 10\\% on English benchmarks while achieving 4.5$\\times$ faster inference speed. These results establish a scalable and efficient methodology for synthetic data generation, advancing the development of safer and more robust multilingual LLM deployments.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13590", "url": null, "sourceid": 220, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=WGBrHMcews", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11283, "modified": "2026-03-29T20:43:04.722465-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=WGBrHMcews", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "58", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13591, "uid": "d757719ed7c2b66dd17dcee2a3cb29f4", "name": "Breaking Data Symmetry is Needed For Generalization in Feature Learning Kernels", "authors": [{"id": 19801, "fullname": "Marcel Bernal", "url": "http://virtual.aistats.org/api/miniconf/users/19801?format=json", "institution": "Universitat Polit\u00e8cnica de Catalunya"}, {"id": 22448, "fullname": "Neil Mallinar", "url": "http://virtual.aistats.org/api/miniconf/users/22448?format=json", "institution": "University of California, San Diego"}, {"id": 22449, "fullname": "Mikhail Belkin", "url": "http://virtual.aistats.org/api/miniconf/users/22449?format=json", "institution": "University of California, San Diego"}], "abstract": "Grokking occurs when a model achieves high training accuracy but generalization to unseen test points happens long after that. This phenomenon was initially observed on a class of algebraic problems, such as learning modular arithmetic (Power et al., 2022). We study grokking on algebraic tasks in a class of feature learning kernels via the Recursive Feature Machine (RFM) algorithm (Radhakrishnan et al., 2024), which iteratively updates feature matrices through the Average Gradient Outer Product (AGOP) of an estimator in order to learn task-relevant features.  Our main experimental finding is that generalization occurs only when a certain symmetry in the training set is broken. Furthermore, we empirically show that RFM generalizes by recovering the underlying invariance group action inherent in the data. We find that the learned feature matrices encode specific elements of the invariance group, explaining the dependence of generalization on symmetry.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13591", "url": null, "sourceid": 1834, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=WC49CwQfyp", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11284, "modified": "2026-03-29T20:43:04.763477-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=WC49CwQfyp", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "35", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13597, "uid": "bc6dc48b743dc5d013b1abaebd2faed2", "name": "PENGUIN: Enhancing Transformer with Periodic-Nested Group Attention for Long-term Time Series Forecasting", "authors": [{"id": 22458, "fullname": "Tian Sun", "url": "http://virtual.aistats.org/api/miniconf/users/22458?format=json", "institution": "Fudan University"}, {"id": 22459, "fullname": "Yuqi Chen", "url": "http://virtual.aistats.org/api/miniconf/users/22459?format=json", "institution": "Xiaohongshu"}, {"id": 22460, "fullname": "Weiwei Sun", "url": "http://virtual.aistats.org/api/miniconf/users/22460?format=json", "institution": "Fudan University"}], "abstract": "Despite advances in the Transformer architecture, their effectiveness for long-term time series forecasting (LTSF) remains controversial. In this paper, we investigate the potential of integrating explicit periodicity modeling into the self-attention mechanism to enhance the performance of Transformerbased architectures for LTSF. Specifically, we propose PENGUIN, a simple yet effective periodic-nested group attention mechanism. Our approach introduces a periodic-aware relative attention bias to directly capture periodic structures and a grouped multi-query attention mechanism to handle multiple coexisting periodicities (e.g., daily and weekly cycles) within time series data. Extensive experiments across diverse benchmarks demonstrate that PENGUIN consistently outperforms both MLP-based and Transformer-based models. Code is available at https://github.com/ysygMhdxw/AISTATS2026_PENGUIN.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13597", "url": null, "sourceid": 323, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=UokZf6B9lQ", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11290, "modified": "2026-03-29T20:43:05.101753-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=UokZf6B9lQ", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "136", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13598, "uid": "67d16d00201083a2b118dd5128dd6f59", "name": "Stationarity-Aware Causal Discovery in Time Series via Minimal Separating Sets", "authors": [{"id": 18655, "fullname": "Shanyun Gao", "url": "http://virtual.aistats.org/api/miniconf/users/18655?format=json", "institution": "Purdue University"}, {"id": 18653, "fullname": "Raghavendra Addanki", "url": "http://virtual.aistats.org/api/miniconf/users/18653?format=json", "institution": "Adobe Systems"}, {"id": 18656, "fullname": "Tong Yu", "url": "http://virtual.aistats.org/api/miniconf/users/18656?format=json", "institution": "Adobe Research"}, {"id": 18667, "fullname": "Ryan Rossi", "url": "http://virtual.aistats.org/api/miniconf/users/18667?format=json", "institution": "Adobe Research"}, {"id": 438, "fullname": "Qifan Song", "url": "http://virtual.aistats.org/api/miniconf/users/438?format=json", "institution": "Purdue University "}, {"id": 19919, "fullname": "Murat Kocaoglu", "url": "http://virtual.aistats.org/api/miniconf/users/19919?format=json", "institution": "Johns Hopkins University"}], "abstract": "Discovering causal relationships from observational time series is a fundamental problem with broad applications in climate science, healthcare, and finance. Causal graphs with time-lagged structure capture the effects of underlying mechanisms over time. Under the causal stationarity assumption, these causal mechanisms remain consistent across time. Existing constraint-based methods leverage stationarity for conditional independence testing and reduce the problem to learning the parents of variables at the final time point, which can then be used to reconstruct the stationary graph. However, their separating set search strategy mimics the PC algorithm and does not take advantage of the stationary structure. We observe that the stationary graph structure and autoregressive edges impose many meaningful constraints on the separating sets between variables at different time lags. After characterizing the behavior of such separating sets, we propose a novel causal discovery algorithm that exploits this structure of minimal separating sets. Extensive evaluations on synthetic and real-world datasets demonstrate the robustness and accuracy of our method.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13598", "url": null, "sourceid": 876, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=UimZ4w1Gdf", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11291, "modified": "2026-03-29T20:43:05.146959-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=UimZ4w1Gdf", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "151", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13601, "uid": "7e9e346dc5fd268b49bf418523af8679", "name": "Retrieval Augmented Time Series Forecasting", "authors": [{"id": 19816, "fullname": "Kutay Tire", "url": "http://virtual.aistats.org/api/miniconf/users/19816?format=json", "institution": "University of Texas at Austin"}, {"id": 22464, "fullname": "Ege Taga", "url": "http://virtual.aistats.org/api/miniconf/users/22464?format=json", "institution": "University of Michigan - Ann Arbor"}, {"id": 11034, "fullname": "Muhammed Emrullah Ildiz", "url": "http://virtual.aistats.org/api/miniconf/users/11034?format=json", "institution": "University of Michigan, Ann Arbor"}, {"id": 12995, "fullname": "Samet Oymak", "url": "http://virtual.aistats.org/api/miniconf/users/12995?format=json", "institution": "University of California, Riverside"}], "abstract": "Retrieval-augmented generation (RAG) is a central component of modern LLM systems, particularly in scenarios where up-to-date information is crucial for accurately responding to user queries or when queries exceed the scope of the training data. The advent of time-series foundation models (TSFM), such as Chronos or Moirai, and the need for effective zero-shot forecasting performance across various time-series domains motivates the question: Do the benefits of RAG similarly carry over to time series forecasting? In this paper, we advocate that the dynamic and event-driven nature of time-series data makes RAG a crucial component of TSFMs and introduce a principled RAG framework for time-series forecasting, called Retrieval Augmented Forecasting (RAF). Within RAF, we develop efficient strategies for retrieving related time-series examples and incorporating them into the forecast. Through experiments and mechanistic studies, we demonstrate that RAF indeed improves the forecasting accuracy across diverse time series domains and TSFMs, with gains that are more pronounced for larger models.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13601", "url": null, "sourceid": 1875, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=UD76JhLswg", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11294, "modified": "2026-03-29T20:43:05.255220-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=UD76JhLswg", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "157", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13604, "uid": "f7664060cc52bc6f3d620bcedc94a4b6", "name": "Explanation Design in Strategic Learning: Sufficient Explanations that Induce Non-harmful Responses", "authors": [{"id": 22475, "fullname": "Kiet Vo", "url": "http://virtual.aistats.org/api/miniconf/users/22475?format=json", "institution": "CISPA Helmholtz Center for Information Security"}, {"id": 22476, "fullname": "Siu Lun Chau", "url": "http://virtual.aistats.org/api/miniconf/users/22476?format=json", "institution": "Nanyang Technological University"}, {"id": 10229, "fullname": "Masahiro Kato", "url": "http://virtual.aistats.org/api/miniconf/users/10229?format=json", "institution": "Mizuho-DL Financial Technology Co., Ltd."}, {"id": 12393, "fullname": "Yixin Wang", "url": "http://virtual.aistats.org/api/miniconf/users/12393?format=json", "institution": "University of Michigan"}, {"id": 17685, "fullname": "Krikamol Muandet", "url": "http://virtual.aistats.org/api/miniconf/users/17685?format=json", "institution": "CISPA Helmholtz Center for Information Security"}], "abstract": "We study the design of explanations in algorithmic decision-making with strategic agents---individuals who may modify their inputs in response to explanations of a decision maker's (DM's) predictive model. While the demand for algorithmic transparency has led much prior work to assume full model disclosure, in practice DMs typically provide only partial information via explanations, which can cause agents to misinterpret the model and take actions that unintentionally reduce their own utility. A central open question is therefore how DMs should communicate explanations that avoid harming strategic agents while still supporting their own goals, e.g., minimising predictive error. In this work, we analyse widely used explanation methods and establish a necessary condition to prevent explanations from inducing self-harming responses. Furthermore, we show that action recommendation-based explanations  (ARexes), which encompass counterfactual explanations, are sufficient to induce all non-harmful responses. Under a conditional homogeneity assumption, this sufficiency extends to ARex-generating methods, echoing the revelation principle in information design. To demonstrate their practical utility, we introduce a simple learning procedure that jointly optimises the predictive model and the explanation-generating policy. Experiments on both synthetic and real-world tasks show that ARexes enable DMs to achieve high predictive performance while preserving agents' utility, offering a principled strategy for safe and effective partial model disclosure.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13604", "url": null, "sourceid": 266, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=TpimCQwkR9", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11297, "modified": "2026-03-29T20:43:05.370765-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=TpimCQwkR9", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "60", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13606, "uid": "c3992e9a68c5ae12bd18488bc579b30d", "name": "Structured Matrix Scaling for Multi-Class Calibration", "authors": [{"id": 22478, "fullname": "Eug\u00e8ne Berta", "url": "http://virtual.aistats.org/api/miniconf/users/22478?format=json", "institution": "INRIA - \u00c9cole Normale Sup\u00e9rieure"}, {"id": 5755, "fullname": "David Holzm\u00fcller", "url": "http://virtual.aistats.org/api/miniconf/users/5755?format=json", "institution": "INRIA Saclay"}, {"id": 366, "fullname": "Michael Jordan", "url": "http://virtual.aistats.org/api/miniconf/users/366?format=json", "institution": "UC Berkeley"}, {"id": 886, "fullname": "Francis Bach", "url": "http://virtual.aistats.org/api/miniconf/users/886?format=json", "institution": "INRIA - Ecole Normale Sup\u00e9rieure"}], "abstract": "Post-hoc recalibration methods are widely used to ensure that classifiers provide faithful probability estimates. We argue that parametric recalibration functions based on logistic regression can be motivated from a simple theoretical setting for both binary and multi-class classification. This insight motivates the use of more expressive calibration methods beyond standard temperature scaling. For multi-class calibration however, a key challenge lies in the increasing number of parameters introduced by more complex models, often coupled with limited calibration data, which can lead to overfitting. Through extensive experiments, we demonstrate that the resulting bias-variance tradeoff can be effectively managed by structured regularization, robust preprocessing and efficient optimization. The resulting methods lead to substantial gains over existing logistic-based calibration techniques. We provide efficient and easy-to-use open-source implementations of our methods, making them an attractive alternative to common temperature, vector, and matrix scaling implementations.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13606", "url": null, "sourceid": 602, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=TQSxZcPgRi", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11299, "modified": "2026-03-29T20:43:05.444443-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=TQSxZcPgRi", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "180", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13608, "uid": "8d5e957f297893487bd98fa830fa6413", "name": "Projection-free Algorithms for Online Convex Optimization with Adversarial Constraints", "authors": [{"id": 22187, "fullname": "Dhruv Sarkar", "url": "http://virtual.aistats.org/api/miniconf/users/22187?format=json", "institution": "Indian Institute of Technology Kharagpur"}, {"id": 22188, "fullname": "Aprameyo Chakrabartty", "url": "http://virtual.aistats.org/api/miniconf/users/22188?format=json", "institution": "Indian Institute of Technology Kharagpur"}, {"id": 22481, "fullname": "Subhamon Supantha", "url": "http://virtual.aistats.org/api/miniconf/users/22481?format=json", "institution": "Chennai Mathematical Institute, Dhirubhai Ambani Institute Of Information and Communication Technology"}, {"id": 22482, "fullname": "Palash Dey", "url": "http://virtual.aistats.org/api/miniconf/users/22482?format=json", "institution": "Indian Institute of Technology Kharagpur"}, {"id": 22483, "fullname": "Abhishek Sinha", "url": "http://virtual.aistats.org/api/miniconf/users/22483?format=json", "institution": "Tata Institute of Fundamental Research"}], "abstract": "We study a generalization of the Online Convex Optimization (OCO) framework with time-varying adversarial constraints. In this setting, at each round, the learner selects an action from a convex decision set $\\mathcal{X}$, after which both a convex cost function and a convex constraint function are revealed. The objective is to design a computationally efficient learning policy that simultaneously achieves low regret with respect to the cost functions and low cumulative constraint violation (CCV) over a horizon of length $T$. A major computational bottleneck in standard OCO algorithms is the projection operation onto the decision set $\\mathcal{X}$. However, for many structured decision sets, linear optimization can be performed efficiently. Motivated by this, we propose a \\emph{projection-free} online conditional gradient (OCG)-based algorithm that requires only a single call to a linear optimization oracle over $\\mathcal{X}$ per round. Our approach improves upon the state of the art for projection-free online learning with adversarial constraints, achieving $\\tilde{O}(T^{3/4})$ bounds for both regret and CCV.  Our algorithm is conceptually simple. It constructs a surrogate cost function as a nonnegative linear combination of the cost and constraint functions, and feeds these surrogate costs into a novel adaptive online conditional gradient subroutine introduced in this paper. We further extend our framework to the bandit setting, where we show that a new form of surrogate loss is necessary to properly handle bandit feedback\u2014an issue overlooked in prior work. Finally, we develop an efficient Follow-the-Perturbed-Leader (FTPL)-based algorithm, particularly well-suited for online combinatorial optimization problems with discrete actions, which also achieves $O(T^{3/4})$ regret and CCV.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13608", "url": null, "sourceid": 147, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=TCCrm6XGi8", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11301, "modified": "2026-03-29T20:43:05.508408-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=TCCrm6XGi8", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "145", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13614, "uid": "7c82fab8c8f89124e2ce92984e04fb40", "name": "FlowPINNs: A Variational Framework for PDE Parameter Inference and Uncertainty Quantification", "authors": [{"id": 20600, "fullname": "David Dalton", "url": "http://virtual.aistats.org/api/miniconf/users/20600?format=json", "institution": "University of Glasgow"}, {"id": 22493, "fullname": "Hao Gao", "url": "http://virtual.aistats.org/api/miniconf/users/22493?format=json", "institution": "University of Glasgow"}, {"id": 22494, "fullname": "Dirk Husmeier", "url": "http://virtual.aistats.org/api/miniconf/users/22494?format=json", "institution": "University of Glasgow"}], "abstract": "Inverse problems for parameter identification in systems governed by partial differential equations (PDEs) are fundamental across numerous domains in science and engineering. While traditionally approached using classical numerical techniques, recent advancements have highlighted the potential of physics-informed neural networks (PINNs) in this context. However, incorporating principled uncertainty quantification (UQ) into PINN frameworks remains a challenge. To address this, we introduce \\textit{flowPINNs}, a probabilistic framework for UQ in PDE parameter inverse problems. The core idea of a flowPINN is to define a variational posterior that combines a normalising flow approximation to the distribution over the PDE parameters with a PINN that represents the corresponding PDE solution. This joint model enables efficient posterior inference by maximisation of the evidence lower bound (ELBO), thereby converting the inverse problem into a tractable optimisation task. Through a series of numerical experiments, we demonstrate that flowPINNs can yield improved predictive performance and more reliable uncertainty estimates compared to existing PINN-based UQ approaches.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13614", "url": null, "sourceid": 1293, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=SqaYpW6poC", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11307, "modified": "2026-03-29T20:43:05.714471-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=SqaYpW6poC", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "60", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13618, "uid": "c4492cbe90fbdbf88a5aec486aa81ed5", "name": "Learning Right Monotone Permutation Matrices for Neural Subsequence Search", "authors": [{"id": 19786, "fullname": "Bhavya Kohli", "url": "http://virtual.aistats.org/api/miniconf/users/19786?format=json", "institution": "National University of Singapore"}, {"id": 22500, "fullname": "Soutrik Sarangi", "url": "http://virtual.aistats.org/api/miniconf/users/22500?format=json", "institution": "Google"}, {"id": 23268, "fullname": "Aziz Shameem", "url": "http://virtual.aistats.org/api/miniconf/users/23268?format=json", "institution": "Indian Institute of Technology, Bombay, Dhirubhai Ambani Institute Of Information and Communication Technology"}, {"id": 22502, "fullname": "Abir De", "url": "http://virtual.aistats.org/api/miniconf/users/22502?format=json", "institution": "Indian Institute of Technology Bombay,"}], "abstract": "Subsequence retrieval seeks relevant segments in a large corpus given a short query. Existing pairwise metric-based methods are computationally intensive, hard to parallelize, and tied to domain-specific metrics. In this work, we introduce a neural framework that casts subsequence matching as end-to-end alignment with permutation matrices satisfying monotonicity used as differentiable approximate subsequence selectors. Our framework yields fixed-dimensional embeddings for variable-length inputs, and we prove these embeddings are compatible with standard Approximate Nearest Neighbor search methods such as Locality-sensitive hashing (LSH), enabling scalable retrieval. We also impose structural priors on admissible subsequences and integrate them directly into the scoring function. The approach is domain-agnostic and operates on pre-trained representations across modalities. Experiments on real-world datasets from two different domains show strong retrieval performance and substantial speedups, with high parallelism on GPU-accelerated hardware.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13618", "url": null, "sourceid": 1735, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=SBgs6trkAD", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11311, "modified": "2026-03-29T20:43:05.848093-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=SBgs6trkAD", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "93", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13619, "uid": "5e9f92a01c986bafcabbafd145520b13", "name": "Efficient Subgroup Analysis via Optimal Trees with Global Parameter Fusion", "authors": [{"id": 19818, "fullname": "Zhongming Xie", "url": "http://virtual.aistats.org/api/miniconf/users/19818?format=json", "institution": "University of California, Berkeley"}, {"id": 22503, "fullname": "Joseph Giorgio", "url": "http://virtual.aistats.org/api/miniconf/users/22503?format=json", "institution": "University of California, Berkeley"}, {"id": 12869, "fullname": "Jingshen Wang", "url": "http://virtual.aistats.org/api/miniconf/users/12869?format=json", "institution": "UC Berkeley"}], "abstract": "Identifying and making statistical inferences on differential treatment effects\u2014commonly known as subgroup analysis in clinical research\u2014is central to precision health. Subgroup analysis allows practitioners to pinpoint populations for whom a treatment is especially beneficial or protective, thereby advancing targeted interventions. Tree-based recursive partitioning methods are widely used for subgroup analysis due to their interpretability. Nevertheless, these approaches encounter significant limitations, including suboptimal partitions induced by greedy heuristics and overfitting from locally estimated splits, especially under limited sample sizes. To address these limitations, we propose a fused optimal causal tree method that leverages mixed-integer optimization (MIO) to facilitate precise subgroup identification. Our approach ensures globally optimal partitions and introduces a parameter-fusion constraint to facilitate information sharing across related subgroups. This design substantially improves subgroup discovery accuracy and enhances statistical efficiency. We provide theoretical guarantees by rigorously establishing out-of-sample risk bounds and comparing them with those of classical tree-based methods. Empirically, our method consistently outperforms popular baselines in simulations. Finally, we demonstrate its practical utility through a case study on the Health and Aging Brain Study\u2013Health Disparities (HABS-HD) dataset, where our approach yields clinically meaningful insights.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13619", "url": null, "sourceid": 645, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=S1M3JR25rd", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11312, "modified": "2026-03-29T20:43:05.884196-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=S1M3JR25rd", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "56", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13620, "uid": "0768281a05da9f27df178b5c39a51263", "name": "Towards Characterizing the Complexity of Riemannian Online Convex Optimization", "authors": [{"id": 20619, "fullname": "Hibiki Fukushima", "url": "http://virtual.aistats.org/api/miniconf/users/20619?format=json", "institution": "The university of Tokyo"}, {"id": 22504, "fullname": "Hiroshi Hirai", "url": "http://virtual.aistats.org/api/miniconf/users/22504?format=json", "institution": "Nagoya University"}, {"id": 18375, "fullname": "Shinji Ito", "url": "http://virtual.aistats.org/api/miniconf/users/18375?format=json", "institution": "The University of Tokyo"}], "abstract": "Online Convex Optimization (OCO) over Riemannian manifolds raises fundamental questions about how geometry affects algorithmic performance. While Riemannian Online Gradient Descent (R-OGD) has been shown to achieve a regret upper bound of $O(DL\\sqrt{\\zeta T})$, where $\\zeta$ depends on the manifold\u2019s curvature, the tightness of this bound remained unclear. We first establish a matching lower bound of $\\Omega(DL\\sqrt{\\zeta T})$ for R-OGD, valid for any predetermined step-size schedules and for certain types of adaptive step-size schedules. This shows that the worst-case regret of R-OGD is $\\Theta(DL\\sqrt{\\zeta T})$, and that the effect of manifold curvature appears as a multiplicative factor of $\\sqrt{\\zeta}$ in the regret. In contrast to the Euclidean setting\u2014where OGD is minimax optimal and regret bounds are independent of feedback models\u2014this result reveals that geometry can substantially degrade the performance of first-order algorithms. Our second contribution shows that this degradation is not unavoidable. In the full-information setting, we analyze a Riemannian extension of Follow-the-Regularized-Leader, which we term R-FTRL. R-FTRL achieves a regret bound of $O(DL\\sqrt{T})$, independent of the curvature. This establishes that curvature-dependent regret is an artifact of limited feedback, not of the problem itself. Our findings suggest a potential separation between first-order and full-information models in non-Euclidean settings, and highlight the subtle interactions between feedback structure, algorithm design, and geometry.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13620", "url": null, "sourceid": 1021, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=Rrbst1HHYL", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11313, "modified": "2026-03-29T20:43:05.937109-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=Rrbst1HHYL", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "168", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13623, "uid": "9431c87f273e507e6040fcb07dcb4509", "name": "Low Rank Based Subspace Inference for the Laplace Approximation of Bayesian Neural Networks", "authors": [{"id": 22509, "fullname": "Josua Faller", "url": "http://virtual.aistats.org/api/miniconf/users/22509?format=json", "institution": "Physikalisch-Technische Bundesanstalt"}, {"id": 22510, "fullname": "J\u00f6rg Martin", "url": "http://virtual.aistats.org/api/miniconf/users/22510?format=json", "institution": "PTB Berlin"}], "abstract": "Subspace inference for neural networks assumes that a subspace of their parameter space suffices to produce a reliable uncertainty quantification. In this work, we underpin the validity of this assumption by using low rank techniques. We derive an expression for a subspace model to a Bayesian inference scenario based on the Laplace approximation that is, in a certain sense, optimal given a specific dataset. We empirically show that a Laplace approximation constructed with a dimensionally reduced covariance matrix closely matches the full Laplace approximation obtained using the exact covariance matrix. Where feasible, this subspace model can serve as a baseline for benchmarking the performance of subspace models.  In addition, we provide a scalable approximation of this subspace construction that is usable in practice and compare it to existing subspace models from the literature. In general, our approximation scheme outperforms previous work. Furthermore, we present a metric to qualitatively compare the approximation quality of different subspace models even if the exact Laplace approximation is unknown.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13623", "url": null, "sourceid": 452, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=RTwTXQX4gq", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11316, "modified": "2026-03-29T20:43:06.078943-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=RTwTXQX4gq", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "101", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13624, "uid": "f033ab37c30201f73f142449d037028d", "name": "Understanding Generalization in Node and Link Prediction", "authors": [{"id": 22511, "fullname": "Antonis Vasileiou", "url": "http://virtual.aistats.org/api/miniconf/users/22511?format=json", "institution": "Rheinisch Westf\u00e4lische Technische Hochschule Aachen"}, {"id": 22512, "fullname": "Timo Stoll", "url": "http://virtual.aistats.org/api/miniconf/users/22512?format=json", "institution": "Rheinisch Westf\u00e4lische Technische Hochschule Aachen"}, {"id": 22513, "fullname": "Christopher Morris", "url": "http://virtual.aistats.org/api/miniconf/users/22513?format=json", "institution": "RWTH Aachen University"}], "abstract": "Using message-passing graph neural networks (MPNNs) for node and link prediction is crucial in various scientific and industrial domains, which has led to the development of diverse MPNN architectures. Besides working well in practical settings, their ability to generalize beyond the training set remains poorly understood. While some studies have explored the generalization of MPNNs in graph-level prediction tasks, much less attention has been given to node- and link-level predictions. Existing works often rely on unrealistic i.i.d. assumptions, overlooking possible correlations between nodes or links, and assuming fixed aggregation and impractical loss functions while neglecting the influence of graph structure. In this work, we introduce a unified framework for analyzing the generalization properties of MPNNs in inductive and transductive node and link prediction settings, incorporating diverse architectural parameters and loss functions, and quantifying the influence of graph structure. Additionally, our proposed generalization framework can be applied beyond graphs to any classification task, regardless of whether it is inductive or transductive. Our empirical study supports our theoretical insights, deepening our understanding of MPNNs' generalization capabilities in these tasks.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13624", "url": null, "sourceid": 80, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=RMDPAATuNw", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11317, "modified": "2026-03-29T20:43:06.119784-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=RMDPAATuNw", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "191", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13625, "uid": "d30960ce77e83d896503d43ba249caf7", "name": "A correlation analysis approach to finding interpretable latent representations via conditional generative models", "authors": [{"id": 22514, "fullname": "James Buenfil", "url": "http://virtual.aistats.org/api/miniconf/users/22514?format=json", "institution": "University of Washington, Seattle"}, {"id": 19802, "fullname": "Eardi Lila", "url": "http://virtual.aistats.org/api/miniconf/users/19802?format=json", "institution": "University of Washington"}], "abstract": "The supervised disentanglement problem, that is, learning interpretable nonlinear latent representations of a target data view while being informed by an auxiliary data view, is a central challenge of interpretable machine learning. We reformulate this problem as a partially linear invertible canonical correlation analysis (PLiCCA). Specifically, given two data views, (i) complex data lying near a potentially high-dimensional manifold, and (ii) auxiliary high-dimensional multivariate data, our approach represents the complex data with latent variables that are maximally correlated with sparse linear combinations of the auxiliary variables. This yields an embedding ordered by interpretability, in contrast to regression-based approaches to supervised disentanglement. We formalize the population PLiCCA problem and provide existence results. We then establish a close theoretical connection between PLiCCA and well-established conditional latent variable models, specifically conditional variational autoencoders and conditional normalizing flows, enabling practical estimation. We demonstrate the utility of our approach on brain morphological data, where our learned embeddings are guided by demographic, psychometric, and behavioral variables, facilitating scientific interpretation and improving generalization.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13625", "url": null, "sourceid": 2149, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=RDEIyjDBOC", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11318, "modified": "2026-03-29T20:43:06.159395-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=RDEIyjDBOC", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "3", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13626, "uid": "b8b4b727d6f5d1b61fff7be687f7970f", "name": "From Counts to Preferences: Preference-Driven Models for Spatio-Temporal Event Data", "authors": [{"id": 22515, "fullname": "Chao Yang", "url": "http://virtual.aistats.org/api/miniconf/users/22515?format=json", "institution": "The Chinese University of Hong Kong, Shenzhen"}, {"id": 22516, "fullname": "Yiling Kuang", "url": "http://virtual.aistats.org/api/miniconf/users/22516?format=json", "institution": "The Chinese University of Hong Kong"}, {"id": 13320, "fullname": "Shuang Li", "url": "http://virtual.aistats.org/api/miniconf/users/13320?format=json", "institution": "The Chinese University of Hong Kong, Shenzhen"}], "abstract": "Spatio-temporal event data---such as crime incidents or shared-mobility usage---are generated by human decisions. Yet most existing models focus on statistical dependencies in time and space, overlooking the cognitive and social factors that shape behavior. We argue that uncovering underlying preferences is essential, as they provide a structured link between observed event data and decision processes. We introduce a preference-driven framework that models event distributions through a two-stage ``consider--then--choose'' process: sparse gating captures limited attention, and utility functions guide selection within the consideration set. To capture heterogeneity, we employ a mixture-of-experts design that reveals distinct preference patterns across groups and contexts. The framework incorporates sparse structural design, and we analyze its theoretical properties by establishing approximation and generalization guarantees. Empirical studies on crime and bike-sharing datasets demonstrate competitive predictive accuracy while providing interpretable insights into behavioral drivers. By shifting the focus from counts to preferences, our approach offers a behaviorally grounded and socially meaningful perspective for modeling event data.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13626", "url": null, "sourceid": 2004, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=QwajHjHvbU", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11319, "modified": "2026-03-29T20:43:06.213706-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=QwajHjHvbU", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "71", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13629, "uid": "069d3bb002acd8d7dd095917f9efe4cb", "name": "Catoni-Style Change Point Detection for Regret Minimization in Piecewise-Stationary Heavy-Tailed Bandits", "authors": [{"id": 12372, "fullname": "Gianmarco Genalti", "url": "http://virtual.aistats.org/api/miniconf/users/12372?format=json", "institution": "Politecnico di Milano"}, {"id": 9491, "fullname": "Sujay Bhatt", "url": "http://virtual.aistats.org/api/miniconf/users/9491?format=json", "institution": "JP Morgan"}, {"id": 12375, "fullname": "Nicola Gatti", "url": "http://virtual.aistats.org/api/miniconf/users/12375?format=json", "institution": "Politecnico di Milano"}, {"id": 118, "fullname": "Alberto Maria Metelli", "url": "http://virtual.aistats.org/api/miniconf/users/118?format=json", "institution": "Politecnico di Milano"}], "abstract": "Regret minimization in stochastic non-stationary bandits gained popularity over the last decade, as it can model a broad class of real-world problems, from advertising to recommendation systems. Existing literature relies on various assumptions about the reward-generating process, such as Bernoulli or subgaussian rewards. However, in settings such as finance and telecommunications, heavy-tailed distributions naturally arise. In this work, we tackle the heavy-tailed piecewise-stationary bandit problem. Heavy-tailed bandits, introduced by Bubeck et al., 2013, operate on the minimal assumption that the finite absolute centered moments of maximum order $1+\\epsilon$ are uniformly bounded by a constant $v<+\\infty$, for some $\\epsilon \\in (0,1]$. We focus on the most popular non-stationary bandit setting, i.e., the piecewise-stationary setting, in which the mean of reward-generating distributions may change at unknown time steps. We provide a novel Catoni-style change-point detection strategy tailored for heavy-tailed distributions that relies on recent advancements in the theory of sequential estimation, which is of independent interest. We introduce Robust-CPD-UCB, which combines this change-point detection strategy with optimistic algorithms for bandits, providing its regret upper bound and an impossibility result on the minimum attainable regret for any policy. Finally, we validate our approach through numerical experiments on synthetic and real-world datasets.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13629", "url": null, "sourceid": 580, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=Q9wdL27LVc", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11322, "modified": "2026-03-29T20:43:06.390362-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=Q9wdL27LVc", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "27", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13635, "uid": "1415db70fe9ddb119e23e9b2808cde38", "name": "Efficient Inference for Coupled Hidden Markov Models in Continuous Time and Discrete Space", "authors": [{"id": 12604, "fullname": "Giosue Migliorini", "url": "http://virtual.aistats.org/api/miniconf/users/12604?format=json", "institution": "University of California, Irvine"}, {"id": 9562, "fullname": "Padhraic Smyth", "url": "http://virtual.aistats.org/api/miniconf/users/9562?format=json", "institution": "University of California, Irvine"}], "abstract": "Systems of interacting continuous time Markov chains are a powerful model class, but inference is typically intractable in high dimensional settings. Auxiliary information, such as noisy observations, is typically only available at discrete times, and incorporating it via a Doob's $h-$transform gives rise to an intractable posterior process that requires approximation. We introduce Latent Interacting Particle Systems, a model class parameterizing the generator of each Markov chain in the system.  Our inference method involves estimating look-ahead functions (twist potentials) that anticipate future information, for which we introduce an efficient parameterization. We incorporate this approximation in a twisted Sequential Monte Carlo sampling scheme. We demonstrate the effectiveness of our approach on a challenging posterior inference task for a latent SIRS model on a graph, and on a neural model for wildfire spread dynamics trained on real data.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13635", "url": null, "sourceid": 1496, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=PFd1nFGbYq", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11328, "modified": "2026-03-29T20:43:06.643955-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=PFd1nFGbYq", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "44", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13637, "uid": "e2ef524fbf3d9fe611d5a8e90fefdc9c", "name": "E-Scores for (In)Correctness Assessment of Generative Model Outputs", "authors": [{"id": 13251, "fullname": "Guneet Singh Dhillon", "url": "http://virtual.aistats.org/api/miniconf/users/13251?format=json", "institution": "University of Oxford"}, {"id": 3744, "fullname": "Javier Gonzalez", "url": "http://virtual.aistats.org/api/miniconf/users/3744?format=json", "institution": "Microsoft Research"}, {"id": 22529, "fullname": "Teodora Pandeva", "url": "http://virtual.aistats.org/api/miniconf/users/22529?format=json", "institution": "Microsoft"}, {"id": 23269, "fullname": "Alicia Curth", "url": "http://virtual.aistats.org/api/miniconf/users/23269?format=json", "institution": "Microsoft Research"}], "abstract": "While generative models, especially large language models (LLMs), are ubiquitous in today's world, principled mechanisms to assess their (in)correctness are limited. Using the conformal prediction framework, previous works construct sets of LLM responses where the probability of including an incorrect response, or error, is capped at a desired user-defined tolerance level. However, since these methods are based on p-values, they are susceptible to p-hacking, i.e., choosing the tolerance level post-hoc can invalidate the guarantees. We therefore leverage e-values to complement generative model outputs with e-scores as a measure of incorrectness. In addition to achieving the same statistical guarantees as before, e-scores provide users flexibility in adaptively choosing tolerance levels after observing the e-scores themselves, by upper bounding a post-hoc notion of error called size distortion. We experimentally demonstrate their efficacy in assessing LLM outputs for different correctness types: mathematical factuality and property constraints satisfaction.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13637", "url": null, "sourceid": 97, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=PCRCLYgiVK", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11330, "modified": "2026-03-29T20:43:06.711875-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=PCRCLYgiVK", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "54", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13638, "uid": "109d2dd3608f669ca17920c511c2a41e", "name": "Efficient Flow Matching using Latent Variables", "authors": [{"id": 22530, "fullname": "Anirban Samaddar", "url": "http://virtual.aistats.org/api/miniconf/users/22530?format=json", "institution": "Argonne National Laboratory"}, {"id": 22531, "fullname": "Yixuan Sun", "url": "http://virtual.aistats.org/api/miniconf/users/22531?format=json", "institution": "Argonne National Laboratory"}, {"id": 19824, "fullname": "Viktor Nilsson", "url": "http://virtual.aistats.org/api/miniconf/users/19824?format=json", "institution": "KTH"}, {"id": 5685, "fullname": "Sandeep Madireddy", "url": "http://virtual.aistats.org/api/miniconf/users/5685?format=json", "institution": "Argonne National Laboratory"}], "abstract": "Flow matching models have shown great potential in image generation tasks among probabilistic generative models. However, most flow matching models in the literature do not explicitly utilize the underlying clustering structure in the target data when learning the flow from a simple source distribution like the standard Gaussian. This leads to inefficient learning, especially for many high-dimensional real-world datasets, which often reside in a low-dimensional manifold.  To this end, we present $\\texttt{Latent-CFM}$, which provides efficient training strategies by conditioning on the features extracted from data using pretrained deep latent variable models. Through experiments on synthetic data from multi-modal distributions and widely used image benchmark datasets, we show that $\\texttt{Latent-CFM}$ exhibits improved generation quality with significantly less training and computation than state-of-the-art flow matching models by adopting pretrained lightweight latent variable models. Beyond natural images, we consider generative modeling of spatial fields stemming from physical processes. Using a 2d Darcy flow dataset, we demonstrate that our approach generates more physically accurate samples than competing approaches. In addition, through latent space analysis, we demonstrate that our approach can be used for conditional image generation conditioned on latent features, which adds interpretability to the generation process.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13638", "url": null, "sourceid": 1528, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=P7Wbb1wYLK", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11331, "modified": "2026-03-29T20:43:06.747084-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=P7Wbb1wYLK", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "62", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13643, "uid": "0a1bf96b7165e962e90cb14648c9462d", "name": "Fast Quasar-Convex Optimization with Constraints", "authors": [{"id": 22355, "fullname": "David Mart\u00ednez-Rubio", "url": "http://virtual.aistats.org/api/miniconf/users/22355?format=json", "institution": "IMDEA Software Institute"}], "abstract": "Quasar-convex functions form a broad nonconvex class with applications to linear dynamical systems, generalized linear models, and Riemannian optimization, among others. Current nearly optimal algorithms work only in affine spaces due to the loss of one degree of freedom when working with general convex constraints. Obtaining an accelerated algorithm that makes nearly optimal $\\bigotilde{1/(\\gamma\\sqrt{\\epsilon})}$ first-order queries to a $\\gamma$-quasar convex smooth function \\emph{with constraints} was independently asked as an open problem in Martinez-Rubio (2022); Lezane, Langer and Koolen, (2024). In this work, we solve this question by designing an inexact accelerated proximal point algorithm that we implement using a first-order method achieving the aforementioned rate and, as a consequence, we improve the complexity of the accelerated geodesically Riemannian optimization solution in Mart\u00ednez-Rubio (2022). We also analyze projected gradient descent and Frank-Wolfe algorithms in this constrained quasar-convex setting. To the best of our knowledge, our work provides the first analyses of first-order methods for quasar-convex smooth functions with general convex constraints.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13643", "url": null, "sourceid": 1455, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=O70rKeVYgZ", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11336, "modified": "2026-03-29T20:43:06.915329-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=O70rKeVYgZ", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "57", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13644, "uid": "4122cb13c7a474c1976c9706ae36521d", "name": "Local Regression on Path Spaces with Signature Metrics", "authors": [{"id": 22547, "fullname": "Davit Gogolashvili", "url": "http://virtual.aistats.org/api/miniconf/users/22547?format=json", "institution": "Weierstrass Institute for Applied Analysis and Stochastics"}, {"id": 22548, "fullname": "Christian Bayer", "url": "http://virtual.aistats.org/api/miniconf/users/22548?format=json", "institution": "Weierstrass Institute for Applied Analysis and Stochastics"}, {"id": 22549, "fullname": "Luca Pelizzari", "url": "http://virtual.aistats.org/api/miniconf/users/22549?format=json", "institution": "Technische Universit\u00e4t Berlin"}], "abstract": "We study nonparametric regression and classification for path-valued data. We introduce a functional Nadaraya-Watson estimator that combines the signature transform from rough path theory with local kernel regression. The signature transform provides a principled way to encode sequential data through iterated integrals, enabling direct comparison of paths in a natural metric space. Our approach leverages signature-induced distances within the classical kernel regression framework, achieving computational efficiency while avoiding the scalability bottlenecks of large-scale kernel matrix operations. We establish finite-sample convergence bounds demonstrating favorable statistical properties of signature-based distances compared to traditional metrics in infinite-dimensional settings. We propose robust signature variants that provide stability against outliers, enhancing practical performance. Applications to both synthetic and real-world data\u2014including stochastic differential equation learning and time series classification\u2014demonstrate competitive accuracy while offering significant computational advantages over existing methods.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13644", "url": null, "sourceid": 1230, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=O6lBk9FFHr", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11337, "modified": "2026-03-29T20:43:06.951442-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=O6lBk9FFHr", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "100", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13648, "uid": "cfecdb276f634854f3ef915e2e980c31", "name": "Differentially Private Minimum Spanning Tree in Euclidean Graphs", "authors": [{"id": 16588, "fullname": "Zongrui Zou", "url": "http://virtual.aistats.org/api/miniconf/users/16588?format=json", "institution": "Nanjing University"}, {"id": 22557, "fullname": "Alessandro Epasto", "url": "http://virtual.aistats.org/api/miniconf/users/22557?format=json", "institution": "Google"}, {"id": 3914, "fullname": "Chenglin Fan", "url": "http://virtual.aistats.org/api/miniconf/users/3914?format=json", "institution": "Seoul National University"}, {"id": 22558, "fullname": "Rudrajit Das", "url": "http://virtual.aistats.org/api/miniconf/users/22558?format=json", "institution": "Google"}], "abstract": "We study differentially private approximation of minimum spanning trees (MST) and hierarchical clustering for Euclidean graph embeddings. Our algorithms achieve an optimal trade-off, providing a $(1+\\eta)$-multiplicative approximation with $\\tilde{O}(n/\\eta^2)$ additive error under $\\rho$-dist privacy. Furthermore, we establish a separation between Euclidean and general graphs by proving a lower bound of $\\Omega(n^{1.5})$ additive error for general graphs under a similar privacy notion, demonstrating that better utility is indeed achievable for geometric data. Our algorithm can also be directly applied to clustering tasks based on specific MST algorithms, incurring only a minimal loss in the approximation guarantee compared to its non-private counterpart.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13648", "url": null, "sourceid": 190, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=Nr5IzX8Bge", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11341, "modified": "2026-03-29T20:43:07.147852-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=Nr5IzX8Bge", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "49", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13651, "uid": "5751ec3e9a4feab575962e78e006250d", "name": "Adaptive Replay Buffer for Offline-to-Online Reinforcement Learning", "authors": [{"id": 19843, "fullname": "Chihyeon Song", "url": "http://virtual.aistats.org/api/miniconf/users/19843?format=json", "institution": "KAIST"}, {"id": 22561, "fullname": "Jaewoo Lee", "url": "http://virtual.aistats.org/api/miniconf/users/22561?format=json", "institution": "Korea Advanced Institute of Science &amp; Technology"}, {"id": 10238, "fullname": "Jinkyoo Park", "url": "http://virtual.aistats.org/api/miniconf/users/10238?format=json", "institution": "KAIST"}], "abstract": "Offline-to-Online Reinforcement Learning (O2O RL) faces a critical dilemma in balancing the use of a fixed offline dataset with newly collected online experiences. Standard methods, often relying on a fixed data-mixing ratio, struggle to manage the trade-off between early learning stability and asymptotic performance. To overcome this, we introduce the Adaptive Replay Buffer (ARB), a novel approach that dynamically prioritizes data sampling based on a lightweight metric we call 'on-policyness'. Unlike prior methods that rely on complex learning procedures or fixed ratios, ARB is designed to be learning-free and simple to implement, seamlessly integrating into existing O2O RL algorithms. It assesses how closely collected trajectories align with the current policy's behavior and assigns a proportional sampling weight to each transition within that trajectory. This strategy effectively leverages offline data for initial stability while progressively focusing learning on the most relevant, high-rewarding online experiences. Our extensive experiments on D4RL benchmarks demonstrate that ARB consistently mitigates early performance degradation and significantly improves the final performance of various O2O RL algorithms, highlighting the importance of an adaptive, behavior-aware replay buffer design.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13651", "url": null, "sourceid": 729, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=NgmNlIBiBz", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11344, "modified": "2026-03-29T20:43:07.355169-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=NgmNlIBiBz", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "13", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13654, "uid": "fccb60fb512d13df5083790d64c4d5dd", "name": "Contraction rates for generalized posteriors based on $f$-divergences: a diffusion process approach", "authors": [{"id": 22566, "fullname": "Enric Boloix", "url": "http://virtual.aistats.org/api/miniconf/users/22566?format=json", "institution": "Basque Center for Applied Mathematics"}, {"id": 20630, "fullname": "Ioar Casado-Telletxea", "url": "http://virtual.aistats.org/api/miniconf/users/20630?format=json", "institution": "Basque Center for Applied Mathematics (BCAM)"}], "abstract": "We study the finite-sample behavior of generalized posteriors defined via $f$-divergences, a broad class of posteriors that includes the standard Bayesian posterior along with most of its generalizations. Our main contribution is to extend the Langevin diffusion representation of the Bayesian posterior to this broader class. With this perspective, we obtain non-asymptotic posterior contraction rates for $f$-divergence-based posteriors by bounding the moments of their associated diffusion. Our results establish nearly optimal rates and clarify how different divergence choices influence posterior concentration. Finally, we illustrate the general framework with concrete examples.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13654", "url": null, "sourceid": 437, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=NDtwR31DJf", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11347, "modified": "2026-03-29T20:43:07.455499-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=NDtwR31DJf", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "37", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13659, "uid": "6ea9ab1baa0efb9e19094440c317e21b", "name": "Counterfactual Credit Guided Bayesian Optimization", "authors": [{"id": 20612, "fullname": "Qiyu Wei", "url": "http://virtual.aistats.org/api/miniconf/users/20612?format=json", "institution": "University of Manchester"}, {"id": 18481, "fullname": "Haowei Wang", "url": "http://virtual.aistats.org/api/miniconf/users/18481?format=json", "institution": "National University of Singapore"}, {"id": 22574, "fullname": "Richard Allmendinger", "url": "http://virtual.aistats.org/api/miniconf/users/22574?format=json", "institution": "The University of Manchester"}, {"id": 9432, "fullname": "Mauricio \u00c1lvarez", "url": "http://virtual.aistats.org/api/miniconf/users/9432?format=json", "institution": "University of Manchester"}], "abstract": "Bayesian optimization has emerged as a prominent methodology for optimizing expensive black-box functions by leveraging Gaussian process surrogates, which focus on capturing the global characteristics of the objective function. However, in numerous practical scenarios, the primary objective is not to construct an exhaustive global surrogate, but rather to quickly pinpoint the global optimum.  Due to the aleatoric nature of the sequential optimization problem and its dependence on the quality of the surrogate model and the initial design, it is restrictive to assume that all observed samples contribute equally to the discovery of the optimum in this context.  In this paper, we introduce Counterfactual Credit Guided Bayesian Optimization (CCGBO), a novel framework that explicitly quantifies the contribution of individual historical observations through counterfactual credit. By incorporating counterfactual credit into the acquisition function, our approach can selectively allocate resources in areas where optimal solutions are most likely to occur. We prove that CCGBO retains sublinear regret. Empirical evaluations on various synthetic and real-world benchmarks demonstrate that CCGBO consistently reduces simple regret and accelerates convergence to the global optimum.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13659", "url": null, "sourceid": 29, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=Mn6w2M5Nvd", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11352, "modified": "2026-03-29T20:43:07.617512-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=Mn6w2M5Nvd", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "41", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13662, "uid": "d4c2e4a3297fe25a71d030b67eb83bfc", "name": "Provable Target Sample Complexity Improvements as Pre\u2011Trained Models Scale", "authors": [{"id": 14314, "fullname": "Kazuto Fukuchi", "url": "http://virtual.aistats.org/api/miniconf/users/14314?format=json", "institution": "University of Tsukuba / RIKEN AIP"}, {"id": 22579, "fullname": "Ryuichiro Hataya", "url": "http://virtual.aistats.org/api/miniconf/users/22579?format=json", "institution": "SB intuitions / Kyoto University"}, {"id": 22580, "fullname": "Kota Matsui", "url": "http://virtual.aistats.org/api/miniconf/users/22580?format=json", "institution": "Kyoto University"}], "abstract": "Pre-trained models have become indispensable for efficiently building models across a broad spectrum of downstream tasks. The advantages of pre-trained models have been highlighted by empirical studies on scaling laws, which demonstrate that larger pre-trained models can significantly reduce the sample complexity of downstream learning. However, existing theoretical investigations of pre-trained models lack the capability to explain this phenomenon. In this paper, we provide a theoretical investigation by introducing a novel framework, caulking, inspired by parameter-efficient fine-tuning (PEFT) methods such as adapter-based fine-tuning, low-rank adaptation, and partial fine-tuning. Our analysis establishes that improved pre-trained models provably decrease the sample complexity of downstream tasks, thereby offering theoretical justification for the empirically observed scaling laws relating pre-trained model size to downstream performance, a relationship not covered by existing results.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13662", "url": null, "sourceid": 728, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=MiuMvce6Ac", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11355, "modified": "2026-03-29T20:43:07.766936-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=MiuMvce6Ac", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "148", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13666, "uid": "66be31e4c40d676991f2405aaecc6934", "name": "Securing Model Weights Against Eavesdropping Adversaries in Federated Learning Using Quantization", "authors": [{"id": 22585, "fullname": "Kushal Chakrabarti", "url": "http://virtual.aistats.org/api/miniconf/users/22585?format=json", "institution": "Tata Consultancy Services Limited, India"}, {"id": 22586, "fullname": "Dipankar Maity", "url": "http://virtual.aistats.org/api/miniconf/users/22586?format=json", "institution": "University of North Carolina, Charlotte"}], "abstract": "While security research in Federated Learning (FL) has predominantly focused on protecting client data, the *confidentiality of the model parameters* themselves represents a critical and underexplored vulnerability. This work addresses model reconstruction attacks by passive eavesdroppers, a threat present in common update strategies like transmitting full models or model increments. To our knowledge, we are the first to repurpose dynamic uniform quantization as a dedicated defense for model confidentiality. Our lightweight, architecture-agnostic approach combines low-bit quantization with an adaptive clipping rule to thwart reconstruction attacks, even under warm adversary initialization. We provide theoretical guarantees establishing that our defense offers persistent, non-zero protection in both protocols. Across extensive experiments on CIFAR-10 and CIFAR-100, with up to 1000 clients in heterogeneous settings, our method reduces the adversary's test accuracy to near-random levels while maintaining global accuracy within 4\\% of the unquantized baseline. Our findings establish that repurposing quantization is a simple yet highly effective strategy for securing the largely overlooked area of model confidentiality in FL.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13666", "url": null, "sourceid": 1716, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=MSoLL2ehax", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11359, "modified": "2026-03-29T20:43:07.909437-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=MSoLL2ehax", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "163", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13668, "uid": "fe131d7f5a6b38b23cc967316c13dae2", "name": "Policy Testing in Markov Decision Processes", "authors": [{"id": 10099, "fullname": "Kaito Ariu", "url": "http://virtual.aistats.org/api/miniconf/users/10099?format=json", "institution": "CyberAgent"}, {"id": 19780, "fullname": "Pai-An WANG", "url": "http://virtual.aistats.org/api/miniconf/users/19780?format=json", "institution": "National Tsing Hua University"}, {"id": 4579, "fullname": "Alexandre Proutiere", "url": "http://virtual.aistats.org/api/miniconf/users/4579?format=json", "institution": "KTH Royal Institute of Technology"}, {"id": 13018, "fullname": "Kenshi Abe", "url": "http://virtual.aistats.org/api/miniconf/users/13018?format=json", "institution": "CyberAgent, Inc."}], "abstract": "We study the policy testing problem in discounted Markov decision processes (MDPs) in the fixed-confidence setting under a generative model with static sampling. The goal is to decide whether the value of a given policy exceeds a specified threshold while minimizing the number of samples. We first derive an instance-dependent lower bound that any reasonable algorithm must satisfy, characterized as the solution to an optimization problem with non-convex constraints. Guided by this formulation, we propose a new algorithm. While this design paradigm is common in pure exploration problems such as best-arm identification, the non-convex constraints that arise in MDPs introduce substantial difficulties. To address them, we reformulate the lower-bound problem by swapping the roles of the objective and the constraints, yielding an alternative problem with a non-convex objective but convex constraints. This reformulation admits an interpretation as a policy optimization task in a newly constructed {\\it reversed MDP}. Leveraging recent advances in policy gradient methods, we solve this problem and design an asymptotically optimal policy testing algorithm. Beyond policy testing, our reformulation and reversed MDP view suggest extensions to other pure exploration tasks in MDPs, including policy evaluation and best policy identification.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13668", "url": null, "sourceid": 255, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=Lh7fVWYDdv", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11361, "modified": "2026-03-29T20:43:07.977913-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=Lh7fVWYDdv", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "137", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13669, "uid": "f5c3dd7514bf620a1b85450d2ae374b1", "name": "Understanding SAM's Robustness to Noisy Labels through Gradient Down-weighting", "authors": [{"id": 22590, "fullname": "Hoang-Chau Luong", "url": "http://virtual.aistats.org/api/miniconf/users/22590?format=json", "institution": "Rochester Institute of Technology"}, {"id": 22591, "fullname": "Thuc Nguyen-Quang", "url": "http://virtual.aistats.org/api/miniconf/users/22591?format=json", "institution": "Ho Chi Minh city University of Science, Vietnam National University"}, {"id": 22592, "fullname": "Dat Tran", "url": "http://virtual.aistats.org/api/miniconf/users/22592?format=json", "institution": "Rowan University"}, {"id": 22593, "fullname": "Minh-Triet Tran", "url": "http://virtual.aistats.org/api/miniconf/users/22593?format=json", "institution": "Ho Chi Minh city University of Science, Vietnam National University"}], "abstract": "Sharpness-Aware Minimization (SAM) was introduced to improve generalization by seeking flat minima, yet it also exhibits robustness to label noise, a phenomenon that remains only partially understood. Prior work has mainly attributed this effect to SAM\u2019s tendency to prolong the learning of clean samples. In this work, we provide a complementary explanation by analyzing SAM at the element-wise level. We show that when noisy gradients dominate a parameter direction, their influence is reduced by the stronger amplification of clean gradients. This slows the memorization of noisy labels while sustaining clean learning, offering a more complete account of SAM\u2019s robustness. Building on this insight, we propose SANER (Sharpness-Aware Noise-Explicit Reweighting), a simple variant of SAM that explicitly magnifies this down-weighting effect. Experiments on benchmark image classification tasks with noisy labels demonstrate that SANER significantly mitigates noisy-label memorization and improves generalization over both SAM and SGD. Moreover, since SANER is designed from the mechanism of SAM, it can also be seamlessly integrated into SAM-like variants, further boosting their robustness.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13669", "url": null, "sourceid": 1818, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=Lh1TsMCIeF", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11362, "modified": "2026-03-29T20:43:08.007414-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=Lh1TsMCIeF", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "184", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13670, "uid": "cb70ab375662576bd1ac5aaf16b3fca4", "name": "Provable Guarantees for Estimating Covariances between Latent Variables with Application to Precision Matrix Estimation", "authors": [{"id": 22594, "fullname": "Haichi Long", "url": "http://virtual.aistats.org/api/miniconf/users/22594?format=json", "institution": "University of Melbourne"}, {"id": 438, "fullname": "Qifan Song", "url": "http://virtual.aistats.org/api/miniconf/users/438?format=json", "institution": "Purdue University "}, {"id": 827, "fullname": "Jean Honorio", "url": "http://virtual.aistats.org/api/miniconf/users/827?format=json", "institution": "University of Melbourne"}], "abstract": "In many scientific fields, key variables of interest are latent---either because they cannot be measured directly or because doing so is prohibitively expensive. As a result, researchers often rely on high-dimensional surrogate observations and must infer relationships among the unobserved quantities. In this work, we address a fundamental challenge: How can one estimate the covariance between variables that are not directly observable? We consider a model where each latent variable elicits high-dimensional observable covariates. Under our model, we propose a method that estimates several spiked covariances from the observed variables and then reconstructs the covariance matrix among the latent variables. Our estimator achieves quadratic-time complexity with respect to the number of latent variables and only requires the sample size to be logarithmic in the number of latent variables. As an immediate application, our procedure can be leveraged to recover the conditional independence structure among the latent variables, providing interpretable insights. Extensive synthetic experiments validate our theory, demonstrating accurate estimation in practice.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13670", "url": null, "sourceid": 243, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=LfVHb2WWIB", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11363, "modified": "2026-03-29T20:43:08.036021-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=LfVHb2WWIB", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "147", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13671, "uid": "215a71a12769b056c3c32e7299f1c5ed", "name": "Lower Bounds for Public-Private Learning under Distribution Shift", "authors": [{"id": 22595, "fullname": "Amrith Setlur", "url": "http://virtual.aistats.org/api/miniconf/users/22595?format=json", "institution": "School of Computer Science, Carnegie Mellon University"}, {"id": 22596, "fullname": "Pratiksha Thaker", "url": "http://virtual.aistats.org/api/miniconf/users/22596?format=json", "institution": "Carnegie Mellon University"}, {"id": 18363, "fullname": "Jonathan Ullman", "url": "http://virtual.aistats.org/api/miniconf/users/18363?format=json", "institution": "University of Michigan - Ann Arbor"}], "abstract": "The most effective differentially private machine learning algorithms in practice rely on an additional source of purportedly public data.  This paradigm is most interesting when the two sources combine to be more than the sum of their parts.  However, there are settings such as mean estimation where we have strong lower bounds, showing that when the two data sources have the same distribution, there is no complementary value to combining the two data sources.  In this work we extend the known lower bounds for public-private learning to setting where the two data sources exhibit significant distribution shift.  Our results apply to both Gaussian mean estimation where the two distributions have different means, and to Gaussian linear regression where the two distributions exhibit parameter shift. We find that when the shift is small (relative to the desired accuracy), either public or private data must be sufficiently abundant to estimate the private parameter. Conversely, when the shift is large, public data provides no benefit.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13671", "url": null, "sourceid": 1397, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=LeLajk8CZ1", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11364, "modified": "2026-03-29T20:43:08.070343-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=LeLajk8CZ1", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "102", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13672, "uid": "d860edd1dd83b36f02ce52bde626c653", "name": "Proof of The TAP Free Energy for High-Dimensional Linear Regression with Spherical Priors at All Temperatures", "authors": [{"id": 18517, "fullname": "Zhiyuan Yu", "url": "http://virtual.aistats.org/api/miniconf/users/18517?format=json", "institution": "University of Illinois at Urbana-Champaign"}, {"id": 1090, "fullname": "Jingbo Liu", "url": "http://virtual.aistats.org/api/miniconf/users/1090?format=json", "institution": "University of Illinois at Urbana-Champaign"}], "abstract": "Approximate inference is central to Bayesian learning, with variational inference (VI) providing a scalable framework for posterior approximation. While mean-field VI often fails in high dimensions, the more refined Bethe approximation, equivalent to the Thouless-Anderson-Palmer (TAP) free energy in statistical physics, has long been conjectured to capture Bayes-optimal behavior. We prove that the TAP formula holds for Bayesian linear regression with a uniform spherical prior at all noise levels ($\\Delta>0$), extending the result of Qiu and Sen (2022) in the high-noise regime. Our argument constructs a ridge regression functional that dominates the TAP free energy, yielding the first rigorous analysis of the global optimizer of the non-concave TAP functional for a planted inference model at an arbitrary noise level. This verifies that TAP, rather than mean-field, is the correct variational description in this setting.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13672", "url": null, "sourceid": 2028, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=LaLaDcywoM", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11365, "modified": "2026-03-29T20:43:08.106085-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=LaLaDcywoM", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "141", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13674, "uid": "4311359ed4969e8401880e3c1836fbe1", "name": "Nonparametric Multi Change Point Detection for Markov Chains via Adaptive Clustering", "authors": [{"id": 21885, "fullname": "Imon Banerjee", "url": "http://virtual.aistats.org/api/miniconf/users/21885?format=json", "institution": "Northwestern University"}, {"id": 22598, "fullname": "Jiaqi Lei", "url": "http://virtual.aistats.org/api/miniconf/users/22598?format=json", "institution": "Northwestern University"}, {"id": 22599, "fullname": "Sanjay Mehrotra", "url": "http://virtual.aistats.org/api/miniconf/users/22599?format=json", "institution": "Northwestern University"}], "abstract": "Offline change point detection tries to detect $\\textit{time points}$ of distribution change in a given data sequence; and is now routinely used in signal processing, speech processing, climatology etc. Despite this broad applicability across economics, computer science, and planetary sciences, rigorous, nonparametric techniques for change point detection with non-independent and identically distributed (i.i.d.) datasets has remained elusive. This paper establishes such guarantees by proposing a non-parametric clustering algorithm which can accurately obtain the change points from a given Markovian dataset of length $n$. It does so by bridging together two different components of mathematical statistics; Rademacher complexities of Markov chains, and adaptive clustering via penalisation. Our first result uses recent advances in Rademacher complexities of regenerating Markov chains to derive a Dvoretzky Kiefer Wolfowitz (DKW) type inequality for the empirical distribution of the Markov chain. We then use this to show that an adaptive clustering algorithm recovers the correct change points for a Markovian sequence. We establish the tightness of our rates by showing that they essentially coincide with the best known rates for i.i.d. data. We end the paper by discussing the computational considerations of the problem.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13674", "url": null, "sourceid": 974, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=LLRXG0wlJ8", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11367, "modified": "2026-03-29T20:43:08.187040-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=LLRXG0wlJ8", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "115", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13680, "uid": "7a53928fa4dd31e82c6ef826f341daec", "name": "FastRank: Fast Tensor Rank Approximation based on Spectral Energy", "authors": [{"id": 19739, "fullname": "Konstantinos Bougiatiotis", "url": "http://virtual.aistats.org/api/miniconf/users/19739?format=json", "institution": "Institute of Informatics and Telecommunications, NCSR &quot;Demokritos&quot;, Athens, Greece"}, {"id": 22615, "fullname": "Georgios Paliouras", "url": "http://virtual.aistats.org/api/miniconf/users/22615?format=json", "institution": "NCSR \u201cDemokritos\u201d"}], "abstract": "Complex multi-dimensional data are often represented as tensors, analyzed through tensor decompositions. A central challenge is selecting the right number of components for the decomposition. In the Canonical Polyadic Decomposition (CPD), this means determining the canonical rank, which directly impacts decomposition quality. Existing methods typically estimate rank by repeatedly computing CPDs, an expensive process. We introduce $FastRank$, a theoretically grounded method that estimates rank without CPD computation. By applying Singular Value Decomposition (SVD) to a sum-reduced matrix of the tensor and analyzing its eigenspectrum, FastRank achieves over $1000\\times$ speedup and surpasses state-of-the-art accuracy. We validate it using both synthetic and real data, including noisy settings, and highlight its scalability in knowledge graph completion, where prior methods fail due to computational limitations.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13680", "url": null, "sourceid": 800, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=KJSycglzzT", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11373, "modified": "2026-03-29T20:43:08.511655-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=KJSycglzzT", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "59", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13686, "uid": "eaae339c4d89fc102edd9dbdb6a28915", "name": "Feature Importance via Sets of Locally Performant Linear Models", "authors": [{"id": 14141, "fullname": "Fatemeh Tohidian", "url": "http://virtual.aistats.org/api/miniconf/users/14141?format=json", "institution": null}, {"id": 5463, "fullname": "Davin Hill", "url": "http://virtual.aistats.org/api/miniconf/users/5463?format=json", "institution": "Northeastern University"}, {"id": 3536, "fullname": "Aria Masoomi", "url": "http://virtual.aistats.org/api/miniconf/users/3536?format=json", "institution": "Northeastern University"}, {"id": 22627, "fullname": "Peter Castaldi", "url": "http://virtual.aistats.org/api/miniconf/users/22627?format=json", "institution": "Harvard University"}, {"id": 399, "fullname": "Jennifer Dy", "url": "http://virtual.aistats.org/api/miniconf/users/399?format=json", "institution": "Northeastern"}], "abstract": "Understanding the contribution of individual features to a model\u2019s prediction is critical in applications such as medicine. While feature importance methods aim to quantify how much a feature contributes to a model\u2019s accuracy, they often overlook heterogeneous patterns in the data and suffer from limited robustness. We propose $\\ell\\text{-MCR}$, a local feature importance method that identifies meaningful neighborhoods around a point of interest, regions where the model or data behavior is locally stable and interpretable. Within these neighborhoods, we estimate feature importance using Model Class Reliance (MCR), which offers robustness by considering the full set of near-optimal models. We also provide a consistency proof for reliably detecting such neighborhoods. Experiments on both synthetic and real-world datasets demonstrate that $\\ell\\text{-MCR}$ captures localized feature importance patterns that global approaches fail to detect.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13686", "url": null, "sourceid": 653, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=JrUKqD2fQ1", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11379, "modified": "2026-03-29T20:43:08.742301-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=JrUKqD2fQ1", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "62", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13697, "uid": "4558dbb6f6f8bb2e16d03b85bde76e2c", "name": "Bad Values but Good Behavior: Learning Highly Misspecified Bandits with Function Approximation", "authors": [{"id": 22653, "fullname": "Debangshu Banerjee", "url": "http://virtual.aistats.org/api/miniconf/users/22653?format=json", "institution": "Hewlett Packard"}, {"id": 882, "fullname": "Aditya Gopalan", "url": "http://virtual.aistats.org/api/miniconf/users/882?format=json", "institution": "Indian Institute of Science (IISc), Bangalore"}], "abstract": "Function approximation with parametric, feature-based reward models is widely used to enable decision-making in bandits with large action spaces. While bandit learning is well understood in the case of little or no misspecification in the reward approximation, real-world applications can often involve significantly high model misspecification. We study whether optimal learning is still possible under arbitrary misspecification. We identify structural, instance-dependent conditions, determined jointly by the problem instance and model class, under which standard algorithms like $\\epsilon$-greedy and LinUCB achieve sublinear regret, despite an arbitrarily large misspecification error in the traditional sense. These results contrast sharply with worst-case analyses that predict linear regret, and show that a broad class of instances remains robust to model error. Our findings offer a theoretical explanation for the empirical success of approximate value-based methods in complex environments.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13697", "url": null, "sourceid": 821, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=IZaQGmyuwn", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11390, "modified": "2026-03-29T20:43:09.225487-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=IZaQGmyuwn", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "16", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13701, "uid": "326a8c055c0d04f5b06544665d8bb3ea", "name": "Beyond the Ideal: Analyzing the Inexact Muon Update", "authors": [{"id": 22677, "fullname": "Egor Shulgin", "url": "http://virtual.aistats.org/api/miniconf/users/22677?format=json", "institution": "KAUST"}, {"id": 22678, "fullname": "AlRashed", "url": "http://virtual.aistats.org/api/miniconf/users/22678?format=json", "institution": "King Abdullah University of Science and Technology"}, {"id": 722, "fullname": "Peter Richtarik", "url": "http://virtual.aistats.org/api/miniconf/users/722?format=json", "institution": "KAUST"}, {"id": 22110, "fullname": "Francesco Orabona", "url": "http://virtual.aistats.org/api/miniconf/users/22110?format=json", "institution": "King Abdullah University of Science and Technology"}], "abstract": "The Muon optimizer has rapidly emerged as a powerful, geometry-aware alternative to AdamW, demonstrating state-of-the-art performance in large-scale training of DNNs. A critical disconnect, however, exists between its theory and practice: Muon's efficiency relies on fast, approximate orthogonalization, yet all prior theoretical work analyzes an idealized, computationally intractable version assuming exact updates. This work moves beyond the ideal by providing the first analysis of the *inexact* orthogonalized update at Muon's core. We develop our analysis within the general framework of Linear Minimization Oracle (LMO)-based optimization, introducing a realistic additive error model to capture the inexactness of practical approximation schemes. Our analysis yields explicit bounds that quantify performance degradation as a function of the LMO inexactness/error, $\\delta$. We reveal a fundamental coupling between this inexactness and the optimal step size and momentum, showing that the training strategy must adapt to the oracle's precision. These findings elevate the approximation procedure (e.g., the number of Newton-Schulz steps) from an implementation detail to a critical parameter that must be *co-tuned* with the learning schedule. Our theoretical insights are validated with experiments on vision and language models.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13701", "url": null, "sourceid": 2350, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=IBRMWPBouf", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11394, "modified": "2026-03-29T20:43:09.377685-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=IBRMWPBouf", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "33", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13702, "uid": "cec6f62cfb44b1be110b7bf70c8362d8", "name": "Hybrid Meta-Learners for Estimating Heterogeneous Treatment Effects", "authors": [{"id": 22679, "fullname": "Zhongyuan Liang", "url": "http://virtual.aistats.org/api/miniconf/users/22679?format=json", "institution": "University of California, Berkeley"}, {"id": 22680, "fullname": "Lars van der Laan", "url": "http://virtual.aistats.org/api/miniconf/users/22680?format=json", "institution": "University of Washington"}, {"id": 22681, "fullname": "Ahmed Alaa", "url": "http://virtual.aistats.org/api/miniconf/users/22681?format=json", "institution": "University of California, Berkeley"}], "abstract": "Estimating conditional average treatment effects (CATE) from observational data involves modeling decisions that differ from supervised learning, particularly concerning how to regularize model complexity. Previous approaches can be grouped into two primary meta-learner paradigms that impose distinct inductive biases. Indirect meta-learners first fit and regularize separate potential outcome (PO) models and then estimate CATE by taking their difference, whereas direct meta-learners construct and directly regularize estimators for the CATE function itself. Neither approach consistently outperforms the other across all scenarios: indirect learners perform well when the PO functions are simple, while direct learners outperform when the CATE is simpler than individual PO functions. In this paper, we introduce the Hybrid Learner (H-learner), a novel regularization strategy that interpolates between the direct and indirect regularizations depending on the dataset at hand. The H-learner achieves this by learning intermediate functions whose difference closely approximates the CATE without necessarily requiring accurate individual approximations of the POs themselves. We demonstrate that intentionally allowing suboptimal fits to the POs improves the bias-variance tradeoff in estimating CATE. Experiments conducted on semi-synthetic and real-world benchmark datasets illustrate that the H-learner consistently operates at the Pareto frontier, effectively combining the strengths of both direct and indirect meta-learners.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13702", "url": null, "sourceid": 1468, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=IBNI3ICqhG", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11395, "modified": "2026-03-29T20:43:09.413582-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=IBNI3ICqhG", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "72", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13703, "uid": "f8bf09f5fceaea80e1f864a1b48938bf", "name": "Contextual ranking and matching. Optimal regret under LST", "authors": [{"id": 22682, "fullname": "Hafedh Ferchichi", "url": "http://virtual.aistats.org/api/miniconf/users/22682?format=json", "institution": "Ecole Nationale de la Statistique et de l&#x27;Administration Economique"}, {"id": 9406, "fullname": "Vianney Perchet", "url": "http://virtual.aistats.org/api/miniconf/users/9406?format=json", "institution": "ENSAE &amp; Criteo AI Lab"}, {"id": 22683, "fullname": "Matthieu LERASLE", "url": "http://virtual.aistats.org/api/miniconf/users/22683?format=json", "institution": "Ecole Nationale de la Statistique et de l&#x27;Administration Economique"}], "abstract": "We address the problem of online matchmaking with contextual information. In each round, a perfect matching between a varying set of players -- with different strengths -- is selected, and the outcomes of the comparisons of the chosen pairs are observed. We assume that matching players incurs dissatisfaction proportional to the \"strength gap\", thereby incentivising the pairing of players with closely matched strengths. Additionally, we assume that the strength of each player can be inferred from some available contextual information through the contextualised linear stochastic transitivity model \\textbf{(LST)}. We propose an algorithm that performs matchmaking by selecting pairs of maximum informativeness among admissible pairs and prove that its regret is optimal up to logarithmic factors.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13703", "url": null, "sourceid": 2064, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=I5teYiPRU1", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11396, "modified": "2026-03-29T20:43:09.447039-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=I5teYiPRU1", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "44", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13704, "uid": "db60b95decdeed944b4cd8685417cfdc", "name": "Optimal Learning in Two-Player Zero-Sum Games under Delayed Feedback", "authors": [{"id": 22684, "fullname": "Ruotong Zhuang", "url": "http://virtual.aistats.org/api/miniconf/users/22684?format=json", "institution": "The University of Tokyo, Tokyo Institute of Technology"}, {"id": 13134, "fullname": "Taira Tsuchiya", "url": "http://virtual.aistats.org/api/miniconf/users/13134?format=json", "institution": "The University of Tokyo"}, {"id": 18375, "fullname": "Shinji Ito", "url": "http://virtual.aistats.org/api/miniconf/users/18375?format=json", "institution": "The University of Tokyo"}], "abstract": "Learning in games is a central topic in both learning theory and game theory, and learning dynamics based on online learning have made significant theoretical and practical advances in recent years. In particular, in two-player zero-sum games, it has been shown that using optimistic follow-the-regularized-leader (OFTRL) as the online learning algorithm allows us to upper bound the individual regret for each player by a constant independent of the time horizon. However, in realistic game scenarios, players are not always able to observe the outcomes of their interactions immediately. Motivated by this, very recently, the problem of learning from delayed feedback in games has been proposed, and it has been shown that by using a variant of OFTRL, one can achieve social regret upper bounds of $\\tilde{O}(D^2 + 1)$ for a fixed delay time $D$. This study investigates the optimal dependence on the delay parameter $D$ in the setting of learning from delayed feedback in games. In particular, we show that a simple algorithm---running $D+1$ independent copies of the standard OFTRL designed for the non-delayed setting---achieves social and individual regret upper bounds of $\\tilde{O}(D + 1)$, thereby improving the existing bounds by a factor of $D$. Moreover, we provide a matching lower bound: for any learning dynamic, there exists a utility function such that the regret of every player is at least $\\Omega(D + 1)$.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13704", "url": null, "sourceid": 2418, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=I46JIcNVlH", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11397, "modified": "2026-03-29T20:43:09.483028-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=I46JIcNVlH", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "128", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13705, "uid": "08d98638c6fcd194a4b1e6992063e944", "name": "LLMs Judging LLMs: A Simplex Perspective", "authors": [{"id": 19819, "fullname": "Patrick Vossler", "url": "http://virtual.aistats.org/api/miniconf/users/19819?format=json", "institution": "University of California, San Francisco"}, {"id": 22685, "fullname": "Fan Xia", "url": "http://virtual.aistats.org/api/miniconf/users/22685?format=json", "institution": "University of California, San Francisco"}, {"id": 22686, "fullname": "Yifan Mai", "url": "http://virtual.aistats.org/api/miniconf/users/22686?format=json", "institution": "Stanford University"}, {"id": 191, "fullname": "Adarsh Subbaswamy", "url": "http://virtual.aistats.org/api/miniconf/users/191?format=json", "institution": "Johns Hopkins University"}, {"id": 12264, "fullname": "Jean Feng", "url": "http://virtual.aistats.org/api/miniconf/users/12264?format=json", "institution": "University of California, San Francisco"}], "abstract": "Given the challenge of automatically evaluating free\u2010form outputs from large language models (LLMs), an increasingly common solution is to use LLMs themselves as the judging mechanism, without any gold-standard scores. Implicitly, this practice accounts for only sampling variability (aleatoric uncertainty) and ignores uncertainty about judge quality (epistemic uncertainty). While this is justified if judges are perfect, it is unclear when such an approach is (i) theoretically valid and (ii) practically robust. We study these questions for the task of ranking LLM candidates from a novel geometric perspective: for $M$-level scoring systems, both LLM judges and candidates can be represented as points on a $(M-1)$-dimensional probability simplex, where geometric concepts (e.g., triangle areas) correspond to key ranking concepts. This perspective yields intuitive theoretical conditions and visual proofs for when rankings are identifiable; for instance, we provide a formal basis for the ``folk wisdom'' that LLM judges are more effective with binary scoring ($M=2$) than with multi-level scoring ($M>2$). Leveraging the simplex, we design geometric Bayesian priors that encode epistemic uncertainty about judge quality and vary the priors to conduct sensitivity analyses. Experiments on LLM benchmarks show that rankings based solely on LLM judges are robust in many but not all datasets, underscoring both their widespread success and the need for caution. Our Bayesian method achieves substantially higher coverage rates than existing procedures, highlighting the importance of modeling epistemic uncertainty.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13705", "url": null, "sourceid": 682, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=HxmC9vLZE3", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11398, "modified": "2026-03-29T20:43:09.523165-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=HxmC9vLZE3", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "98", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13710, "uid": "7bd28f15a49d5e5848d6ec70e584e625", "name": "Auto-Regressive Masked Diffusion Models", "authors": [{"id": 22695, "fullname": "Mahdi Karami", "url": "http://virtual.aistats.org/api/miniconf/users/22695?format=json", "institution": "Research, Google"}, {"id": 22696, "fullname": "Ali Ghodsi", "url": "http://virtual.aistats.org/api/miniconf/users/22696?format=json", "institution": "University of Waterloo"}], "abstract": "Masked diffusion models (MDMs) have emerged as a promising approach for language modeling, yet they face a performance gap compared to autoregressive models (ARMs) and require more training iterations. In this work, we present the Auto-Regressive Masked Diffusion (ARMD) model, an architecture designed to bridge this gap by unifying the training efficiency of autoregressive models with the strengths of diffusion-based learning. Our key insight is to interpret masked diffusion process as a block-wise causal model. This allows us to design a strictly causal, permutation-equivariant, attention-based architecture that computes all conditional probabilities across multiple denoising steps in a single, parallel forward pass. The resulting architecture supports efficient, autoregressive-style decoding and a progressive permutation training scheme, allowing the model to learn both canonical left-to-right and random token orderings. On standard language model- ing benchmarks, ARMD achieves state-of-the- art performance, outperforming established diffusion-based methods while requiring significantly fewer training steps.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13710", "url": null, "sourceid": 1571, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=HV3C5WNwf2", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11403, "modified": "2026-03-29T20:43:09.744082-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=HV3C5WNwf2", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "27", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13711, "uid": "a666587afda6e89aec274a3657558a27", "name": "High-Dimensional Analysis of Bootstrap Ensemble Classifiers", "authors": [{"id": 22697, "fullname": "Hamza Cherkaoui", "url": "http://virtual.aistats.org/api/miniconf/users/22697?format=json", "institution": "Telecom SudParis"}, {"id": 22698, "fullname": "Malik Tiomoko", "url": "http://virtual.aistats.org/api/miniconf/users/22698?format=json", "institution": "Huawei Technologies Ltd."}, {"id": 617, "fullname": "Mohamed El Amine Seddik", "url": "http://virtual.aistats.org/api/miniconf/users/617?format=json", "institution": "Technology Innovation Institute"}, {"id": 22699, "fullname": "Cosme Louart", "url": "http://virtual.aistats.org/api/miniconf/users/22699?format=json", "institution": "The Chinese University of Hong Kong"}, {"id": 22700, "fullname": "Ekkehard Schnoor", "url": "http://virtual.aistats.org/api/miniconf/users/22700?format=json", "institution": "Fraunhofer HHI"}, {"id": 23272, "fullname": "Bal\u00e1zs K\u00e9gl", "url": "http://virtual.aistats.org/api/miniconf/users/23272?format=json", "institution": "Meta"}], "abstract": "Bootstrap methods have long been the cornerstone of ensemble learning in machine learning. This paper presents a theoretical analysis of bootstrap techniques applied to the Least Square Support Vector Machine (LSSVM) ensemble in the context of large and growing sample sizes and feature dimensionalities. Using tools from Random Matrix Theory, we investigate the performance of this classifier that aggregates decision functions from multiple weak classifiers, each trained on different subsets of the data. We provide insights into the use of bootstrap methods in high-dimensional settings, enhancing our understanding of their impact. Based on these findings, we propose strategies to select the number of subsets and the regularization parameter that maximize the performance of the LSSVM. Empirical experiments on synthetic and real-world datasets validate our theoretical results.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13711", "url": null, "sourceid": 326, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=HRNBXXmTwh", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11404, "modified": "2026-03-29T20:43:09.787567-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=HRNBXXmTwh", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "80", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13713, "uid": "feab05aa91085b7a8012516bc3533958", "name": "Three-operator splitting with stale gradients for faster non-linear optimal transport", "authors": [{"id": 22703, "fullname": "Jacob Lindb\u00e4ck", "url": "http://virtual.aistats.org/api/miniconf/users/22703?format=json", "institution": "KTH Royal Institute of Technology"}, {"id": 17777, "fullname": "David Alvarez-Melis", "url": "http://virtual.aistats.org/api/miniconf/users/17777?format=json", "institution": "School of Engineering and Applied Sciences, Harvard University"}, {"id": 22704, "fullname": "Mikael Johansson", "url": "http://virtual.aistats.org/api/miniconf/users/22704?format=json", "institution": "KTH Royal Institute of Technology, Stockholm, Sweden"}], "abstract": "Scalable optimization for non-linear optimal transport (OT) poses unique challenges; it requires efficient memory management of large matrices, effective parallelization strategies suited for modern accelerators like GPUs, and theoretical guarantees that support practical implementation patterns. To address these challenges, we introduce a new algorithm based on three-operator splitting that reduces gradient computation costs by allowing gradient evaluations to run asynchronously and in parallel with other computations. Using monotone operator theory, we establish new convergence guarantees for this asynchronous adaptation and extend existing results to important non-convex problem classes, including Gromov\u2013Wasserstein as a notable example. We validate our method through a series of experiments demonstrating improved accuracy and faster convergence for a broad range of problems", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13713", "url": null, "sourceid": 1145, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=HLw6b0S8PN", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11406, "modified": "2026-03-29T20:43:09.855572-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=HLw6b0S8PN", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "161", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13722, "uid": "01894d6f048493d2cacde3c579c315a3", "name": "Predictive Deep Sets", "authors": [{"id": 22718, "fullname": "Alex H\u00e4m\u00e4l\u00e4inen", "url": "http://virtual.aistats.org/api/miniconf/users/22718?format=json", "institution": "Aalto University"}, {"id": 22719, "fullname": "Sammie Katt", "url": "http://virtual.aistats.org/api/miniconf/users/22719?format=json", "institution": "Aalto University"}, {"id": 3700, "fullname": "Samuel Kaski", "url": "http://virtual.aistats.org/api/miniconf/users/3700?format=json", "institution": "Aalto University and University of Manchester"}], "abstract": "Amortized meta-learning methods, such as neural processes, promise near-instantaneous inference on new labeled datasets encountered during downstream tasks. Recent adaptations of the transformer architecture have propelled these approaches to impressive performance in tasks like function estimation, parameter inference, and decision-making. Curiously, their success still primarily stems from the expressiveness of transformers, lacking a bias for modeling the functional structures between features and labels shared across datasets. We argue and show that this leads to training sample inefficiency and sub-optimal performance, and address this by introducing a novel set encoding technique called Predictive Deep Sets. Our approach exploits a strong bias for functional structures by meta-learning an RKHS that captures domain-critical functional patterns, and by representing datasets as optimal fit functions within this space. Besides providing theoretical justification for this approach, we empirically demonstrate orders of magnitude increases in training data sample efficiency compared to strong baselines across various settings.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13722", "url": null, "sourceid": 2213, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=G5ofLaGCit", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11415, "modified": "2026-03-29T20:43:10.244863-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=G5ofLaGCit", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "139", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13726, "uid": "b534ba68236ba543ae44b22bd110a1d6", "name": "A Recovery Theory for Diffusion Priors: Deterministic Analysis of the Implicit Prior Algorithm", "authors": [{"id": 22727, "fullname": "Oscar Leong", "url": "http://virtual.aistats.org/api/miniconf/users/22727?format=json", "institution": "UCLA, University of California, Los Angeles"}, {"id": 22728, "fullname": "Yann Traonmilin", "url": "http://virtual.aistats.org/api/miniconf/users/22728?format=json", "institution": "CNRS, University of Bordeaux"}], "abstract": "Recovering high-dimensional signals from corrupted measurements is a central challenge in inverse problems. Recent advances in generative diffusion models have shown remarkable empirical success in providing strong data-driven priors, but rigorous recovery guarantees remain limited. In this work, we develop a theoretical framework for analyzing deterministic diffusion-based algorithms for inverse problems, focusing on a deterministic version of the algorithm proposed by Kadkhodaie \\& Simoncelli \\cite{kadkhodaie2021stochastic}. First, we show that when the underlying data distribution concentrates on a low-dimensional model set, the associated noise-convolved scores can be interpreted as time-varying projections onto such a set. This leads to interpreting previous algorithms using diffusion priors for inverse problems as generalized projected gradient descent methods with varying projections. When the sensing matrix satisfies a restricted isometry property over the model set, we can derive quantitative convergence rates that depend explicitly on the noise schedule. We apply our framework to two instructive data distributions: uniform distributions over low-dimensional compact, convex sets and low-rank Gaussian mixture models. In the latter setting, we can establish global convergence guarantees despite the nonconvexity of the underlying model set.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13726", "url": null, "sourceid": 496, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=FwSlWVkW7i", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11419, "modified": "2026-03-29T20:43:10.393997-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=FwSlWVkW7i", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "8", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13728, "uid": "28dc6b0e1b33769b4b94685e4f4d1e5c", "name": "Finite-Time Analysis of Gradient Descent for Shallow Transformers", "authors": [{"id": 20622, "fullname": "Enes Arda", "url": "http://virtual.aistats.org/api/miniconf/users/20622?format=json", "institution": "Ohio State University"}, {"id": 22320, "fullname": "Semih Cayci", "url": "http://virtual.aistats.org/api/miniconf/users/22320?format=json", "institution": "Rheinisch Westf\u00e4lische Technische Hochschule Aachen"}, {"id": 4393, "fullname": "Atilla Eryilmaz", "url": "http://virtual.aistats.org/api/miniconf/users/4393?format=json", "institution": "Ohio State University"}], "abstract": "Understanding why Transformers perform so well remains challenging due to their non-convex optimization landscape. In this work, we analyze a shallow Transformer with $m$ independent heads trained by projected gradient descent in the kernel regime. Our analysis reveals two main findings: (i) the width required for nonasymptotic guarantees scales only logarithmically with the sample size $n$, and (ii) the optimization error is independent of the sequence length $T$. This contrasts sharply with recurrent architectures, where the optimization error can grow exponentially with $T$. The trade-off is memory: to keep the full context, the Transformer's memory requirement grows with the sequence length. We validate our theoretical results numerically in a teacher\u2013student setting and compare Transformers with recurrent architectures on an autoregressive task.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13728", "url": null, "sourceid": 1721, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=FmLMB2B0qH", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11421, "modified": "2026-03-29T20:43:10.464387-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=FmLMB2B0qH", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "69", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13729, "uid": "847cc55b7032108eee6dd897f3bca8a5", "name": "Local Causal Discovery for Statistically Efficient Causal Inference", "authors": [{"id": 14438, "fullname": "M\u00e1ty\u00e1s Schubert", "url": "http://virtual.aistats.org/api/miniconf/users/14438?format=json", "institution": "University of Amsterdam"}, {"id": 18684, "fullname": "Tom Claassen", "url": "http://virtual.aistats.org/api/miniconf/users/18684?format=json", "institution": "Radboud University Nijmegen"}, {"id": 2119, "fullname": "Sara Magliacane", "url": "http://virtual.aistats.org/api/miniconf/users/2119?format=json", "institution": "Saarland University, University of Amsterdam"}], "abstract": "Causal discovery methods can identify valid adjustment sets for causal effect estimation for a pair of target variables, even when the underlying causal graph is unknown. Global causal discovery methods focus on learning the whole causal graph and therefore enable the recovery of optimal adjustment sets, i.e., sets with the lowest asymptotic variance, but they quickly become computationally prohibitive as the number of variables grows. Local causal discovery methods offer a more scalable alternative by focusing on the local neighborhood of the target variables, but are restricted to statistically suboptimal adjustment sets. In this work, we propose Local Optimal Adjustments Discovery (LOAD), a sound and complete causal discovery approach that combines the computational efficiency of local methods with the statistical optimality of global methods. First, LOAD identifies the causal relation between the targets and tests if the causal effect is identifiable by using only local information. If it is identifiable, it then finds the optimal adjustment set by leveraging local causal discovery to infer the mediators and their parents. Otherwise, it returns the locally valid parent adjustment sets based on the learned local structure. In our experiments on synthetic and realistic data LOAD outperforms global methods in scalability, while providing more accurate effect estimation than local methods.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13729", "url": null, "sourceid": 857, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=FlWl20PFd7", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11422, "modified": "2026-03-29T20:43:10.498102-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=FlWl20PFd7", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "99", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13731, "uid": "ca46c1b9512a7a8315fa3c5a946e8265", "name": "On Computational Limits of FlowAR Models: Expressivity and Efficiency", "authors": [{"id": 22732, "fullname": "Yang Cao", "url": "http://virtual.aistats.org/api/miniconf/users/22732?format=json", "institution": "Wyoming Seminary"}, {"id": 22733, "fullname": "Chengyue Gong", "url": "http://virtual.aistats.org/api/miniconf/users/22733?format=json", "institution": "ByteDance Inc."}, {"id": 22734, "fullname": "Yekun Ke", "url": "http://virtual.aistats.org/api/miniconf/users/22734?format=json", "institution": "Harbin Institute of Technology"}, {"id": 18660, "fullname": "Xiaoyu Li", "url": "http://virtual.aistats.org/api/miniconf/users/18660?format=json", "institution": "University of New South Wales"}, {"id": 18751, "fullname": "Yingyu Liang", "url": "http://virtual.aistats.org/api/miniconf/users/18751?format=json", "institution": "The University of Hong Kong"}, {"id": 14757, "fullname": "Zhizhou Sha", "url": "http://virtual.aistats.org/api/miniconf/users/14757?format=json", "institution": "Tsinghua university"}, {"id": 18162, "fullname": "Zhenmei Shi", "url": "http://virtual.aistats.org/api/miniconf/users/18162?format=json", "institution": "MongoDB"}, {"id": 17715, "fullname": "Zhao Song", "url": "http://virtual.aistats.org/api/miniconf/users/17715?format=json", "institution": "University of California, Berkeley"}], "abstract": "The expressive power and computational complexity of deep visual generative models, such as flow-based and autoregressive (AR) models, have gained considerable interest for their wide-ranging applications in generative tasks. However, the theoretical characterization of their expressiveness through the lens of circuit complexity remains underexplored, particularly for the state-of-the-art architecture like FlowAR proposed by [Ren et al., 2024], which integrates flow-based and autoregressive mechanisms.  This gap limits our understanding of their inherent computational limits and practical efficiency. In this study, we address this gap by analyzing the circuit complexity of the FlowAR architecture. We demonstrate that when the largest feature map produced by the FlowAR model has dimensions $n \\times n \\times c$, the FlowAR model is simulable by a family of threshold circuits $\\mathsf{TC}^0$, which have constant depth $O(1)$ and polynomial width $\\mathrm{poly}(n)$. This is the first study to rigorously highlight the limitations in the expressive power of FlowAR models. Furthermore, we identify the conditions under which the FlowAR model computations can achieve almost quadratic time. To validate our theoretical findings, we present efficient model variant constructions based on low-rank approximations that align with the derived criteria. Our work provides a foundation for future comparisons with other generative paradigms and guides the development of more efficient and expressive implementations.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13731", "url": null, "sourceid": 214, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=FZqLS3Y2t2", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11424, "modified": "2026-03-29T20:43:10.576592-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=FZqLS3Y2t2", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "120", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13735, "uid": "e655c7716a4b3ea67f48c6322fc42ed6", "name": "Adaptive Combinatorial Experimental Design: Pareto Optimality for Decision-Making and Inference", "authors": [{"id": 19341, "fullname": "Hongrui Xie", "url": "http://virtual.aistats.org/api/miniconf/users/19341?format=json", "institution": "University of Science and Technology of China"}, {"id": 17789, "fullname": "Junyu Cao", "url": "http://virtual.aistats.org/api/miniconf/users/17789?format=json", "institution": "University of Texas, Austin"}, {"id": 22737, "fullname": "Kan Xu", "url": "http://virtual.aistats.org/api/miniconf/users/22737?format=json", "institution": "Arizona State University"}], "abstract": "In this paper, we provide the first investigation into adaptive combinatorial experimental design, focusing on the trade-off between regret minimization and statistical power in combinatorial multi-armed bandits (CMAB). While minimizing regret requires repeated exploitation of high-reward arms, accurate inference on reward gaps requires sufficient exploration of suboptimal actions. We formalize this trade-off through the concept of Pareto optimality and establish equivalent conditions for Pareto-efficient learning in CMAB. We consider two relevant cases under different information structures, i.e., full-bandit feedback and semi-bandit feedback, and propose two algorithms MixCombKL and MixCombUCB respectively for these two cases. We provide theoretical guarantees showing that both algorithms are Pareto optimal, achieving finite-time guarantees on both regret and estimation error of arm gaps. Our results further reveal that richer feedback significantly tightens the attainable Pareto frontier, with the primary gains arising from improved estimation accuracy under our proposed methods. Taken together, these findings establish a principled framework for adaptive combinatorial experimentation in multi-objective decision-making.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13735", "url": null, "sourceid": 1492, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=FLiPfTIzuD", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11428, "modified": "2026-03-29T20:43:10.752883-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=FLiPfTIzuD", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "9", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13741, "uid": "0966289037ad9846c5e994be2a91bafa", "name": "Rashomon Effect for Visualizing High-Dimensional Data", "authors": [{"id": 22749, "fullname": "Yiyang Sun", "url": "http://virtual.aistats.org/api/miniconf/users/22749?format=json", "institution": "Duke University"}, {"id": 22750, "fullname": "Haiyang Huang", "url": "http://virtual.aistats.org/api/miniconf/users/22750?format=json", "institution": "Google"}, {"id": 22751, "fullname": "Gaurav Rajesh Parikh", "url": "http://virtual.aistats.org/api/miniconf/users/22751?format=json", "institution": "Duke University"}, {"id": 4499, "fullname": "Cynthia Rudin", "url": "http://virtual.aistats.org/api/miniconf/users/4499?format=json", "institution": "Duke"}], "abstract": "Dimension reduction (DR) is inherently non-unique: multiple embeddings can preserve the structure of high-dimensional data equally well while differing in layout or geometry. In this paper, we formally define the Rashomon set for DR\u2014the collection of `good' embeddings\u2014and show how embracing this multiplicity leads to more powerful and trustworthy representations. Specifically, we pursue three goals. First, we introduce PCA-informed alignment to steer embeddings toward principal components, making axes interpretable without distorting local neighborhoods. Second, we design concept-alignment regularization that aligns an embedding dimension with external knowledge, such as class labels or user-defined concepts. Third, we propose a method to extract common knowledge across the Rashomon set by identifying trustworthy and persistent nearest-neighbor relationships, which we use to construct refined embeddings with improved local structure while preserving global relationships. By moving beyond a single embedding and leveraging the Rashomon set, we provide a flexible framework for building interpretable, robust, and goal-aligned visualizations.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13741", "url": null, "sourceid": 1396, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=En02edjVRs", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11434, "modified": "2026-03-29T20:43:11.160674-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=En02edjVRs", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "136", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13748, "uid": "eddea82ad2755b24c4e168c5fc2ebd40", "name": "Generalizing Behavior via Inverse Reinforcement Learning with Closed-Form Reward Centroids", "authors": [{"id": 22760, "fullname": "Filippo Lazzati", "url": "http://virtual.aistats.org/api/miniconf/users/22760?format=json", "institution": "Politecnico di Milano"}, {"id": 118, "fullname": "Alberto Maria Metelli", "url": "http://virtual.aistats.org/api/miniconf/users/118?format=json", "institution": "Politecnico di Milano"}], "abstract": "We study the problem of generalizing an expert agent's behavior, provided through demonstrations, to new environments and/or additional constraints. Inverse Reinforcement Learning (IRL) offers a promising solution by seeking to recover the expert's underlying reward function, which, if used for planning in the new setting, would reproduce the desired behavior. However, IRL is inherently ill-posed: multiple reward functions, forming the so-called feasible set, can explain the same observed behavior. Since these rewards may induce different policies in the new setting, in the absence of additional information, a decision criterion is needed to select which policy to deploy. In this paper, we propose a novel, principled criterion that selects the \"average\" policy among those induced by the rewards in a certain bounded subset of the feasible set. Remarkably, we show that this policy can be obtained by planning with the reward centroid of that subset, for which we derive a closed-form expression. We then present a provably efficient algorithm for estimating this centroid using only an offline dataset of expert demonstrations. Finally, we conduct numerical simulations that illustrate the relationship between the expert's behavior and the behavior produced by our method.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13748", "url": null, "sourceid": 309, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=EWMpEO82AX", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11441, "modified": "2026-03-29T20:43:11.430572-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=EWMpEO82AX", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "73", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13756, "uid": "428fca9bc1921c25c5121f9da7815cde", "name": "Sample Average Approximation for Alpha-Divergence Minimization with Exponential Convergence Guarantees", "authors": [{"id": 22780, "fullname": "Fran\u00e7ois Bertholom", "url": "http://virtual.aistats.org/api/miniconf/users/22780?format=json", "institution": "Telecom SudParis"}, {"id": 22781, "fullname": "Fran\u00e7ois Roueff", "url": "http://virtual.aistats.org/api/miniconf/users/22781?format=json", "institution": "T\u00e9l\u00e9com Paris"}, {"id": 22782, "fullname": "randal douc", "url": "http://virtual.aistats.org/api/miniconf/users/22782?format=json", "institution": "Telecom Sudparis"}], "abstract": "We study the problem of approximating an unnormalized target distribution using probability densities from an exponential family. Specifically, we establish convergence guarantees for a monotonic alpha-divergence minimization algorithm, which decreases the alpha-divergence at each iteration. To illustrate our theoretical results, we propose an implementable Sample Average Approximation algorithm that solves a discrete approximation of the original problem. Through a detailed analysis of the loss landscape and the algorithm's dynamics, we provide practical design guidelines which suffice to ensure its convergence.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13756", "url": null, "sourceid": 463, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=E1wwfKrIkv", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11449, "modified": "2026-03-29T20:43:11.811511-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=E1wwfKrIkv", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "163", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13758, "uid": "5e76bef6e019b2541ff53db39f407a98", "name": "Free Random Projection for In-Context Reinforcement Learning", "authors": [{"id": 1617, "fullname": "Tomohiro Hayase", "url": "http://virtual.aistats.org/api/miniconf/users/1617?format=json", "institution": "AIST"}, {"id": 22785, "fullname": "Benoit Collins", "url": "http://virtual.aistats.org/api/miniconf/users/22785?format=json", "institution": "Kyoto University"}, {"id": 10172, "fullname": "Nakamasa Inoue", "url": "http://virtual.aistats.org/api/miniconf/users/10172?format=json", "institution": "Tokyo Institute of Technology"}], "abstract": "Hierarchical inductive biases are hypothesized to promote generalizable policies in reinforcement learning, as demonstrated by explicit hyperbolic latent representations and architectures. Therefore, a more flexible approach is to have these biases emerge naturally from the algorithm. We introduce Free Random Projection, an input mapping grounded in free probability theory that constructs random orthogonal matrices where hierarchical structure arises inherently. The free random projection integrates seamlessly into existing in-context reinforcement learning frameworks by encoding hierarchical organization within the input space without requiring explicit architectural modifications. Empirical results on multi-environment benchmarks show that free random projection consistently outperforms the standard random projection, leading to improvements in generalization. Furthermore, analyses within linearly solvable Markov decision processes and investigations of the spectrum of kernel random matrices reveal the theoretical underpinnings of free random projection's enhanced performance, highlighting its capacity for effective adaptation in hierarchically structured state spaces.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13758", "url": null, "sourceid": 1456, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=E0fx0pMyPf", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11451, "modified": "2026-03-29T20:43:11.893073-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=E0fx0pMyPf", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "62", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13764, "uid": "f3d9de86462c28781cbe5c47ef22c3e5", "name": "On the Normalization of Confusion Matrices: Methods and Geometric Interpretations", "authors": [{"id": 22803, "fullname": "Johan Erbani", "url": "http://virtual.aistats.org/api/miniconf/users/22803?format=json", "institution": "Institut National des Sciences Appliqu\u00e9es de Lyon"}, {"id": 22804, "fullname": "Sonia Ben Mokhtar", "url": "http://virtual.aistats.org/api/miniconf/users/22804?format=json", "institution": "CNRS"}, {"id": 22805, "fullname": "Pierre-Edouard Portier", "url": "http://virtual.aistats.org/api/miniconf/users/22805?format=json", "institution": "Caisse d&#x27;Epargne Rh\u00f4ne Alpes"}, {"id": 22806, "fullname": "El\u00f6d Egyed-Zsigmond", "url": "http://virtual.aistats.org/api/miniconf/users/22806?format=json", "institution": "LIRIS / Institut National des Sciences Appliqu\u00e9es de Lyon"}, {"id": 22807, "fullname": "Diana Nurbakova", "url": "http://virtual.aistats.org/api/miniconf/users/22807?format=json", "institution": "INSA Lyon"}], "abstract": "The confusion matrix is a standard tool for evaluating classifiers, providing a detailed view of model errors. In heterogeneous settings, its entries are influenced by two main factors: class similarity, reflecting how easily the model confuses certain classes, and distribution bias, stemming from imbalanced training or test distributions. Because confusion matrix values jointly reflect both factors, it is difficult to disentangle their individual effects. To address this issue, we introduce bi-normalization via Iterative Proportional Fitting, a generalization of row and column normalization. Unlike standard approaches, this method recovers the underlying structure of class similarity. By disentangling error sources, it enables a more precise diagnosis of model behavior and facilitates classifier improvement. We further establish connections between normalization, importance sampling, and class representations in the model\u2019s latent space, thus offering a clearer interpretation of normalization schemes. Our implementation is publicly available.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13764", "url": null, "sourceid": 2214, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=DH3fFrJDs4", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11457, "modified": "2026-03-29T20:43:12.104893-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=DH3fFrJDs4", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "124", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13768, "uid": "459a4ddcb586f24efd9395aa7662bc7c", "name": "Interpretable DNA Sequence Classification via Dynamic Feature Generation in Decision Trees", "authors": [{"id": 12489, "fullname": "Nicolas HUYNH", "url": "http://virtual.aistats.org/api/miniconf/users/12489?format=json", "institution": "University of Cambridge"}, {"id": 13190, "fullname": "Krzysztof Kacprzyk", "url": "http://virtual.aistats.org/api/miniconf/users/13190?format=json", "institution": "University of Cambridge"}, {"id": 22812, "fullname": "Ryan Sheridan", "url": "http://virtual.aistats.org/api/miniconf/users/22812?format=json", "institution": "University of Colorado Anschutz Medical Campus"}, {"id": 22813, "fullname": "David Bentley", "url": "http://virtual.aistats.org/api/miniconf/users/22813?format=json", "institution": "University of Colorado Anschutz Medical Campus"}, {"id": 863, "fullname": "Mihaela van der Schaar", "url": "http://virtual.aistats.org/api/miniconf/users/863?format=json", "institution": "University of Cambridge"}], "abstract": "The analysis of DNA sequences has become critical in numerous fields, from evolutionary biology to understanding gene regulation and disease mechanisms. While deep neural networks can achieve remarkable predictive performance, they typically operate as black boxes. Contrasting these black boxes, axis-aligned decision trees offer a promising direction for interpretable DNA sequence analysis, yet they suffer from a fundamental limitation: considering individual raw features in isolation at each split limits their expressivity, which results in prohibitive tree depths that hinder both interpretability and generalization performance. We address this challenge by introducing DEFT, a novel framework that adaptively generates high-level sequence features during tree construction. DEFT leverages large language models to propose biologically-informed features tailored to the local sequence distributions at each node and to iteratively refine them with a reflection mechanism. Empirically, we demonstrate that DEFT discovers human-interpretable and highly predictive sequence features across a diverse range of genomic tasks.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13768", "url": null, "sourceid": 1292, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=CuMJbiVXnG", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11461, "modified": "2026-03-29T20:43:12.250428-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=CuMJbiVXnG", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "84", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13769, "uid": "bd5af7cd922fd2603be4ee3dc43b0b77", "name": "Modeling Multi-Objective Tradeoffs with Monotonic Utility Functions", "authors": [{"id": 19877, "fullname": "Edward Chen", "url": "http://virtual.aistats.org/api/miniconf/users/19877?format=json", "institution": "Stanford University"}, {"id": 22814, "fullname": "Natalie Dullerud", "url": "http://virtual.aistats.org/api/miniconf/users/22814?format=json", "institution": "Stanford University"}, {"id": 22815, "fullname": "Thomas Niedermayr", "url": "http://virtual.aistats.org/api/miniconf/users/22815?format=json", "institution": "Stanford University"}, {"id": 22816, "fullname": "Elizabeth Kidd", "url": "http://virtual.aistats.org/api/miniconf/users/22816?format=json", "institution": "Stanford University"}, {"id": 22817, "fullname": "Ransalu Senanayake", "url": "http://virtual.aistats.org/api/miniconf/users/22817?format=json", "institution": "Arizona State University"}, {"id": 22818, "fullname": "Pang Wei Koh", "url": "http://virtual.aistats.org/api/miniconf/users/22818?format=json", "institution": "University of Washington"}, {"id": 18018, "fullname": "Sanmi Koyejo", "url": "http://virtual.aistats.org/api/miniconf/users/18018?format=json", "institution": "Stanford University &amp; Virtue AI"}, {"id": 22819, "fullname": "Carlos Guestrin", "url": "http://virtual.aistats.org/api/miniconf/users/22819?format=json", "institution": "Stanford University"}], "abstract": "Countless science and engineering applications in multi-objective optimization (MOO) necessitate that decision-makers (DMs) select a Pareto-optimal (PO) solution which aligns with their preferences. Evaluating individual solutions is often expensive, and the high-dimensional trade-off space makes exhaustive exploration of the full Pareto frontier (PF) infeasible. We introduce a novel, principled two-step process for obtaining a compact set of PO points that aligns with user preferences, which are specified a priori as general monotonic utility functions (MFs). Our process (1) densely samples the user's region of interest on the PF, then (2) sparsifies the results into a small, diverse set for the DM. We instantiate this framework with soft-hard functions (SHFs), an intuitive class of MFs that operationalizes the common expert heuristic of imposing soft and hard bounds. We provide extensive empirical validation of our framework instantiated with SHFs on diverse domains, including brachytherapy, engineering design, and large language models. For brachytherapy, our approach returns a compact set of points with over 3% greater SHF-defined utility than the next best approach. Among the other domains, our approach consistently leads in utility, as a final compact set of just 5 points captures over 99% of the utility offered by the entire dense set.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13769", "url": null, "sourceid": 2128, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=Cu3coRhlM8", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11462, "modified": "2026-03-29T20:43:12.287974-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=Cu3coRhlM8", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "102", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13775, "uid": "eb160de1de89d9058fcb0b968dbbbd68", "name": "Orientability of Causal Relations in Time Series using Summary Causal Graphs and Faithful Distributions", "authors": [{"id": 22834, "fullname": "Timoth\u00e9e Loranchet", "url": "http://virtual.aistats.org/api/miniconf/users/22834?format=json", "institution": "Institut national de la sant\u00e9 et de la recherche m\u00e9dicale"}, {"id": 9985, "fullname": "Charles ASSAAD", "url": "http://virtual.aistats.org/api/miniconf/users/9985?format=json", "institution": "EasyVista"}], "abstract": "Understanding causal relations between temporal variables is a central challenge in time series analysis, particularly when the full causal structure is unknown. Even when the full causal structure cannot be fully specified, experts often succeed in providing a high-level abstraction of the causal graph, known as a summary causal graph, which captures the main causal relations between different time series while abstracting away micro-level details. In this work, we present conditions that guarantee the orientability of micro-level edges between temporal variables given the background knowledge encoded in a summary causal graph and assuming having access to a faithful and causally sufficient distribution with respect to the true unknown graph. Our results provide theoretical guarantees for edge orientation at the micro-level, even in the presence of cycles or bidirected edges at the macro-level. These findings offer practical guidance for leveraging SCGs to inform causal discovery in complex temporal systems and highlight the value of incorporating expert knowledge to improve causal inference from observational time series data.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13775", "url": null, "sourceid": 117, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=CORGtiRxxo", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11468, "modified": "2026-03-29T20:43:12.554900-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=CORGtiRxxo", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "133", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13776, "uid": "03f544613917945245041ea1581df0c2", "name": "From Transformers to State Spaces: GeoMamba-SE(3) for Fast and Accurate Molecular Learning", "authors": [{"id": 22835, "fullname": "Jiayu Qin", "url": "http://virtual.aistats.org/api/miniconf/users/22835?format=json", "institution": "State University of New York at Buffalo"}, {"id": 22836, "fullname": "Zhengquan Luo", "url": "http://virtual.aistats.org/api/miniconf/users/22836?format=json", "institution": "Mohamed bin Zayed University of Artificial Intelligence"}, {"id": 22837, "fullname": "Jian Chen", "url": "http://virtual.aistats.org/api/miniconf/users/22837?format=json", "institution": "Dolby Laboratories"}, {"id": 22838, "fullname": "Xuhui Li", "url": "http://virtual.aistats.org/api/miniconf/users/22838?format=json", "institution": "Mohamed bin Zayed University of Artificial Intelligence"}, {"id": 22839, "fullname": "Jiayi Chen", "url": "http://virtual.aistats.org/api/miniconf/users/22839?format=json", "institution": "Fujian Normal University"}, {"id": 803, "fullname": "Zhiqiang Xu", "url": "http://virtual.aistats.org/api/miniconf/users/803?format=json", "institution": "Baidu"}], "abstract": "Transformers play an important role in molecular representation learning, enabling unsupervised learning from large scale unlabeled molecule datasets. However, existing Transformer based methods suffer from heavy training computation and slow inference. To accelerate the computation and relieve the burdensome pre-training, we propose a Mamba-based framework that leverages selective state space models to learn molecular representations more efficiently. Unlike conventional methods, our model, GeoMamba-SE(3), offers streamlined computation with linear-time complexity. However, naively applying Mamba to molecules struggles with SE(3) symmetry, representations can drift under rotations/translations\u2014leading to chemically inconsistent features. To address this, we introduce a geometry and statistics aware design: (i) complete local frames at atoms by converting geometric vectors into scalar channels suitable for SSMs; (ii) multi-stream Mamba blocks are modulated by SE(3)-invariant scalars to preserve geometric stability; and (iii) we impose statistical symmetry constraints via orbit-kernel losses and invariant risk minimization, treating SE(3) actions and conformers as environments. This yields practical SE(3) stability without heavy high-order tensor representations. Experiments show that our method achieves new state-of-the-art performance benchmarks on the MoleculeNet datasets, while using only one-sixth of the training computation and 57\\% less computation for inference.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13776", "url": null, "sourceid": 1089, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=CG2BZY2HIe", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11469, "modified": "2026-03-29T20:43:12.596877-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=CG2BZY2HIe", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "66", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13777, "uid": "a223c6b3710f85df22e9377d6c4f7553", "name": "Graph Learning is Suboptimal in Causal Bandits", "authors": [{"id": 22840, "fullname": "Mohammad Shahverdikondori", "url": "http://virtual.aistats.org/api/miniconf/users/22840?format=json", "institution": "EPFL - EPF Lausanne"}, {"id": 19736, "fullname": "Jalal Etesami", "url": "http://virtual.aistats.org/api/miniconf/users/19736?format=json", "institution": "TU Munich"}, {"id": 935, "fullname": "Negar Kiyavash", "url": "http://virtual.aistats.org/api/miniconf/users/935?format=json", "institution": "\u00c9cole Polytechnique F\u00e9d\u00e9rale de Lausanne"}], "abstract": "We study regret minimization in causal bandits under causal sufficiency where the underlying causal structure is not known to the agent. Previous work has focused on identifying the reward\u2019s parents and then applying classic bandit methods to them, or jointly learning the parents while minimizing regret.  We investigate whether such strategies are optimal. Somewhat counterintuitively, our results show that learning the parent set is suboptimal. We do so by proving that there exist instances where regret minimization and parent identification are fundamentally conflicting objectives. We further analyze both the known and unknown parent set size regimes, establish novel regret lower bounds that capture the combinatorial structure of the action space. Building on these insights, we propose nearly optimal algorithms that bypass graph and parent recovery, demonstrating that parent identification is indeed unnecessary for regret minimization. Experiments confirm that there exists a large performance gap between our method and existing baselines in various environments.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13777", "url": null, "sourceid": 1367, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=CF8vvfItK3", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11470, "modified": "2026-03-29T20:43:12.639891-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=CF8vvfItK3", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "71", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13779, "uid": "49af6c4e558a7569d80eee2e035e2bd7", "name": "High-Probability Bounds for Heterogeneous Local Differential Privacy", "authors": [{"id": 18367, "fullname": "Maryam Aliakbarpour", "url": "http://virtual.aistats.org/api/miniconf/users/18367?format=json", "institution": "Rice University"}, {"id": 22845, "fullname": "Alireza Fallah", "url": "http://virtual.aistats.org/api/miniconf/users/22845?format=json", "institution": "Rice University"}, {"id": 22846, "fullname": "Swaha Roy", "url": "http://virtual.aistats.org/api/miniconf/users/22846?format=json", "institution": "Rice University"}, {"id": 19807, "fullname": "Ria Stevens", "url": "http://virtual.aistats.org/api/miniconf/users/19807?format=json", "institution": "Rice University"}], "abstract": "We study statistical estimation under local differential privacy (LDP) when users may hold heterogeneous privacy levels and accuracy must be guaranteed with high probability. Departing from the common in-expectation analyses, and for one-dimensional and multi-dimensional mean estimation problems, we develop finite sample upper bounds in $\\ell_2$-norm that hold with probability at least $1-\\beta$. We complement these results with matching minimax lower bounds, establishing the optimality (up to constants) of our guarantees in the heterogeneous LDP regime. We further study distribution learning in $\\ell_\\infty$-distance, designing an algorithm with high-probability guarantees under heterogeneous privacy demands. Our techniques offer principled guidance for designing mechanisms in settings with user-specific privacy levels.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13779", "url": null, "sourceid": 1587, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=CD1TaNT1yi", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11472, "modified": "2026-03-29T20:43:12.717699-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=CD1TaNT1yi", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "81", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13786, "uid": "c6e19e830859f2cb9f7c8f8cacb8d2a6", "name": "On the Latent Information Geometry of the Grassmann Manifold", "authors": [{"id": 19840, "fullname": "Lorenzo Cazzella", "url": "http://virtual.aistats.org/api/miniconf/users/19840?format=json", "institution": "Politecnico di Milano"}, {"id": 489, "fullname": "Soren Hauberg", "url": "http://virtual.aistats.org/api/miniconf/users/489?format=json", "institution": "Technical University of Denmark, Denmark"}, {"id": 22597, "fullname": "Georgios Arvanitidis", "url": "http://virtual.aistats.org/api/miniconf/users/22597?format=json", "institution": "Technical University of Denmark"}, {"id": 22871, "fullname": "Matteo Matteucci", "url": "http://virtual.aistats.org/api/miniconf/users/22871?format=json", "institution": "Politecnico di Milano"}], "abstract": "Modeling linear subspaces and relations among them naturally arises in several applications in signal processing, computer vision, and system identification. In this paper, we investigate the latent information geometry of deep generative models that output linear subspaces. Such subspaces are members of the Grassmann manifold, which we model with a matrix Bingham distribution as a likelihood. We derive the Fisher-Rao metric on the statistical manifold of the matrix Bingham parameters, and propose pulling this back to the latent space to achieve uncertainty-aware and identifiable latent representations. We provide numerical results assessing the meaningfulness of the achieved latent subspace representations on a relevant vehicular wireless communications scenario.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13786", "url": null, "sourceid": 581, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=BhLHFZwMEr", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11479, "modified": "2026-03-29T20:43:12.983850-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=BhLHFZwMEr", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "127", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13789, "uid": "74071a673307ca7459bcf75fbd024e09", "name": "Provable Effects of Data Replay in Continual Learning: A Feature Learning Perspective", "authors": [{"id": 19912, "fullname": "Meng Ding", "url": "http://virtual.aistats.org/api/miniconf/users/19912?format=json", "institution": "University at Buffalo"}, {"id": 22875, "fullname": "Jinhui Xu", "url": "http://virtual.aistats.org/api/miniconf/users/22875?format=json", "institution": "University of Science and Technology of China"}, {"id": 22876, "fullname": "Kaiyi Ji", "url": "http://virtual.aistats.org/api/miniconf/users/22876?format=json", "institution": "State University of New York at Buffalo"}], "abstract": "Continual learning (CL) aims to train models on a sequence of tasks while retaining performance on previously learned ones. A core challenge in this setting is \\textit{catastrophic forgetting}, where new learning interferes with past knowledge. Among various mitigation strategies, data-replay methods\u2014where past samples are periodically revisited\u2014are considered simple yet effective, especially when memory constraints are relaxed. However, the theoretical effectiveness of full data replay, where all past data is accessible during training, remains largely unexplored. In this paper, we present the first theoretical framework for analyzing full data-replay training in continual learning from a feature learning perspective. Adopting a multi-view data model, we identify the signal-to-noise ratio (SNR) as a critical factor affecting forgetting. Focusing on task-incremental binary classification across $M$ tasks, our analysis verifies two key conclusions: (1) forgetting can still occur under full replay when the cumulative noise from later tasks dominates the signal from earlier ones; and (2) with sufficient signal accumulation, data replay can recover earlier tasks-even if their initial learning was poor. Notably, we uncover a novel insight into task ordering: prioritizing higher-signal tasks not only facilitates learning of lower-signal tasks but also helps prevent catastrophic forgetting\u2014highlighting the importance of order-aware replay strategies. We validate our theoretical findings through synthetic experiments that visualize the interplay between signal learning and noise memorization across varying SNRs and task correlation regimes.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13789", "url": null, "sourceid": 477, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=BPYDrYmGWs", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11482, "modified": "2026-03-29T20:43:13.092785-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=BPYDrYmGWs", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "146", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13790, "uid": "0d0871f0806eae32d30983b62252da50", "name": "Best Policy Learning from Trajectory Preference Feedback", "authors": [{"id": 22877, "fullname": "Akhil Agnihotri", "url": "http://virtual.aistats.org/api/miniconf/users/22877?format=json", "institution": "University of Southern California"}, {"id": 1223, "fullname": "Rahul Jain", "url": "http://virtual.aistats.org/api/miniconf/users/1223?format=json", "institution": "University of Southern California"}, {"id": 22878, "fullname": "Deepak Ramachandran", "url": "http://virtual.aistats.org/api/miniconf/users/22878?format=json", "institution": "Google"}, {"id": 22879, "fullname": "Zheng Wen", "url": "http://virtual.aistats.org/api/miniconf/users/22879?format=json", "institution": "OpenAI"}], "abstract": "Reinforcement Learning from Human Feedback (RLHF) has emerged as a powerful approach for aligning generative models, but its reliance on learned reward models makes it vulnerable to mis-specification and reward hacking. Preference-based Reinforcement Learning (PbRL) offers a more robust alternative by directly leveraging noisy binary comparisons over trajectories. We study the best policy identification problem in PbRL, motivated by post-training optimization of generative models, for example, during multi-turn interactions. Learning in this setting combines an offline preference dataset\u2014potentially biased or out-of-distribution and collected from a rater of subpar 'competence'\u2014with online pure exploration, making systematic online learning essential. To this end, we propose Posterior Sampling for Preference Learning ($\\mathsf{PSPL}$), a novel algorithm inspired by Top-Two Thompson Sampling that maintains posteriors over the reward model and dynamics. We provide the first Bayesian simple regret guarantees for PbRL and introduce an efficient approximation that outperforms existing baselines on simulation and image generation benchmarks.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13790", "url": null, "sourceid": 929, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=BLyDBJkina", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11483, "modified": "2026-03-29T20:43:13.134661-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=BLyDBJkina", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "18", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13791, "uid": "6395ebd0f4b478145ecfbaf939454fa4", "name": "Time-Aware Synthetic Control", "authors": [{"id": 18498, "fullname": "Saeyoung Rho", "url": "http://virtual.aistats.org/api/miniconf/users/18498?format=json", "institution": "Columbia University"}, {"id": 22880, "fullname": "Cyrus Illick", "url": "http://virtual.aistats.org/api/miniconf/users/22880?format=json", "institution": "Columbia University"}, {"id": 19471, "fullname": "Samhitha Narasipura", "url": "http://virtual.aistats.org/api/miniconf/users/19471?format=json", "institution": "Columbia University"}, {"id": 19632, "fullname": "Alberto Abadie", "url": "http://virtual.aistats.org/api/miniconf/users/19632?format=json", "institution": "MIT"}, {"id": 17670, "fullname": "Daniel Hsu", "url": "http://virtual.aistats.org/api/miniconf/users/17670?format=json", "institution": "Columbia University"}, {"id": 18509, "fullname": "Vishal Misra", "url": "http://virtual.aistats.org/api/miniconf/users/18509?format=json", "institution": ", Columbia University"}], "abstract": "The synthetic control (SC) framework is widely used for observational causal inference with time-series panel data. Despite its success across diverse applications, existing SC methods typically treat pre-intervention time indices as exchangeable, meaning they may fail to exploit temporal structure when strong trends are present. We propose Time-Aware Synthetic Control (TASC), a method that addresses this limitation by adopting a state-space model with a constant trend component while preserving the low-rank structure of the signal. TASC uses the Kalman filter and the Rauch\u2013Tung\u2013Striebel smoother in two steps: it first fits a generative time-series model with expectation\u2013maximization and then performs counterfactual inference. We evaluate TASC on simulated and real-world datasets spanning policy evaluation and sports prediction. Our results demonstrate that TASC offers advantages in settings with high observation noise and long prediction horizons.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13791", "url": null, "sourceid": 954, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=BKni5bxkA9", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11484, "modified": "2026-03-29T20:43:13.173215-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=BKni5bxkA9", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "166", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13792, "uid": "fd2c5e4680d9a01dba3aada5ece22270", "name": "SiGHT: A Self-Supervised Graph-based Hallucination DeTection Framework for Domain-Specific LLMs", "authors": [{"id": 19492, "fullname": "Chen Ziying", "url": "http://virtual.aistats.org/api/miniconf/users/19492?format=json", "institution": "Department of Computer Science, National Yang Ming Chiao Tung University"}, {"id": 18248, "fullname": "Meng-Fen Chiang", "url": "http://virtual.aistats.org/api/miniconf/users/18248?format=json", "institution": "National Yang Ming Chiao Tung University"}, {"id": 22881, "fullname": "Wen-Chih Peng", "url": "http://virtual.aistats.org/api/miniconf/users/22881?format=json", "institution": "National Yang Ming Chiao Tung University"}], "abstract": "Factual reliability in domain-specific Large Language Models (LLMs) is paramount in high-stakes applications where incorrect outputs carry significant risks. Current detection methodologies often rely on expensive retrieval validation or labor-intensive manual annotation, creating substantial barriers to scalable deployment. To bridge the gap, we propose SiGHT, a self-supervised graph framework designed for efficient hallucination detection in specialized contexts. SiGHT introduces an automated training pipeline that leverages prompt strategies to synthesize plausible hallucinated content from structured knowledge, effectively eliminating the need for human labeling. By mapping texts to high-resolution word-level relational graphs, the framework employs a Graph Attention Network (GAT) to model fine-grained semantic dependencies and identify structural inconsistencies. Empirical evaluations on the MSMARCO-QnA and RAGTruth-QA benchmarks demonstrate that SiGHT achieves a 46.94% relative F1 gain over prior graph baselines. Notably, SiGHT remains competitive with state of the art detectors while utilizing only 0.03M parameters and incurring a minimal inference latency of 0.342 seconds per instance. Dominating the accuracy--efficiency frontier, SiGHT delivers a robust and scalable architecture for real-time hallucination monitoring in high-stakes specialized pipelines.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13792", "url": null, "sourceid": 1135, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=BExAfKElad", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11485, "modified": "2026-03-29T20:43:13.217418-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=BExAfKElad", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "150", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13795, "uid": "cc1aa436277138f61cda703991069eaf", "name": "Recency Biased Causal Attention for Time-series Forecasting", "authors": [{"id": 22886, "fullname": "Kareem Hegazy", "url": "http://virtual.aistats.org/api/miniconf/users/22886?format=json", "institution": "University of California, Berkeley"}, {"id": 1341, "fullname": "Michael Mahoney", "url": "http://virtual.aistats.org/api/miniconf/users/1341?format=json", "institution": "&quot;University of California, Berkeley&quot;"}, {"id": 12654, "fullname": "N. Benjamin Erichson", "url": "http://virtual.aistats.org/api/miniconf/users/12654?format=json", "institution": "Berkeley Lab and ICSI"}], "abstract": "Recency bias is a useful inductive prior for sequential modeling: it emphasizes nearby observations and can still allow longer-range dependencies. Standard Transformer attention lacks this property, relying on all-to-all interactions that overlook the causal and often local structure of temporal data. We propose a simple mechanism to introduce recency bias by reweighting attention scores with a smooth heavy-tailed decay. This adjustment strengthens local temporal dependencies without sacrificing the flexibility to capture broader and data-specific correlations. We show that recency-biased attention consistently improves sequential modeling, aligning Transformer more closely with the read\u2013ignore\u2013write operations of RNNs. Finally, we demonstrate that our approach achieves competitive and often superior performance on challenging time-series forecasting benchmarks.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13795", "url": null, "sourceid": 977, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=BAimLZ9Wzx", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11488, "modified": "2026-03-29T20:43:13.329078-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=BAimLZ9Wzx", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "138", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13797, "uid": "621461af90cadfdaf0e8d4cc25129f91", "name": "Efficient and Accurate Tensor Compression via Recursive Sketching", "authors": [{"id": 22890, "fullname": "Amit Sharma", "url": "http://virtual.aistats.org/api/miniconf/users/22890?format=json", "institution": "Indian Institute of Technology, Hyderabad"}, {"id": 19690, "fullname": "Mohammad Azhar Khan", "url": "http://virtual.aistats.org/api/miniconf/users/19690?format=json", "institution": "IIT Hyderabad"}, {"id": 22544, "fullname": "Rameshwar Pratap", "url": "http://virtual.aistats.org/api/miniconf/users/22544?format=json", "institution": "Indian Institute of Technology, Hyderabad"}], "abstract": "The computation of inner products between high-order tensor data points is a fundamental task in numerous machine learning and scientific applications. However, the naive approach to these computations incurs exponential time complexity with respect to the number of modes. The work of  Rakhshan and Rabusseau (AISTAT, 2020) introduced an extension of the random projection tailored for tensor datasets, which compresses large tensors into compact vectors (\\textit{a.k.a} sketches). Their approach provides unbiased estimates of the original pairwise inner products. However, the variance of their estimates grows exponentially with the number of modes, making their estimates less reliable for small sketch sizes. In this work, we propose improved sketching algorithms that provide unbiased estimates for pairwise inner products, with significantly lower variance - independent of the number of modes\u2014compared to that of  Rakhshan and Rabusseau (AISTAT, 2020). Furthermore, our approach offers asymptotically improved time complexity. Our sketching algorithm builds on the framework of Ahle et al. (SODA 2020), which proposed sketching techniques for high-degree \\textit{polynomial kernels}.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13797", "url": null, "sourceid": 907, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=AvXsY6fCMS", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11490, "modified": "2026-03-29T20:43:13.423275-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=AvXsY6fCMS", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "43", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13798, "uid": "f387624df552cea2f369918c5e1e12bc", "name": "Efficient Learning of Stationary Diffusions with Stein-type Discrepancies", "authors": [{"id": 19391, "fullname": "Fabian Bleile", "url": "http://virtual.aistats.org/api/miniconf/users/19391?format=json", "institution": "TU Munich"}, {"id": 22891, "fullname": "Sarah Lumpp", "url": "http://virtual.aistats.org/api/miniconf/users/22891?format=json", "institution": "Technical University of Munich"}, {"id": 9554, "fullname": "Mathias Drton", "url": "http://virtual.aistats.org/api/miniconf/users/9554?format=json", "institution": "Technical University of Munich"}], "abstract": "Learning a stationary diffusion amounts to estimating the parameters of a stochastic differential equation whose stationary distribution matches a target distribution. We build on the recently introduced kernel deviation from stationarity (KDS), which enforces stationarity by evaluating expectations of the diffusion's generator in a reproducing kernel Hilbert space. Leveraging the connection between KDS and Stein discrepancies, we introduce the Stein-type KDS (SKDS) as an alternative formulation. We prove that a vanishing SKDS guarantees alignment of the learned diffusion\u2019s stationary distribution with the target. Furthermore, under broad parametrizations, SKDS is convex with an empirical version that is $\\epsilon$-quasiconvex with high probability. Empirically, learning with SKDS attains comparable accuracy to KDS while substantially reducing computational cost, and yields improvements over the majority of competitive baselines.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13798", "url": null, "sourceid": 553, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=AoYF9Xgcac", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11491, "modified": "2026-03-29T20:43:13.462806-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=AoYF9Xgcac", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "55", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13807, "uid": "d9731321ef4e063ebbee79298fa36f56", "name": "Uncertainty Quantification for Named Entity Recognition via Conformal Prediction", "authors": [{"id": 22906, "fullname": "Matthew Singer", "url": "http://virtual.aistats.org/api/miniconf/users/22906?format=json", "institution": "North Carolina State University"}, {"id": 22907, "fullname": "Karl Pazdernik", "url": "http://virtual.aistats.org/api/miniconf/users/22907?format=json", "institution": "Pacific Northwest National Laboratory / North Carolina State University"}, {"id": 22908, "fullname": "Srijan Sengupta", "url": "http://virtual.aistats.org/api/miniconf/users/22908?format=json", "institution": "North Carolina State University"}], "abstract": "Named Entity Recognition (NER) is a foundational component in many language tasks, such as knowledge graph construction, information extraction, and question answering. However, existing NER models typically output a single predicted label sequence without any quantification of uncertainty, leaving downstream applications vulnerable to cascading errors. We introduce a conformal prediction framework for NER that produces prediction sets over full label sequences with finite-sample coverage guarantees, serving an analogous role to confidence intervals in classical statistics. To improve efficiency, we propose three innovations: (i) hybrid probability-index nonconformity scores, (ii) conditional calibration across strata such as sentence length and language, and (iii) an adaptation of the RAPS procedure to sequence labeling. These techniques mitigate the problem of overly large prediction sets while maintaining valid coverage. Experiments on CoNLL++, CoNLL-Reduced, and WikiNEuRal benchmarks demonstrate that our methods consistently achieve the target confidence while producing efficient prediction sets across diverse base models. This work establishes a statistically principled approach to uncertainty-aware NER with direct benefits for downstream knowledge-driven NLP systems.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13807", "url": null, "sourceid": 1398, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=A9VYjTIVi7", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11500, "modified": "2026-03-29T20:43:13.843917-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=A9VYjTIVi7", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "173", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13809, "uid": "556f391937dfd4398cbac35e050a2177", "name": "Fair Clustering via Hierarchical Fair-Dirichlet Prior", "authors": [{"id": 19420, "fullname": "Abhisek Chakraborty", "url": "http://virtual.aistats.org/api/miniconf/users/19420?format=json", "institution": "Eli Lilly and Company"}, {"id": 1069, "fullname": "Anirban Bhattacharya", "url": "http://virtual.aistats.org/api/miniconf/users/1069?format=json", "institution": "Texas A&amp;M University"}, {"id": 1071, "fullname": "Debdeep Pati", "url": "http://virtual.aistats.org/api/miniconf/users/1071?format=json", "institution": "Texas A&amp;M University"}], "abstract": "The advent of ML-driven decision-making has led to an increasing focus on algorithmic fairness. The widespread utility of clustering  has naturally prompted proliferation of literature on fair clustering. A popular notion of fairness in clustering mandates the clusters to be balanced, i.e., each level of a protected attribute must be approximately equally represented in each cluster.  In this article, we offer a novel model-based formulation of fair clustering, complementing the existing literature which is almost exclusively based on optimizing appropriate objective functions. We first rigorously define a notion of fair clustering in the population level and develop a Bayesian methodology equipped with a novel hierarchical prior specification that targets the population level objective by enforcing the notion of balance in the resulting clusters.  In addition, we devise a scheme for principled performance evaluation of competing algorithms leveraging on a concrete notion of optimal recovery. An efficient collapsed Gibbs sampler is developed to sample from the posterior by integrating a novel scheme for non-uniform sampling from the space of binary matrices with fixed margin with a proposal guided by optimal transport. Superior empirical performance of the proposed methodology, compared to the state-of-the-art, is demonstrated across numerical experiments, benchmark data-sets, and  gender-neutral fair clustering in distress analysis interview corpus.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13809", "url": null, "sourceid": 684, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=9ws2fiGpLX", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11502, "modified": "2026-03-29T20:43:13.920409-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=9ws2fiGpLX", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "61", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13810, "uid": "7f39f8317fbdb1988ef4c628eba02591", "name": "On the Expressivity of Selective State-Space Layers: A Multivariate Polynomial Approach", "authors": [{"id": 22911, "fullname": "Edo Cohen-Karlik", "url": "http://virtual.aistats.org/api/miniconf/users/22911?format=json", "institution": "Tel Aviv University"}, {"id": 12519, "fullname": "Itamar Zimerman", "url": "http://virtual.aistats.org/api/miniconf/users/12519?format=json", "institution": "Tel-Aviv University"}, {"id": 22912, "fullname": "Liane Galanti", "url": "http://virtual.aistats.org/api/miniconf/users/22912?format=json", "institution": "Princeton University"}, {"id": 22913, "fullname": "Ido Atad", "url": "http://virtual.aistats.org/api/miniconf/users/22913?format=json", "institution": "Tel Aviv University"}, {"id": 22914, "fullname": "Amir Globerson", "url": "http://virtual.aistats.org/api/miniconf/users/22914?format=json", "institution": "Tel Aviv University"}, {"id": 22915, "fullname": "Lior Wolf", "url": "http://virtual.aistats.org/api/miniconf/users/22915?format=json", "institution": "Tel Aviv University"}], "abstract": "Recent advances in efficient sequence modeling have introduced selective state-space layers, a key component of the Mamba architecture, which have demonstrated remarkable success in a wide range of NLP and vision tasks. While Mamba\u2019s empirical performance has matched or surpassed SoTA transformers on such diverse benchmarks, the theoretical foundations underlying its powerful representational capabilities remain less explored. In this work, we investigate the expressivity of selective state-space layers using multivariate polynomials, and prove that they surpass linear transformers in expressiveness. Consequently, our findings reveal that Mamba offers superior representational power over linear attention-based models for long-sequences, while not sacrificing their generalization. Our theoretical insights are validated by a comprehensive set of empirical experiments on various datasets.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13810", "url": null, "sourceid": 61, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=9vVYKBQF1N", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11503, "modified": "2026-03-29T20:43:13.958440-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=9vVYKBQF1N", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "124", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13814, "uid": "c45147dee729311ef5b5c3003946c48f", "name": "Preference-based Conditional Treatment Effects and Policy Learning", "authors": [{"id": 22921, "fullname": "Dovid Parnas", "url": "http://virtual.aistats.org/api/miniconf/users/22921?format=json", "institution": "Technion - Israel Institute of Technology"}, {"id": 13031, "fullname": "Mathieu Even", "url": "http://virtual.aistats.org/api/miniconf/users/13031?format=json", "institution": "ENS - Inria"}, {"id": 12293, "fullname": "julie Josse", "url": "http://virtual.aistats.org/api/miniconf/users/12293?format=json", "institution": "Polytechnique/INRIA"}, {"id": 18006, "fullname": "Uri Shalit", "url": "http://virtual.aistats.org/api/miniconf/users/18006?format=json", "institution": "Technion"}], "abstract": "This paper introduces a preference-based framework for conditional treatment effect estimation and policy learning, by defining the Conditional Preference-based Treatment Effect (CPTE). CPTE only requires that different outcomes can be compared, as opposed to traditional conditional average treatment effects (CATE) that quantify the conditional magnitude of an intervention on a real-valued outcome. This allows for flexible modeling of heterogeneous effects even when outcomes are multivariate, ordinal, or preference-driven. We show how CPTE enables interpretable estimands, and propose estimation strategies based on matching, quantile regression, and distributional methods. Building on this, we develop efficient influence function\u2013based estimators to correct plug-in biases and optimize policy values. Through synthetic and semi-synthetic experiments, we demonstrate that our approach produces personalized policies that are robust to outcome heterogeneity and align with ground-truth preferences, improving decision quality over CATE-based baselines.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13814", "url": null, "sourceid": 116, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=9QwXTpmwvt", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11507, "modified": "2026-03-29T20:43:14.118213-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=9QwXTpmwvt", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "140", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13816, "uid": "95d309f0b035d97f69902e7972c2b2e6", "name": "In-memory Training on Analog Devices with Limited Conductance States via Multi-tile Residual Learning", "authors": [{"id": 22925, "fullname": "Jindan Li", "url": "http://virtual.aistats.org/api/miniconf/users/22925?format=json", "institution": "Cornell University"}, {"id": 22926, "fullname": "Zhaoxian Wu", "url": "http://virtual.aistats.org/api/miniconf/users/22926?format=json", "institution": "Cornell University"}, {"id": 22927, "fullname": "Gaowen Liu", "url": "http://virtual.aistats.org/api/miniconf/users/22927?format=json", "institution": "Cisco Systems"}, {"id": 22928, "fullname": "Tayfun Gokmen", "url": "http://virtual.aistats.org/api/miniconf/users/22928?format=json", "institution": "IBM, International Business Machines"}, {"id": 5616, "fullname": "Tianyi Chen", "url": "http://virtual.aistats.org/api/miniconf/users/5616?format=json", "institution": "Cornell University / RPI"}], "abstract": "Analog in-memory computing (AIMC) accelerators enable efficient deep neural network computation directly within memory using resistive crossbar arrays, where model parameters are represented by the conductance states of memristive devices.  However, effective in-memory training typically requires at least 8-bit conductance states to match digital baselines. Realizing such fine-grained states is costly and often requires complex noise mitigation techniques that increase circuit complexity and energy consumption. In practice, many promising memristive devices such as ReRAM offer only about 4-bit resolution due to fabrication constraints, and this limited update precision substantially degrades training accuracy. To enable on-chip training with these limited-state devices, this paper proposes a \\emph{multi-tile residual learning} framework that sequentially learns on multiple crossbar tiles to compensate the residual errors from low-precision weight updates.  Our theoretical analysis shows that the optimality gap shrinks with the number of tiles and achieves a linear convergence rate. Experiments on standard image classification benchmarks demonstrate that our method consistently outperforms state-of-the-art in-situ analog training strategies under limited-state settings, while incurring only moderate hardware overhead as confirmed by our cost analysis.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13816", "url": null, "sourceid": 1578, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=9GEOJ3KgVM", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11509, "modified": "2026-03-29T20:43:14.211863-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=9GEOJ3KgVM", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "85", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13817, "uid": "ee8374ec4e4ad797d42350c904d73077", "name": "Tractable Shapley Values and Interactions via Tensor Networks", "authors": [{"id": 19519, "fullname": "Farzaneh Heidari", "url": "http://virtual.aistats.org/api/miniconf/users/19519?format=json", "institution": null}, {"id": 37, "fullname": "Chao Li", "url": "http://virtual.aistats.org/api/miniconf/users/37?format=json", "institution": "RIKEN AIP"}, {"id": 946, "fullname": "Guillaume Rabusseau", "url": "http://virtual.aistats.org/api/miniconf/users/946?format=json", "institution": "Mila, Universit\u00e9 de Montr\u00e9al"}], "abstract": "We show how to replace the $O(2^n)$ coalition enumeration  over $n$ features behind Shapley values and Shapley-style interaction indices with a few-evaluation scheme on a tensor-network (TN) surrogate: TN-SHAP. The key idea is to represent a predictor\u2019s local behavior as a factorized multilinear map, so that coalitional quantities become linear probes of a coefficient tensor. TN-SHAP replaces exhaustive coalition sweeps with just a small number of targeted evaluations to extract order$-k$ Shapley interactions.  In particular, both order-1 (single-feature) and order-2 (pairwise) computations have cost $O\\!\\big(n\\,\\mathrm{poly}(\\chi) + n^2\\big)$, where $\\chi$ is the TN\u2019s maximal cut rank.  We provide theoretical guarantees on the approximation error and tractability of TN-SHAP.  On UCI datasets, our method matches enumeration on the fitted surrogate while reducing evaluation by orders of magnitude and achieves \\textbf{25--1000$\\times$} wall-clock speedups over KernelSHAP-IQ at comparable accuracy, while amortizing training across local cohorts.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13817", "url": null, "sourceid": 1353, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=9FHX6mjV4B", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11510, "modified": "2026-03-29T20:43:14.256725-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=9FHX6mjV4B", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "170", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13818, "uid": "61b4a64be663682e8cb037d9719ad8cd", "name": "Sequential Off-Policy Learning with Logarithmic Smoothing", "authors": [{"id": 22929, "fullname": "Maxime Haddouche", "url": "http://virtual.aistats.org/api/miniconf/users/22929?format=json", "institution": "INRIA"}, {"id": 22930, "fullname": "Otmane Sakhi", "url": "http://virtual.aistats.org/api/miniconf/users/22930?format=json", "institution": "Criteo"}], "abstract": "Off-policy learning enables training policies from logged interaction data. Most prior work considers the batch setting, where a policy is learned from data generated by a single behavior policy. In real systems, however, policies are updated and redeployed repeatedly, each time training on all previously collected data while generating new interactions for future updates. This sequential off-policy learning setting is common in practice but remains largely unexplored theoretically. In this work, we present and study a simple algorithm for *sequential off-policy learning*, combining Logarithmic Smoothing (LS) estimation with online PAC-Bayesian tools. We further show that a principled adjustment to LS improves performance and accelerates convergence under mild conditions. The algorithms introduced generalize previous work: they match state-of-the-art offline approaches in the batch case and substantially outperform them when policies are updated sequentially. Empirical evaluations highlight both the benefits of the sequential framework and the strength of the proposed algorithms.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13818", "url": null, "sourceid": 786, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=93CdBSCWoR", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11511, "modified": "2026-03-29T20:43:14.295979-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=93CdBSCWoR", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "147", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13821, "uid": "51d92be1c60d1db1d2e5e7a07da55b26", "name": "Weighted quantization using MMD: From mean field to mean shift using gradient flows", "authors": [{"id": 22935, "fullname": "Ayoub Belhadji", "url": "http://virtual.aistats.org/api/miniconf/users/22935?format=json", "institution": "Massachusetts Institute of Technology"}, {"id": 22936, "fullname": "Daniel Sharp", "url": "http://virtual.aistats.org/api/miniconf/users/22936?format=json", "institution": "Massachusetts Institute of Technology"}, {"id": 12863, "fullname": "Youssef Marzouk", "url": "http://virtual.aistats.org/api/miniconf/users/12863?format=json", "institution": "Massachusetts Institute of Technology"}], "abstract": "Approximating a probability distribution using a set of particles is a fundamental problem in machine learning and statistics, with applications including clustering and quantization. Formally, we seek a weighted mixture of Dirac measures that best approximates the target distribution. While much existing work relies on the Wasserstein distance to quantify approximation errors, maximum mean discrepancy (MMD) has received comparatively less attention, especially when allowing for variable particle weights. We argue that a _Wasserstein--Fisher--Rao_ gradient flow is well-suited for designing quantizations optimal under MMD. We show that a system of interacting particles satisfying a set of ODEs discretizes this flow. We further derive a new fixed-point algorithm called _mean shift interacting particles_ (MSIP). We show that MSIP extends the classical mean shift algorithm, widely used for identifying modes in kernel density estimators. Moreover, we show that MSIP can be interpreted as preconditioned gradient descent and that it acts as a relaxation of Lloyd's algorithm for clustering. Our unification of gradient flows, mean shift, and MMD-optimal quantization yields algorithms that are more robust than state-of-the-art methods, as demonstrated via high-dimensional and multi-modal numerical experiments.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13821", "url": null, "sourceid": 462, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=8oRiidpWSK", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11514, "modified": "2026-03-29T20:43:14.448231-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=8oRiidpWSK", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "195", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13822, "uid": "1385974ed5904a438616ff7bdb3f7439", "name": "Dualformer: Time-Frequency Dual Domain Learning for Long-term Time Series Forecasting", "authors": [{"id": 22937, "fullname": "Jingjing Bai", "url": "http://virtual.aistats.org/api/miniconf/users/22937?format=json", "institution": "University of Osaka/ RIKEN AIP"}, {"id": 14504, "fullname": "Yoshinobu Kawahara", "url": "http://virtual.aistats.org/api/miniconf/users/14504?format=json", "institution": "The University of Osaka / RIKEN"}], "abstract": "Transformer-based models, despite their promise for long-term time series forecasting (LTSF), suffer from an inherent low-pass filtering effect that limits their effectiveness. This issue arises due to undifferentiated propagation of frequency components across layers, causing a progressive attenuation of high-frequency information crucial for capturing fine-grained temporal variations. To address this limitation, we propose Dualformer, a principled dual-domain framework that rethinks frequency modeling from a layer-wise perspective. Dualformer introduces three key components: (1) a dual-branch architecture that concurrently models complementary temporal patterns in both time and frequency domains; (2) a hierarchical frequency sampling module that allocates distinct frequency bands to different layers, preserving high-frequency details in lower layers while modeling low-frequency trends in deeper layers; and (3) a periodicity-aware weighting mechanism that dynamically balances contributions from the dual branches based on the harmonic energy ratio of inputs, supported theoretically by a derived lower bound. This design enables structured frequency modeling and adaptive integration of time-frequency features, effectively preserving high-frequency information and enhancing generalization. Extensive experiments conducted on eight widely used benchmarks demonstrate Dualformer\u2019s robustness and superior performance, particularly on heterogeneous or weakly periodic data.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13822", "url": null, "sourceid": 140, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=8eSCtjH5Jl", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11515, "modified": "2026-03-29T20:43:14.488836-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=8eSCtjH5Jl", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "52", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13823, "uid": "19b650660b253761af189682e03501dd", "name": "Differentially Private E-Values", "authors": [{"id": 14512, "fullname": "Daniel Csillag", "url": "http://virtual.aistats.org/api/miniconf/users/14512?format=json", "institution": "FGV EMAp"}, {"id": 10049, "fullname": "Diego Mesquita", "url": "http://virtual.aistats.org/api/miniconf/users/10049?format=json", "institution": "Getulio Vargas Foundation (FGV EMAp)"}], "abstract": "E-values have gained prominence as flexible tools for statistical inference and risk control, enabling anytime- and post-hoc-valid procedures under minimal assumptions. However, many applications fundamentally rely on sensitive data, which can be leaked through e-values. To ensure their safe release, we propose a general framework for differentially private e-values that transforms any non-private e-value into a differentially private one. Towards this end, we develop a novel biased multiplicative noise mechanism that ensures our differentially private e-values remain statistically valid. We show that our differentially private e-values attain strong statistical power, and are asymptotically as powerful as their non-private counterparts. Experiments across online risk monitoring, private healthcare, and conformal e-prediction demonstrate our approach's effectiveness and broad applicability.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13823", "url": null, "sourceid": 863, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=8Tvn60M7jV", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11516, "modified": "2026-03-29T20:43:14.537784-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=8Tvn60M7jV", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "39", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13824, "uid": "d9fc5b73a8d78fad3d6dffe419384e70", "name": "Structural Alignment Improves Graph Test-Time Adaptation", "authors": [{"id": 19806, "fullname": "Hans Hao-Hsun Hsu", "url": "http://virtual.aistats.org/api/miniconf/users/19806?format=json", "institution": "Georgia Institute of Technology"}, {"id": 22938, "fullname": "Shikun Liu", "url": "http://virtual.aistats.org/api/miniconf/users/22938?format=json", "institution": "Georgia Institute of Technology"}, {"id": 22939, "fullname": "Han Zhao", "url": "http://virtual.aistats.org/api/miniconf/users/22939?format=json", "institution": "University of Illinois, Urbana Champaign"}, {"id": 22940, "fullname": "Pan Li", "url": "http://virtual.aistats.org/api/miniconf/users/22940?format=json", "institution": "Georgia Institute of Technology"}], "abstract": "Graph-based learning excels at capturing interaction patterns in diverse domains like recommendation, fraud detection, and particle physics. However, its performance often degrades under distribution shifts, especially those altering network connectivity. Current methods to address these shifts typically require retraining with the source dataset, which is often infeasible due to computational or privacy limitations. We introduce Test-Time Structural Alignment (TSA), a novel algorithm for Graph Test-Time Adaptation (GTTA) that adapts a pretrained model to align graph structures during inference without the cost of retraining. Grounded in a theoretical understanding of graph data distribution shifts, TSA employs three synergistic strategies: uncertainty-aware neighborhood weighting to accommodate neighbor label distribution shifts, adaptive balancing of self-node and aggregated neighborhood representations based on their signal-to-noise ratio, and decision boundary refinement to correct residual label and feature shifts. Extensive experiments on synthetic and real-world datasets demonstrate TSA's consistent outperformance of both non-graph TTA methods and state-of-the-art GTTA baselines.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13824", "url": null, "sourceid": 388, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=8Q3qQxmlkJ", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11517, "modified": "2026-03-29T20:43:14.580145-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=8Q3qQxmlkJ", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "178", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13827, "uid": "217eedd1ba8c592db97d0dbe54c7adfc", "name": "Split-Flows: Measure Transport and Information Loss Across Molecular Resolutions", "authors": [{"id": 19248, "fullname": "Sander Hummerich", "url": "http://virtual.aistats.org/api/miniconf/users/19248?format=json", "institution": "Heidelberg University"}, {"id": 13241, "fullname": "Ullrich K\u00f6the", "url": "http://virtual.aistats.org/api/miniconf/users/13241?format=json", "institution": "University of Heidelberg"}, {"id": 22947, "fullname": "Tristan Bereau", "url": "http://virtual.aistats.org/api/miniconf/users/22947?format=json", "institution": "Heidelberg University"}], "abstract": "By reducing resolution, coarse-grained models greatly accelerate molecular simulations, unlocking access to long-timescale phenomena, though at the expense of microscopic information. Recovering this fine-grained detail is essential for tasks that depend on atomistic accuracy, making backmapping a central challenge in molecular modeling. We introduce split-flows, a novel flow-based approach that reinterprets backmapping as a continuous-time measure transport across resolutions. Unlike existing generative strategies, split-flows establish a direct probabilistic link between resolutions, enabling expressive conditional sampling of atomistic structures and\u2014for the first time\u2014a tractable route to computing mapping entropies, an information-theoretic measure of the irreducible detail lost in coarse-graining. We demonstrate these capabilities on diverse molecular systems, including Chignolin, a lipid bilayer, and alanine dipeptide, highlighting split-flows as a principled framework for accurate backmapping and systematic evaluation of coarse-grained models.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13827", "url": null, "sourceid": 738, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=8AK8PUSiju", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11520, "modified": "2026-03-29T20:43:14.707720-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=8AK8PUSiju", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "175", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13828, "uid": "7f975a56c761db6506eca0b37ce6ec87", "name": "Injecting Measurement Information Yields a Fast and Noise-Robust Diffusion-Based Inverse Problem Solver", "authors": [{"id": 22948, "fullname": "Jonathan Patsenker", "url": "http://virtual.aistats.org/api/miniconf/users/22948?format=json", "institution": "Yale University"}, {"id": 22949, "fullname": "Henry Li", "url": "http://virtual.aistats.org/api/miniconf/users/22949?format=json", "institution": "Google"}, {"id": 22950, "fullname": "Myeongseob Ko", "url": "http://virtual.aistats.org/api/miniconf/users/22950?format=json", "institution": "Virginia Polytechnic Institute and State University"}, {"id": 10010, "fullname": "Ruoxi Jia", "url": "http://virtual.aistats.org/api/miniconf/users/10010?format=json", "institution": "Virginia Tech"}, {"id": 4066, "fullname": "Yuval Kluger", "url": "http://virtual.aistats.org/api/miniconf/users/4066?format=json", "institution": "Yale School of Medicine"}], "abstract": "Diffusion models have been firmly established as principled zero-shot solvers for linear and nonlinear inverse problems, owing to their powerful image prior and iterative sampling algorithm. These approaches often rely on Tweedie's formula, which relates the diffusion variate $\\mathbf{x}_t$ to the posterior mean $\\mathbb{E} [\\mathbf{x}_0 | \\mathbf{x}_t]$, in order to guide the diffusion trajectory with an estimate of the final denoised sample $\\mathbf{x}_0$. However, this does not consider information from the measurement $\\mathbf{y}$, which must then be integrated downstream. In this work, we propose to estimate the conditional posterior mean $\\mathbb{E} [\\mathbf{x}_0 | \\mathbf{x}_t, \\mathbf{y}]$, which can be formulated as the solution to a lightweight, single-parameter maximum likelihood estimation problem. The resulting prediction can be integrated into any standard sampler, resulting in a fast and memory-efficient inverse solver. Our optimizer is amenable to a noise-aware likelihood-based stopping criteria that is robust to measurement noise in $\\mathbf{y}$. We demonstrate comparable or improved performance against a wide selection of contemporary inverse solvers across multiple datasets and tasks.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13828", "url": null, "sourceid": 1011, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=85ZcmV25jV", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11521, "modified": "2026-03-29T20:43:14.747068-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=85ZcmV25jV", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "83", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13832, "uid": "2647c1dba23bc0e0f9cdf75339e120d2", "name": "Visual Prompting Reimagined: The Power of Activation Prompts", "authors": [{"id": 22957, "fullname": "Yihua Zhang", "url": "http://virtual.aistats.org/api/miniconf/users/22957?format=json", "institution": "Michigan State University"}, {"id": 22958, "fullname": "Hongkang Li", "url": "http://virtual.aistats.org/api/miniconf/users/22958?format=json", "institution": "University of Pennsylvania, University of Pennsylvania"}, {"id": 22959, "fullname": "Yuguang Yao", "url": "http://virtual.aistats.org/api/miniconf/users/22959?format=json", "institution": "Intuit"}, {"id": 22960, "fullname": "Aochuan Chen", "url": "http://virtual.aistats.org/api/miniconf/users/22960?format=json", "institution": "The Hong Kong University of Science and Technology (Guangzhou)"}, {"id": 22961, "fullname": "Shuai Zhang", "url": "http://virtual.aistats.org/api/miniconf/users/22961?format=json", "institution": "New Jersey Institute of Technology"}, {"id": 22962, "fullname": "Pin-Yu Chen", "url": "http://virtual.aistats.org/api/miniconf/users/22962?format=json", "institution": "International Business Machines"}, {"id": 22963, "fullname": "Meng Wang", "url": "http://virtual.aistats.org/api/miniconf/users/22963?format=json", "institution": "Rensselaer Polytechnic Institute"}, {"id": 969, "fullname": "Sijia Liu", "url": "http://virtual.aistats.org/api/miniconf/users/969?format=json", "institution": "Michigan State University &amp; IBM Research"}], "abstract": "Visual prompting (VP) has emerged as a popular method to repurpose large pretrained models for downstream vision tasks. Unlike many parameter-efficient finetuning (PEFT) techniques that modify model parameters, VP introduces a universal perturbation directly into the input data to facilitate task-specific finetuning while keeping the pretrained model intact. However, there exists a noticeable performance gap between VP and conventional finetuning methods, highlighting an unexplored realm in theory and practice to understand and advance VP to close its performance gap. Towards this end, we introduce a novel concept, termed activation prompt (AP), which extends the scope of input-level VP by enabling universal perturbations to be applied to activation maps within the intermediate layers of the model. With the aid of AP, we show that VP, by its input perturbation design, has intrinsic limitations in both performance and efficiency. By contrast, AP shares a natural connection to normalization tuning, e.g., batch normalization for convolutional neural networks (CNNs) and layer normalization for vision transformers (ViTs). This illuminates the reason behind the observed better accuracy of normalization tuning than VP in the literature. Furthermore, we show that the choice of prompting exhibits a distinct preference for layer depth, with conclusions varying significantly between CNNs and ViTs. We theoretically elucidate the rationale behind such preference by analyzing global features across layers. By conducting extensive experiments across 29 datasets and various model architectures, we provide a thorough performance analysis of AP, comparing it with VP and PEFT baselines. Our experimental results demonstrate that AP significantly surpasses the input-level VP in terms of both accuracy and efficiency, considering factors like time, parameters, memory usage, and throughout. These results further support our new insights into the incapabilities of VP and the capabilities of AP.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13832", "url": null, "sourceid": 1824, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=7onqhBs7NV", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11525, "modified": "2026-03-29T20:43:14.911292-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=7onqhBs7NV", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "189", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13835, "uid": "82f2b308c3b01637c607ce05f52a2fed", "name": "The Majority Vote Paradigm Shift: When Popular Meets Optimal", "authors": [{"id": 19791, "fullname": "Antonio Purificato", "url": "http://virtual.aistats.org/api/miniconf/users/19791?format=json", "institution": "Sapienza University of Rome &amp; Amazon"}, {"id": 20598, "fullname": "Maria Sofia Bucarelli", "url": "http://virtual.aistats.org/api/miniconf/users/20598?format=json", "institution": "CNRS, I3S, INRIA"}, {"id": 22969, "fullname": "Anil Nelakanti", "url": "http://virtual.aistats.org/api/miniconf/users/22969?format=json", "institution": "International Institute of Information Technology Hyderabad"}, {"id": 22970, "fullname": "Andrea Bacciu", "url": "http://virtual.aistats.org/api/miniconf/users/22970?format=json", "institution": "Amazon.com"}, {"id": 18031, "fullname": "Fabrizio Silvestri", "url": "http://virtual.aistats.org/api/miniconf/users/18031?format=json", "institution": "Sapienza University of Rome"}, {"id": 22971, "fullname": "Amin Mantrach", "url": "http://virtual.aistats.org/api/miniconf/users/22971?format=json", "institution": "Amazon"}], "abstract": "Reliably labelling data typically requires annotations from multiple human workers. However, humans are far from being perfect.  Hence, it is a common practice to aggregate labels gathered from multiple annotators to make a more confident estimate of the true label. Among many aggregation methods, the simple and well-known Majority Vote (MV)  selects the class label polling the highest number of votes.  However, despite its importance, the optimality of MV\u2019s label aggregation has not been extensively studied. We address this gap in our work by characterising the conditions under which MV achieves the theoretically optimal lower bound on label estimation error.  Our results capture the tolerable limits on annotation noise under which MV can optimally recover labels for a given class distribution. This certificate of optimality provides a more principled approach to model selection for label aggregation as an alternative to otherwise inefficient practices that sometimes include higher experts, gold labels, etc.,  that are all marred by the same human uncertainty despite huge time and monetary costs. Experiments on both synthetic and real-world data corroborate our theoretical findings.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13835", "url": null, "sourceid": 725, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=7f09dQVENn", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11528, "modified": "2026-03-29T20:43:15.030117-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=7f09dQVENn", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "183", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13838, "uid": "c21002f464c5fc5bee3b98ced83963b8", "name": "Tight Lower Bounds and Optimal Algorithms for Stochastic Nonconvex Optimization with Heavy-Tailed Noise", "authors": [{"id": 22974, "fullname": "Adrien Fradin", "url": "http://virtual.aistats.org/api/miniconf/users/22974?format=json", "institution": "Institut Polytechnique de Paris"}, {"id": 22975, "fullname": "Abdurakhmon Sadiev", "url": "http://virtual.aistats.org/api/miniconf/users/22975?format=json", "institution": "King Abdullah University of Science and Technology"}, {"id": 22976, "fullname": "Laurent Condat", "url": "http://virtual.aistats.org/api/miniconf/users/22976?format=json", "institution": "KAUST"}, {"id": 722, "fullname": "Peter Richtarik", "url": "http://virtual.aistats.org/api/miniconf/users/722?format=json", "institution": "KAUST"}], "abstract": "We study stochastic nonconvex optimization under heavy-tailed noise. In this setting, the stochastic gradients only have bounded p\u2013th central moment ($p$\u2013BCM) for some $p \\in (1,2]$. Building on the foundational work of Arjevani et al. (2022) in stochastic optimization, we establish tight sample complexity lower bounds for all first-order methods under relaxed mean-squared smoothness ($q$-WAS) and $\\delta$-similarity ($(q,\\delta)$-S) assumptions, allowing any exponent $q\\in[1,2]$ instead of the standard $q= 2$. These results substantially broaden the scope of existing lower bounds. To complement them, we show that Normalized Stochastic Gradient Descent with Momentum Variance Reduction (NSGD-MVR), a known algorithm, matches these bounds in expectation. Beyond expectation guarantees, we introduce a new algorithm, Double-Clipped NSGD-MVR, which allows the derivation of high-probability convergence rates under weaker assumptions than in previous works. Finally, for second-order methods with stochastic Hessians satisfying bounded $q$-th central moment assumptions for some exponent $q \\in[1,2] $ (allowing $q\\neq p$), we establish sharper lower bounds than previous works while improving over Sadiev et al. (2025) (where only $p=q$ is considered) and yielding stronger convergence exponents. Together, these results provide a nearly complete complexity characterization of stochastic nonconvex optimization in heavy-tailed regimes.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13838", "url": null, "sourceid": 1125, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=7Oyy8esOwq", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11531, "modified": "2026-03-29T20:43:15.147345-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=7Oyy8esOwq", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "163", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13840, "uid": "b23975176653284f1f7356ba5539cfcb", "name": "Regularizing attention scores with the bootstrapping", "authors": [{"id": 19600, "fullname": "Maxim Laletin", "url": "http://virtual.aistats.org/api/miniconf/users/19600?format=json", "institution": "University of Warsaw"}, {"id": 22977, "fullname": "Neo Christopher Chung", "url": "http://virtual.aistats.org/api/miniconf/users/22977?format=json", "institution": "University of Warsaw"}], "abstract": "Vision transformers (ViT) rely on attention mechanism to weigh input features, and therefore attention scores have naturally been considered as explanations for its decision-making process. However, attention scores are almost always non-zero, resulting in noisy attention maps and limiting interpretability. Can we quantify uncertainty measures of attention scores and obtain regularized attention scores? To this end, we consider attention scores of ViT in a statistical framework where, e.g., noise would lead to insignificant yet non-zero scores. Leveraging statistical learning techniques, we introduce the bootstrapping for attention scores which generates a baseline distribution of attention scores by resampling input features. Such a bootstrap distribution is then used to estimate significances and posterior probabilities of attention scores. In natural and medical images, the proposed Attention Regularization approach demonstrates a straightforward removal of spurious attention arising from noise, drastically improving shrinkage and sparsity. Quantitative evaluations are conducted using both simulation and real-world datasets. Our study highlights bootstrapping as a practical regularization tool when using attention scores as explanations for ViT.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13840", "url": null, "sourceid": 2347, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=7H5UoEUJHU", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11533, "modified": "2026-03-29T20:43:15.255334-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=7H5UoEUJHU", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "155", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13841, "uid": "ba2fd310dcaa8781a9a652a31baf3c68", "name": "Boltzmann Exploration for Heavy-Tailed Bandits", "authors": [{"id": 22978, "fullname": "Hyeon-jun Park", "url": "http://virtual.aistats.org/api/miniconf/users/22978?format=json", "institution": "Chung-Ang University"}, {"id": 22979, "fullname": "Yoon-Sik Cho", "url": "http://virtual.aistats.org/api/miniconf/users/22979?format=json", "institution": "Chung-Ang University"}, {"id": 22980, "fullname": "Kyungjae Lee", "url": "http://virtual.aistats.org/api/miniconf/users/22980?format=json", "institution": "Korea University"}], "abstract": "We consider the stochastic multi\u2011armed bandit problem with heavy-tailed rewards, assuming only that each arm's reward distribution has a finite $p$-th moment for $p\\in(1,2]$. Although prior work has proposed algorithms robust to heavy-tailed rewards, these methods do not admit closed-form action-selection probabilities, hindering efficient offline evaluation and potentially introducing bias in inverse propensity weighting (IPW) estimators. We propose heavy Boltzmann exploration (H-BE), a Boltzmann-style randomized policy whose action-selection probabilities are available in closed form even under heavy-tailed noise. Theoretically, we establish that H-BE attains the minimax-optimal gap-independent regret bound $O(\\nu^{\\frac{1}{p}} K^{1-\\frac{1}{p}} T^{\\frac{1}{p}})$ and a gap-dependent regret bound $O(\\sum_{i:\\Delta_i>0}{\\log(T \\Delta_i^{\\frac{p}{p-1}}/K)}/{\\Delta_i^{\\frac{1}{p-1}}})$ where $\\nu$ bounds the $p$-th moment, $K$ is the number of arms, $T$ is the horizon, and $\\Delta_i$ denotes the suboptimality gap of arm $i$. Empirically, H-BE shows competitive cumulative regret relative to state-of-the-art baselines, and its explicit propensities enable more stable and efficient IPW-based offline evaluation.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13841", "url": null, "sourceid": 524, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=7FJV7uc3bB", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11534, "modified": "2026-03-29T20:43:15.307645-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=7FJV7uc3bB", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "25", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13842, "uid": "c399862d3b9d6b76c8436e924a68c45b", "name": "General Weighted Averaging in Stochastic Gradient Descent: CLT and Adaptive Optimality", "authors": [{"id": 22863, "fullname": "Ziyang Wei", "url": "http://virtual.aistats.org/api/miniconf/users/22863?format=json", "institution": "University of Chicago"}, {"id": 22864, "fullname": "Wanrong Zhu", "url": "http://virtual.aistats.org/api/miniconf/users/22864?format=json", "institution": "University of California, Irvine"}, {"id": 22866, "fullname": "Wei Biao Wu", "url": "http://virtual.aistats.org/api/miniconf/users/22866?format=json", "institution": "University of Chicago"}], "abstract": "Stochastic Gradient Descent (SGD) is a cornerstone of machine learning, prized for its efficiency in large-scale optimization. This paper revisits SGD by introducing a general weighted averaging framework that significantly enhances its applicability. We establish asymptotic normality for a wide range of weighted averaged SGD solutions under minimal assumptions, providing a groundbreaking necessary condition for the central limit theorem in certain settings. This enables asymptotically valid online inference, empowering real-time confidence interval construction. Furthermore, we propose an adaptive averaging scheme, inspired by optimal weights for linear models, which achieves optimal superior non-asymptotic bounds. Our theoretical advances and empirical validations redefine SGD\u2019s capabilities, offering transformative insights for statistical learning and optimization.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13842", "url": null, "sourceid": 534, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=72PIwEnWbZ", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11535, "modified": "2026-03-29T20:43:15.348333-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=72PIwEnWbZ", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "71", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13843, "uid": "a869ccbcbd9568808b8497e28275c7c8", "name": "Differentially Private Clipped-SGD: High-Probability Convergence with Arbitrary Clipping Level", "authors": [{"id": 22981, "fullname": "Saleh Khah", "url": "http://virtual.aistats.org/api/miniconf/users/22981?format=json", "institution": "Iran University of Science and Technology"}, {"id": 22192, "fullname": "Savelii Chezhegov", "url": "http://virtual.aistats.org/api/miniconf/users/22192?format=json", "institution": "Moscow Independent Research Institute of Artificial Intelligence"}, {"id": 22982, "fullname": "Shahrokh Farahmand", "url": "http://virtual.aistats.org/api/miniconf/users/22982?format=json", "institution": "Iran University of Science and Technology (IUST)"}, {"id": 13274, "fullname": "Samuel Horvath", "url": "http://virtual.aistats.org/api/miniconf/users/13274?format=json", "institution": "Mohamed bin Zayed University of Artificial Intelligence"}, {"id": 5637, "fullname": "Eduard Gorbunov", "url": "http://virtual.aistats.org/api/miniconf/users/5637?format=json", "institution": "Mohamed bin Zayed University of Artificial Intelligence"}], "abstract": "Gradient clipping is a fundamental tool in Deep Learning, improving the high-probability convergence of stochastic first-order methods like SGD, AdaGrad, and Adam under heavy-tailed noise, which is common in training large language models. It is also a crucial component of Differential Privacy (DP) mechanisms. However, existing high-probability convergence analyses typically require the clipping threshold to increase with the number of optimization steps, which is incompatible with standard DP mechanisms like the Gaussian mechanism. In this work, we close this gap by providing the first high-probability convergence analysis for DP-Clipped-SGD with a fixed clipping level, applicable to both convex and non-convex smooth optimization under heavy-tailed noise, characterized by a bounded central $\\alpha$-th moment assumption, $\\alpha \\in (1,2]$. Our results show that, with a fixed clipping level, the method converges to a neighborhood of the optimal solution with a \\emph{faster rate} than the existing ones. The neighborhood can be balanced against the noise introduced by DP, providing a refined trade-off between convergence speed and privacy guarantees.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13843", "url": null, "sourceid": 2131, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=6yoPpbYECH", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11536, "modified": "2026-03-29T20:43:15.389601-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=6yoPpbYECH", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "53", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13844, "uid": "9c19a2aa1d84e04b0bd4bc888792bd1e", "name": "Auditing Pay-Per-Token in Large Language Models", "authors": [{"id": 22162, "fullname": "Ander Artola Velasco", "url": "http://virtual.aistats.org/api/miniconf/users/22162?format=json", "institution": "Max Planck Institute for Software Systems"}, {"id": 22160, "fullname": "Stratis Tsirtsis", "url": "http://virtual.aistats.org/api/miniconf/users/22160?format=json", "institution": "Hasso Plattner Institute"}, {"id": 22164, "fullname": "Manuel Gomez Rodriguez", "url": "http://virtual.aistats.org/api/miniconf/users/22164?format=json", "institution": "MPI-SWS"}], "abstract": "Millions of users rely on a market of cloud-based services to obtain access to state-of-the-art large language models. However, it has been very recently shown that the de facto pay-per-token pricing mechanism used by providers creates a financial incentive for them to strategize and misreport the (number of) tokens a model used to generate an output. In this paper, we develop an auditing framework based on martingale theory that enables a trusted third-party auditor who sequentially queries a provider to detect token misreporting. Crucially, we show that our framework is guaranteed to always detect token misreporting, regardless of the provider's (mis-)reporting policy, and not falsely flag a faithful provider as unfaithful with high probability. To validate our auditing framework, we conduct experiments across a wide range of (mis-)reporting policies using several large language models from the $\\texttt{Llama}$, $\\texttt{Gemma}$ and $\\texttt{Ministral}$ families, and input prompts from a popular crowdsourced benchmarking platform. The results show that our framework detects an unfaithful provider after observing fewer than $\\sim$$70$ reported outputs, while maintaining the probability of falsely flagging a faithful provider below $\\alpha = 0.05$.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13844", "url": null, "sourceid": 1598, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=6lboj007YA", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11537, "modified": "2026-03-29T20:43:15.429533-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=6lboj007YA", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "26", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13846, "uid": "e5f6ad6ce374177eef023bf5d0c018b6", "name": "Improved Algorithms for Clustering with Noisy Distance Oracles", "authors": [{"id": 19792, "fullname": "Pinki Pradhan", "url": "http://virtual.aistats.org/api/miniconf/users/19792?format=json", "institution": "NISER, Bhubaneswar"}, {"id": 19879, "fullname": "Anup Bhattacharya", "url": "http://virtual.aistats.org/api/miniconf/users/19879?format=json", "institution": "NISER"}, {"id": 22984, "fullname": "Ragesh Jaiswal", "url": "http://virtual.aistats.org/api/miniconf/users/22984?format=json", "institution": "Indian Institute of Technology, Delhi"}], "abstract": "Bateni *et al.* has recently introduced the *weak-strong distance oracle model* to study clustering problems in settings with limited distance information. Given query access to the strong-oracle and weak-oracle in the weak-strong oracle model, the authors design approximation algorithms for $k$-means and $k$-center clustering problems. In this work, we design algorithms with improved guarantees for $k$-means and $k$-center clustering problems in the weak-strong oracle model. The $k$-means++ algorithm is routinely used to solve $k$-means in settings where complete distance information is available. One of the main contributions of this work is to show that $k$-means++ algorithm can be adapted to work in the weak-strong oracle model using only a small number of strong-oracle queries, which is the critical resource in this model. In particular, our $k$-means++ based algorithm gives a constant approximation for $k$-means and uses $O(k^2 \\log^2{n})$ strong-oracle queries. This improves on the algorithm of Bateni *et al.* that uses $O(k^2 \\log^4n \\log^2 \\log n)$ strong-oracle queries for a constant factor approximation of $k$-means. For the $k$-center problem, we give a simple  *ball-carving* based $6(1 + \\epsilon)$-approximation algorithm that uses $O(k^3 \\log^2{n} \\log{\\frac{\\log{n}}{\\epsilon}})$ strong-oracle queries. This is an improvement over the $14(1 + \\epsilon)$-approximation algorithm of Bateni *et al.* that uses $O(k^2 \\log^4{n} \\log^2{\\frac{\\log{n}}{\\epsilon}})$ strong-oracle queries.  To show the effectiveness of our algorithms, we perform empirical evaluations on real-world datasets and show that our algorithms significantly outperform the algorithms of Bateni *et al.*", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13846", "url": null, "sourceid": 573, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=6lFHvZzieB", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11539, "modified": "2026-03-29T20:43:15.502890-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=6lFHvZzieB", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "85", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13847, "uid": "40008b9a5380fcacce3976bf7c08af5b", "name": "Rate-optimal Design for Anytime Best Arm Identification", "authors": [{"id": 10054, "fullname": "Junpei Komiyama", "url": "http://virtual.aistats.org/api/miniconf/users/10054?format=json", "institution": "New York University"}, {"id": 5305, "fullname": "Kyoungseok Jang", "url": "http://virtual.aistats.org/api/miniconf/users/5305?format=json", "institution": "Chung-Ang University"}, {"id": 10246, "fullname": "Junya Honda", "url": "http://virtual.aistats.org/api/miniconf/users/10246?format=json", "institution": "Kyoto University / RIKEN"}], "abstract": "We consider the best arm identification problem, where the goal is to identify the arm with the highest mean reward from a set of $K$ arms under a limited sampling budget. This problem models many practical scenarios such as A/B testing. We consider a class of algorithms for this problem, which is provably minimax optimal up to a constant factor. This idea is a generalization of existing works in fixed-budget best arm identification, which are limited to a particular choice of risk measures. Based on the framework, we propose Almost Tracking, a closed-form algorithm that has a provable guarantee on the popular risk measure. Unlike existing algorithms, Almost Tracking does not require the total budget in advance nor does it need to discard a significant part of samples, which gives a practical advantage. Through experiments on synthetic and real-world datasets, we show that our algorithm outperforms existing anytime algorithms as well as fixed-budget algorithms. Our recommended algorithm for practitioners is found in the final section.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13847", "url": null, "sourceid": 340, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=6hmBuvqNye", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11540, "modified": "2026-03-29T20:43:15.542142-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=6hmBuvqNye", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "150", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13849, "uid": "cf1f78fe923afe05f7597da2be7a3da8", "name": "Entropic Projection Alignment: Estimating, Explaining, and Improving Model Performance Under Tabular Distribution Shift", "authors": [{"id": 22985, "fullname": "Salim I. Amoukou", "url": "http://virtual.aistats.org/api/miniconf/users/22985?format=json", "institution": "J.P. Morgan Chase"}, {"id": 22986, "fullname": "Emanuele Albini", "url": "http://virtual.aistats.org/api/miniconf/users/22986?format=json", "institution": "Imperial College London"}, {"id": 21948, "fullname": "Tom Bewley", "url": "http://virtual.aistats.org/api/miniconf/users/21948?format=json", "institution": "Mistral AI"}, {"id": 22987, "fullname": "Saumitra Mishra", "url": "http://virtual.aistats.org/api/miniconf/users/22987?format=json", "institution": "J.P. Morgan Chase"}, {"id": 21950, "fullname": "Manuela Veloso", "url": "http://virtual.aistats.org/api/miniconf/users/21950?format=json", "institution": "School of Computer Science, Carnegie Mellon University"}], "abstract": "We propose a unified framework for addressing three key challenges of distribution shift: (1) estimating a model\u2019s performance on an unlabeled target domain, (2) explaining the shift by identifying the features responsible, and (3) improving the target domain performance. Our method, Entropic Projection Alignment (EPA), aligns the source distribution to the target by matching carefully selected moments while simultaneously minimising the KL divergence from the source. This formulation yields a unique closed-form solution for importance weights, achieving robustness through implicit variance control. Drawing on domain adaptation theory, we establish that moment matching is sufficient for reliable estimation and adaptation, avoiding the need for full density ratio recovery. Extensive experiments, together with strong theoretical guarantees, demonstrate that EPA consistently outperforms state-of-the-art baselines while offering substantial computational efficiency.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13849", "url": null, "sourceid": 1365, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=6PuvcdqLzT", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11542, "modified": "2026-03-29T20:43:15.609530-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=6PuvcdqLzT", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "48", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13855, "uid": "a424ed4bd3a7d6aea720b86d4a360f75", "name": "Tight Analysis of Decentralized SGD: a Markov Chain Perspective", "authors": [{"id": 19752, "fullname": "Lucas Versini", "url": "http://virtual.aistats.org/api/miniconf/users/19752?format=json", "institution": "Ecole Polytechnique"}, {"id": 13050, "fullname": "Paul Mangold", "url": "http://virtual.aistats.org/api/miniconf/users/13050?format=json", "institution": "\u00c9cole polytechnique, France"}, {"id": 3809, "fullname": "Aymeric Dieuleveut", "url": "http://virtual.aistats.org/api/miniconf/users/3809?format=json", "institution": "\u00c9cole polytechnique"}], "abstract": "We propose a novel analysis of the Decentralized Stochastic Gradient Descent (DSGD) algorithm with constant step size, interpreting the iterates of the algorithm as a Markov chain. We show that DSGD converges to a stationary distribution, with its bias, to first order, decomposable into two components: one due to decentralization (growing with the graph's spectral gap and heterogeneity) and one due to stochasticity. Remarkably, the variance of local parameters is, at the first-order, inversely proportional to the number of agents, regardless of the network topology and even when clients' iterates are not averaged at the end. As a consequence of our analysis, we obtain non-asymptotic convergence bounds for clients' local iterates, confirming that DSGD has linear speed-up in the number of clients, and that the network topology only impacts higher-order terms.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13855", "url": null, "sourceid": 1425, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=5ob5u8lZeL", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11548, "modified": "2026-03-29T20:43:16.041030-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=5ob5u8lZeL", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "162", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13856, "uid": "4921f95baf824205e1b13f22d60357a1", "name": "Parameter-Efficient Multi-Task Learning via Progressive Task-Specific Adaptation", "authors": [{"id": 19756, "fullname": "Neeraj Gangwar", "url": "http://virtual.aistats.org/api/miniconf/users/19756?format=json", "institution": "University of Illinois Urbana-Champaign"}, {"id": 12886, "fullname": "Anshuka Rangi", "url": "http://virtual.aistats.org/api/miniconf/users/12886?format=json", "institution": "Amazon"}, {"id": 19774, "fullname": "Rishabh Deshmukh", "url": "http://virtual.aistats.org/api/miniconf/users/19774?format=json", "institution": "Amazon.com LLC"}, {"id": 13586, "fullname": "Holakou Rahmanian", "url": "http://virtual.aistats.org/api/miniconf/users/13586?format=json", "institution": "Amazon"}, {"id": 22994, "fullname": "Yesh Dattatreya", "url": "http://virtual.aistats.org/api/miniconf/users/22994?format=json", "institution": "Amazon"}, {"id": 22954, "fullname": "Nickvash Kani", "url": "http://virtual.aistats.org/api/miniconf/users/22954?format=json", "institution": "University of Illinois at Urbana-Champaign"}], "abstract": "Parameter-efficient fine-tuning methods have emerged as a promising solution for adapting pre-trained models to various downstream tasks. While these methods perform well in single-task learning, extending them to multi-task learning exacerbates common issues, such as task interference and negative transfer, due to the limited number of trainable parameters. To address these challenges, we introduce progressive task-specific multi-task adaptation, a novel parameter-efficient approach for multi-task learning. Our approach introduces adapter modules that are shared in early layers and become increasingly task-specific in later layers. Additionally, we propose a gradient-based approach for computing task similarity and use this measure to allocate similar tasks to the shared adapter modules. To evaluate our approach, we adapt Swin and Pyramid Vision Transformers on PASCAL and NYUD-v2. On both datasets, our approach outperforms prior parameter-efficient multi-task methods while using fewer trainable parameters.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13856", "url": null, "sourceid": 1692, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=5mfoGYUUTf", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11549, "modified": "2026-03-29T20:43:16.078189-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=5mfoGYUUTf", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "135", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13860, "uid": "d490d7b4576290fa60eb31b5fc917ad1", "name": "Revisiting Social Welfare in Bandits: UCB is (Nearly) All You Need", "authors": [{"id": 22187, "fullname": "Dhruv Sarkar", "url": "http://virtual.aistats.org/api/miniconf/users/22187?format=json", "institution": "Indian Institute of Technology Kharagpur"}, {"id": 22998, "fullname": "Nishant Pandey", "url": "http://virtual.aistats.org/api/miniconf/users/22998?format=json", "institution": "Indian Institute of Technology Kanpur"}, {"id": 19810, "fullname": "Sayak Ray Chowdhury", "url": "http://virtual.aistats.org/api/miniconf/users/19810?format=json", "institution": "Indian Institute of Technology Kanpur"}], "abstract": "Regret in stochastic multi-armed bandits traditionally measures the difference between the highest reward and either the arithmetic mean of accumulated rewards or the final reward. These conventional metrics often fail to address fairness among agents receiving rewards, particularly in settings where rewards are distributed across a population, such as patients in clinical trials. To address this, a recent body of work has introduced Nash regret, which evaluates performance via the geometric mean of accumulated rewards, aligning with the Nash social welfare function known for satisfying fairness axioms.   To minimize Nash regret, existing approaches require specialized algorithm designs and strong assumptions, such as multiplicative concentration inequalities and bounded, non-negative rewards, making them unsuitable for even Gaussian reward distributions. We demonstrate that an initial uniform exploration phase followed by a standard Upper Confidence Bound (UCB) algorithm achieves near-optimal Nash regret, while relying only on additive Hoeffding bounds, and naturally extending to sub-Gaussian rewards. Furthermore, we generalize the algorithm to a broad class of fairness metrics called the $p$-mean regret, proving (nearly) optimal regret bounds uniformly across all $p$ values. This is in contrast to prior work, which made extremely restrictive assumptions on the bandit instances and even then achieved suboptimal regret bounds. Numerical simulations validate our method\u2019s practical efficacy, broadening the accessibility of fairness in bandit algorithms.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13860", "url": null, "sourceid": 600, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=5ZnD2yY7EE", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11553, "modified": "2026-03-29T20:43:16.253794-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=5ZnD2yY7EE", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "157", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13864, "uid": "632cee946db83e7a52ce5e8d6f0fed35", "name": "Reinforcement Learning Using Known Invariances", "authors": [{"id": 23003, "fullname": "Alexandru Cioba", "url": "http://virtual.aistats.org/api/miniconf/users/23003?format=json", "institution": "Mediatek Research"}, {"id": 14549, "fullname": "Aya Kayal", "url": "http://virtual.aistats.org/api/miniconf/users/14549?format=json", "institution": "American University of Beirut"}, {"id": 4284, "fullname": "Laura Toni", "url": "http://virtual.aistats.org/api/miniconf/users/4284?format=json", "institution": "UCL"}, {"id": 308, "fullname": "Sattar Vakili", "url": "http://virtual.aistats.org/api/miniconf/users/308?format=json", "institution": "MediaTek Research"}, {"id": 10177, "fullname": "Alberto Bernacchia", "url": "http://virtual.aistats.org/api/miniconf/users/10177?format=json", "institution": "MediaTek Research"}], "abstract": "In many real-world reinforcement learning (RL) problems, the environment exhibits inherent symmetries that can be exploited to improve learning efficiency. This paper develops a theoretical and algorithmic framework for incorporating known group symmetries into kernel-based RL. We propose a symmetry-aware variant of optimistic least-squares value iteration (LSVI), which leverages invariant kernels to encode invariance in both rewards and transition dynamics. Our analysis establishes new bounds on the maximum information gain and covering numbers for invariant RKHSs, explicitly quantifying the sample efficiency gains from symmetry. Empirical results on a customized Frozen Lake environment and a 2D placement design problem confirm the theoretical improvements, demonstrating that symmetry-aware RL achieves significantly better performance than their standard kernel counterparts. These findings highlight the value of structural priors in designing more sample-efficient reinforcement learning algorithms.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13864", "url": null, "sourceid": 823, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=5InIWpmXwm", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11557, "modified": "2026-03-29T20:43:16.425286-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=5InIWpmXwm", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "140", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13866, "uid": "8ce6790cc6a94e65f17f908f462fae85", "name": "Towards Sensitivity-Aware Language Models", "authors": [{"id": 19750, "fullname": "Dren Fazlija", "url": "http://virtual.aistats.org/api/miniconf/users/19750?format=json", "institution": "L3S Research Center Hannover"}, {"id": 23005, "fullname": "Iyiola E. Olatunji", "url": "http://virtual.aistats.org/api/miniconf/users/23005?format=json", "institution": "University of Luxembourg"}, {"id": 23006, "fullname": "Daniel Kudenko", "url": "http://virtual.aistats.org/api/miniconf/users/23006?format=json", "institution": "L3S Research Center"}, {"id": 23007, "fullname": "Sandipan Sikdar", "url": "http://virtual.aistats.org/api/miniconf/users/23007?format=json", "institution": "Universit\u00e4t Hannover"}], "abstract": "With LLMs increasingly deployed in corporate data management, it is crucial to ensure that these models do not leak sensitive information. In the context of corporate data management, the concept of sensitivity awareness has been introduced, enabling LLMs to adhere to predefined access rights rules. However, it remains unclear how sensitivity awareness relates to established notions of privacy, such as differential privacy (DP), thereby making it difficult to deploy meaningfully in real-world applications. In this work, we formalize the notion of sensitivity awareness and theoretically establish its connection to DP. Additionally, we develop a supervised fine-tuning recipe to make existing, four-bit quantized LLMs more sensitivity-aware. With a performance boost of up to 21.7%, the finetuned LLMs not only substantially improve over their baseline but also outperform other full-precision open-source and commercial models of similar size in achieving sensitivity awareness, demonstrating the effectiveness of our proposed approach. At the same time, our method also largely preserves the models' performance on other tasks, such as general instruction-following, mathematical, and common-sense reasoning.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13866", "url": null, "sourceid": 1142, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=5EK40FvtbQ", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11559, "modified": "2026-03-29T20:43:16.504743-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=5EK40FvtbQ", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "169", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13867, "uid": "4d2e7bd33c475784381a64e43e50922f", "name": "Statistical Inference for Explainable Boosting Machines", "authors": [{"id": 23008, "fullname": "Haimo Fang", "url": "http://virtual.aistats.org/api/miniconf/users/23008?format=json", "institution": "Fudan University"}, {"id": 22065, "fullname": "Kevin Tan", "url": "http://virtual.aistats.org/api/miniconf/users/22065?format=json", "institution": "Wharton Statistics Department, The Wharton School"}, {"id": 23009, "fullname": "Jonathan Pipping", "url": "http://virtual.aistats.org/api/miniconf/users/23009?format=json", "institution": "The Wharton School, University of Pennsylvania"}, {"id": 23010, "fullname": "Giles Hooker", "url": "http://virtual.aistats.org/api/miniconf/users/23010?format=json", "institution": "University of Pennsylvania"}], "abstract": "Explainable boosting machines (EBMs) are popular ``glass-box'' models that learn a set of univariate functions using boosting trees. These achieve explainability through visualizations of each feature\u2019s effect. However, unlike linear model coefficients, uncertainty quantification for the learned univariate functions requires computationally intensive bootstrapping, making it hard to know which features truly matter. We provide an alternative using recent advances in statistical inference for gradient boosting, deriving methods for statistical inference as well as end-to-end theoretical guarantees. Using a moving average instead of a sum of trees (Boulevard regularization) allows the boosting process to converge to a feature-wise kernel ridge regression. This produces asymptotically normal predictions that achieve the minimax-optimal MSE for fitting Lipschitz GAMs with $p$ features of $O(p n^{-2/3})$, successfully avoiding the curse of dimensionality. We then construct prediction intervals for the response and confidence intervals for each learned univariate function with a runtime independent of the number of datapoints, enabling further explainability within EBMs.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13867", "url": null, "sourceid": 1199, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=5C4jVuFkA8", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11560, "modified": "2026-03-29T20:43:16.550020-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=5C4jVuFkA8", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "152", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13868, "uid": "beb22fb694d513edcf5533cf006dfeae", "name": "Corruption-robust Offline Multi-agent Reinforcement Learning from Human Feedback", "authors": [{"id": 5595, "fullname": "Andi Nika", "url": "http://virtual.aistats.org/api/miniconf/users/5595?format=json", "institution": "MPI-SWS"}, {"id": 10953, "fullname": "Debmalya Mandal", "url": "http://virtual.aistats.org/api/miniconf/users/10953?format=json", "institution": "University of Warwick"}, {"id": 18094, "fullname": "Parameswaran Kamalaruban", "url": "http://virtual.aistats.org/api/miniconf/users/18094?format=json", "institution": "Featurespace"}, {"id": 942, "fullname": "Adish Singla", "url": "http://virtual.aistats.org/api/miniconf/users/942?format=json", "institution": "MPI-SWS"}, {"id": 9167, "fullname": "Goran Radanovic", "url": "http://virtual.aistats.org/api/miniconf/users/9167?format=json", "institution": "Max Planck Institute for Software Systems"}], "abstract": "We consider robustness against data corruption in offline multi-agent reinforcement learning from human feedback (MARLHF) under a strong\u2010contamination model: given a dataset $D$ of trajectory\u2013preference tuples (each preference being an $n$-dimensional binary label vector representing each of the $n$ agents\u2019 preferences), an $\\epsilon$-fraction of the samples may be arbitrarily corrupted. We model the problem using the framework of linear Markov games. First, under a \\emph{uniform coverage} assumption\u2014where every policy of interest is sufficiently represented in the clean (prior to corruption) data\u2014we introduce a robust estimator that guarantees an $O(\\epsilon^{1-o(1)})$ bound on the Nash\u2010equilibrium gap. Next, we move to the more challenging unilateral coverage setting, in which only a Nash equilibrium and its single\u2010player deviations are covered: here our proposed algorithm achieves an $O(\\sqrt{\\epsilon})$ Nash\u2010gap bound. Both of these procedures, however, suffer from intractable computation. To address this, we relax our solution concept to coarse correlated equilibria (CCE). Under the same unilateral\u2010coverage regime, we then derive a quasi-polynomial\u2010time algorithm whose CCE gap scales as $O(\\sqrt{\\epsilon})$.  To the best of our knowledge, this is the first systematic treatment of adversarial data corruption in offline MARLHF.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13868", "url": null, "sourceid": 797, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=58cTQWAnNW", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11561, "modified": "2026-03-29T20:43:16.592970-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=58cTQWAnNW", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "32", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13739, "uid": "6eb6e75fddec0218351dc5c0c8464104", "name": "We Still Don\u2019t Understand High-Dimensional Bayesian Optimization", "authors": [{"id": 22744, "fullname": "Colin Doumont", "url": "http://virtual.aistats.org/api/miniconf/users/22744?format=json", "institution": "T\u00fcbingen AI Center"}, {"id": 21912, "fullname": "Donney Fan", "url": "http://virtual.aistats.org/api/miniconf/users/21912?format=json", "institution": "University of British Columbia"}, {"id": 22745, "fullname": "Natalie Maus", "url": "http://virtual.aistats.org/api/miniconf/users/22745?format=json", "institution": "Massachusetts Institute of Technology"}, {"id": 22746, "fullname": "Jacob Gardner", "url": "http://virtual.aistats.org/api/miniconf/users/22746?format=json", "institution": "University of Pennsylvania"}, {"id": 23285, "fullname": "Henry Moss", "url": "http://virtual.aistats.org/api/miniconf/users/23285?format=json", "institution": "Lancaster University"}, {"id": 13157, "fullname": "Geoff Pleiss", "url": "http://virtual.aistats.org/api/miniconf/users/13157?format=json", "institution": "University of British Columbia, Vector Institute"}], "abstract": "High-dimensional spaces have historically challenged Bayesian optimization (BO). Existing methods aim to overcome this curse of dimensionality by carefully encoding structural assumptions, from locality to sparsity to smoothness, into the optimization procedure. Surprisingly, we demonstrate that these approaches are outperformed by arguably the simplest method imaginable: Bayesian linear regression. After applying a geometric transformation to avoid boundary-seeking behaviour, Gaussian processes with linear kernels yield state-of-the-art performance on tasks with 60- to 6,000-dimensional search spaces. Linear models offer numerous advantages over their non-parametric counterparts: they afford closed-form acquisition function optimization, they yield asymptotically lower regret, and their computation scales linearly with data, a fact we exploit on molecular optimization tasks with >20,000 observations. Coupled with empirical and theoretical analyses, our results suggest the need to depart from past intuitions about BO methods in high-dimensional spaces.", "topic": null, "keywords": [], "decision": "Accept (Oral)", "session": "", "eventtype": "Oral", "event_type": "Oral", "room_name": null, "virtualsite_url": "/virtual/2026/oral/13739", "url": null, "sourceid": -1180, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=2026-Oral--1180-f2f65888", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": null, "parent_id": null, "eventmedia": [{"id": 11432, "modified": "2026-03-29T20:43:11.038315-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=2026-Oral--1180-f2f65888", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": null, "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13445, "uid": "db576a7d2453575f29eab4bac787b919", "name": "Creator Incentives in Recommender Systems: A Cooperative Game-Theoretic Approach for Stable and Fair Collaboration in Multi-Agent Bandits", "authors": [{"id": 19800, "fullname": "Ramakrishnan Krishnamurthy", "url": "http://virtual.aistats.org/api/miniconf/users/19800?format=json", "institution": "New York University, Courant Institute."}, {"id": 22121, "fullname": "Arpit Agarwal", "url": "http://virtual.aistats.org/api/miniconf/users/22121?format=json", "institution": "IIT Bombay"}, {"id": 22122, "fullname": "Lakshmi Subramanian", "url": "http://virtual.aistats.org/api/miniconf/users/22122?format=json", "institution": "New York University"}, {"id": 22123, "fullname": "Maximilian Nickel", "url": "http://virtual.aistats.org/api/miniconf/users/22123?format=json", "institution": "Facebook"}], "abstract": "We study a collaborative variant of the stochastic linear bandit problem in a multi-agent setting, motivated by real-world recommender systems where multiple content creators indirectly influence each other\u2019s outcomes. In our formulation, agents interact with a shared environment but may choose to form coalitions, sharing observations to improve collective learning efficiency. We formalize this setup as a transferable utility (TU) cooperative game, where the value of a coalition is defined as the negative sum of cumulative regrets incurred by its members. This framework allows us to examine how algorithmic design and structural assumptions about agents---such as identical vs. heterogeneous action sets---affect collaboration incentives. We show that under some algorithmic conditions, the induced TU game exhibits desirable properties: for identical agents, the game is convex and admits a non-empty core containing the Shapley value, ensuring stable and equitable collaboration. For heterogeneous agents, we demonstrate core non-emptiness and propose a simple, implementable payoff mechanism that satisfies all but one Shapley value axioms. Experimental results on problem instances derived from MovieLens-100k dataset further illustrate how the empirical payout aligns and diverges from ideal cooperative outcome for different settings. Our results offer a principled lens for designing collaborative learning systems that are both effective and incentive-aligned.", "topic": null, "keywords": [], "decision": "Accept (Oral)", "session": "Oral Session 4: RL, Bandits & Online Decision-Making", "eventtype": "Oral", "event_type": "Oral", "room_name": null, "virtualsite_url": "/virtual/2026/oral/13445", "url": null, "sourceid": -1054, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=2026-Oral--1054-eaa208ed", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Oral%20Session%204:%20RL,%20Bandits%20&%20Online%20Decision-Making?format=json", "parent_id": 11475, "eventmedia": [{"id": 11138, "modified": "2026-03-29T20:42:59.198141-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=2026-Oral--1054-eaa208ed", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": null, "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13627, "uid": "7fa732b517cbed14a48843d74526c11a", "name": "Complexity-Aware Deep Symbolic Regression with Robust Risk-Seeking Policy Gradients", "authors": [{"id": 22517, "fullname": "Zachary Bastiani", "url": "http://virtual.aistats.org/api/miniconf/users/22517?format=json", "institution": "University of Utah"}, {"id": 22518, "fullname": "Mike Kirby", "url": "http://virtual.aistats.org/api/miniconf/users/22518?format=json", "institution": ", University of Utah"}, {"id": 22519, "fullname": "Jacob Hochhalter", "url": "http://virtual.aistats.org/api/miniconf/users/22519?format=json", "institution": "University of Utah"}, {"id": 552, "fullname": "Shandian Zhe", "url": "http://virtual.aistats.org/api/miniconf/users/552?format=json", "institution": "University of Utah"}], "abstract": "We propose a novel deep symbolic regression (DSR) approach to enhance the robustness and interpretability of data-driven mathematical expression discovery. Existing DSR methods are built on recurrent neural networks, solely guided by data fitness, and potentially meet tail barriers that can zero out the policy gradient, causing inefficient model updates. To address these issues, we design a decoder-only architecture that performs attention in the frequency domain and introduce a dual-indexed position encoding to conduct layer-wise generation. Second, we propose a Bayesian information criterion (BIC)-based reward function that can automatically adjust the trade-off between expression complexity and data fitness, without the need for explicit manual tuning. Third, we develop a  ranking-based weighted policy update method that eliminates the tail barriers and enhances training effectiveness. Extensive benchmarks and systematic experiments demonstrate the  advantages of our approach. We have released our implementation at https://github.com/ZakBastiani/CADSR.", "topic": null, "keywords": [], "decision": "Accept (Oral)", "session": "Oral Session 5: Deep Architectures, Transformers & Representation Learning", "eventtype": "Oral", "event_type": "Oral", "room_name": null, "virtualsite_url": "/virtual/2026/oral/13627", "url": null, "sourceid": -925, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=2026-Oral--925-36cc4527", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Oral%20Session%205:%20Deep%20Architectures,%20Transformers%20&%20Representation%20Learning?format=json", "parent_id": 11482, "eventmedia": [{"id": 11320, "modified": "2026-03-29T20:43:06.270624-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=2026-Oral--925-36cc4527", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": null, "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13862, "uid": "f976b57bb9dd27aa2e7e7df2825893a6", "name": "Orthogonal Representation Learning for Estimating Causal Quantities", "authors": [{"id": 19881, "fullname": "Valentyn Melnychuk", "url": "http://virtual.aistats.org/api/miniconf/users/19881?format=json", "institution": "LMU Munich"}, {"id": 23000, "fullname": "Dennis Frauen", "url": "http://virtual.aistats.org/api/miniconf/users/23000?format=json", "institution": "Ludwig-Maximilians-Universit\u00e4t M\u00fcnchen"}, {"id": 23001, "fullname": "Jonas Schweisthal", "url": "http://virtual.aistats.org/api/miniconf/users/23001?format=json", "institution": "Ludwig-Maximilians-Universit\u00e4t M\u00fcnchen"}, {"id": 9166, "fullname": "Stefan Feuerriegel", "url": "http://virtual.aistats.org/api/miniconf/users/9166?format=json", "institution": "LMU Munich"}], "abstract": "End-to-end representation learning has become a powerful tool for estimating causal quantities from high-dimensional observational data, but its efficiency remained unclear. Here, we face a central tension: End-to-end representation learning methods often work well in practice but lack asymptotic optimality in the form of the quasi-oracle efficiency. In contrast, two-stage Neyman-orthogonal learners provide such a theoretical optimality property but do not explicitly benefit from the strengths of representation learning. In this work, we step back and ask two research questions: (1) When do representations strengthen existing Neyman-orthogonal learners? and (2) Can a balancing constraint \u2013 commonly proposed technique in the representation learning literature \u2013 provide improvements to Neyman-orthogonality? We address these two questions through our theoretical and empirical analysis, where we introduce a unifying framework that connects representation learning with Neyman-orthogonal learners (namely, OR-learners). In particular, we show that, under the low-dimensional manifold hypothesis, the OR-learners can strictly improve the estimation error of the standard Neyman-orthogonal learners. At the same time, we find that the balancing constraint requires an additional inductive bias and cannot generally compensate for the lack of Neyman-orthogonality of the end-to-end approaches. Building on these insights, we offer guidelines for how users can effectively combine representation learning with the classical Neyman-orthogonal learners to achieve both practical performance and theoretical guarantees.", "topic": null, "keywords": [], "decision": "Accept (Oral)", "session": "Oral Session 7: Causality, Kernels & Statistical Testing", "eventtype": "Oral", "event_type": "Oral", "room_name": null, "virtualsite_url": "/virtual/2026/oral/13862", "url": null, "sourceid": -2306, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=2026-Oral--2306-576c4939", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Oral%20Session%207:%20Causality,%20Kernels%20&%20Statistical%20Testing?format=json", "parent_id": 11474, "eventmedia": [{"id": 11555, "modified": "2026-03-29T20:43:16.337478-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=2026-Oral--2306-576c4939", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": null, "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13416, "uid": "2d1b2a5ff364606ff041650887723470", "name": "Panprediction: Optimal Predictions for Any Downstream Task and Loss", "authors": [{"id": 3646, "fullname": "Sivaraman Balakrishnan", "url": "http://virtual.aistats.org/api/miniconf/users/3646?format=json", "institution": "Carnegie Mellon University"}, {"id": 12830, "fullname": "Nika Haghtalab", "url": "http://virtual.aistats.org/api/miniconf/users/12830?format=json", "institution": "University of California, Berkeley"}, {"id": 17670, "fullname": "Daniel Hsu", "url": "http://virtual.aistats.org/api/miniconf/users/17670?format=json", "institution": "Columbia University"}, {"id": 22040, "fullname": "Brian Lee", "url": "http://virtual.aistats.org/api/miniconf/users/22040?format=json", "institution": "University of California, Berkeley"}, {"id": 1571, "fullname": "Eric Zhao", "url": "http://virtual.aistats.org/api/miniconf/users/1571?format=json", "institution": "Nvidia Research"}], "abstract": "Machine learning is classically formulated as optimizing a model for a known task and loss function. However, an emerging paradigm instead views model training as extracting enough information from a dataset so that the model\u2014after light post-processing\u2014can minimize any loss on any downstream task. This paper formalizes a mathematical framework for this paradigm, which we call panprediction. Panprediction generalizes the problem of omniprediction (Gopalan et al., 2021) and sits upstream from the problem of multi-group learning (Rothblum and Yona, 2021), which respectively focus on predictions that generalize to many downstream losses or many downstream tasks, but not both. We show that panprediction admits a nearly lossless reduction to a statistically efficient notion of calibration, called step calibration. Using this reduction, we design deterministic and randomized panpredictors that can be learned with $\\tilde{O}(1/\\varepsilon^3)$ and $\\tilde{O}(1/\\varepsilon^2)$ samples, respectively. This improves the best known sample complexity guarantee of deterministic omniprediction, and matches the best known sample complexity guarantees of randomized omniprediction and both deterministic and randomized multi-group learning. Our results demonstrate that simultaneously minimizing infinitely many losses on infinitely many tasks can be as statistically easy as minimizing one loss on one task.", "topic": null, "keywords": [], "decision": "Accept (Oral)", "session": "Oral Session 8: Robustness, Calibration, Privacy & Evaluation (Applied)", "eventtype": "Oral", "event_type": "Oral", "room_name": null, "virtualsite_url": "/virtual/2026/oral/13416", "url": null, "sourceid": -1888, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=2026-Oral--1888-0499cd83", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Oral%20Session%208:%20Robustness,%20Calibration,%20Privacy%20&%20Evaluation%20(Applied)?format=json", "parent_id": 11492, "eventmedia": [{"id": 11109, "modified": "2026-03-29T20:42:58.042987-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=2026-Oral--1888-0499cd83", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": null, "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13877, "uid": "d072677d210ac4c03ba046120f0802ec", "name": "Conformal Margin Risk Minimization: An Envelope Framework for Robust Learning under Label Noise", "authors": [{"id": 23034, "fullname": "Yuanjie Shi", "url": "http://virtual.aistats.org/api/miniconf/users/23034?format=json", "institution": "Washington State University"}, {"id": 23035, "fullname": "Peihong Li", "url": "http://virtual.aistats.org/api/miniconf/users/23035?format=json", "institution": "Washington State University"}, {"id": 23036, "fullname": "Zijian Zhang", "url": "http://virtual.aistats.org/api/miniconf/users/23036?format=json", "institution": "Washington State University"}, {"id": 22453, "fullname": "Jana Doppa", "url": "http://virtual.aistats.org/api/miniconf/users/22453?format=json", "institution": "Washington State University, Pullman"}, {"id": 23037, "fullname": "Yan Yan", "url": "http://virtual.aistats.org/api/miniconf/users/23037?format=json", "institution": "Washington State University, Pullman"}], "abstract": "Training reliable classifiers under label noise is a challenging task. Existing methods often rely on restrictive assumptions about the noise distribution, model design, or access to clean data. Such assumptions rarely hold in practice, especially under severe or heterogeneous noise. We propose Conformal Margin Risk Minimization (CMRM), an uncertainty-aware envelope framework to improve the robustness of prior methods with noisy labeled data. Specifically, CMRM computes the confidence margin as the gap between confidence scores of observed label and other labels, and then a conformal quantile estimated over a batch of examples provides a statistically valid proxy for the set-level quantile. Minimizing the conformal margin risk allows the training to focus on low uncertainty (high margin) samples while filtering out high uncertainty (low margin) samples below the quantile, as mislabeled samples.  We derive a learning bound for CMRM under arbitrary label noise with weaker assumptions than prior work. Experiments show that CMRM consistently improves accuracy and robustness of prior methods across different classification benchmarks without prior knowledge of noise.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13877", "url": null, "sourceid": 1864, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=3wGVIAhxty", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11570, "modified": "2026-03-29T20:43:16.957631-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=3wGVIAhxty", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "43", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13857, "uid": "a2557a7b2e94197ff767970b67041697", "name": "DeepRV: Accelerating spatiotemporal inference with pre-trained neural priors", "authors": [{"id": 21992, "fullname": "Jhonathan Navott", "url": "http://virtual.aistats.org/api/miniconf/users/21992?format=json", "institution": "University of Oxford  / Imperial College London"}, {"id": 21991, "fullname": "Daniel Jenson", "url": "http://virtual.aistats.org/api/miniconf/users/21991?format=json", "institution": "Oxford University"}, {"id": 21996, "fullname": "Seth Flaxman", "url": "http://virtual.aistats.org/api/miniconf/users/21996?format=json", "institution": "University of Oxford"}, {"id": 5135, "fullname": "Elizaveta Semenova", "url": "http://virtual.aistats.org/api/miniconf/users/5135?format=json", "institution": "Oxford University"}], "abstract": "Gaussian Processes (GPs) provide a flexible and statistically principled foundation for modelling spatiotemporal phenomena, but their $\\mathcal{O}(N^3)$ scaling makes them intractable for large datasets.  Approximate methods such as variational inference (VI), inducing points (sparse GPs), low-rank factorizations (RFFs), local factorizations and approximations (INLA), improve scalability but trade off accuracy or flexibility. We introduce DeepRV, a neural-network surrogate that closely matches full GP accuracy including hyperparameter estimates, while reducing computational complexity to $\\mathcal{O}(N^2)$, increasing scalability and inference speed. DeepRV serves as a drop-in replacement for GP prior realisations in e.g.~MCMC-based probabilistic programming pipelines, preserving full model flexibility. Across simulated benchmarks, non-separable spatiotemporal GPs, and a real-world application to education deprivation in London (n = 4,994 locations), DeepRV achieves the highest fidelity to exact GPs while substantially accelerating inference.  Code is provided in the [anonymized] Python package, with all experiments run on single consumer-grade GPU to ensure accessibility for practitioners.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13857", "url": null, "sourceid": 189, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=5lM1So6mQN", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11550, "modified": "2026-03-29T20:43:16.112873-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=5lM1So6mQN", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "43", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13788, "uid": "4e62e752ae53fb6a6eebd0f6146aa702", "name": "Convex Markov Games and Beyond: New Proof of Existence, Characterization and Learning Algorithms for Nash Equilibria", "authors": [{"id": 2253, "fullname": "Anas Barakat", "url": "http://virtual.aistats.org/api/miniconf/users/2253?format=json", "institution": "ETH Zurich"}, {"id": 3969, "fullname": "Ioannis Panageas", "url": "http://virtual.aistats.org/api/miniconf/users/3969?format=json", "institution": "UC Irvine"}, {"id": 22874, "fullname": "Antonios Varvitsiotis", "url": "http://virtual.aistats.org/api/miniconf/users/22874?format=json", "institution": "Singapore University of Technology and Design"}], "abstract": "Convex Markov Games (cMGs) were recently introduced as a broad class of multi-agent learning problems that generalize Markov games to settings where strategic agents optimize general utilities beyond additive rewards. While cMGs expand the modeling frontier, their theoretical foundations, particularly the structure of Nash equilibria (NE) and guarantees for learning algorithms, are not yet well understood. In this work, we address these gaps for an extension of cMGs, which we term General Utility Markov Games (GUMGs), capturing new applications requiring coupling between agents' occupancy measures. We prove that in GUMGs, Nash equilibria coincide with the fixed points of projected pseudo-gradient dynamics (i.e. first-order stationary points), enabled by a novel agent-wise gradient domination property. This insight also yields a simple proof of NE existence using Brouwer\u2019s fixed-point theorem. We further show the existence of Markov perfect equilibria. Building on this characterization, we establish a policy gradient theorem for GUMGs and design a model-free policy gradient algorithm. For potential GUMGs, we establish iteration complexity guarantees for computing approximate-NE under exact gradients and provide sample complexity bounds in both the generative model and on-policy settings. Our results extend beyond prior work restricted to zero-sum cMGs, providing the first theoretical analysis of common-interest cMGs.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13788", "url": null, "sourceid": 2136, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=BYqJneaXKQ", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11481, "modified": "2026-03-29T20:43:13.053776-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=BYqJneaXKQ", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "45", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13677, "uid": "67c6a1e7ce56d3d6fa748ab6d9af3fd7", "name": "Design-Based Finite-Sample Analysis for Regression Adjustment", "authors": [{"id": 22611, "fullname": "Dogyoon Song", "url": "http://virtual.aistats.org/api/miniconf/users/22611?format=json", "institution": "University of California, Davis"}], "abstract": "In randomized experiments, regression adjustment can improve the precision of average treatment effect (ATE) estimation using covariates without requiring a correctly specified outcome model.  Although well studied in low-dimensional settings, its behavior in high-dimensional regimes, where the number of covariates $p$ may exceed the number of observations $n$, remains underexplored.  Moreover, existing analyses are largely asymptotic, providing limited guidance for finite-sample inference.  We develop a design-based, non-asymptotic framework for analyzing the regression-adjusted ATE estimator under complete randomization.  This yields finite-sample-valid confidence intervals with explicit, instance-adaptive widths, even when $p > n$.  While these intervals rely on oracle (population-level) quantities, we also outline data-driven envelopes computable from observed data.  Our approach hinges on a refined swap sensitivity analysis of an estimator: stochastic fluctuation is controlled via a variance-adaptive Doob martingale and Freedman's inequality, and design bias is bounded by Stein's method of exchangeable pairs.  The analysis elucidates how covariate geometry governs concentration and bias of the adjusted estimator, suggesting when and how regression adjustment can be effective.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13677", "url": null, "sourceid": 47, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=KaE2RO5PL8", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11370, "modified": "2026-03-29T20:43:08.337166-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=KaE2RO5PL8", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "45", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13913, "uid": "53adaf494dc89ef7196d73636eb2451b", "name": "Efficient Uncoupled Learning Dynamics with $\\tilde{O}\\left(T^{-1/4}\\right)$ Last-Iterate Convergence in Bilinear Saddle-Point Problems over Convex Sets under Bandit Feedback", "authors": [{"id": 5518, "fullname": "ARNAB MAITI", "url": "http://virtual.aistats.org/api/miniconf/users/5518?format=json", "institution": "University of Washington"}, {"id": 23079, "fullname": "Claire Jie Zhang", "url": "http://virtual.aistats.org/api/miniconf/users/23079?format=json", "institution": "Department of Computer Science, University of Washington"}, {"id": 1244, "fullname": "Kevin Jamieson", "url": "http://virtual.aistats.org/api/miniconf/users/1244?format=json", "institution": "University of Washington"}, {"id": 12739, "fullname": "Jamie Morgenstern", "url": "http://virtual.aistats.org/api/miniconf/users/12739?format=json", "institution": "U Washington"}, {"id": 3969, "fullname": "Ioannis Panageas", "url": "http://virtual.aistats.org/api/miniconf/users/3969?format=json", "institution": "UC Irvine"}, {"id": 4128, "fullname": "Lillian Ratliff", "url": "http://virtual.aistats.org/api/miniconf/users/4128?format=json", "institution": "University of Washington"}], "abstract": "In this paper, we study last-iterate convergence of learning algorithms in bilinear saddle-point problems, a preferable notion of convergence that captures the day-to-day behavior of learning dynamics. We focus on the challenging setting where players select actions from compact convex sets and receive only bandit feedback.  Our main contribution is the design of an uncoupled learning algorithm that guarantees last-iterate convergence to the Nash equilibrium with high probability. We establish a convergence rate of $\\tilde{O}(T^{-1/4})$ up to polynomial factors in problem parameters. Crucially, our proposed algorithm is computationally efficient, requiring only an efficient linear optimization oracle over the players' compact action sets. The algorithm is obtained by combining techniques from experimental design and the classic Follow-The-Regularized-Leader (FTRL) framework, with a carefully chosen regularizer function tailored to the geometry of the action set of each learner.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13913", "url": null, "sourceid": 1068, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=08gFRz36dc", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11606, "modified": "2026-03-29T20:43:18.398950-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=08gFRz36dc", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "46", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13687, "uid": "51174add1c52758f33d414ceaf3fe6ba", "name": "Counterfactually Fair Conformal Prediction", "authors": [{"id": 19798, "fullname": "Ozgur Guldogan", "url": "http://virtual.aistats.org/api/miniconf/users/19798?format=json", "institution": "University of California, Santa Barbara"}, {"id": 22628, "fullname": "Neeraj Sarna", "url": "http://virtual.aistats.org/api/miniconf/users/22628?format=json", "institution": "Munich RE"}, {"id": 22629, "fullname": "Yuanyuan Li", "url": "http://virtual.aistats.org/api/miniconf/users/22629?format=json", "institution": "Munich Re"}, {"id": 22630, "fullname": "Michael Berger", "url": "http://virtual.aistats.org/api/miniconf/users/22630?format=json", "institution": "Munich Re"}], "abstract": "While counterfactual fairness of point predictors is well studied, its extension to prediction *sets*---central to fair decision-making under uncertainty---remains underexplored. On the other hand, conformal prediction (CP) provides efficient, distribution-free, finite-sample valid prediction sets, yet does not ensure counterfactual fairness. We close this gap by developing *Counterfactually Fair Conformal Prediction* (CF-CP) that produces counterfactually fair prediction sets. Through symmetrization of conformity scores across protected-attribute interventions, we prove that CF-CP results in counterfactually fair prediction sets while maintaining the marginal coverage property. Furthermore, we empirically demonstrate that on both synthetic and real datasets, across regression and classification tasks, CF-CP achieves the desired counterfactual fairness and meets the target coverage rate with minimal increase in prediction set size. CF-CP offers a simple, training-free route to counterfactually fair uncertainty quantification.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13687", "url": null, "sourceid": 2114, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=JoBnhpd9PQ", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11380, "modified": "2026-03-29T20:43:08.778975-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=JoBnhpd9PQ", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "47", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13555, "uid": "7b13b2203029ed80337f27127a9f1d28", "name": "Enforcing Fair Predicted Scores on Intervals of Percentiles by Difference-of-Convex Constraints", "authors": [{"id": 22357, "fullname": "Yutian He", "url": "http://virtual.aistats.org/api/miniconf/users/22357?format=json", "institution": "University of Iowa"}, {"id": 22358, "fullname": "Yankun Huang", "url": "http://virtual.aistats.org/api/miniconf/users/22358?format=json", "institution": "Arizona State University"}, {"id": 22359, "fullname": "Yao Yao", "url": "http://virtual.aistats.org/api/miniconf/users/22359?format=json", "institution": "University of Minnesota - Twin Cities"}, {"id": 9682, "fullname": "Qihang Lin", "url": "http://virtual.aistats.org/api/miniconf/users/9682?format=json", "institution": "University of Iowa"}], "abstract": "Fairness in machine learning has become a critical concern, particularly in high-stakes applications. Existing approaches often focus on achieving full fairness across all score ranges generated by predictive models, ensuring fairness in both high and low-scoring populations. However, this stringent requirement can compromise predictive performance and may not align with the practical fairness concerns of stakeholders. In this work, we propose a novel framework for building partially fair machine learning models, which enforce fairness within a specific score range of interest, such as the middle range where decisions are most contested, while maintaining flexibility in other regions. We introduce two statistical metrics to rigorously evaluate partial fairness within a given score range, such as the top 20%\u201340% of scores. To achieve partial fairness, we propose an in-processing method by formulating the model training problem as constrained optimization with difference-of-convex constraints, which can be solved by an inexact difference-of-convex algorithm (IDCA). We provide the complexity analysis of IDCA for finding a nearly KKT point. Through numerical experiments on real-world datasets, we demonstrate that our framework achieves high predictive performance while enforcing partial fairness where it matters most.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13555", "url": null, "sourceid": 993, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=ZY2jNXUeja", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11248, "modified": "2026-03-29T20:43:03.347622-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=ZY2jNXUeja", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "47", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13802, "uid": "9cf81d8026a9018052c429cc4e56739b", "name": "Differentially Private Clustering in Data Streams", "authors": [{"id": 22557, "fullname": "Alessandro Epasto", "url": "http://virtual.aistats.org/api/miniconf/users/22557?format=json", "institution": "Google"}, {"id": 22895, "fullname": "Tamalika Mukherjee", "url": "http://virtual.aistats.org/api/miniconf/users/22895?format=json", "institution": "Max Planck Institute for Security and Privacy"}, {"id": 22896, "fullname": "Peilin Zhong", "url": "http://virtual.aistats.org/api/miniconf/users/22896?format=json", "institution": "Meta"}], "abstract": "Clustering tasks such as k-means and k-median are central in unsupervised learning, and streaming algorithms for these tasks are widely used to handle large or evolving datasets. When applied in sensitive domains, however, such algorithms must also provide rigorous privacy guarantees.  In this work, we provide the first differentially private (DP) algorithms for k-means and k-median clustering of d-dimensional Euclidean data points over a stream of length at most T, using space that is sublinear in T, in the continual release setting where the algorithm is required to output a clustering at every timestep.  We achieve (1) an O(1)-multiplicative approximation with ~O(k^{1.5} poly(d, log T)) space and poly(k,d,log T) additive error, or (2) a (1+gamma)-multiplicative approximation with ~O_gamma(poly(k, 2^{O_gamma(d)}, log T)) space for any gamma>0, with additive error poly(k, 2^{O_gamma(d)}, log T). Our main technical contribution is a DP clustering framework for data streams that only requires an offline DP coreset or clustering algorithm as a blackbox.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13802", "url": null, "sourceid": 604, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=AfjgkX1M1c", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11495, "modified": "2026-03-29T20:43:13.625448-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=AfjgkX1M1c", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "48", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13552, "uid": "54f5f4071faca32ad5285fef87b78646", "name": "Deep Polynomial Chaos Expansion", "authors": [{"id": 22352, "fullname": "Johannes Exenberger", "url": "http://virtual.aistats.org/api/miniconf/users/22352?format=json", "institution": "Technische Universit\u00e4t Wien"}, {"id": 22353, "fullname": "Sascha Ranftl", "url": "http://virtual.aistats.org/api/miniconf/users/22353?format=json", "institution": "Purdue University"}, {"id": 5697, "fullname": "Robert Peharz", "url": "http://virtual.aistats.org/api/miniconf/users/5697?format=json", "institution": "Graz University of Technology"}], "abstract": "Polynomial chaos expansion (PCE) is a classical and widely used surrogate modeling technique in physical simulation and uncertainty quantification. By taking a linear combination of a set of basis polynomials - orthonormal with respect to the distribution of uncertain input parameters - PCE enables tractable inference of key statistical quantities such as (conditional) means, variances, covariances, and Sobol sensitivity indices, which are essential for understanding the modeled system and identifying influential parameters and their interactions. The applicability of PCE to high-dimensional problems is limited by poor scalability, as the number of basis functions grows exponentially with the number of parameters. In this paper, we address this challenge by combining PCE with ideas from tractable probabilistic circuits, resulting in *deep polynomial chaos expansion* (DeepPCE) - a deep generalization of PCE that scales effectively to high-dimensional input spaces. DeepPCE achieves predictive performance comparable to that of multilayer perceptrons (MLPs), while retaining PCE's ability to compute *exact* statistical inferences via simple forward passes. In contrast, such computations in MLPs require costly and often inaccurate approximations, such as Monte Carlo integration.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13552", "url": null, "sourceid": 1740, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=aDrWwCb8gn", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11245, "modified": "2026-03-29T20:43:03.214532-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=aDrWwCb8gn", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "49", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13530, "uid": "e077e1a544eec4f0307cf5c3c721d944", "name": "Ergodic and Subhomogeneous Dynamics in Hyperbolic Neural Networks", "authors": [{"id": 19794, "fullname": "Nico Alvarado", "url": "http://virtual.aistats.org/api/miniconf/users/19794?format=json", "institution": "CENIA"}, {"id": 22299, "fullname": "Sebastian Burgos", "url": "http://virtual.aistats.org/api/miniconf/users/22299?format=json", "institution": "Pennsylvania State University"}], "abstract": "We analyze the long term behavior of hyperbolic neural networks through subhomogeneous layer maps, focusing on stability, growth control, and robustness under stochastic perturbations. This work unifies the standard hyperbolic models via explicit isometries and M\u00f6bius operations, allowing statements to be transported across representations without loss of geometric meaning. Within this model invariant view, we study iterated, noise perturbed transformations and develop an ergodic theoretic framework that characterizes their asymptotic behavior, including conditions that promote stability and convergence of averaged iterates. Beyond theory, these insights inform practical design choices for training procedures that remain well-behaved in the presence of noise and avoid unbounded parameter growth, thereby supporting more reliable use of hyperbolic representations in hierarchical and graph structured learning tasks.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13530", "url": null, "sourceid": 1331, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=cMN3o7KYGp", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11223, "modified": "2026-03-29T20:43:02.425244-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=cMN3o7KYGp", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "49", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13733, "uid": "23c97e9cb93576e45d2feaf00d0e8502", "name": "Deliberate-When-Needed: Flow-Reasoner for Neuro-Symbolic Continuous Thought", "authors": [{"id": 22735, "fullname": "Wenjie Shen", "url": "http://virtual.aistats.org/api/miniconf/users/22735?format=json", "institution": "University of Science and Technology of China"}, {"id": 22736, "fullname": "Boyang Li", "url": "http://virtual.aistats.org/api/miniconf/users/22736?format=json", "institution": "The Chinese University of Hong Kong"}, {"id": 22515, "fullname": "Chao Yang", "url": "http://virtual.aistats.org/api/miniconf/users/22515?format=json", "institution": "The Chinese University of Hong Kong, Shenzhen"}, {"id": 13320, "fullname": "Shuang Li", "url": "http://virtual.aistats.org/api/miniconf/users/13320?format=json", "institution": "The Chinese University of Hong Kong, Shenzhen"}], "abstract": "We present Flow-Reasoner, a Deliberate-When-Needed neuro-symbolic model that integrates continuous latent cognition with selective symbolic reasoning.  The mental module is a latent state vector evolving smoothly under a first-order ordinary differential equation (ODE), capturing continuous thought that drifts and decays between interventions. The action module is a temporal point process whose intensities are modulated by symbolic rules. Crucially, reasoning is not constant: it is triggered only at irregular instants\u2014when an observed action arrives or when a latent state crosses a threshold\u2014at which point a bounded differentiable forward-chaining procedure updates beliefs and adjusts event likelihoods. Between these triggers, cognition evolves autonomously under the ODE without symbolic intervention. This design yields a model that (i) unifies continuous-time dynamics with selective logical reasoning, (ii) predicts both the type and timing of future actions, and (iii) produces concise rule traces that explain predictions. Empirical studies on synthetic benchmarks and real-world behavioral datasets demonstrate that Flow-Reasoner consistently outperforms strong temporal point process baselines, while providing interpretable, cognitively inspired explanations of decision dynamics.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13733", "url": null, "sourceid": 2140, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=FQe2n8OAKT", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11426, "modified": "2026-03-29T20:43:10.647589-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=FQe2n8OAKT", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "50", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13806, "uid": "d6ef5f7fa914c19931a55bb262ec879c", "name": "Gaussian Equivalence for Self-Attention: Asymptotic Spectral Analysis of Attention Matrix", "authors": [{"id": 1617, "fullname": "Tomohiro Hayase", "url": "http://virtual.aistats.org/api/miniconf/users/1617?format=json", "institution": "AIST"}, {"id": 22785, "fullname": "Benoit Collins", "url": "http://virtual.aistats.org/api/miniconf/users/22785?format=json", "institution": "Kyoto University"}, {"id": 390, "fullname": "Ryo Karakida", "url": "http://virtual.aistats.org/api/miniconf/users/390?format=json", "institution": "National Institute of Advanced Industrial Science and Technology"}], "abstract": "Self-attention layers have become fundamental building blocks of modern deep neural networks, yet their theoretical understanding remains limited, particularly from the perspective of random matrix theory. In this work, we provide a rigorous analysis of the singular value spectrum of the attention matrix and establish the first Gaussian equivalence result for attention. In a natural regime where the inverse temperature remains of constant order, we show that the singular value distribution of the attention matrix is asymptotically characterized by a tractable linear model. We further demonstrate that the distribution of squared singular values deviates from the Marchenko\u2013Pastur law, which has been believed in previous work. Our proof relies on two key ingredients: precise control of fluctuations in the normalization term and a refined linearization that leverages favorable Taylor expansions of the exponential. This analysis also identifies a threshold for linearization and elucidates why attention, despite not being an entrywise operation, admits a rigorous Gaussian equivalence in this regime.", "topic": null, "keywords": [], "decision": "Accept (Oral)", "session": "Oral Session 2: Learning Theory & High-Dimensional Statistics", "eventtype": "Oral", "event_type": "Oral", "room_name": null, "virtualsite_url": "/virtual/2026/oral/13806", "url": null, "sourceid": -1114, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=2026-Oral--1114-6cb908ed", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Oral%20Session%202:%20Learning%20Theory%20&%20High-Dimensional%20Statistics?format=json", "parent_id": 11476, "eventmedia": [{"id": 11499, "modified": "2026-03-29T20:43:13.787195-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=2026-Oral--1114-6cb908ed", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": null, "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13465, "uid": "0f840be9b8db4d3fbd5ba2ce59211f55", "name": "Evaluation of Large Language Models via Coupled Token Generation", "authors": [{"id": 22159, "fullname": "Nina Corvelo Benz", "url": "http://virtual.aistats.org/api/miniconf/users/22159?format=json", "institution": "Max Planck Institute of Biochemistry"}, {"id": 22160, "fullname": "Stratis Tsirtsis", "url": "http://virtual.aistats.org/api/miniconf/users/22160?format=json", "institution": "Hasso Plattner Institute"}, {"id": 19839, "fullname": "Eleni Straitouri", "url": "http://virtual.aistats.org/api/miniconf/users/19839?format=json", "institution": "Max Planck Institute for Software Systems"}, {"id": 22161, "fullname": "Ivi Chatzi", "url": "http://virtual.aistats.org/api/miniconf/users/22161?format=json", "institution": "MPI-SWS"}, {"id": 22162, "fullname": "Ander Artola Velasco", "url": "http://virtual.aistats.org/api/miniconf/users/22162?format=json", "institution": "Max Planck Institute for Software Systems"}, {"id": 22163, "fullname": "Suhas Thejaswi", "url": "http://virtual.aistats.org/api/miniconf/users/22163?format=json", "institution": "Aalto University"}, {"id": 22164, "fullname": "Manuel Gomez Rodriguez", "url": "http://virtual.aistats.org/api/miniconf/users/22164?format=json", "institution": "MPI-SWS"}], "abstract": "State-of-the-art large language models rely on randomization to respond to a prompt. Consequently, a model may respond differently to the same prompt if asked multiple times. In this work, we argue that the evaluation and ranking of large language models should control for this randomization. Our starting point is the development of a causal model for coupled autoregressive generation, which allows different large language models to sample responses with the same source of randomness. Building upon our causal model, we first show that, on evaluations based on benchmark datasets, coupled autoregressive generation leads to the same conclusions as vanilla autoregressive generation but using provably fewer samples. However, we further show that, on evaluations based on pairwise comparisons, the two approaches can surprisingly lead to different rankings when comparing more than two models. To complement our theoretical results, we conduct experiments with several models from the $\\texttt{Llama}$, $\\texttt{Mistral}$ and $\\texttt{Qwen}$ families. We find that, across multiple benchmark datasets, coupled autoregressive generation requires up to $75$\\% fewer samples to reach the same conclusions as vanilla autoregressive generation. Further, we find that the win-rates derived from pairwise comparisons by a strong large language model to prompts from the LMSYS Chatbot Arena platform differ under coupled and vanilla autoregressive generation.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13465", "url": null, "sourceid": 959, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=jVqfL2jHUG", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11158, "modified": "2026-03-29T20:42:59.997691-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=jVqfL2jHUG", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "50", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13369, "uid": "142949df56ea8ae0be8b5306971900a4", "name": "Doctor Rashomon and the UNIVERSE of Madness: Variable Importance with Unobserved Confounding and the Rashomon Effect", "authors": [{"id": 21957, "fullname": "Jon Donnelly", "url": "http://virtual.aistats.org/api/miniconf/users/21957?format=json", "institution": "Duke University"}, {"id": 12624, "fullname": "Srikar Katta", "url": "http://virtual.aistats.org/api/miniconf/users/12624?format=json", "institution": "Duke University"}, {"id": 21958, "fullname": "Emanuele Borgonovo", "url": "http://virtual.aistats.org/api/miniconf/users/21958?format=json", "institution": "Bocconi University"}, {"id": 4499, "fullname": "Cynthia Rudin", "url": "http://virtual.aistats.org/api/miniconf/users/4499?format=json", "institution": "Duke"}], "abstract": "Variable importance (VI) methods are often used for hypothesis generation, feature selection, and scientific validation. In the standard VI pipeline, an analyst estimates VI for a single predictive model with only the observed features. However, the importance of a feature depends heavily on which other variables are included in the model, and essential variables are often omitted from observational datasets. Moreover, the VI estimated for one model is often not the same as the VI estimated for another equally-good model \u2013 a phenomenon known as the Rashomon Effect. We address these gaps by introducing UNobservables and Inference for Variable importancE using Rashomon SEts (UNIVERSE). Our approach adapts Rashomon sets \u2013 the sets of near-optimal models in a dataset \u2013 to produce bounds on the true VI even with missing features. We theoretically guarantee the robustness of our approach, show strong performance on semi-synthetic simulations, and demonstrate its utility in a credit risk task.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13369", "url": null, "sourceid": 376, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=uRBSFQfEdO", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11062, "modified": "2026-03-29T20:42:56.065853-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=uRBSFQfEdO", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "51", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13450, "uid": "372d3f309fef061977fb2f7ba36d74d2", "name": "Exact and Approximate MCMC for Doubly-intractable Probabilistic Graphical Models Leveraging the Underlying Independence Model", "authors": [{"id": 22131, "fullname": "Yujie Chen", "url": "http://virtual.aistats.org/api/miniconf/users/22131?format=json", "institution": "Purdue University"}, {"id": 22129, "fullname": "Antik Chakraborty", "url": "http://virtual.aistats.org/api/miniconf/users/22129?format=json", "institution": "Purdue University"}, {"id": 22130, "fullname": "Anindya Bhadra", "url": "http://virtual.aistats.org/api/miniconf/users/22130?format=json", "institution": "Purdue University"}], "abstract": "Bayesian inference for doubly-intractable pairwise exponential graphical models typically involves variations of the exchange algorithm or approximate Markov chain Monte Carlo (MCMC) samplers. However, existing methods for both classes of algorithms require either perfect samplers or sequential samplers for complex models, which are often either not available, or suffer from poor mixing, especially in high dimensions. We develop a method that does not require perfect or sequential sampling, and can be applied to both classes of methods: exact and approximate MCMC. The key to our approach is to utilize the tractable independence model underlying the intractable probabilistic graphical model for the purpose of constructing a finite sample unbiased Monte Carlo (and not MCMC) estimate of the Metropolis--Hastings ratio. This innovation turns out to be crucial for scalability in high dimensions. The method is demonstrated on the Ising model. Gradient-based alternatives to construct a proposal, such as Langevin and Hamiltonian Monte Carlo approaches, also arise as a natural corollary to our general procedure, and are demonstrated as well.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13450", "url": null, "sourceid": 1465, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=lDnMftmNhP", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11143, "modified": "2026-03-29T20:42:59.419835-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=lDnMftmNhP", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "52", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13889, "uid": "2a9d121cd9c3a1832bb6d2cc6bd7a8a7", "name": "Explore-then-Commit for Nonstationary Linear Bandits with Latent Dynamics", "authors": [{"id": 19384, "fullname": "Sunmook Choi", "url": "http://virtual.aistats.org/api/miniconf/users/19384?format=json", "institution": "Cornell University"}, {"id": 23060, "fullname": "Yahya Sattar", "url": "http://virtual.aistats.org/api/miniconf/users/23060?format=json", "institution": "Cornell University"}, {"id": 21879, "fullname": "Yassir Jedra", "url": "http://virtual.aistats.org/api/miniconf/users/21879?format=json", "institution": "Imperial College London"}, {"id": 1196, "fullname": "Maryam Fazel", "url": "http://virtual.aistats.org/api/miniconf/users/1196?format=json", "institution": "University of Washington"}, {"id": 12737, "fullname": "Sarah Dean", "url": "http://virtual.aistats.org/api/miniconf/users/12737?format=json", "institution": "Cornell University"}], "abstract": "We study a nonstationary bandit problem where rewards depend on both actions and latent states, the latter governed by unknown linear dynamics. Crucially, the state dynamics also depend on the actions, resulting in tension between short-term and long-term rewards. We propose an explore-then-commit algorithm for a finite horizon $T$. During the exploration phase, random Rademacher actions enable estimation of the Markov parameters of the linear dynamics, which characterize the action-reward relationship. In the commit phase, the algorithm uses the estimated parameters to design an optimized action sequence for long-term reward. Our proposed algorithm achieves $\\tilde{\\mathcal{O}}(pT^{2/3})$ regret where $p$ is the action dimension. Our analysis handles two key challenges: learning from temporally correlated rewards, and designing action sequences with optimal long-term reward. We address the first challenge by providing near-optimal sample complexity and error bounds for system identification using bilinear rewards. We address the second challenge by proving an equivalence with indefinite quadratic optimization over a hypercube, a known NP-hard problem. We provide a sub-optimality guarantee for this problem, enabling our regret upper bound. Lastly, we propose a semidefinite relaxation with Goemans-Williamson rounding as a practical approach.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13889", "url": null, "sourceid": 912, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=2gbhbWlmQq", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11582, "modified": "2026-03-29T20:43:17.394272-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=2gbhbWlmQq", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "53", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13853, "uid": "71a3cb155f8dc89bf3d0365288219936", "name": "Duality-based Residual Estimation for Fully Offline Value-based Reinforcement Learning", "authors": [{"id": 22992, "fullname": "Kohei Miyaguchi", "url": "http://virtual.aistats.org/api/miniconf/users/22992?format=json", "institution": "LY Corporation"}], "abstract": "Value-based reinforcement learning (RL) efficiently handles high-dimensional state spaces, but existing methods lack a principled method for hyperparameter tuning without online interaction, limiting use in safety-critical and data-scarce domains. We propose the **Duality-based Residual Estimator (DRE)**, a simple offline validation metric for value-based offline RL. DRE is compatible with standard value-based Off-Policy Evaluation (OPE) and enables automatic hyperparameter selection, which is formalized through an adaptive extension of the Probably Approximately Correct (PAC) guarantee for Q-function selection. Our results address a key theoretical bottleneck toward *fully offline* value-based RL, which enables deployment without extensive online tuning.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13853", "url": null, "sourceid": 677, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=5ymJ1ylWNd", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11546, "modified": "2026-03-29T20:43:15.763372-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=5ymJ1ylWNd", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "53", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13837, "uid": "3875115bacc48cca24ac51ee4b0e7975", "name": "Direct Preference Optimization With Unobserved Preference Heterogeneity", "authors": [{"id": 22972, "fullname": "Keertana Chidambaram", "url": "http://virtual.aistats.org/api/miniconf/users/22972?format=json", "institution": "Stanford University"}, {"id": 22973, "fullname": "Karthik Seetharaman", "url": "http://virtual.aistats.org/api/miniconf/users/22973?format=json", "institution": "Stanford University"}, {"id": 12717, "fullname": "Vasilis Syrgkanis", "url": "http://virtual.aistats.org/api/miniconf/users/12717?format=json", "institution": "Stanford University"}], "abstract": "Reinforcement Learning from Human Feedback (RLHF) has become central to aligning large language models with human values, typically by first learning a reward model from preference data which is then used to update the model with reinforcement learning. Recent alternatives such as Direct Preference Optimization (DPO) simplify this pipeline by directly optimizing on preferences. However, both approaches often assume uniform annotator preferences and rely on binary comparisons, overlooking two key limitations: the diversity of human evaluators and the limitations of pairwise feedback. In this work, we address both these issues. First, we connect preference learning in RLHF with the econometrics literature and show that binary comparisons are insufficient for identifying latent user preferences from finite user data and infinite users, while (even incomplete) rankings over three or more responses ensure identifiability. Second, we introduce methods to incorporate heterogeneous preferences into alignment algorithms. We develop an Expectation-Maximization adaptation of DPO that discovers latent annotator types and trains a mixture of LLMs accordingly. Then we propose an aggregation algorithm using a min-max regret fairness criterion to produce a single generative policy with equitable performance guarantees. Together, these contributions establish a theoretical and algorithmic framework for fairness and personalization for diverse users in generative model alignment.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13837", "url": null, "sourceid": 2046, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=7RFcY3ljMO", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11530, "modified": "2026-03-29T20:43:15.111294-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=7RFcY3ljMO", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "54", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13793, "uid": "6d9c547cf146054a5a720606a7694467", "name": "Fact-Augmented Lookahead for LLM Agents: Simple Online Memory, No Finetuning", "authors": [{"id": 10025, "fullname": "Samuel Holt", "url": "http://virtual.aistats.org/api/miniconf/users/10025?format=json", "institution": "University of Cambridge"}, {"id": 22882, "fullname": "Max Ruiz Luyten", "url": "http://virtual.aistats.org/api/miniconf/users/22882?format=json", "institution": "University of Cambridge"}, {"id": 22883, "fullname": "Thomas Pouplin", "url": "http://virtual.aistats.org/api/miniconf/users/22883?format=json", "institution": "University of Cambridge"}, {"id": 863, "fullname": "Mihaela van der Schaar", "url": "http://virtual.aistats.org/api/miniconf/users/863?format=json", "institution": "University of Cambridge"}], "abstract": "Large Language Models (LLMs) are increasingly capable but often require targeted guidance or extensive interaction history to plan effectively in complex, interactive environments. We introduce an LLM agent framework that enhances planning through in-context learning, facilitated by \\emph{atomic fact} augmentation and a recursive, depth-limited lookahead. The agent extracts task-critical facts from its trajectories, validates candidates with a lightweight predictive-consistency filter (and optionally compresses them), and uses the resulting fact set to condition action proposal, single-step latent world-model simulation, and state-value estimation. Planning proceeds by simulating and evaluating candidate trajectories with the accumulated facts and recent history, enabling online improvement without weight updates. We provide abstraction-style motivation\u2014treating facts as reducing state aliasing (proxy $\\epsilon_{\\mathrm{sim}}$) and fact-conditioned simulation as lowering one-step error (proxy $\\delta_{\\mathrm{model}}$)\u2014without claiming formal guarantees. Empirically, on text FrozenLake variants, CrafterMini, and ALFWorld, the approach improves cumulative return over ReAct/Reflexion and search-only baselines.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13793", "url": null, "sourceid": 1408, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=BCkwhqeRJV", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11486, "modified": "2026-03-29T20:43:13.245749-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=BCkwhqeRJV", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "54", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13517, "uid": "3fb451ca2e89b3a13095b059d8705b15", "name": "Discrete State Diffusion Models: A Sample Complexity Perspective", "authors": [{"id": 22282, "fullname": "Aadithya Srikanth", "url": "http://virtual.aistats.org/api/miniconf/users/22282?format=json", "institution": "Purdue University"}, {"id": 22283, "fullname": "Mudit", "url": "http://virtual.aistats.org/api/miniconf/users/22283?format=json", "institution": "Purdue University"}, {"id": 1130, "fullname": "Vaneet Aggarwal", "url": "http://virtual.aistats.org/api/miniconf/users/1130?format=json", "institution": "KAUST"}], "abstract": "Diffusion models have demonstrated remarkable performance in generating high-dimensional samples across domains such as vision, language, and the sciences. Although continuous-state diffusion models have been extensively studied both empirically and theoretically, discrete-state diffusion models, essential for applications involving text, sequences, and combinatorial structures, remain significantly less understood from a theoretical standpoint. In particular, all existing analyses of discrete-state models assume score estimation error bounds without studying sample complexity results. In this work, we present a principled theoretical framework for discrete-state diffusion, providing the first sample complexity bound of $\\widetilde{\\mathcal{O}}(\\epsilon^{-2})$. Our structured decomposition of the score estimation error into statistical, approximation, optimization, and clipping components offers critical insights into how discrete-state models can be trained efficiently. This analysis addresses a fundamental gap in the literature and establishes the theoretical tractability and practical relevance of discrete-state diffusion models.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13517", "url": null, "sourceid": 1522, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=dTneiZmFre", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11210, "modified": "2026-03-29T20:43:01.927675-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=dTneiZmFre", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "55", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13400, "uid": "ad3019b856147c17e82a5bead782d2a8", "name": "Fast and Robust Simulation-Based Inference With Optimization Monte Carlo", "authors": [{"id": 22009, "fullname": "Vasilis Gkolemis", "url": "http://virtual.aistats.org/api/miniconf/users/22009?format=json", "institution": "\u201cAthena\u201d Research Centre Artemidos 6 &amp; Epidavrou GR-151 25, Athens - GREECE VAT NUMBER: EL999723442"}, {"id": 22010, "fullname": "Christos Diou", "url": "http://virtual.aistats.org/api/miniconf/users/22010?format=json", "institution": "Harokopio University of Athens"}, {"id": 22011, "fullname": "Michael Gutmann", "url": "http://virtual.aistats.org/api/miniconf/users/22011?format=json", "institution": "University of Edinburgh"}], "abstract": "Bayesian parameter inference for complex stochastic simulators is challenging due to intractable likelihood functions. Existing simulation-based inference methods often require large number of simulations and become costly to use in high-dimensional parameter spaces or in problems with partially uninformative outputs. We propose a new method for differentiable simulators that delivers accurate posterior inference with substantially reduced runtimes. Building on the Optimization Monte Carlo framework, our approach reformulates stochastic simulation as deterministic optimization problems. Gradient-based methods are then applied to efficiently navigate toward high-density posterior regions and avoid wasteful simulations in low-probability areas. A JAX-based implementation further enhances the performance through vectorization of key method components. Extensive experiments, including high-dimensional parameter inference, uninformative outputs, multiple observations, multimodal posteriors, and real-world applications, show that our method consistently matches the accuracy of state-of-the-art approaches and reduces the runtime by a substantial margin.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13400", "url": null, "sourceid": 1393, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=qKH5MSuTU9", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11093, "modified": "2026-03-29T20:42:57.371087-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=qKH5MSuTU9", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "56", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13347, "uid": "130f1a8e9e102707f3f91b010f151b0b", "name": "Disentangling Federated Learning Heterogeneity: A Dual-Perspective Analysis of Quantifying Skew versus Scarcity", "authors": [{"id": 23284, "fullname": "Wenkai Zeng", "url": "http://virtual.aistats.org/api/miniconf/users/23284?format=json", "institution": "University of Sydney, University of Sydney"}, {"id": 21908, "fullname": "NAN YANG", "url": "http://virtual.aistats.org/api/miniconf/users/21908?format=json", "institution": "University of Sydney"}, {"id": 21909, "fullname": "Zhiyu Zhu", "url": "http://virtual.aistats.org/api/miniconf/users/21909?format=json", "institution": "University of Technology Sydney"}, {"id": 21910, "fullname": "Zhibo Jin", "url": "http://virtual.aistats.org/api/miniconf/users/21910?format=json", "institution": "University of Technology Sydney"}, {"id": 21911, "fullname": "Dong Yuan", "url": "http://virtual.aistats.org/api/miniconf/users/21911?format=json", "institution": "University of Sydney"}], "abstract": "Federated Learning faces significant challenges due to data heterogeneity, which manifests as Label Distribution Skew and label missingness. We propose Skew-Scarcity Disentanglement Indicator (SSDI), a novel metric that decomposes heterogeneity into two disentangled components: Label Distribution Skew (LDS) (quantity skew of present labels) and Label Coverage Deficiency (LCD) (deviation due to missing labels). Using a PAC-Bayesian framework, we derive a generalization bound indicating that Label Coverage Deficiency becomes the dominant risk factor as the number of clients increases, severely degrading accuracy on rare labels. Our study reveals that, for a fixed number of labels, increasing clients is a primary driver of per-label accuracy variance by exacerbating Label Coverage Deficiency. Moreover, a higher global missing rate intensifies this divergence effect and can precipitate severe performance breakdown at a lower critical threshold of clients. Experiments on vision benchmarks confirm that SSDI accurately captures the severity of performance divergence. The SSDI framework provides a principled tool for diagnosing heterogeneity and guiding targeted mitigation strategies. The code for the SSDI-controlled client-label matrix generation used in our experiments is available at https://github.com/wkzeng/SSDI.git.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13347", "url": null, "sourceid": 2296, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=xaIfTT9rsK", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11040, "modified": "2026-03-29T20:42:55.229735-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=xaIfTT9rsK", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "56", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13882, "uid": "a01610228fe998f515a72dd730294d87", "name": "Faster Parallel MCMC: Metropolis Adjustment Is Best Served Warm", "authors": [{"id": 23044, "fullname": "Jakob Robnik", "url": "http://virtual.aistats.org/api/miniconf/users/23044?format=json", "institution": "University of California, Berkeley"}, {"id": 23045, "fullname": "Uros Seljak", "url": "http://virtual.aistats.org/api/miniconf/users/23045?format=json", "institution": "University of California Berkeley"}], "abstract": "Despite the enormous success of Hamiltonian Monte Carlo and related Markov Chain Monte Carlo (MCMC) methods, sampling often still represents the computational bottleneck in scientific applications. Availability of parallel resources can significantly speed up MCMC inference by running a large number of chains in parallel, each collecting a single sample. However, the parallel approach converges slowly  if the chains are not initialized close to the target distribution (cold start). Theoretically this can be resolved by initially running MCMC without Metropolis-Hastings adjustment to quickly converge to the vicinity of the target distribution and then turn on adjustment to achieve fine convergence. However, no practical scheme uses this strategy, due to the difficulty of automatically selecting the step size during the unadjusted phase. We here develop Late Adjusted Parallel Sampler (LAPS), which is precisely such a scheme and is applicable out of the box. LAPS takes advantage of ensemble-based hyperparameter adaptation to estimate the bias at each iteration and converts it to the appropriate step size. We show that LAPS consistently and significantly outperforms ensemble adjusted methods such as MEADS or ChESS and the optimization-based initializer Pathfinder on a variety of standard benchmark problems. LAPS typically achieves two orders of magnitude lower wall-clock time than the corresponding sequential algorithms such as NUTS.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13882", "url": null, "sourceid": 1212, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=3Whbbm5f79", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11575, "modified": "2026-03-29T20:43:17.148512-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=3Whbbm5f79", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "58", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13852, "uid": "0353ab4cbed5beae847a7ff6e220b5cf", "name": "Examining the Bias of In-Batch Sampling in Similarity Learning with Two-Tower Models", "authors": [{"id": 19796, "fullname": "Yaxu Liu", "url": "http://virtual.aistats.org/api/miniconf/users/19796?format=json", "institution": "National Taiwan University"}, {"id": 22991, "fullname": "Li-Chung Lin", "url": "http://virtual.aistats.org/api/miniconf/users/22991?format=json", "institution": "Department of computer science and informational engineering, National Taiwan University"}, {"id": 22323, "fullname": "Chih-Jen Lin", "url": "http://virtual.aistats.org/api/miniconf/users/22323?format=json", "institution": "National Taiwan Univ / MBZUAI"}], "abstract": "Two-tower models are widely used for applications involving learning similarities between pairs of entities, such as user-item pairs in recommender systems. These models are commonly trained using stochastic gradient methods. However, uniformly sampling data often leads to problematic batches that lack positive pairs, especially when positives are a minority of the dataset\u2014a situation particularly common in similarity learning. Instead, a strategy known as in-batch sampling is widely adopted to ensure the presence of positive pairs and training efficiency. Nevertheless, in-batch sampling introduces its own issues, such as mistaking positives for negatives and oversampling popular pairs, resulting in significant performance degradation. In this work, we provide the first systematic analysis of these issues, showing that they all arise from the inconsistency between the expected objective under in-batch sampling and the full-data objective. We refer to this inconsistency as the bias of in-batch sampling. To validate our analysis, we design an unbiased batch loss and conduct rigorous experiments directly comparing unbiased and biased losses. The results provide strong empirical confirmation of our theoretical findings.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13852", "url": null, "sourceid": 461, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=5zbvctrkEF", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11545, "modified": "2026-03-29T20:43:15.730008-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=5zbvctrkEF", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "59", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13388, "uid": "05a70454516ecd9194c293b0e415777f", "name": "DRAUN: An Optimization-Agnostic Data Reconstruction Attack on Federated Unlearning", "authors": [{"id": 20616, "fullname": "Hithem Lamri", "url": "http://virtual.aistats.org/api/miniconf/users/20616?format=json", "institution": "NYU Abu Dhabi"}, {"id": 21986, "fullname": "Manaar Alam", "url": "http://virtual.aistats.org/api/miniconf/users/21986?format=json", "institution": "New York University, Abu Dhabi"}, {"id": 21987, "fullname": "Haiyan Jiang", "url": "http://virtual.aistats.org/api/miniconf/users/21987?format=json", "institution": "New York University"}, {"id": 21988, "fullname": "Michail Maniatakos", "url": "http://virtual.aistats.org/api/miniconf/users/21988?format=json", "institution": "New York University Abu Dhabi"}], "abstract": "Federated Unlearning (FU) enables clients to remove the influence of specific data from a collaboratively trained shared global model, addressing regulatory requirements such as GDPR and CCPA. However, this unlearning process introduces a new privacy risk: A malicious server may exploit unlearning updates to reconstructthe data requested for removal, a form of Data Reconstruction Attack (DRA). While DRAs for machine unlearning have been studied extensively in centralized Machine Learning-as-a-Service (MLaaS) settings, their applicability to FU remains unclear due to the decentralized, client-driven nature of FU. This work presents DRAUN, the first attack framework to reconstruct unlearned data in FU systems. DRAUN targets optimization-based unlearning methods, which are widely adopted for their efficiency. We theoretically demonstrate why existing DRAs targeting machine unlearning in MLaaS fail in FU and show how DRAUN overcomes these limitations. We validate our approach through extensive experiments on five datasets and five model architectures, evaluating its performance against five popular unlearning methods, effectively demonstrating that state-of-the-art FU methods remain vulnerable to DRAs.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13388", "url": null, "sourceid": 2212, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=rvIOVztkmc", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11081, "modified": "2026-03-29T20:42:56.888714-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=rvIOVztkmc", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "59", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13439, "uid": "b710915795b9e9c02cf10d6d2bdb688c", "name": "Dual Averaging Converges for Nonconvex Smooth Stochastic Optimization", "authors": [{"id": 22107, "fullname": "Tuo Liu", "url": "http://virtual.aistats.org/api/miniconf/users/22107?format=json", "institution": "King Abdullah University of Science and Technology"}, {"id": 22108, "fullname": "El Mehdi Saad", "url": "http://virtual.aistats.org/api/miniconf/users/22108?format=json", "institution": "UM6P"}, {"id": 22109, "fullname": "Wojciech Kotlowski", "url": "http://virtual.aistats.org/api/miniconf/users/22109?format=json", "institution": "Poznan University of Technology"}, {"id": 22110, "fullname": "Francesco Orabona", "url": "http://virtual.aistats.org/api/miniconf/users/22110?format=json", "institution": "King Abdullah University of Science and Technology"}], "abstract": "Dual averaging and gradient descent with their stochastic variants stand as the two canonical recipe books for first-order optimization: Every modern variant can be viewed as a descendant of one or the other. In the convex regime, these algorithms have been deeply studied, and we know that the two classes are essentially equivalent in terms of theoretical guarantees. On the other hand, in the non-convex setting, the situation is drastically different: While it is provable that SGD can minimize the gradient norm of non-convex smooth functions, no finite-time complexity guarantee for Stochastic Dual Averaging (SDA) was known in the same setting. In this paper, we close this gap by a reduction that views SDA as SGD applied to a sequence of implicitly regularized objectives. We show that a tuned SDA exhibits a rate of convergence $\\mathcal{O}(1 / T + \\sigma \\log T/ \\sqrt{T})$, similar to that of SGD under the same assumptions. To our best knowledge, this is the first complete convergence theory for dual averaging on non-convex smooth stochastic problems without restrictive assumptions, closing a long-standing open problem in the field. Beyond the base algorithm, we also discuss ADA-DA, a variant that marries SDA with AdaGrad's auto-scaling, which achieves the same rate without requiring knowledge of the noise variance.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13439", "url": null, "sourceid": 1582, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=mHyCz0KHOu", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11132, "modified": "2026-03-29T20:42:58.959432-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=mHyCz0KHOu", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "60", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13700, "uid": "1baff70e2669e8376347efd3a874a341", "name": "Formally Exploring Time-Series Anomaly Detection Evaluation Metrics", "authors": [{"id": 22657, "fullname": "Dennis Wagner", "url": "http://virtual.aistats.org/api/miniconf/users/22657?format=json", "institution": "Universit\u00e4t Kaiserslautern"}, {"id": 22658, "fullname": "Arjun Nair", "url": "http://virtual.aistats.org/api/miniconf/users/22658?format=json", "institution": "Universit\u00e4t Kaiserslautern"}, {"id": 22659, "fullname": "Billy Franks", "url": "http://virtual.aistats.org/api/miniconf/users/22659?format=json", "institution": "Rheinland-Pf\u00e4lzische Technische Universit\u00e4t"}, {"id": 22660, "fullname": "Justus Arweiler", "url": "http://virtual.aistats.org/api/miniconf/users/22660?format=json", "institution": "Rheinland-Pf\u00e4lzische Technische Universit\u00e4t Kaiserslautern"}, {"id": 22661, "fullname": "Aparna Muraleedharan", "url": "http://virtual.aistats.org/api/miniconf/users/22661?format=json", "institution": "Technische Universit\u00e4t M\u00fcnchen"}, {"id": 22662, "fullname": "Indra Jungjohann", "url": "http://virtual.aistats.org/api/miniconf/users/22662?format=json", "institution": "Rheinland-Pf\u00e4lzische Technische Universit\u00e4t"}, {"id": 22663, "fullname": "Fabian Hartung", "url": "http://virtual.aistats.org/api/miniconf/users/22663?format=json", "institution": "BASF"}, {"id": 22664, "fullname": "Andriy Balinskyy", "url": "http://virtual.aistats.org/api/miniconf/users/22664?format=json", "institution": "RPTU Kaiserslautern-Landau"}, {"id": 13503, "fullname": "Saurabh Varshneya", "url": "http://virtual.aistats.org/api/miniconf/users/13503?format=json", "institution": "RPTU Kaiserslautern"}, {"id": 22665, "fullname": "Mayank Ahuja", "url": "http://virtual.aistats.org/api/miniconf/users/22665?format=json", "institution": "Rheinland-Pf\u00e4lzische Technische Universit\u00e4t"}, {"id": 22666, "fullname": "Nabeel Hussain Syed", "url": "http://virtual.aistats.org/api/miniconf/users/22666?format=json", "institution": "Rheinland-Pf\u00e4lzische Technische Universit\u00e4t"}, {"id": 22079, "fullname": "Mayank Nagda", "url": "http://virtual.aistats.org/api/miniconf/users/22079?format=json", "institution": "RPTU Kaiserslautern-Landau"}, {"id": 13346, "fullname": "Philipp Liznerski", "url": "http://virtual.aistats.org/api/miniconf/users/13346?format=json", "institution": "RPTU"}, {"id": 22667, "fullname": "Steffen Reithermann", "url": "http://virtual.aistats.org/api/miniconf/users/22667?format=json", "institution": "Rheinland-Pf\u00e4lzische Technische Universit\u00e4t"}, {"id": 22668, "fullname": "Maja Waldron", "url": "http://virtual.aistats.org/api/miniconf/users/22668?format=json", "institution": "University of Wisconsin - Madison"}, {"id": 22669, "fullname": "Sebastian Vollmer", "url": "http://virtual.aistats.org/api/miniconf/users/22669?format=json", "institution": "German Research Center for AI"}, {"id": 22670, "fullname": "Ralf Schulz", "url": "http://virtual.aistats.org/api/miniconf/users/22670?format=json", "institution": "Rheinland-Pf\u00e4lzische Technische Universit\u00e4t"}, {"id": 22671, "fullname": "Torsten Katz", "url": "http://virtual.aistats.org/api/miniconf/users/22671?format=json", "institution": "BASF"}, {"id": 1383, "fullname": "Stephan Mandt", "url": "http://virtual.aistats.org/api/miniconf/users/1383?format=json", "institution": "University of California, Irivine"}, {"id": 22672, "fullname": "Michael Bortz", "url": "http://virtual.aistats.org/api/miniconf/users/22672?format=json", "institution": "Rheinland-Pf\u00e4lzische Technische Universit\u00e4t"}, {"id": 22673, "fullname": "Heike Leitte", "url": "http://virtual.aistats.org/api/miniconf/users/22673?format=json", "institution": "Rheinland-Pf\u00e4lzische Technische Universit\u00e4t"}, {"id": 22674, "fullname": "Daniel Neider", "url": "http://virtual.aistats.org/api/miniconf/users/22674?format=json", "institution": "TU Dortmund University"}, {"id": 22675, "fullname": "Jakob Burger", "url": "http://virtual.aistats.org/api/miniconf/users/22675?format=json", "institution": "Technische Universit\u00e4t M\u00fcnchen"}, {"id": 22083, "fullname": "Fabian Jirasek", "url": "http://virtual.aistats.org/api/miniconf/users/22083?format=json", "institution": "RPTU Kaiserslautern"}, {"id": 22676, "fullname": "Hans Hasse", "url": "http://virtual.aistats.org/api/miniconf/users/22676?format=json", "institution": "Rheinland-Pf\u00e4lzische Technische Universit\u00e4t"}, {"id": 22084, "fullname": "Sophie Fellenz", "url": "http://virtual.aistats.org/api/miniconf/users/22084?format=json", "institution": "RPTU Kaiserslautern"}, {"id": 4921, "fullname": "Marius Kloft", "url": "http://virtual.aistats.org/api/miniconf/users/4921?format=json", "institution": "RPTU"}], "abstract": "Detecting anomalies in time series is vital to ensure safety and reliability in many real-world applications. Despite the staggering number of anomaly detection methods, it remains unclear which methods perform best, largely due to flawed evaluation practices.  Without rigorous analysis, evaluations yield unintuitive or misleading comparisons. Existing evaluation metrics often focus on specifics and, therefore, fail to capture essential aspects of the anomaly detection task. In this work, we formalize the problem by introducing verifiable properties of evaluation metrics that individually reflect important aspects of anomaly detection in time series. By formalizing requirements and analyzing them systematically, we outline a theoretical framework for evaluating time-series anomaly detection that can support principled evaluations and reliable comparisons. We analyze 37 known metrics and prove that most satisfy only few and none satisfy all properties, explaining many observed inconsistencies in evaluations. To address this gap, we introduce a new flexible evaluation metric LARM that provably satisfies all properties. We illustrate the adaptability of this approach by refining the properties to satisfy stricter requirements and adapting LARM to these advanced properties yielding ALARM.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13700", "url": null, "sourceid": 1489, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=INJj1SB5Uw", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11393, "modified": "2026-03-29T20:43:09.341137-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=INJj1SB5Uw", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "61", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13782, "uid": "e2c0be24560d78c5e599c2a9c9d0bbd2", "name": "FedCCA: Federated Canonical Correlation Analysis", "authors": [{"id": 22836, "fullname": "Zhengquan Luo", "url": "http://virtual.aistats.org/api/miniconf/users/22836?format=json", "institution": "Mohamed bin Zayed University of Artificial Intelligence"}, {"id": 22852, "fullname": "Kai Fong Ernest Chong", "url": "http://virtual.aistats.org/api/miniconf/users/22852?format=json", "institution": "Singapore University of Technology and Design"}, {"id": 1227, "fullname": "Pengfei Wei", "url": "http://virtual.aistats.org/api/miniconf/users/1227?format=json", "institution": "National University of Singapore"}, {"id": 22853, "fullname": "Changyou Chen", "url": "http://virtual.aistats.org/api/miniconf/users/22853?format=json", "institution": "State University of New York at Buffalo"}, {"id": 22854, "fullname": "Peilin Zhao", "url": "http://virtual.aistats.org/api/miniconf/users/22854?format=json", "institution": "Shanghai Jiao Tong University"}, {"id": 22855, "fullname": "Renmin Han", "url": "http://virtual.aistats.org/api/miniconf/users/22855?format=json", "institution": "Shandong University"}, {"id": 22856, "fullname": "Chunlai Zhou", "url": "http://virtual.aistats.org/api/miniconf/users/22856?format=json", "institution": "Renmin University of China"}, {"id": 22857, "fullname": "Yunlong Wang", "url": "http://virtual.aistats.org/api/miniconf/users/22857?format=json", "institution": "Institute of automation, Chinese academy of science, Chinese Academy of Sciences"}, {"id": 803, "fullname": "Zhiqiang Xu", "url": "http://virtual.aistats.org/api/miniconf/users/803?format=json", "institution": "Baidu"}], "abstract": "Canonical Correlation Analysis (CCA) is a key tool for cross-modal learning, but centralized solutions are impractical due to the heavy cost of high-dimensional covariance operations and the privacy sensitivity of distributed data. To address these challenges, we propose FedCCA, a federated framework that replaces explicit inverses and inner least-squares solves with a truncated von Neumann series, reducing matrix inversions to lightweight matrix\u2013vector multiplications while retaining provable convergence. This series formulation not only improves efficiency, but also provides explicit and tunable control of truncation error, and its structure naturally splits into client-side multiplications and a server-side projection step, making it particularly suitable for federated deployment. Building on this foundation, we incorporate Gaussian differential privacy and derive practical upper and lower bounds on the required noise variance, which yield end-to-end $(\\varepsilon,\\delta)$ guarantees together with convergence stability. Empirical results on five datasets confirm that FedCCA achieves accuracy comparable to centralized CCA and consistently outperforms ALS/TALS baselines in both sub-optimality gap and convergence speed, all while maintaining rigorous privacy protection.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13782", "url": null, "sourceid": 203, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=C1IA9PNO0Q", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11475, "modified": "2026-03-29T20:43:12.826679-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=C1IA9PNO0Q", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "63", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13663, "uid": "8a1e808b55fde9455cb3d8857ed88389", "name": "From Cells to Sentences: An End-to-End Framework for Table Understanding", "authors": [{"id": 19846, "fullname": "Deepak Vijaykeerhty", "url": "http://virtual.aistats.org/api/miniconf/users/19846?format=json", "institution": "IBM"}, {"id": 22581, "fullname": "Arvind Agarwal", "url": "http://virtual.aistats.org/api/miniconf/users/22581?format=json", "institution": "International Business Machines"}], "abstract": "Real-world tables are messy: column headers are inconsistent, cells contain errors or missing values, and crucial information is scattered across multiple tables and documents. These issues cause even state-of-the-art language models to fail at seemingly simple questions. We present a robust framework for table understanding that explicitly handles these challenges through three coordinated mechanisms: structure-aware encoders that learn invariance to common corruptions, trainable slots that compress evidence to a fixed-size representation, and grounding modules that align each slot to supporting text passages. Unlike prior work that treats tables as flat text or relies on clean schemas, our approach maintains strong performance even under schema corruption and structural perturbations. Across eight benchmarks spanning question answering, fact verification, and text generation, we achieve the best performance among methods without external tools on five tasks and remain competitive with systems using much larger models or SQL executors. Under schema corruption and row/column permutations, our method degrades by less than 2 points, while baselines drop by up to 6-22 points, confirming that explicit denoising and grounding are essential for robust table understanding.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13663", "url": null, "sourceid": 1076, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=MiaWNN3Qyl", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11356, "modified": "2026-03-29T20:43:07.811753-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=MiaWNN3Qyl", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "63", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13556, "uid": "5531a5834816222280f20d1ef9e95f69", "name": "Efficient Model Performance Evaluation Using a Combination of Expert and Crowd-sourced Labels", "authors": [{"id": 22360, "fullname": "Sam Corbett-Davies", "url": "http://virtual.aistats.org/api/miniconf/users/22360?format=json", "institution": "Meta"}, {"id": 22361, "fullname": "Viet-An Nguyen", "url": "http://virtual.aistats.org/api/miniconf/users/22361?format=json", "institution": "Meta"}, {"id": 22362, "fullname": "Udi Weinsberg", "url": "http://virtual.aistats.org/api/miniconf/users/22362?format=json", "institution": "Tel Aviv University"}], "abstract": "As models, particularly large language models (LLMs), are deployed on increasingly challenging tasks, correctly evaluating their performance is growing in importance and difficulty. Expert human labelers are high-quality but scarce and resource-intensive to obtain, while crowd-sourced labels are more readily accessible at scale but lower in quality. We propose Maven (Model And Voter EvaluatioN), a hierarchical Bayesian model that combines these two label sources to produce model performance estimates on binary tasks that are less biased than using crowd-sourced labels alone and have lower variance than using expert labels alone. By modeling the ranking of model scores, Maven is robust to a range of prediction distributions and achieves constant inference time regardless of dataset size.  We validate our approach on both simulated and real-world data, and deploy it to measure production models at Meta.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13556", "url": null, "sourceid": 2023, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=ZRYBWabi25", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11249, "modified": "2026-03-29T20:43:03.385733-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=ZRYBWabi25", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "63", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13535, "uid": "b1a59b315fc9a3002ce38bbe070ec3f5", "name": "FedDuA: Doubly Adaptive Federated Learning", "authors": [{"id": 20615, "fullname": "Shokichi Takakura", "url": "http://virtual.aistats.org/api/miniconf/users/20615?format=json", "institution": "LY Corporation"}, {"id": 22137, "fullname": "Seng Pei Liew", "url": "http://virtual.aistats.org/api/miniconf/users/22137?format=json", "institution": "SB Intuitions"}, {"id": 22138, "fullname": "Satoshi Hasegawa", "url": "http://virtual.aistats.org/api/miniconf/users/22138?format=json", "institution": "LY Corporation"}], "abstract": "Federated learning is a distributed learning framework where clients collaboratively train a global model without sharing their raw data.   FedAvg is a popular algorithm for federated learning, but it often suffers from slow convergence   due to the heterogeneity of local datasets and anisotropy in the parameter space.   In this work, we formalize the central server optimization procedure through the lens of mirror descent and propose a novel framework,   called FedDuA, which adaptively selects the global learning rate based on both inter-client and coordinate-wise heterogeneity in the local updates.   We prove that our proposed doubly adaptive step-size rule is minimax optimal and provide a convergence analysis for convex objectives.   Although the proposed method does not require additional communication or computational cost on clients,   extensive numerical experiments show that our proposed framework outperforms baselines in various settings and is robust to the choice of hyperparameters.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13535", "url": null, "sourceid": 261, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=boYXnBaRmB", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11228, "modified": "2026-03-29T20:43:02.592468-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=boYXnBaRmB", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "64", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13454, "uid": "ea159dc9788ffac311592613b7f71fbb", "name": "Eliciting Truthful Feedback for Preference-Based Learning via the VCG Mechanism", "authors": [{"id": 19476, "fullname": "Leo Landolt", "url": "http://virtual.aistats.org/api/miniconf/users/19476?format=json", "institution": "ETH Z\u00fcrich"}, {"id": 11108, "fullname": "Anna Maddux", "url": "http://virtual.aistats.org/api/miniconf/users/11108?format=json", "institution": "EPFL"}, {"id": 22139, "fullname": "Andreas Schlaginhaufen", "url": "http://virtual.aistats.org/api/miniconf/users/22139?format=json", "institution": "EPFL - EPF Lausanne"}, {"id": 22140, "fullname": "Saurabh Vaishampayan", "url": "http://virtual.aistats.org/api/miniconf/users/22140?format=json", "institution": "EPFL - EPF Lausanne"}, {"id": 13281, "fullname": "Maryam Kamgarpour", "url": "http://virtual.aistats.org/api/miniconf/users/13281?format=json", "institution": "EPFL"}], "abstract": "We study resource allocation problems in which a central planner allocates resources among strategic agents with private cost functions in order to minimize a social cost, defined as an aggregate of the agents\u2019 costs. This setting poses two main challenges: (i) the agents\u2019 cost functions may be unknown to them or difficult to specify explicitly, and (ii) agents may misreport their costs strategically. To address these challenges, we propose an algorithm that combines preference-based learning with Vickrey\u2013Clarke\u2013Groves (VCG) payments to incentivize truthful reporting. Our algorithm selects informative preference queries via D-optimal design, estimates cost parameters through maximum likelihood, and computes VCG allocations and payments based on these estimates. In a one-shot setting, we prove that the mechanism is approximately truthful, individually rational, and efficient up to an error of $\\tilde{\\mathcal O}(K^{-1/2})$ for $K$ preference queries per agent. In an online setting, these guarantees hold asymptotically with sublinear regret at a rate of $\\tilde{\\mathcal O}(T^{2/3})$ after $T$ rounds. Finally, we validate our approach through a numerical case study on demand response in local electricity markets.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13454", "url": null, "sourceid": 2384, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=kyBEsq2xw1", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11147, "modified": "2026-03-29T20:42:59.573226-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=kyBEsq2xw1", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "64", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13664, "uid": "cd89fef7ffdd490db800357f47722b20", "name": "From Token Imbalance to Balanced Routing: An ELBO-Regularized Probabilistic Framework for Contrastive Multimodal Learning", "authors": [{"id": 22582, "fullname": "Habibeh Naderi", "url": "http://virtual.aistats.org/api/miniconf/users/22582?format=json", "institution": "Dalhousie University"}, {"id": 22583, "fullname": "Behrouz Haji Soleimani", "url": "http://virtual.aistats.org/api/miniconf/users/22583?format=json", "institution": "Kinaxis"}, {"id": 22584, "fullname": "Stan Matwin", "url": "http://virtual.aistats.org/api/miniconf/users/22584?format=json", "institution": "Dalhousie University"}], "abstract": "We introduce CoPRIME (Contrastive Probabilistic Routing for IMbalanced tokens with ELBO-regularized mixture of experts), a probabilistic routing framework for multimodal representation learning that generalizes multimodal representation learning beyond vision-text by tackling the fundamental challenge of extreme token imbalance across modalities. This imbalanced-ness is particularly pronounced between spectrogram-tokenized audio and text. CoPRIME augments contrastive pretraining with an ELBO-regularized routing objective that jointly promotes 1) expert specialization, requiring experts to explain the tokens they receive, and 2) diverse utilization via KL regularization to a uniform prior. To stabilize routing, we further replace standard CoV-based regularizers with entropy-based importance and load losses, yielding smoother gradients and flexible, modality-aware routing without rigid uniformity constraints. On MOSEI and IEMOCAP datasets, CoPRIME achieves state-of-the-art zero- and few-shot emotion and sentiment results, outperforming dense Transformers and prior multimodal MoE variants while retaining the efficiency of sparse conditional computation. Ablations isolate the role of each loss and show that ELBO is the primary driver of stable specialization under modality imbalance, with entropy-based regularizers further improving convergence and utilization.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13664", "url": null, "sourceid": 1062, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=Ma3CTyCf2V", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11357, "modified": "2026-03-29T20:43:07.842077-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=Ma3CTyCf2V", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "65", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13472, "uid": "beed13602b9b0e6ecb5b568ff5058f07", "name": "FIELDING: Clustered Federated Learning with Data Drift", "authors": [{"id": 22176, "fullname": "Minghao Li", "url": "http://virtual.aistats.org/api/miniconf/users/22176?format=json", "institution": "Harvard University"}, {"id": 22177, "fullname": "Dmitrii Avdiukhin", "url": "http://virtual.aistats.org/api/miniconf/users/22177?format=json", "institution": "Northwestern University"}, {"id": 22178, "fullname": "Rana Shahout", "url": "http://virtual.aistats.org/api/miniconf/users/22178?format=json", "institution": "Harvard University"}, {"id": 22179, "fullname": "Nikita Ivkin", "url": "http://virtual.aistats.org/api/miniconf/users/22179?format=json", "institution": "Amazon"}, {"id": 3929, "fullname": "Vladimir Braverman", "url": "http://virtual.aistats.org/api/miniconf/users/3929?format=json", "institution": "Johns Hopkins University"}, {"id": 22180, "fullname": "Minlan Yu", "url": "http://virtual.aistats.org/api/miniconf/users/22180?format=json", "institution": "Harvard University"}], "abstract": "Federated Learning (FL) trains deep models across edge devices without centralizing raw data. However, client heterogeneity slows down convergence and limits global model accuracy. Clustered FL (CFL) mitigates this by grouping clients with similar representations and training a separate model for each cluster. In practice, client data evolves over time -- a phenomenon we refer to as data drift -- which breaks cluster homogeneity and degrades performance. Data drift can take different forms depending on whether changes occur in the output values, the input features, or the relationship between them. We propose FIELDING, a CFL framework for handling diverse types of data drift with low overhead. FIELDING detects drift at individual clients and performs selective re-clustering to balance cluster quality and model performance, while remaining robust to varying levels of heterogeneity. Experiments show that FIELDING improves final model accuracy by 2.4\u20136.9% and achieves target accuracy 1.38x\u20133.10x faster than existing state-of-the-art CFL methods.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13472", "url": null, "sourceid": 383, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=i4q2xjAuld", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11165, "modified": "2026-03-29T20:43:00.229042-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=i4q2xjAuld", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "65", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13826, "uid": "884d247c6f65a96a7da4d1105d584ddd", "name": "FocusViT: Faithful Explanations for Vision Transformers via Gradient-Guided Layer-Skipping", "authors": [{"id": 19416, "fullname": "Mohsin Ali", "url": "http://virtual.aistats.org/api/miniconf/users/19416?format=json", "institution": "University of Essex "}, {"id": 22944, "fullname": "Haider Raza", "url": "http://virtual.aistats.org/api/miniconf/users/22944?format=json", "institution": "University of Essex"}, {"id": 22945, "fullname": "John Gan", "url": "http://virtual.aistats.org/api/miniconf/users/22945?format=json", "institution": "University of Essex"}, {"id": 22946, "fullname": "Muhammad Haris Khan", "url": "http://virtual.aistats.org/api/miniconf/users/22946?format=json", "institution": "Mohamed Bin Zayed University of Artificial Intelligence"}], "abstract": "Vision Transformers (ViTs) have emerged as powerful alternatives to CNNs for various vision tasks, yet their token-based, attention-driven architecture makes interpreting their predictions challenging. Existing explainability methods, such as Grad-CAM and Attention Rollout, either fail to capture hierarchical semantic information or assume attention directly reflects importance, often leading to misleading explanations. We propose FocusViT, a novel explainability framework that integrates gradient-weighted attention attribution with validation-based, faithfulness-driven layer aggregation. By fusing attention maps with class-specific gradients and introducing per-head dynamic weighting, FocusViT highlights not only where the model attends but also how sensitive the prediction is to those attentions. Furthermore, our adaptive layer-skipping strategy ensures that only semantically meaningful layers contribute to the final explanation, enhancing both faithfulness and clarity. Extensive quantitative and qualitative evaluations on diverse benchmarks demonstrate that FocusViT improves over existing methods in faithfulness and sparsity, achieving competitive robustness and class sensitivity, and provides sharper, more reliable visual explanations for ViTs. The official implementation is publicly available at: https://github.com/game-sys/focusvit-aistats2026.git", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13826", "url": null, "sourceid": 650, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=8FwDBlYUFB", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11519, "modified": "2026-03-29T20:43:14.672072-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=8FwDBlYUFB", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "67", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13767, "uid": "ffeabd223de0d4eacb9a3e6e53e5448d", "name": "Explicit Density Approximation for Neural Implicit Samplers Using a Bernstein-Based Convex Divergence", "authors": [{"id": 10835, "fullname": "Jos\u00e9 Manuel de Frutos", "url": "http://virtual.aistats.org/api/miniconf/users/10835?format=json", "institution": "Universidad Carlos III Madrid"}, {"id": 5540, "fullname": "Pablo Mart\u00ednez Olmos", "url": "http://virtual.aistats.org/api/miniconf/users/5540?format=json", "institution": "Universidad Carlos III Madrid"}, {"id": 22811, "fullname": "Manuel V\u00e1zquez", "url": "http://virtual.aistats.org/api/miniconf/users/22811?format=json", "institution": "Universidad Carlos III de Madrid"}, {"id": 13067, "fullname": "Joaqu\u00edn M\u00edguez", "url": "http://virtual.aistats.org/api/miniconf/users/13067?format=json", "institution": "Universidad Carlos III de Madrid"}], "abstract": "Rank-based objectives such as the invariant statistical loss (ISL) are robust, likelihood-free tools for training implicit generative models. We propose \\emph{dual-ISL}, obtained by interchanging the roles of the target $p$ and model density $\\tilde p$ within ISL, which induces a \\emph{convex} optimization problem over model densities. We show that the associated rank-based discrepancy $d_K$ is \\emph{continuous} under weak and $L^1$ convergence and \\emph{convex} in its first argument, properties not shared by classical divergences such as KL or Wasserstein distances. Additionally, we prove that $d_K$ admits an $L^2$ interpretation: it is the projection of the density ratio $q=p/\\tilde p$ onto a Bernstein polynomial basis. This yields explicit truncation-error bounds, sharp convergence rates, and a closed-form expression for the truncated density approximation. To handle multivariate data, we further introduce a sliced dual-ISL via random one-dimensional projections that preserves both continuity and convexity. Empirically, across several benchmarks, dual-ISL delivers faster and smoother convergence than standard ISL and offers competitive, often superior, mode coverage relative to state-of-the-art implicit models (modern GAN baselines, including multi-critic setups), while providing an explicit density approximation.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13767", "url": null, "sourceid": 575, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=CyQG7D7FwT", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11460, "modified": "2026-03-29T20:43:12.207361-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=CyQG7D7FwT", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "67", "schedule_html": "", "latitude": null, "longitude": null, "related_events": ["http://virtual.aistats.org/api/miniconf/events/13918?format=json"], "related_events_ids": [13918]}, {"id": 13561, "uid": "6a2feef8ed6a9fe76d6b3f30f02150b4", "name": "Generalized Correlation Shifting for Lasso", "authors": [{"id": 22371, "fullname": "Izuru Miyazaki", "url": "http://virtual.aistats.org/api/miniconf/users/22371?format=json", "institution": "Graduate University for Advanced Studies"}, {"id": 22372, "fullname": "Hironori Fujisawa", "url": "http://virtual.aistats.org/api/miniconf/users/22372?format=json", "institution": "The Institute of Statistical Mathematics"}], "abstract": "The Lasso has been widely used in a high-dimensional setting, but its estimation accuracy may become inadequate when the covariates are highly correlated or when the number of covariates is extremely large. To overcome this problem, we propose a novel preconditioner that adaptively induces a low-rank structure in the design matrix. The proposed preconditioner achieves a higher probability of sign correctness under some conditions. We establish theoretical guarantees showing that our method dominates the standard Lasso, and we further demonstrate its superiority over the correlation shifting. To validate its practical effectiveness, we conducted numerical experiments on synthetic and semi-real datasets, and the proposed method presented better performance than existing methods.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13561", "url": null, "sourceid": 1092, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=YpQnKBXBkj", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11254, "modified": "2026-03-29T20:43:03.615066-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=YpQnKBXBkj", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "67", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13500, "uid": "abd815286ba1007abfbb8415b83ae2cf", "name": "From Guess2Graph: When and How Can Unreliable Experts Safely Boost Causal Discovery in Finite Samples?", "authors": [{"id": 22247, "fullname": "Sujai Hiremath", "url": "http://virtual.aistats.org/api/miniconf/users/22247?format=json", "institution": "Cornell University"}, {"id": 815, "fullname": "Dominik Janzing", "url": "http://virtual.aistats.org/api/miniconf/users/815?format=json", "institution": "Amazon"}, {"id": 13103, "fullname": "Philipp M. Faller", "url": "http://virtual.aistats.org/api/miniconf/users/13103?format=json", "institution": "Karlsruhe Institute of Technology"}, {"id": 816, "fullname": "Patrick Bloebaum", "url": "http://virtual.aistats.org/api/miniconf/users/816?format=json", "institution": "Amazon Research Tuebingen"}, {"id": 3571, "fullname": "Elke Kirschbaum", "url": "http://virtual.aistats.org/api/miniconf/users/3571?format=json", "institution": "Amazon Research"}, {"id": 12622, "fullname": "Shiva Kasiviswanathan", "url": "http://virtual.aistats.org/api/miniconf/users/12622?format=json", "institution": "Amazon AWS AI"}, {"id": 11076, "fullname": "Kyra Gan", "url": "http://virtual.aistats.org/api/miniconf/users/11076?format=json", "institution": "Cornell Tech, Cornell University"}], "abstract": "Causal discovery algorithms often perform poorly with limited samples. While integrating expert knowledge (including from LLMs)  as constraints promises to improve performance,  guarantees for existing methods require perfect predictions or uncertainty estimates, making them unreliable for practical use. We propose the Guess2Graph (G2G) framework, which uses expert guesses to guide the sequence of statistical tests rather than replacing them.  This maintains statistical consistency while enabling performance improvements. We develop two instantiations of G2G: PC-Guess, which augments the PC algorithm, and gPC-Guess, a learning-augmented variant designed to better leverage high-quality expert input. Theoretically, both preserve correctness regardless of expert error, with gPC-Guess provably outperforming its non-augmented counterpart in finite samples when experts are \"better than random\". Empirically, both show monotonic improvement with expert accuracy, with gPC-Guess achieving significantly stronger gains.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13500", "url": null, "sourceid": 632, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=fvVfoB1oko", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11193, "modified": "2026-03-29T20:43:01.306123-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=fvVfoB1oko", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "68", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13358, "uid": "71a58e8cb75904f24cde464161c3e766", "name": "Fast Private Adaptive Query Answering for Large Data Domains", "authors": [{"id": 11122, "fullname": "Miguel Fuentes", "url": "http://virtual.aistats.org/api/miniconf/users/11122?format=json", "institution": "University of Massachusetts Amherst"}, {"id": 21927, "fullname": "Brett Mullins", "url": "http://virtual.aistats.org/api/miniconf/users/21927?format=json", "institution": "University of Massachusetts at Amherst"}, {"id": 21928, "fullname": "Yingtai Xiao", "url": "http://virtual.aistats.org/api/miniconf/users/21928?format=json", "institution": "ByteDance Inc."}, {"id": 21929, "fullname": "Daniel Kifer", "url": "http://virtual.aistats.org/api/miniconf/users/21929?format=json", "institution": "Pennsylvania State University"}, {"id": 21930, "fullname": "Cameron Musco", "url": "http://virtual.aistats.org/api/miniconf/users/21930?format=json", "institution": "University of Massachusetts, Amherst"}, {"id": 1210, "fullname": "Daniel Sheldon", "url": "http://virtual.aistats.org/api/miniconf/users/1210?format=json", "institution": "University of Massachusetts, Amherst"}], "abstract": "Privately releasing marginals of a tabular dataset is a foundational problem in differential privacy.  However, state-of-the-art mechanisms suffer from a computational bottleneck when marginal estimates are reconstructed from noisy measurements. Recently, residual queries were introduced and shown to lead to highly efficient reconstruction in the batch query answering setting. We introduce new techniques to integrate residual queries into state-of-the-art adaptive mechanisms such as AIM. Our contributions include a novel conceptual framework for residual queries using multi-dimensional arrays, lazy updating strategies, and adaptive optimization of the per-round privacy budget allocation. Together these contributions reduce error, improve speed, and simplify residual query operations. We integrate these innovations into a new mechanism (AIM+GReM), which improves AIM by using fast residual-based reconstruction instead of a graphical model approach.  Our mechanism is orders of magnitude faster than the original framework and demonstrates competitive error and greatly improved scalability.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13358", "url": null, "sourceid": 1615, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=wkYTC6w2pK", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11051, "modified": "2026-03-29T20:42:55.685726-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=wkYTC6w2pK", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "68", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13507, "uid": "21fe5b8ba755eeaece7a450849876228", "name": "Frequency-Based Hyperparameter Selection in Games", "authors": [{"id": 19737, "fullname": "Aniket Sanyal", "url": "http://virtual.aistats.org/api/miniconf/users/19737?format=json", "institution": "Technical University of Munich"}, {"id": 22270, "fullname": "Baraah Sidahmed", "url": "http://virtual.aistats.org/api/miniconf/users/22270?format=json", "institution": "Universit\u00e4t des Saarlandes"}, {"id": 22271, "fullname": "Rebekka Burkholz", "url": "http://virtual.aistats.org/api/miniconf/users/22271?format=json", "institution": "CISPA Helmholtz Center for Information Security"}, {"id": 89, "fullname": "Tatjana Chavdarova", "url": "http://virtual.aistats.org/api/miniconf/users/89?format=json", "institution": "TU Wien"}], "abstract": "Learning in smooth games fundamentally differs from standard minimization due to rotational dynamics, which invalidate classical hyperparameter tuning strategies. Despite their practical importance, effective methods for tuning in games remain underexplored. A notable example is LookAhead (LA), which achieves strong empirical performance but introduces additional parameters that critically influence performance. We propose a principled approach to hyperparameter selection in games by leveraging frequency estimation of oscillatory dynamics. Specifically, we analyze oscillations both in continuous-time trajectories and through the spectrum of the discrete dynamics in the associated frequency-based space. Building on this analysis, we introduce *Modal LookAhead (MoLA)*, an extension of LA that selects the hyperparameters adaptively to a given problem. We provide convergence guarantees and demonstrate in experiments that MoLA accelerates training in both purely rotational games and mixed regimes, all with minimal computational overhead.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13507", "url": null, "sourceid": 1633, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=ehITpk83z7", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11200, "modified": "2026-03-29T20:43:01.584680-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=ehITpk83z7", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "70", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13707, "uid": "c54e7837e0cd0ced286cb5995327d1ab", "name": "Hyperbolic Learning with Supervision from any Granularity", "authors": [{"id": 21937, "fullname": "Mina GhadimiAtigh", "url": "http://virtual.aistats.org/api/miniconf/users/21937?format=json", "institution": "University of Amsterdam"}, {"id": 22687, "fullname": "Max van Spengler", "url": "http://virtual.aistats.org/api/miniconf/users/22687?format=json", "institution": "University of Amsterdam"}, {"id": 22688, "fullname": "Teng Long", "url": "http://virtual.aistats.org/api/miniconf/users/22688?format=json", "institution": "University of Amsterdam, University of Amsterdam"}, {"id": 22689, "fullname": "Melika Ayoughi", "url": "http://virtual.aistats.org/api/miniconf/users/22689?format=json", "institution": "University of Amsterdam, ILLC"}, {"id": 22690, "fullname": "Tejaswi Kasarla", "url": "http://virtual.aistats.org/api/miniconf/users/22690?format=json", "institution": "University of Amsterdam"}, {"id": 21938, "fullname": "Pascal Mettes", "url": "http://virtual.aistats.org/api/miniconf/users/21938?format=json", "institution": "University of Amsterdam"}], "abstract": "Supervised classification commonly follows a one-vs-rest paradigm where each sample belongs to one category from a set of independent classes. In real-world settings, classes are typically not independent, but organized hierarchically from coarse-grained to fine-grained. More pressingly, people naturally annotate at different levels of granularity, depending on their expertise, biases, or data quality. What should be the correct label of a picture of a bird? Is it \\emph{animal}, \\emph{bird}, \\emph{albatross}, or \\emph{Laysan albatross}? What if one annotator is an ornithologist and the other has little bird knowledge? Similarly, if two pictures of a \\emph{Laysan albatross} differ in blurriness, we tend to annotate blurry ones more generically, as we are unsure of details that differentiate classes at the finest levels. Currently, many annotations are removed, ignored, or reassigned because they do not match the required granularity. Instead of viewing the world as a flat, independent collection of concepts, this paper strives to perform supervised learning with labels at any granularity. We propose a hyperbolic embedding space, where classes are hierarchically organized as prototypes. We introduce a coarse-to-fine Busemann approach, where images are optimized to the correct region of the hyperbolic embedding space by projecting their labels -- which can be as precise or generic as desired -- to ideal prototypes on the boundary of the Poincar\u00e9 ball. Experiments show that our approach improves multi-granular classification and beats the current state-of-the-art, which views different granularities as independent, instead of a connected tree.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13707", "url": null, "sourceid": 1198, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=Hi2H3Logzx", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11400, "modified": "2026-03-29T20:43:09.599544-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=Hi2H3Logzx", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "73", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13395, "uid": "06a15eb1c3836723b53e4abca8d9b879", "name": "From Restless to Contextual: A Thresholding Bandit Reformulation for Finite-horizon Improvement", "authors": [{"id": 20636, "fullname": "Jiamin Xu", "url": "http://virtual.aistats.org/api/miniconf/users/20636?format=json", "institution": "Cornell University"}, {"id": 21999, "fullname": "Ivan Nazarov", "url": "http://virtual.aistats.org/api/miniconf/users/21999?format=json", "institution": "Causal Foundry"}, {"id": 22000, "fullname": "Aditya Rastogi", "url": "http://virtual.aistats.org/api/miniconf/users/22000?format=json", "institution": "Causal Foundry"}, {"id": 22001, "fullname": "Africa Perianez Santiago", "url": "http://virtual.aistats.org/api/miniconf/users/22001?format=json", "institution": "RIKEN"}, {"id": 11076, "fullname": "Kyra Gan", "url": "http://virtual.aistats.org/api/miniconf/users/11076?format=json", "institution": "Cornell Tech, Cornell University"}], "abstract": "This paper addresses the poor finite-horizon performance of existing online *restless bandit* (RB) algorithms, which stems from the prohibitive sample complexity of learning a full  *Markov decision process* (MDP) for each agent. We argue that superior finite-horizon performance requires *rapid convergence* to a *high-quality* policy. Thus motivated, we introduce a reformulation of online RBs as a *budgeted thresholding contextual bandit*, which simplifies the learning problem by encoding long-term state transitions into a scalar reward. We prove the first non-asymptotic optimality of an oracle policy for a simplified finite-horizon setting. We  propose a practical learning policy under a heterogeneous-agent, multi-state setting, and show that it achieves a sublinear regret, achieving *faster convergence* than existing methods. This directly translates to higher cumulative reward, as empirically validated by significant gains over state-of-the-art algorithms in large-scale heterogeneous environments. The code is provided in [github](https://github.com/jamie01713/EGT). Our work provides a new pathway for achieving practical, sample-efficient learning in finite-horizon RBs.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13395", "url": null, "sourceid": 1844, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=qSoOUukzag", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11088, "modified": "2026-03-29T20:42:57.191500-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=qSoOUukzag", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "73", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13390, "uid": "db957c626a8cd7a27231adfbf51e20eb", "name": "Functional Properties of the Focal-Entropy", "authors": [{"id": 19920, "fullname": "Jaimin Shah", "url": "http://virtual.aistats.org/api/miniconf/users/19920?format=json", "institution": "University of Minnesota, Twin cities"}, {"id": 21989, "fullname": "Martina Cardone", "url": "http://virtual.aistats.org/api/miniconf/users/21989?format=json", "institution": "University of Minnesota - Twin Cities"}, {"id": 21990, "fullname": "Alex Dytso", "url": "http://virtual.aistats.org/api/miniconf/users/21990?format=json", "institution": "Qualcomm Inc, QualComm"}], "abstract": "The focal-loss has become a widely used alternative to cross-entropy in class-imbalanced classification problems, particularly in computer vision. Despite its empirical success, a systematic information-theoretic study of the focal-loss remains incomplete. In this work, we adopt a distributional viewpoint and study the focal-entropy, a focal-loss analogue of the cross-entropy. Our analysis establishes conditions for finiteness, convexity, and continuity of the focal-entropy, and provides various asymptotic characterizations. We prove the existence and uniqueness of the focal-entropy minimizer, describe its structure, and show that it can depart significantly from the data distribution. In particular, we rigorously show that the focal-loss amplifies mid-range probabilities, suppresses high-probability outcomes, and, under extreme class imbalance, induces an over-suppression regime in which very small probabilities are further diminished. These results, which are also experimentally validated, offer a theoretical foundation for understanding the focal-loss and clarify the trade-offs that it introduces when applied to imbalanced learning tasks.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13390", "url": null, "sourceid": 1788, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=ronNpzeyCF", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11083, "modified": "2026-03-29T20:42:56.964461-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=ronNpzeyCF", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "74", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13361, "uid": "3d779cae2d46cf6a8a99a35ba4167977", "name": "Hyperbolic Part-Whole Image Segmentation", "authors": [{"id": 21936, "fullname": "Mikhail Vlasenko", "url": "http://virtual.aistats.org/api/miniconf/users/21936?format=json", "institution": "University of Amsterdam"}, {"id": 21937, "fullname": "Mina GhadimiAtigh", "url": "http://virtual.aistats.org/api/miniconf/users/21937?format=json", "institution": "University of Amsterdam"}, {"id": 21938, "fullname": "Pascal Mettes", "url": "http://virtual.aistats.org/api/miniconf/users/21938?format=json", "institution": "University of Amsterdam"}], "abstract": "Semantic segmentation typically focuses on pixel-level classification at the object level. Yet, objects naturally decompose into parts and subparts, mirroring human visual perception. In this work, we introduce a hyperbolic prototypical segmentation framework capable of simultaneously representing multiple granularity levels within a unified embedding space. Leveraging hyperbolic geometry's unique capacity to model hierarchies effectively, we propose to embed class prototypes within the Poincar\u00e9 ball. We introduce a tree-aware prototype initialization strategy and a distortion-*p* loss that together yield improved hierarchical embeddings. Furthermore, we derive an optimized formulation of the hyperbolic distance function, enabling tractable inference for dense prediction tasks. A shared transformer encoder paired with separate hyperbolic heads allows efficient multi-level segmentation from a single model. Experiments on the recently introduced SubPartImageNet show that our approach (i) improves over the state-of-the-art, especially at the *subpart* and *part* levels, at a fraction of the number of parameters, (ii) enables zero-shot generalization, and (iii) allows for transfer from part- to object-level predictions without object-level supervision. All code will be made publicly available.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13361", "url": null, "sourceid": 1336, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=wNwJTwpLio", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11054, "modified": "2026-03-29T20:42:55.790554-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=wNwJTwpLio", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "74", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13344, "uid": "9cfdf10e8fc047a44b08ed031e1f0ed1", "name": "Graphon Mixtures", "authors": [{"id": 21905, "fullname": "Sevvandi Kandanaarachchi", "url": "http://virtual.aistats.org/api/miniconf/users/21905?format=json", "institution": "CSIRO - Australia&#x27;s National Science Agency"}, {"id": 9726, "fullname": "Cheng Soon Ong", "url": "http://virtual.aistats.org/api/miniconf/users/9726?format=json", "institution": "CSIRO and Australian National University"}], "abstract": "Social networks have a small number of large hubs, and a large number of small dense communities. We propose a generative model that captures both hub and dense structures. Based on recent results about graphons on line graphs, our model is a graphon mixture,   enabling us to generate sequences of graphs where each graph is a combination of sparse and dense graphs. We propose a new condition on sparse graphs (the max-degree), which enables us to identify hubs. We show theoretically that we can estimate the normalized degree of the hubs, as well as estimate the graphon corresponding to sparse components of graph mixtures. We illustrate our approach on synthetic data and real-world networks, showing the benefits of explicitly modeling sparse graphs.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13344", "url": null, "sourceid": 226, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=xy9kqlOHAw", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11037, "modified": "2026-03-29T20:42:55.077306-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=xy9kqlOHAw", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "75", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13881, "uid": "7647966b7343c29048673252e490f736", "name": "Happiness as a Measure of Fairness", "authors": [{"id": 11048, "fullname": "Georg Pichler", "url": "http://virtual.aistats.org/api/miniconf/users/11048?format=json", "institution": "TU Wien"}, {"id": 23043, "fullname": "Marco Romanelli", "url": "http://virtual.aistats.org/api/miniconf/users/23043?format=json", "institution": "Hofstra University"}, {"id": 19338, "fullname": "Pablo Piantanida", "url": "http://virtual.aistats.org/api/miniconf/users/19338?format=json", "institution": "MILA &amp; ILLS, CNRS Universit\u00e9 Paris-Saclay"}], "abstract": "In this paper, we propose a novel fairness framework grounded in the concept of _happiness_, a measure of the utility each group gains from decision outcomes. By capturing fairness through this intuitive lens, we not only offer a more human-centered approach, but also one that is mathematically rigorous: In order to compute the optimal, fair post-processing strategy, only a linear program needs to be solved. This makes our method both efficient and scalable with existing optimization tools. Furthermore, it unifies and extends several well-known fairness definitions, and our empirical results highlight its practical strengths across diverse scenarios.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13881", "url": null, "sourceid": 89, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=3e6j5mzClz", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11574, "modified": "2026-03-29T20:43:17.110877-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=3e6j5mzClz", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "76", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13502, "uid": "754dda4b1ba34c6fa89716b85d68532b", "name": "i-IF-Learn: Iterative Feature Selection and Unsupervised Learning for High-Dimensional Complex Data", "authors": [{"id": 22251, "fullname": "Chen Ma", "url": "http://virtual.aistats.org/api/miniconf/users/22251?format=json", "institution": "Southern University of Science and Technology"}, {"id": 22252, "fullname": "Wanjie Wang", "url": "http://virtual.aistats.org/api/miniconf/users/22252?format=json", "institution": "National University of Singapore"}, {"id": 22253, "fullname": "Shuhao Fan", "url": "http://virtual.aistats.org/api/miniconf/users/22253?format=json", "institution": "national university of singaore, National University of Singapore"}], "abstract": "Unsupervised learning of high-dimensional data is challenging due to irrelevant or noisy features obscuring underlying structures. It's common that only a few features, called the influential features, meaningfully define the clusters. Recovering these influential features is helpful in data interpretation and clustering. We propose i-IF-Learn, an iterative unsupervised framework that jointly performs feature selection and clustering. Our core innovation is an adaptive feature selection statistic that effectively combines pseudo-label supervision with unsupervised signals, dynamically adjusting based on intermediate label reliability to mitigate error propagation common in iterative frameworks. Leveraging low-dimensional embeddings (PCA or Laplacian eigenmaps) followed by $k$-means, i-IF-Learn simultaneously outputs influential feature subset and clustering labels. Numerical experiments on gene microarray and single-cell RNA-seq datasets show that i-IF-Learn significantly surpasses classical and deep clustering baselines. Furthermore, using our selected influential features as preprocessing substantially enhances downstream deep models such as DeepCluster, UMAP, and VAE, highlighting the importance and effectiveness of targeted feature selection.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13502", "url": null, "sourceid": 1091, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=fgj6X8ADL3", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11195, "modified": "2026-03-29T20:43:01.385042-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=fgj6X8ADL3", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "76", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13407, "uid": "d947bf06a885db0d477d707121934ff8", "name": "Harnessing the Power of Reinforcement Learning for Adaptive MCMC", "authors": [{"id": 14350, "fullname": "Congye Wang", "url": "http://virtual.aistats.org/api/miniconf/users/14350?format=json", "institution": "Newcastle University"}, {"id": 22024, "fullname": "Matthew Fisher", "url": "http://virtual.aistats.org/api/miniconf/users/22024?format=json", "institution": "Newcastle University, UK"}, {"id": 1703, "fullname": "Heishiro Kanagawa", "url": "http://virtual.aistats.org/api/miniconf/users/1703?format=json", "institution": "Newcastle University"}, {"id": 18344, "fullname": "Wilson Ye Chen", "url": "http://virtual.aistats.org/api/miniconf/users/18344?format=json", "institution": "University of Sydney, Australia"}, {"id": 612, "fullname": "Chris Oates", "url": "http://virtual.aistats.org/api/miniconf/users/612?format=json", "institution": "Newcastle"}], "abstract": "Sampling algorithms drive probabilistic machine learning, and recent years have seen an explosion in the diversity of tools for this task.  However, the increasing sophistication of sampling algorithms is correlated with an increase in the tuning burden.  There is now a greater need than ever to treat the tuning of samplers as a learning task in its own right.  In a conceptual breakthrough, Wang et al. (2025) formulated Metropolis-Hastings as a Markov decision process, opening up the possibility for adaptive tuning using reinforcement learning (RL).  Their emphasis was on theoretical foundations; realising the practical benefit of reinforcement learning Metropolis-Hastings (RLMH) was left for subsequent work. The purpose of this paper is twofold:  First, we observe the surprising result that natural choices of reward, such as the acceptance rate, or the expected squared jump distance, provide insufficient signal for training RLMH.  Instead, we propose a novel reward based on the contrastive divergence, whose superior performance in the context of RLMH is demonstrated.  Second, we explore the potential of RLMH and present adaptive gradient-based samplers that balance flexibility of the Markov transition kernel with learnability of the associated RL task.  A comprehensive simulation study using the posteriordb benchmark supports the practical effectiveness of RLMH.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13407", "url": null, "sourceid": 274, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=pLmDmkVYxi", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11100, "modified": "2026-03-29T20:42:57.654340-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=pLmDmkVYxi", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "77", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13708, "uid": "05311655a15b75fab86956663e1819cd", "name": "In-Context Function Learning in Large Language Models", "authors": [{"id": 22691, "fullname": "Elif Akata", "url": "http://virtual.aistats.org/api/miniconf/users/22691?format=json", "institution": "Helmholtz Zentrum M\u00fcnchen"}, {"id": 22692, "fullname": "Konstantinos Voudouris", "url": "http://virtual.aistats.org/api/miniconf/users/22692?format=json", "institution": "UK Department for Science, Innovation and Technology"}, {"id": 14254, "fullname": "Vincent Fortuin", "url": "http://virtual.aistats.org/api/miniconf/users/14254?format=json", "institution": "Helmholtz AI &amp; TU Munich"}, {"id": 22693, "fullname": "Eric Schulz", "url": "http://virtual.aistats.org/api/miniconf/users/22693?format=json", "institution": "Max Planck Institute for Biological Cybernetics"}], "abstract": "Large language models (LLMs) can learn from a few demonstrations provided at inference time. We study this in-context learning phenomenon through the lens of Gaussian processes (GPs). We build controlled experiments where models observe sequences of function samples drawn from known GP priors. We evaluate prediction error in relation to the number of demonstrations and compare against two principled references: (i) an empirical GP-regression learner that gives a lower bound on achievable error, and (ii) the expected error of a 1-nearest-neighbor (1-NN) rule, which gives a data-driven upper bound. Across model sizes, we find that LLM learning curves are strongly influenced by the function-generating kernels and approach the GP lower bound as the number of demonstrations increases. We then study the inductive biases of these models using a likelihood-based analysis. We find that LLM predictions are most likely under less smooth GP kernels. Finally, we explore whether post-training can shift these inductive biases and improve sample-efficiency on functions sampled from GPs with smoother kernels. We find that both reinforcement learning and supervised fine-tuning can effectively shift inductive biases in the direction of the training data. Together, our framework quantifies the extent to which LLMs behave like GP learners and provides tools for steering their inductive biases for continuous function learning tasks.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13708", "url": null, "sourceid": 1338, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=HemJLYgBIi", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11401, "modified": "2026-03-29T20:43:09.639016-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=HemJLYgBIi", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "78", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13373, "uid": "f9d1152547c0bde01830b7e8bd60024c", "name": "Gradient Descent with Provably Tuned Learning-rate Schedules", "authors": [{"id": 21963, "fullname": "Dravyansh Sharma", "url": "http://virtual.aistats.org/api/miniconf/users/21963?format=json", "institution": "Toyota Technological Institute at Chicago"}], "abstract": "Gradient-based iterative optimization methods are the workhorse of modern machine learning. They crucially rely on careful tuning of parameters like learning rate and momentum. However, one typically sets them using heuristic approaches without formal near-optimality guarantees. Recent work studies how to learn a good step-size in gradient descent. However, like most of the literature with theoretical guarantees for gradient-based optimization, their results rely on strong assumptions on the function class including convexity and smoothness which do not hold in typical applications. In this work, we develop novel analytical tools for provably tuning hyperparameters in gradient-based algorithms that apply to non-convex and non-smooth functions. We obtain matching sample complexity bounds for learning the step-size in gradient descent shown for smooth, convex functions in prior work (up to logarithmic factors) but for a much broader class of functions. Our analysis applies to gradient descent on neural networks with commonly used activation functions (including ReLU, sigmoid and tanh). We extend our framework to tuning multiple hyperparameters, including tuning the learning rate schedule, simultaneously tuning momentum and step-size, and pre-training the initialization vector. Our approach can be used to bound the sample complexity for minimizing both the validation loss as well as the number of gradient descent iterations.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13373", "url": null, "sourceid": 1861, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=u0hZEMV8tY", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11066, "modified": "2026-03-29T20:42:56.203738-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=u0hZEMV8tY", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "78", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13494, "uid": "6832a7b24bc06775d02b7406880b93fc", "name": "GRANITE: A Generalized Regional Framework for Identifying Agreement in Feature-Based Explanations", "authors": [{"id": 14488, "fullname": "Julia Herbinger", "url": "http://virtual.aistats.org/api/miniconf/users/14488?format=json", "institution": "Leibniz Institute for Agricultural Engineering and Bioeconomy"}, {"id": 22233, "fullname": "Gabriel Laberge", "url": "http://virtual.aistats.org/api/miniconf/users/22233?format=json", "institution": "Thales"}, {"id": 22234, "fullname": "Maximilian Muschalik", "url": "http://virtual.aistats.org/api/miniconf/users/22234?format=json", "institution": "Institute of Computer Science, Ludwig-Maximilians-Universit\u00e4t M\u00fcnchen"}, {"id": 22235, "fullname": "Yann Pequignot", "url": "http://virtual.aistats.org/api/miniconf/users/22235?format=json", "institution": "Universite Laval, Universit\u00e9 Laval"}, {"id": 9239, "fullname": "Marvin N. Wright", "url": "http://virtual.aistats.org/api/miniconf/users/9239?format=json", "institution": "Leibniz Institute for Prevention Research and Epidemiology - BIPS"}, {"id": 13226, "fullname": "Fabian Fumagalli", "url": "http://virtual.aistats.org/api/miniconf/users/13226?format=json", "institution": "Bielefeld University"}], "abstract": "Feature-based explanation methods aim to quantify how features influence the model's behavior, either locally or globally, but different methods often disagree, producing conflicting explanations. This disagreement arises primarily from two sources: how feature interactions are handled and how feature dependencies are incorporated. We propose GRANITE, a generalized regional explanation framework that partitions the feature space into regions where interaction and distribution influences are minimized. This approach aligns different explanation methods, yielding more consistent and interpretable explanations. GRANITE unifies existing regional approaches, extends them to feature groups, and introduces a recursive partitioning algorithm to estimate such regions. We demonstrate its effectiveness on real-world datasets, providing a practical tool for consistent and interpretable feature explanations.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13494", "url": null, "sourceid": 2198, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=gAO7AFSTJD", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11187, "modified": "2026-03-29T20:43:01.057077-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=gAO7AFSTJD", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "79", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13834, "uid": "3a0772443a0739141292a5429b952fe6", "name": "High-dimensional Learning with Noisy Labels", "authors": [{"id": 22968, "fullname": "Aymane Firdoussi", "url": "http://virtual.aistats.org/api/miniconf/users/22968?format=json", "institution": "Ecole Normale Sup\u00e9rieure de Cachan"}, {"id": 617, "fullname": "Mohamed El Amine Seddik", "url": "http://virtual.aistats.org/api/miniconf/users/617?format=json", "institution": "Technology Innovation Institute"}], "abstract": "This paper provides theoretical insights into high-dimensional binary classification with class-conditional noisy labels.      Specifically, we study the behavior of a linear classifier with a label noisiness aware loss function, when both the dimension of data $p$ and the sample size $n$ are large and comparable.     Relying on random matrix theory by supposing a Gaussian mixture data model, the performance of the linear classifier when $p,n\\to \\infty$ is shown to converge towards a limit, involving scalar statistics of the data.      Importantly, our findings show that the low-dimensional intuitions to handle label noise do not hold in high-dimension, in the sense that the optimal classifier in low-dimension dramatically fails in high-dimension.      Based on our derivations, we design an optimized method that is shown to be provably more efficient in handling noisy labels in high dimensions.     Our theoretical conclusions are further confirmed by experiments on real datasets, where we show that our optimized approach outperforms the considered baselines.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13834", "url": null, "sourceid": 559, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=7hCzeXsTrs", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11527, "modified": "2026-03-29T20:43:14.989349-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=7hCzeXsTrs", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "81", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13794, "uid": "1e1d184167ca7676cf665225e236a3d2", "name": "Incoherence in goal-conditioned autoregressive models", "authors": [{"id": 22884, "fullname": "Jacek Karwowski", "url": "http://virtual.aistats.org/api/miniconf/users/22884?format=json", "institution": "University of Oxford"}, {"id": 22885, "fullname": "Raymond Douglas", "url": "http://virtual.aistats.org/api/miniconf/users/22885?format=json", "institution": "ACS Research"}], "abstract": "We investigate mathematically the notion of incoherence: a structural issue with reinforcement learning policies derived by naive goal-conditioning of autoregressive models. We focus on the process of re-training models on their own actions, that is, fine-tuning offline-learned policies with online RL. We prove that it decreases incoherence and leads to an improvement in return, and we aim to characterise the resulting trajectory of policies. By re-framing standard notions of control-as-inference and soft Q learning, we establish a three-way correspondence with two other ways of understanding the iterative re-training process: as folding the posterior into the reward and, in the deterministic case, as decreasing the temperature parameter; the correspondence has computational content via the training-inference trade-off. Through soft-conditioning generative models, we discuss the link between incoherence and the effective horizon of Laidlaw et al. (2024).", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13794", "url": null, "sourceid": 1169, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=BBsvuE0KV1", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11487, "modified": "2026-03-29T20:43:13.278858-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=BBsvuE0KV1", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "81", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13762, "uid": "2321994d85d661d792223f647000c65f", "name": "How to approximate inference with subtractive mixture models", "authors": [{"id": 22791, "fullname": "Lena Zellinger", "url": "http://virtual.aistats.org/api/miniconf/users/22791?format=json", "institution": "University of Edinburgh"}, {"id": 23260, "fullname": "Nicola Branchini", "url": "http://virtual.aistats.org/api/miniconf/users/23260?format=json", "institution": "The University of Warwick"}, {"id": 22792, "fullname": "Lennert De Smet", "url": "http://virtual.aistats.org/api/miniconf/users/22792?format=json", "institution": "KU Leuven"}, {"id": 4628, "fullname": "Victor Elvira", "url": "http://virtual.aistats.org/api/miniconf/users/4628?format=json", "institution": "University of Edinburgh"}, {"id": 10899, "fullname": "Nikolay Malkin", "url": "http://virtual.aistats.org/api/miniconf/users/10899?format=json", "institution": "Mila"}, {"id": 5113, "fullname": "Antonio Vergari", "url": "http://virtual.aistats.org/api/miniconf/users/5113?format=json", "institution": "University of Edinburgh"}], "abstract": "Classical mixture models (MMs) are widely used tractable proposals for approximate inference settings such as variational inference (VI) and importance sampling (IS). Recently, mixture models with negative coefficients, called subtractive mixture models (SMMs), have been proposed as a potentially more expressive alternative. However, how to effectively use SMMs for VI and IS is still an open question as they do not provide latent variable semantics and therefore cannot use sampling schemes for classical MMs.  In this work, we study how to circumvent this issue by designing several expectation estimators for IS and learning schemes for VI with SMMs, and we empirically evaluate them for distribution approximation. Finally, we discuss the additional challenges in estimation stability and learning efficiency that they carry and propose ways to overcome them.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13762", "url": null, "sourceid": 2342, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=DXEx6DYinL", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11455, "modified": "2026-03-29T20:43:12.032176-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=DXEx6DYinL", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "82", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13720, "uid": "08b255a5d42b89b0585260b6f2360bdd", "name": "High-dimensional Level Set Estimation with Trust Regions and Double Acquisition Functions", "authors": [{"id": 19897, "fullname": "Giang Ngo", "url": "http://virtual.aistats.org/api/miniconf/users/19897?format=json", "institution": "Deakin Applied Artificial Intelligence Initiative"}, {"id": 22716, "fullname": "Dat Phan Trong", "url": "http://virtual.aistats.org/api/miniconf/users/22716?format=json", "institution": "Deakin University"}, {"id": 22717, "fullname": "Dang Nguyen", "url": "http://virtual.aistats.org/api/miniconf/users/22717?format=json", "institution": "Deakin University"}, {"id": 4443, "fullname": "Sunil Gupta", "url": "http://virtual.aistats.org/api/miniconf/users/4443?format=json", "institution": "Deakin University, Australia"}], "abstract": "Level set estimation (LSE) classifies whether an unknown function's value exceeds a specified threshold for given inputs, a fundamental problem in many real-world applications. In active learning settings with limited initial data, we aim to iteratively acquire informative points to construct an accurate classifier for this task. In high-dimensional spaces, this becomes challenging where the search volume grows exponentially with increasing dimensionality. We propose TRLSE, an algorithm for high-dimensional LSE, which identifies and refines regions near the threshold boundary with dual acquisition functions operating at both global and local levels. We provide a theoretical analysis of TRLSE's accuracy and show its superior sample efficiency against existing methods through extensive evaluations on multiple synthetic and real-world LSE problems.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13720", "url": null, "sourceid": 590, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=GAg3V73w1D", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11413, "modified": "2026-03-29T20:43:10.180734-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=GAg3V73w1D", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "82", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13675, "uid": "1cc3633c579a90cfdd895e64021e2163", "name": "Information-theoretic error bounds for source localization in neural sensing", "authors": [{"id": 22600, "fullname": "Leighton Barnes", "url": "http://virtual.aistats.org/api/miniconf/users/22600?format=json", "institution": "CMU, Carnegie Mellon University"}, {"id": 22601, "fullname": "Yuxin Guo", "url": "http://virtual.aistats.org/api/miniconf/users/22601?format=json", "institution": "Carnegie Mellon University"}, {"id": 21990, "fullname": "Alex Dytso", "url": "http://virtual.aistats.org/api/miniconf/users/21990?format=json", "institution": "Qualcomm Inc, QualComm"}, {"id": 22602, "fullname": "Pulkit Grover", "url": "http://virtual.aistats.org/api/miniconf/users/22602?format=json", "institution": "Carnegie Mellon University"}], "abstract": "We formulate a point-source localization problem in $d$ dimensions, where a source inside the ball of radius $R$ emits a signal that is picked up by various sensors located at the surface of the ball. For $d=3$, this can model problems in neural sensing, where a net of electroencephalogram (EEG) or magnetoencephalogram (MEG) sensors try to locate the source of a distinct neural event such as a seizure. For a power law decay model with exponent $\\alpha>0$ for the sensors, we obtain a lower bound on the minimax risk for localizing the source that is asymptotically $\\frac{d^2\\sigma^2R^{2\\alpha+2}}{n\\alpha^2PK}$ under mean-squared error loss, where $\\sigma^2$ is the noise variance, $P$ is the signal power, $K$ is the number of sensors, and $n$ is the number of independent measurements. In the case $d\\leq 2(\\alpha+1)$ with uniformly distributed sensor locations, we then give a matching upper bound, including getting the exact constant correct, for the asymptotic minimax rate in a neighborhood of the origin. We show that there is a phase transition at $d=2(\\alpha+2)$, above which a certain Fisher information quantity is minimized at the boundary of the ball, and below which it is minimized at the origin. At the critical dimension $d=2(\\alpha+2)$, the Fisher information is constant throughout the entire parameter space. For the special case $d=3$, we supplement and compare this information-theoretic analysis with a simulated forward EEG model that uses a realistic head model derived from population-averaged magnetic resonance imaging data.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13675", "url": null, "sourceid": 930, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=L98g0z9wbR", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11368, "modified": "2026-03-29T20:43:08.225400-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=L98g0z9wbR", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "82", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13749, "uid": "f0fcf351df4eb6786e9bb6fc4e2dee02", "name": "Improving Adaptive Moment Optimization via Preconditioner Diagonalization", "authors": [{"id": 14523, "fullname": "Son Nguyen", "url": "http://virtual.aistats.org/api/miniconf/users/14523?format=json", "institution": "The University of Texas at Austin"}, {"id": 18631, "fullname": "Bo Liu", "url": "http://virtual.aistats.org/api/miniconf/users/18631?format=json", "institution": "University of Texas, Austin"}, {"id": 18644, "fullname": "Lizhang Chen", "url": "http://virtual.aistats.org/api/miniconf/users/18644?format=json", "institution": "University of Texas at Austin"}, {"id": 18306, "fullname": "qiang liu", "url": "http://virtual.aistats.org/api/miniconf/users/18306?format=json", "institution": "University of Texas, Austin"}], "abstract": "Modern adaptive optimization methods, such as Adam and its variants, have emerged as the most widely used tools in deep learning over recent years. These algorithms offer automatic mechanisms for dynamically adjusting the update step based on estimates of gradient statistics. Compared to traditional algorithms like Stochastic Gradient Descent, these adaptive methods are typically more robust to model scale and hyperparameter tuning. However, the gradient statistics employed by these methods often do not leverage sufficient gradient covariance information, leading to suboptimal updates in certain directions of the parameter space and potentially slower convergence. In this work, we keep track of such covariance statistics in the form of a structured preconditioner matrix. Unlike other works, our approach does not apply direct approximations to estimate this matrix. We instead implement an invertible transformation that maps the preconditioner matrix into a new space where it becomes approximately diagonal. This enables a diagonal approximation of the preconditioner matrix in the transformed space, offering several computational advantages. Empirical results show that our approach can substantially enhance the convergence speed of the modern adaptive optimizers. Notably, for large language models like LLaMA, we can achieve a speedup of 2x compared to the baseline Adam. Additionally, our method can be integrated with memory-efficient optimizers like Adafactor to manage computational overhead.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13749", "url": null, "sourceid": 1884, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=EVYuPAAPQe", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11442, "modified": "2026-03-29T20:43:11.468759-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=EVYuPAAPQe", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "84", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13565, "uid": "b4a528955b84f584974e92d025a75d1f", "name": "Implicit Updates for Average-Reward Temporal Difference Learning", "authors": [{"id": 19427, "fullname": "Hwanwoo Kim", "url": "http://virtual.aistats.org/api/miniconf/users/19427?format=json", "institution": "Duke University"}, {"id": 22384, "fullname": "Dongkyu Cho", "url": "http://virtual.aistats.org/api/miniconf/users/22384?format=json", "institution": "Duke University"}, {"id": 21915, "fullname": "Eric Laber", "url": "http://virtual.aistats.org/api/miniconf/users/21915?format=json", "institution": "Duke University"}], "abstract": "Temporal difference (TD) learning is a cornerstone of reinforcement learning. In the average-reward setting, standard TD($\\lambda$) is  highly sensitive to the choice of step-size and thus requires careful tuning to maintain numerical stability. We introduce  average-reward implicit TD($\\lambda$), which employs an implicit fixed point update to provide data-adaptive stabilization while  preserving the per iteration computational complexity of standard average-reward TD($\\lambda$). In contrast to prior finite-time  analyses of average-reward TD($\\lambda$), which impose restrictive step-size conditions, we establish finite-time error bounds for the  implicit variant under substantially weaker step-size requirements. Empirically, average-reward implicit TD($\\lambda$) operates  reliably over a much broader range of step-sizes and exhibits markedly improved numerical stability. This enables more efficient  policy evaluation and policy learning, highlighting its effectiveness as a robust alternative to average-reward TD($\\lambda$).", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13565", "url": null, "sourceid": 701, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=YXVd22hh0r", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11258, "modified": "2026-03-29T20:43:03.763474-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=YXVd22hh0r", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "84", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13515, "uid": "4e8412ad48562e3c9934f45c3e144d48", "name": "Kernel Treatment Effects with Adaptively Collected Data", "authors": [{"id": 14599, "fullname": "Houssam Zenati", "url": "http://virtual.aistats.org/api/miniconf/users/14599?format=json", "institution": "UCL Gatsby"}, {"id": 16297, "fullname": "Bariscan Bozkurt", "url": "http://virtual.aistats.org/api/miniconf/users/16297?format=json", "institution": "Gatsby Computational Neuroscience Unit"}, {"id": 3537, "fullname": "Arthur Gretton", "url": "http://virtual.aistats.org/api/miniconf/users/3537?format=json", "institution": "Gatsby Computational Neuroscience Unit"}], "abstract": "Adaptive experiments improve efficiency by adjusting treatment assignments based on past outcomes, but this adaptivity breaks the i.i.d. assumptions that underpins classical asymptotics. At the same time, many questions of interest are distributional, extending beyond average effects. Kernel treatment effects (KTE) provide a flexible framework by representing counterfactual outcome distributions in an RKHS and comparing them via kernel distances. We present the first kernel-based framework for distributional inference under adaptive data collection. Our method combines doubly robust scores with variance stabilization to ensure asymptotic normality via a Hilbert-space martingale CLT, and introduces a sample-fitted stabilized test with valid type-I error. Experiments show it is well calibrated and effective for both mean shifts and higher-moment differences, outperforming adaptive baselines limited to scalar effects.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13515", "url": null, "sourceid": 1481, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=di2b5POqcE", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11208, "modified": "2026-03-29T20:43:01.865313-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=di2b5POqcE", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "85", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13616, "uid": "85fc37b18c57097425b52fc7afbb6969", "name": "Improving Coverage in Combined Prediction Sets with Weighted p-values", "authors": [{"id": 22496, "fullname": "Gina Wong", "url": "http://virtual.aistats.org/api/miniconf/users/22496?format=json", "institution": "Johns Hopkins University"}, {"id": 22497, "fullname": "Drew Prinster", "url": "http://virtual.aistats.org/api/miniconf/users/22497?format=json", "institution": "Johns Hopkins University"}, {"id": 1121, "fullname": "Suchi Saria", "url": "http://virtual.aistats.org/api/miniconf/users/1121?format=json", "institution": "Johns Hopkins University"}, {"id": 22498, "fullname": "Rama Chellappa", "url": "http://virtual.aistats.org/api/miniconf/users/22498?format=json", "institution": "Johns Hopkins University"}, {"id": 9722, "fullname": "Anqi Liu", "url": "http://virtual.aistats.org/api/miniconf/users/9722?format=json", "institution": "Johns Hopkins University"}], "abstract": "Conformal prediction quantifies the uncertainty of machine learning models by augmenting point predictions with valid prediction sets. For complex scenarios involving multiple trials, models, or data sources, conformal prediction sets can be aggregated to create a prediction set that captures the overall uncertainty, often improving precision. However, aggregating multiple prediction sets with individual $1-\\alpha$ coverage inevitably weakens the overall guarantee, typically resulting in $1-2\\alpha$ worst-case coverage. In this work, we propose a framework for the *weighted aggregation of prediction sets*, where weights are assigned to each prediction set based on their contribution. Our framework offers flexible control over how the sets are aggregated, achieving tighter coverage bounds that interpolate between the $1-2\\alpha$ guarantee of the combined models and the $1-\\alpha$ guarantee of an individual model depending on the distribution of weights. Importantly, our framework generalizes to data-dependent weights, as we derive a procedure for weighted aggregation that maintains finite-sample validity even when the weights depend on the data. This extension makes our framework broadly applicable to settings where weights are learned, such as mixture-of-experts (MoE), and we demonstrate through experiments in the MoE setting that our methods achieve adaptive coverage.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13616", "url": null, "sourceid": 621, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=SqFPcmIslQ", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11309, "modified": "2026-03-29T20:43:05.781052-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=SqFPcmIslQ", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "86", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13354, "uid": "978d76676f5e7918f81d28e7d092ca0d", "name": "Is Supervised Learning Really That Different from Unsupervised?", "authors": [{"id": 21921, "fullname": "Oskar Allerbo", "url": "http://virtual.aistats.org/api/miniconf/users/21921?format=json", "institution": "KTH Royal Institute of Technology"}, {"id": 21922, "fullname": "Thomas Sch\u00f6n", "url": "http://virtual.aistats.org/api/miniconf/users/21922?format=json", "institution": "Uppsala University"}], "abstract": "We demonstrate how supervised learning can be decomposed into a two-stage procedure, where (1) all model parameters are selected in an unsupervised manner, and (2) the outputs y are added to the model, without changing the parameter values. This is achieved by a new model selection criterion that - in contrast to cross-validation - can be used also without access to y. For linear ridge regression, we bound the asymptotic out-of-sample risk of our method in terms of the optimal asymptotic risk. We also demonstrate that versions of linear and kernel ridge regression, smoothing splines, k-nearest neighbors, random forests, and neural networks, trained without access to y, perform similarly to their standard y-based counterparts. Hence, our results suggest that the difference between supervised and unsupervised learning is less fundamental than it may appear.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13354", "url": null, "sourceid": 2099, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=x09RDZBfdc", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11047, "modified": "2026-03-29T20:42:55.498253-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=x09RDZBfdc", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "87", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13804, "uid": "ddb30680a691d157187ee1cf9e896d03", "name": "Incorporating Expert Knowledge into Bayesian Causal Discovery of Mixtures of Directed Acyclic Graphs", "authors": [{"id": 22898, "fullname": "Zachris Bj\u00f6rkman", "url": "http://virtual.aistats.org/api/miniconf/users/22898?format=json", "institution": "Aalto University"}, {"id": 22899, "fullname": "Jorge Loria", "url": "http://virtual.aistats.org/api/miniconf/users/22899?format=json", "institution": "Aalto University"}, {"id": 22900, "fullname": "Sophie Wharrie", "url": "http://virtual.aistats.org/api/miniconf/users/22900?format=json", "institution": "University of Melbourne"}, {"id": 3700, "fullname": "Samuel Kaski", "url": "http://virtual.aistats.org/api/miniconf/users/3700?format=json", "institution": "Aalto University and University of Manchester"}], "abstract": "Bayesian causal discovery benefits from prior information elicited from domain experts, and in heterogeneous domains any prior knowledge would be badly needed. However, so far prior elicitation approaches have assumed a single causal graph and hence are not suited to heterogeneous domains. We propose a causal elicitation strategy for heterogeneous settings, based on Bayesian experimental design (BED) principles, and a _variational mixture structure learning_ (VaMSL) method\u2014extending the earlier _differentiable Bayesian structure learning_ (DiBS) method\u2014to iteratively infer mixtures of causal Bayesian networks (CBNs). We construct an informative graph prior incorporating elicited expert feedback in the inference of mixtures of CBNs. Our proposed method successfully produces a set of alternative causal models (mixture components or clusters), and achieves an improved structure learning performance on heterogeneous synthetic data when informed by a simulated expert. Finally, we demonstrate that our approach is capable of capturing complex distributions in a breast cancer database.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13804", "url": null, "sourceid": 435, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=ALmU95pl4m", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11497, "modified": "2026-03-29T20:43:13.697722-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=ALmU95pl4m", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "88", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13850, "uid": "df12ecd077efc8c23881028604dbb8cc", "name": "KQ-SVD: Compressing the KV Cache with Provable Guarantees on Attention Fidelity", "authors": [{"id": 19754, "fullname": "Damien Lesens", "url": "http://virtual.aistats.org/api/miniconf/users/19754?format=json", "institution": "ENS de Lyon"}, {"id": 22988, "fullname": "Beheshteh T. Rakhshan", "url": "http://virtual.aistats.org/api/miniconf/users/22988?format=json", "institution": "Montreal Institute for Learning Algorithms, University of Montreal, Universit\u00e9 de Montr\u00e9al"}, {"id": 946, "fullname": "Guillaume Rabusseau", "url": "http://virtual.aistats.org/api/miniconf/users/946?format=json", "institution": "Mila, Universit\u00e9 de Montr\u00e9al"}], "abstract": "The Key\u2013Value (KV) cache is central to the efficiency of transformer-based large language models (LLMs), storing previously computed vectors to accelerate inference. Yet, as sequence length and batch size grow, the cache becomes a major memory bottleneck. Prior compression methods typically apply low-rank decomposition to keys alone or attempt to jointly embed queries and keys, but both approaches neglect that attention fundamentally depends on their inner products. In this work, we prove that such strategies are sub-optimal for approximating the attention matrix. We introduce KQ-SVD, a simple and computationally efficient method that directly performs an optimal low-rank decomposition of the attention matrix via a closed-form solution. By targeting the true source of redundancy, KQ-SVD preserves attention outputs with higher fidelity under compression. Extensive evaluations on LLaMA and Mistral models demonstrate that our approach consistently delivers superior projection quality.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13850", "url": null, "sourceid": 1523, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=6PltpCBNmO", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11543, "modified": "2026-03-29T20:43:15.658552-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=6PltpCBNmO", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "89", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13673, "uid": "ff49cc40a8890e6a60f40ff3026d2730", "name": "Learning Geometry and Topology via Multi-Chart Flows", "authors": [{"id": 10906, "fullname": "Hanlin Yu", "url": "http://virtual.aistats.org/api/miniconf/users/10906?format=json", "institution": "University of Helsinki"}, {"id": 489, "fullname": "Soren Hauberg", "url": "http://virtual.aistats.org/api/miniconf/users/489?format=json", "institution": "Technical University of Denmark, Denmark"}, {"id": 4017, "fullname": "Marcelo Hartmann", "url": "http://virtual.aistats.org/api/miniconf/users/4017?format=json", "institution": "University of Helsinki"}, {"id": 4019, "fullname": "Arto Klami", "url": "http://virtual.aistats.org/api/miniconf/users/4019?format=json", "institution": "University of Helsinki"}, {"id": 22597, "fullname": "Georgios Arvanitidis", "url": "http://virtual.aistats.org/api/miniconf/users/22597?format=json", "institution": "Technical University of Denmark"}], "abstract": "Real world data often lie on low-dimensional Riemannian manifolds embedded in high-dimensional spaces. This motivates learning degenerate normalizing flows that map between the ambient space and a low-dimensional latent space. However, if the manifold has a non-trivial topology, it can never be correctly learned using a single flow. Instead multiple flows must be `glued together'. In this paper, we first propose the general training scheme for learning such a collection of flows, and secondly we develop the first numerical algorithms for computing geodesics on such manifolds. Empirically, we demonstrate that this leads to highly significant improvements in topology estimation.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13673", "url": null, "sourceid": 1333, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=LY0Nz2AMo5", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11366, "modified": "2026-03-29T20:43:08.139106-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=LY0Nz2AMo5", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "90", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13417, "uid": "15d4e891d784977cacbfcbb00c48f133", "name": "Information Hidden in Gradients of Regression with Target Noise", "authors": [{"id": 19834, "fullname": "Arash Jamshidi", "url": "http://virtual.aistats.org/api/miniconf/users/19834?format=json", "institution": "University of Helsinki"}, {"id": 19788, "fullname": "Katsiaryna Haitsiukevich", "url": "http://virtual.aistats.org/api/miniconf/users/19788?format=json", "institution": "University of Helsinki"}, {"id": 12368, "fullname": "Kai Puolam\u00e4ki", "url": "http://virtual.aistats.org/api/miniconf/users/12368?format=json", "institution": "University of Helsinki"}], "abstract": "Second-order information---such as curvature or data covariance---is critical for optimisation, diagnostics, and robustness. However, in many modern settings, only the gradients are observable. We show that the gradients alone can reveal the Hessian, equalling the data covariance $\\Sigma$ for the linear regression. Our key insight is a simple variance calibration: injecting Gaussian noise so that the total target noise variance equals the batch size ensures that the empirical gradient covariance closely approximates the Hessian, even when evaluated far from the optimum. We provide non-asymptotic operator-norm guarantees under sub-Gaussian inputs. We also show that without such calibration, recovery can fail by an $\\Omega(1)$ factor. The proposed method is practical (a ``set target-noise variance to $n$\u2019\u2019 rule) and robust (variance $\\mathcal{O}(n)$ suffices to recover $\\Sigma$ up to scale). Applications include preconditioning for faster optimisation, adversarial risk estimation, and gradient-only training, for example, in distributed systems. We support our theoretical results with experiments on synthetic and real data.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13417", "url": null, "sourceid": 441, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=od6Ol5hdaN", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11110, "modified": "2026-03-29T20:42:58.086740-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=od6Ol5hdaN", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "90", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13602, "uid": "2ba8698b79439589fdd2b0f7218d8b07", "name": "Learning How Deep to Go: Self-Scaling Deep Reinforcement Learning", "authors": [{"id": 22465, "fullname": "Michelangelo Vegli\u00f2", "url": "http://virtual.aistats.org/api/miniconf/users/22465?format=json", "institution": "Campus Bio-Medico University of Rome"}, {"id": 22466, "fullname": "Marco Fantozzi", "url": "http://virtual.aistats.org/api/miniconf/users/22466?format=json", "institution": "University of Parma"}, {"id": 22467, "fullname": "Antonio Di Cecco", "url": "http://virtual.aistats.org/api/miniconf/users/22467?format=json", "institution": "Campus Bio-Medico University of Rome"}, {"id": 22468, "fullname": "Carlo Metta", "url": "http://virtual.aistats.org/api/miniconf/users/22468?format=json", "institution": "CNR"}, {"id": 22469, "fullname": "Flora Angileri", "url": "http://virtual.aistats.org/api/miniconf/users/22469?format=json", "institution": "University of Roma &quot;Tor Vergata&quot;"}, {"id": 22470, "fullname": "Simone Treccani", "url": "http://virtual.aistats.org/api/miniconf/users/22470?format=json", "institution": "University of Parma"}, {"id": 22471, "fullname": "Adrienne Macazar", "url": "http://virtual.aistats.org/api/miniconf/users/22471?format=json", "institution": "University of Parma"}, {"id": 22472, "fullname": "Silvia Galfre&#x27;", "url": "http://virtual.aistats.org/api/miniconf/users/22472?format=json", "institution": "University of Pisa"}, {"id": 22473, "fullname": "Maurizio Parton", "url": "http://virtual.aistats.org/api/miniconf/users/22473?format=json", "institution": "University of Chieti-Pescara, Italy"}, {"id": 22474, "fullname": "Francesco Morandin", "url": "http://virtual.aistats.org/api/miniconf/users/22474?format=json", "institution": "University of Pisa"}], "abstract": "Deep Reinforcement Learning (DRL) has achieved remarkable results in complex sequential decision-making tasks, often using very deep neural networks. However, these architectures incur substantial computational and energy costs, and selecting the optimal network depth in advance remains an open challenge. In this paper, we introduce SCALE-RL, a self-scaling DRL framework that dynamically adjusts its architectural depth during training, allowing the network to automatically adapt its depth to the task. Integrated into an AlphaZero-style pipeline for Othello, our approach matches the playing strength of the baseline agent while reducing network depth by 50\\%. This process not only translates into substantial savings in computation and energy but also enhances model interpretability through the additive decomposition of decision-making across layers. Our results suggest that enabling DRL models to discover the complexity they require, rather than relying on fixed, over-parameterized architectures, makes it possible to develop more efficient, interpretable, and sustainable DRL agents.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13602", "url": null, "sourceid": 1249, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=U9dlwZfURP", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11295, "modified": "2026-03-29T20:43:05.294346-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=U9dlwZfURP", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "91", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13685, "uid": "2d405b367158e3f12d7c1e31a96b3af3", "name": "Learning Markov Processes as Sum-of-Square Forms for Analytical Belief Propagation", "authors": [{"id": 22625, "fullname": "Peter Amorese", "url": "http://virtual.aistats.org/api/miniconf/users/22625?format=json", "institution": "University of Colorado at Boulder"}, {"id": 22626, "fullname": "Morteza Lahijanian", "url": "http://virtual.aistats.org/api/miniconf/users/22626?format=json", "institution": "University of Colorado at Boulder"}], "abstract": "Harnessing the predictive capability of Markov process models requires propagating probability density functions (beliefs) through the model. For many existing models however, belief propagation is analytically infeasible, requiring approximation or sampling to generate predictions. This paper proposes a functional modeling framework leveraging sparse Sum-of-Squares forms for valid (conditional) density estimation. We show that such an architecture enables generalized simultaneous learning of basis functions and coefficients, while preserving analytical integrability. We study the theoretical underpinnings of the proposed model with respect to (i) representational capacity, (ii) analytical marginalization, and (iii) sparse parameter representation. In addition, we propose a training method that allows for exact adherence to the normalization and non-negativity constraints. Our results show that the proposed method achieves accuracy comparable to state-of-the-art approaches while requiring significantly less memory in low-dimensional spaces, and it further scales to 12D systems when existing methods fail beyond 2D.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13685", "url": null, "sourceid": 2062, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=JtBpCUK0Ac", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11378, "modified": "2026-03-29T20:43:08.708424-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=JtBpCUK0Ac", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "92", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13378, "uid": "9597353e41e6957b5e7aa79214fcb256", "name": "Laplace approximation for Bayesian variable selection via Le Cam's one-step procedure", "authors": [{"id": 21970, "fullname": "Tianrui Hou", "url": "http://virtual.aistats.org/api/miniconf/users/21970?format=json", "institution": "Boston University, Boston University"}, {"id": 12461, "fullname": "Aguemon Atchade", "url": "http://virtual.aistats.org/api/miniconf/users/12461?format=json", "institution": "Boston University"}], "abstract": "Relevant feature selection in high-dimensional settings is a central challenge in modern scientific research and decision-making. While many existing methods offer strong statistical guarantees, they are often computationally intractable in high-dimensional problems. To address this issue, we introduce a novel Laplace approximation method based on Le Cam\u2019s one-step procedure, termed \\textsf{OLAP}. This approach is specifically designed to alleviate computational burdens while maintaining statistical rigor. Under standard high-dimensional assumptions, we establish that \\textsf{OLAP} achieves consistent variable selection. Moreover, the method yields a posterior distribution that can be efficiently explored in polynomial time via a simple Gibbs sampling algorithm. We demonstrate the effectiveness of OLAP through applications to logistic and Poisson regression models, using both simulated and real data.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13378", "url": null, "sourceid": 1607, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=tYcwtcFygR", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11071, "modified": "2026-03-29T20:42:56.434068-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=tYcwtcFygR", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "92", "schedule_html": "", "latitude": null, "longitude": null, "related_events": ["http://virtual.aistats.org/api/miniconf/events/13917?format=json"], "related_events_ids": [13917]}, {"id": 13836, "uid": "69dafe8b58066478aea48f3d0f384820", "name": "Learning to Explore with Lagrangians for Bandits under Unknown Constraints", "authors": [{"id": 20587, "fullname": "Udvas Das", "url": "http://virtual.aistats.org/api/miniconf/users/20587?format=json", "institution": "INRIA Lille"}, {"id": 12340, "fullname": "Debabrota Basu", "url": "http://virtual.aistats.org/api/miniconf/users/12340?format=json", "institution": "Inria, CNRS, Univ. Lille"}], "abstract": "Pure exploration in bandits formalises multiple real-world problems, such as tuning hyper-parameters or conducting user studies to test a set of items, where different safety, resource, and fairness constraints on the decision space naturally appear. We study these problems as pure exploration in multi-armed bandits with unknown linear constraints, where the aim is to identify an *$r$-optimal and feasible policy* as fast as possible with a given level of confidence. First, we propose a Lagrangian relaxation of the sample complexity lower bound for pure exploration under constraints. Second, we leverage properties of convex optimisation in the Lagrangian lower bound to propose two computationally efficient extensions of Track-and-Stop and Gamified Explorer, namely LATS and LAGEX. Then, we propose a constraint-adaptive stopping rule, and while tracking the lower bound, use optimistic estimate of the feasible set at each step. We show that LAGEX achieves asymptotically optimal sample complexity upper bound, while LATS shows asymptotic optimality up to *novel* constraint-dependent constants. Finally, we conduct numerical experiments with different reward distributions and constraints that validate efficient performance of LATS and LAGEX.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13836", "url": null, "sourceid": 2312, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=7VNrNTXuK5", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11529, "modified": "2026-03-29T20:43:15.071343-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=7VNrNTXuK5", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "94", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13636, "uid": "afda332245e2af431fb7b672a68b659d", "name": "Learning to Choose or Choosing to Learn: Best-of-N vs. Supervised Fine-Tuning for Bit String Generation", "authors": [{"id": 16163, "fullname": "Seamus Seamus", "url": "http://virtual.aistats.org/api/miniconf/users/16163?format=json", "institution": "University of Michigan"}, {"id": 22527, "fullname": "Vinod Raman", "url": "http://virtual.aistats.org/api/miniconf/users/22527?format=json", "institution": "Google DeepMind"}, {"id": 22528, "fullname": "Unique Subedi", "url": "http://virtual.aistats.org/api/miniconf/users/22528?format=json", "institution": "University of Michigan - Ann Arbor"}, {"id": 17761, "fullname": "Yuekai Sun", "url": "http://virtual.aistats.org/api/miniconf/users/17761?format=json", "institution": "University of Michigan"}], "abstract": "Using the bit string generation problem as a case study, we theoretically compare two standard methods for adapting large language models to new tasks. The first, referred to as *supervised fine-tuning*, involves training a new next token predictor on good generations. The second method, *Best-of-N*, trains a reward model to select good responses from a collection generated by an unaltered base model. If the learning setting is realizable, we find that supervised fine-tuning outperforms BoN through a better dependence on the response length in its rate of convergence. If realizability fails, then depending on the failure mode, BoN can enjoy a better rate of convergence in either $n$ or a rate of convergence with better dependence on the response length.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13636", "url": null, "sourceid": 822, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=PEAGyyk5kg", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11329, "modified": "2026-03-29T20:43:06.679329-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=PEAGyyk5kg", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "94", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13408, "uid": "2ba596643cbbbc20318224181fa46b28", "name": "Learning When Not to Learn: Risk-Sensitive Abstention in Bandits with Unbounded Rewards", "authors": [{"id": 19866, "fullname": "Sarah Liaw", "url": "http://virtual.aistats.org/api/miniconf/users/19866?format=json", "institution": "Harvard University"}, {"id": 22025, "fullname": "Benjamin Plaut", "url": "http://virtual.aistats.org/api/miniconf/users/22025?format=json", "institution": "University of California, Berkeley"}], "abstract": "In high-stakes AI applications, even a single action can cause irreparable damage. However, nearly all of sequential decision-making theory assumes that all errors are recoverable (e.g., by bounding rewards). Standard bandit algorithms that explore aggressively may cause irreparable damage when this assumption fails. Some prior work avoids irreparable errors by asking for help from a mentor, but a mentor may not always be available. In this work, we formalize a model of learning with unbounded rewards without a mentor as a two-action contextual bandit with an abstain option: at each round the agent observes an input and chooses either to abstain (always 0 reward) or to commit (execute a preexisting task policy). Committing yields rewards that are upper-bounded but can be arbitrarily negative, and the commit reward is assumed Lipschitz in the input. We propose a caution-based algorithm that learns when not to learn: it chooses a trusted region and commits only where the available evidence does not already certify harm. Under these conditions and i.i.d. inputs, we establish sublinear regret guarantees, theoretically demonstrating the effectiveness of cautious exploration for deploying learning agents safely in high-stakes environments.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13408", "url": null, "sourceid": 957, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=pLRYWrIpZu", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11101, "modified": "2026-03-29T20:42:57.688949-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=pLRYWrIpZu", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "95", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13546, "uid": "ff7d0f525b3be596a51fb919492c099c", "name": "LLM-as-a-Judge on a Budget", "authors": [{"id": 3557, "fullname": "Aadirupa Saha", "url": "http://virtual.aistats.org/api/miniconf/users/3557?format=json", "institution": "UIC"}, {"id": 22335, "fullname": "Aniket Wagde", "url": "http://virtual.aistats.org/api/miniconf/users/22335?format=json", "institution": "University of Illinois at Chicago"}, {"id": 22336, "fullname": "Branislav Kveton", "url": "http://virtual.aistats.org/api/miniconf/users/22336?format=json", "institution": "Adobe Research"}], "abstract": "LLM-as-a-judge has emerged as a cornerstone technique for automatic evaluation of large language models (LLMs). The key idea is to leverage the reasoning capabilities of LLMs to evaluate prompt-response pairs using rationales paired with numeric scores, and thus combining the comprehensiveness of human evaluation with automated metrics. The rationales and scores are sampled from an LLM and thus random. To get a more precise estimate of the mean score generated by the LLM, a common practice is to evaluate each prompt-response pair multiple times. Therefore, practitioners face the following critical challenge: given a fixed computational budget, how to optimally allocate LLM judgments across prompt-response pairs to estimate the mean scores as precisely as possible? We present a principled variance-adaptive approach that addresses this fundamental problem by leveraging insights from multi-armed bandit (MAB) theory and concentration inequalities. Our method dynamically allocates LLM judgements based on the estimated variance of LLM scores for each prompt-response pair, concentrating computational resources on the most uncertain scores. We prove that our algorithm achieves a near-optimal sample complexity for minimizing the worst-case estimation error across prompt-response pairs, providing theoretical guarantees for practitioners working under budget constraints. Extensive experiments on two popular evaluation datasets, *Summarize from Feedback* and *HelpSteer2*, show that our method significantly reduces the worst-case estimation error of the estimated mean scores while maintaining a fixed query budget. Our results establish a novel theoretical foundation for efficient LLM judges and provide a practical guidance for deploying such evaluation pipelines at scale, with broad implications to AI safety, model development, and automated assessment in our increasingly LLM-driven world.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13546", "url": null, "sourceid": 1912, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=asVCSlgfJO", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11239, "modified": "2026-03-29T20:43:02.984001-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=asVCSlgfJO", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "96", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13534, "uid": "a7d8ae4569120b5bec12e7b6e9648b86", "name": "Linear Reasoning Vs. Proof by Cases: Obstacles for Large Language Models in FOL Problem Solving", "authors": [{"id": 18245, "fullname": "Yuliang Ji", "url": "http://virtual.aistats.org/api/miniconf/users/18245?format=json", "institution": "Nanjing University of Science and Technology"}, {"id": 22304, "fullname": "Fuchen Shen", "url": "http://virtual.aistats.org/api/miniconf/users/22304?format=json", "institution": "Westlake University"}, {"id": 18222, "fullname": "Jian Wu", "url": "http://virtual.aistats.org/api/miniconf/users/18222?format=json", "institution": "Tokyo Institute of Technology, Tokyo Institute of Technology"}, {"id": 22305, "fullname": "Qiujie Xie", "url": "http://virtual.aistats.org/api/miniconf/users/22305?format=json", "institution": "Westlake University"}, {"id": 22306, "fullname": "Yue Zhang", "url": "http://virtual.aistats.org/api/miniconf/users/22306?format=json", "institution": "Westlake University"}], "abstract": "To comprehensively evaluate the mathematical reasoning capabilities of Large Language Models (LLMs), researchers have introduced  abundant mathematical reasoning datasets. However, most existing datasets primarily focus on linear reasoning, neglecting other parts such as proof by contradiction and proof by cases, which are crucial for investigating LLMs\u2019 reasoning abilities. To address this limitation, we first introduce a novel first-order logic (FOL) dataset named PC-FOL, annotated by professional mathematicians, focusing on case-based reasoning problems. All instances in this dataset are equipped with a manually written natural language proof, clearly distinguishing it from conventional linear reasoning datasets. Our experimental results over leading LLMs demonstrate a substantial performance gap between linear reasoning and case-based reasoning problems. To further investigate this phenomenon, we provide a theoretical analysis grounded in graphical model, which provides an explanation for the observed disparity between the two types of reasoning problems. We hope this work can reveal the core challenges in the field of automated natural language mathematical proof generation, paving the way for future research.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13534", "url": null, "sourceid": 1176, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=bvBrOP8DNC", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11227, "modified": "2026-03-29T20:43:02.561152-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=bvBrOP8DNC", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "96", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13706, "uid": "c7635bfd99248a2cdef8249ef7bfbef4", "name": "Lipschitz Multiscale Deep Equilibrium Models: A Theoretically Guaranteed and Accelerated Approach", "authors": [{"id": 19778, "fullname": "Naoki Sato", "url": "http://virtual.aistats.org/api/miniconf/users/19778?format=json", "institution": "Meiji University"}, {"id": 9802, "fullname": "Hideaki Iiduka", "url": "http://virtual.aistats.org/api/miniconf/users/9802?format=json", "institution": "Meiji University"}], "abstract": "Deep equilibrium models (DEQs) achieve infinitely deep network representations without stacking layers by exploring fixed points of layer transformations in neural networks. Such models constitute an innovative approach that achieves performance comparable to state-of-the-art methods in many large-scale numerical experiments, despite requiring significantly less memory. However, DEQs face the challenge of requiring vastly more computational time for training and inference than conventional methods, as they repeatedly perform fixed-point iterations with no convergence guarantee upon each input. Therefore, this study explored an approach to improve fixed-point convergence and consequently reduce computational time by restructuring the model architecture to guarantee fixed-point convergence. Our proposed approach for image classification, Lipschitz multiscale DEQ, has theoretically guaranteed fixed-point convergence for both forward and backward passes by hyperparameter adjustment, achieving up to a 4.75$\\times$ speedup in numerical experiments on CIFAR-10 at the cost of a minor drop in accuracy.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13706", "url": null, "sourceid": 1124, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=HkPWADS3hs", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11399, "modified": "2026-03-29T20:43:09.560785-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=HkPWADS3hs", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "97", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13524, "uid": "54229abfcfa5649e7003b83dd4755294", "name": "Linearly Separable Features in Shallow Nonlinear Networks: Width Scales Polynomially with Intrinsic Data Dimension", "authors": [{"id": 19766, "fullname": "Alec Xu", "url": "http://virtual.aistats.org/api/miniconf/users/19766?format=json", "institution": "University of Michigan"}, {"id": 21906, "fullname": "Can Yaras", "url": "http://virtual.aistats.org/api/miniconf/users/21906?format=json", "institution": "University of Michigan - Ann Arbor"}, {"id": 5178, "fullname": "Peng Wang", "url": "http://virtual.aistats.org/api/miniconf/users/5178?format=json", "institution": "University of Michigan"}, {"id": 12582, "fullname": "Qing Qu", "url": "http://virtual.aistats.org/api/miniconf/users/12582?format=json", "institution": "University of Michigan"}], "abstract": "Deep neural networks have attained remarkable success across diverse classification tasks. Recent empirical studies have shown that deep networks learn features that are linearly separable across classes. However, these findings often lack rigorous justifications, even under relatively simple settings. In this work, we address this gap by examining the linear separation capabilities of shallow nonlinear networks. Specifically, inspired by the low intrinsic dimensionality of image data, we model inputs as a union of low-dimensional subspaces (UoS) and demonstrate that a single nonlinear layer can transform such data into linearly separable sets. Theoretically, we show that this transformation occurs with high probability when using random weights and quadratic activations. Notably, we prove this can be achieved when the network width scales polynomially with the intrinsic dimension of the data rather than the ambient dimension. Experimental results corroborate these theoretical findings and demonstrate that similar linear separation properties hold in practical scenarios beyond our analytical scope. This work bridges the gap between empirical observations and theoretical understanding of the separation capacity of nonlinear networks, offering deeper insights into model interpretability and generalization.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13524", "url": null, "sourceid": 91, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=ckbVWAqb28", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11217, "modified": "2026-03-29T20:43:02.178406-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=ckbVWAqb28", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "97", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13859, "uid": "2a50e9c2d6b89b95bcb416d6857f8b45", "name": "LLMPhy: Parameter-Identifiable Physical Reasoning Combining Large Language Models and Physics Engines", "authors": [{"id": 19923, "fullname": "Anoop Cherian", "url": "http://virtual.aistats.org/api/miniconf/users/19923?format=json", "institution": "Mitsubishi Electric Research Labs (MERL)"}, {"id": 22995, "fullname": "Radu Corcodel", "url": "http://virtual.aistats.org/api/miniconf/users/22995?format=json", "institution": "Mitsubishi Electric Research Labs"}, {"id": 22996, "fullname": "Siddarth Jain", "url": "http://virtual.aistats.org/api/miniconf/users/22996?format=json", "institution": "Mitsubishi Electric Research Labs"}, {"id": 22997, "fullname": "Diego Romeres", "url": "http://virtual.aistats.org/api/miniconf/users/22997?format=json", "institution": "Mitsubishi Electric Research Labs"}], "abstract": "Most learning-based approaches to complex physical reasoning sidestep the crucial problem of parameter identification (e.g., mass, friction) that governs scene dynamics--despite its importance in real-world applications; e.g., collision avoidance, robotic manipulation. In this paper, we present LLMPhy, a black-box optimization framework that integrates large language models (LLMs) with physics simulators for physical reasoning. The core insight of LLMPhy is to bridge the textbook physical knowledge embedded in LLMs with the world models implemented in modern physics engines, thereby enabling the construction of digital twins of input scenes through the estimation of latent parameters. Specifically, LLMPhy decomposes digital twin construction into two subproblems: a continuous one of estimating physical parameters and a discrete one of estimating scene layout. For each subproblem, LLMPhy iteratively prompts the LLM to generate programs embedding parameter estimates, executes them in the physics engine to reconstruct the scene, and then uses the resulting reconstruction error as feedback to refine the LLM\u2019s predictions.   As existing physical reasoning benchmarks rarely account for parameter identifiability, we introduce three new datasets\u2014including one real-world task\u2014specifically designed to evaluate this capability in a zero-shot setting. Our results show that LLMPhy achieves state-of-the-art performance on these tasks, recovers physical parameters more accurately, and converges more reliably than popular black-box methods.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13859", "url": null, "sourceid": 1320, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=5aLtvjoQpn", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11552, "modified": "2026-03-29T20:43:16.217723-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=5aLtvjoQpn", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "98", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13894, "uid": "5487315b1286f907165907aa8fc96619", "name": "Lloyd's $K$-Means Clustering Algorithm is Frank-Wolfe in Disguise", "authors": [{"id": 23071, "fullname": "Michael Pokojovy", "url": "http://virtual.aistats.org/api/miniconf/users/23071?format=json", "institution": "Old Dominion University"}, {"id": 23072, "fullname": "J. Marcus Jobe", "url": "http://virtual.aistats.org/api/miniconf/users/23072?format=json", "institution": "Miami University of Ohio"}, {"id": 533, "fullname": "Simon Lacoste-Julien", "url": "http://virtual.aistats.org/api/miniconf/users/533?format=json", "institution": "Mila, Universit\u00e9 de Montr\u00e9al"}], "abstract": "Lloyd's $K$-means algorithm, also known as na\u00efve $K$-means, is a widely used *ad hoc* optimization heuristic, designed to minimize the sum of squared errors (SSE) across all $K$-partitions of a dataset via iterative cluster refinement. In this work, we establish a novel connection between Lloyd's algorithm and the Frank-Wolfe (FW) algorithm, a prominent first-order method for projection-free optimization. We demonstrate that Lloyd's algorithm is a special case of FW. Leveraging recent advances in FW methods for concave objectives, we derive a non-asymptotic $\\mathcal{O}(1/t)$ convergence rate to a local minimum of the SSE objective. To account for empty clusters, an outcome possible under Lloyd's greedy assignment, we develop an FW variant for semismooth objectives while retaining the same convergence rate that is solely controlled by the initial SSE value. We illustrate our findings with a simulation study for spherical Gaussian mixtures and a real-world image segmentation dataset.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13894", "url": null, "sourceid": 694, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=27Xl7DnW8D", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11587, "modified": "2026-03-29T20:43:17.612209-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=27Xl7DnW8D", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "99", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13520, "uid": "42a3964579017f3cb42b26605b9ae8ef", "name": "MineGrad: Gradient Inversion Attacks on LoRA Fine-Tuning", "authors": [{"id": 9784, "fullname": "Hasin Us Sami", "url": "http://virtual.aistats.org/api/miniconf/users/9784?format=json", "institution": "University of California, Riverside"}, {"id": 22288, "fullname": "Swapneel Sen", "url": "http://virtual.aistats.org/api/miniconf/users/22288?format=json", "institution": "University of California, Riverside"}, {"id": 9785, "fullname": "Basak Guler", "url": "http://virtual.aistats.org/api/miniconf/users/9785?format=json", "institution": "University of California, Riverside"}], "abstract": "Parameter-efficient fine-tuning (PEFT), such as low-rank adaptation (LoRA), has recently been adopted in federated learning to reduce communication and computation costs. In this setup, users download a pretrained model from the server prior to fine-tuning, and then fine-tune lightweight LoRA modules locally while keeping the pretrained model frozen, sharing only the gradients of the fine-tuning parameters with the server. Despite its growing popularity, robustness of federated fine-tuning against an adversarial server remains underexplored, where the server maliciously tampers with the training protocol to breach the privacy of users\u2019 data. In this work, we investigate gradient inversion attacks on LoRA fine-tuning.  We propose an analytical attack that enables a malicious server to recover private user data by leveraging a poisoned pretrained model and fine-tuning parameters. Our design embeds fine-tuning data within the shared gradients, to allow the server to analytically reconstruct user data. Unlike prior works, our attack is applicable to both language and vision tasks, does not rely on computationally expensive (adversarial) pretraining with public datasets or require the number of training tokens to be less than the rank of LoRA modules. Experimental results on both language and vision tasks demonstrate high-fidelity data recovery across multiple baselines, revealing several critical vulnerabilities.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13520", "url": null, "sourceid": 1635, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=dD9XOZUpNc", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11213, "modified": "2026-03-29T20:43:02.022911-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=dD9XOZUpNc", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "99", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13549, "uid": "f31b20466ae89669f9741e047487eb37", "name": "Model Selection for Average Reward RL with Application to Utility Maximization in Repeated Games", "authors": [{"id": 22348, "fullname": "Alireza Masoumian", "url": "http://virtual.aistats.org/api/miniconf/users/22348?format=json", "institution": "University of Alberta"}, {"id": 22349, "fullname": "James Wright", "url": "http://virtual.aistats.org/api/miniconf/users/22349?format=json", "institution": "University of Alberta"}], "abstract": "In standard RL, the structure of the Markov Decision Process (e.g. state space) is known. In online model selection, a learner attempts to learn an optimal policy for an MDP knowing only that it belongs to one of $M >1$ model classes of varying complexity. Recent results have shown that this can be feasibly accomplished in episodic online RL. In this work, we propose $\\textsf{MRBEAR}$, an online model selection algorithm for the average reward RL setting which is based on the idea of regret balancing and elimination. The regret of the algorithm is in $\\tilde O(M C_{m*}^2 B_{m*}(T,\\delta))$ where $C_{m*}$ represents the complexity of the simplest well-specified model class and $B_{m^*}(T,\\delta)$ is its corresponding regret bound. This result shows that in average reward RL, the additional cost of model selection scales only linearly in $M$, the number of model classes. As an application, in a simultaneous general-sum repeated game, where the opponent follows a fixed unknown limited memory strategy, the learner can maximize its utility using $\\textsf{MRBEAR}$. By proving a lower bound, we showed the learner's regret is tight in opponent's memory order. In addition, the algorithm's performance is demonstrated through experiments.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13549", "url": null, "sourceid": 1877, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=aUdNLHo16h", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11242, "modified": "2026-03-29T20:43:03.106606-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=aUdNLHo16h", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "100", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13744, "uid": "77f959f119f4fb2321e9ce801e2f5163", "name": "Loss Gaps Parity for Fairness in Heterogeneous Federated Learning", "authors": [{"id": 22753, "fullname": "Brahim Erraji", "url": "http://virtual.aistats.org/api/miniconf/users/22753?format=json", "institution": "INRIA"}, {"id": 22754, "fullname": "Micha\u00ebl Perrot", "url": "http://virtual.aistats.org/api/miniconf/users/22754?format=json", "institution": "Inria Centre at the University of Lille"}, {"id": 535, "fullname": "Aur\u00e9lien Bellet", "url": "http://virtual.aistats.org/api/miniconf/users/535?format=json", "institution": "INRIA"}], "abstract": "While clients may join federated learning to improve performance on data they rarely observe locally, they often remain self-interested, expecting the global model to perform well on their own data. This motivates an objective that ensures all clients achieve a similar \\emph{loss gap}\u2014the difference in performance between the global model and the best model they could train using only their local data. To this end, we propose EAGLE, a novel federated learning algorithm that explicitly regularizes the global model to minimize disparities in loss gaps across clients. Our approach is particularly effective in heterogeneous settings, where clients' optimal local models may be misaligned. Unlike existing methods that encourage loss parity, potentially degrading performance for many clients, EAGLE targets fairness in relative improvements. We provide theoretical convergence guarantees for EAGLE under non-convex loss functions, and characterize how its iterates perform relative to the standard federated learning objective using a novel heterogeneity measure. Empirically, we demonstrate that EAGLE reduces the disparity in loss gaps among clients by prioritizing those furthest from their local optimal loss, while maintaining competitive utility in both convex and non-convex cases compared to strong baselines.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13744", "url": null, "sourceid": 1502, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=EdJndEEi5V", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11437, "modified": "2026-03-29T20:43:11.274099-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=EdJndEEi5V", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "101", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13398, "uid": "bc7f621451b4f5df308a8e098112185d", "name": "Model-independent O(1/k)-convergence rate for TD(0) with linear function approximation, universal learning steps and i.i.d. samples", "authors": [{"id": 22006, "fullname": "Ziad Kobeissi", "url": "http://virtual.aistats.org/api/miniconf/users/22006?format=json", "institution": "INRIA"}, {"id": 22007, "fullname": "Elo\u00efse Berthier", "url": "http://virtual.aistats.org/api/miniconf/users/22007?format=json", "institution": "ENSTA, Institut Polytechnique de Paris"}], "abstract": "In this paper, we study the finite-time behaviour of the TD(0) temporal-difference method with linear function approximation (LFA). We consider on-policy i.i.d. samples, a constant learning step, and the Polyak-Juditsky averaging method. We establish a new convergence rate that is (i) optimal in the number of iterations $k$ (i.e., of order $1/k$) and (ii) is model-independent: it does not depend on the choice of the linear parametrisation and is robust to ill-conditioning. This resolves a question posed by Lakshminarayanan and Szepesvari (2018) about the attainability of such a rate, open for more than seven years. Our analysis extends to TD(0) the results by Bach and Moulines (2013), who obtained a similar rate for Stochastic Gradient Descent (SGD) on least-square regression problems.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13398", "url": null, "sourceid": 2399, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=qNt7M2cCbf", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11091, "modified": "2026-03-29T20:42:57.299961-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=qNt7M2cCbf", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "101", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13419, "uid": "42e77b63637ab381e8be5f8318cc28a2", "name": "Low-Complexity and Consistent Graphon Estimation from Multiple Networks", "authors": [{"id": 19784, "fullname": "Roland Boniface SOGAN", "url": "http://virtual.aistats.org/api/miniconf/users/19784?format=json", "institution": "Sorbonne Universit\u00e9"}, {"id": 22047, "fullname": "Tabea Rebafka", "url": "http://virtual.aistats.org/api/miniconf/users/22047?format=json", "institution": "AgroParisTech"}], "abstract": "Recovering the random graph model from an observed collection of networks is known to present significant  challenges in the setting, where  the  networks do not share a common node set and have different sizes. More specifically, the goal is the estimation of the  graphon function that parametrizes the nonparametric exchangeable random graph model. Existing methods typically suffer from either limited accuracy or high computational complexity. We introduce a new histogram-based estimator with low algorithmic complexity that achieves high accuracy by jointly aligning the nodes of all graphs, in contrast to most conventional methods that order nodes graph by graph. Consistency results of the proposed graphon estimator are established.  A numerical study shows that the proposed estimator outperforms  existing methods in terms of accuracy, especially when the dataset comprises only small and  variable-size networks. Moreover, the computing time of the new method is considerably shorter than that of other consistent methodologies. Additionally, when applied to a graph neural network classification task, the proposed estimator enables more effective data augmentation, yielding improved performance across diverse real-world datasets.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13419", "url": null, "sourceid": 628, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=oajeob6ffw", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11112, "modified": "2026-03-29T20:42:58.159789-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=oajeob6ffw", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "102", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13743, "uid": "eb163727917cbba1eea208541a643e74", "name": "Low-Rank Bias, Weight Decay, and Model Merging in Neural Networks", "authors": [{"id": 490, "fullname": "Ilja Kuzborskij", "url": "http://virtual.aistats.org/api/miniconf/users/490?format=json", "institution": "Google DeepMind"}, {"id": 3755, "fullname": "Yasin Abbasi-Yadkori", "url": "http://virtual.aistats.org/api/miniconf/users/3755?format=json", "institution": "DeepMind"}], "abstract": "We explore the low-rank structure of the weight matrices in neural networks   at the stationary points (limiting solutions of optimization algorithms)   with $L2$ regularization (also known as weight decay). We show several   properties of such deep neural networks, induced by $L2$ regularization.   In particular, for a stationary point we show alignment of the   parameters and the gradient, norm preservation across layers, and low-rank   bias: properties previously known in the context of solutions of gradient descent/flow type algorithms.  Experiments   show that the assumptions made in the analysis only mildly affect the   observations.    In addition, we investigate a multitask learning phenomenon enabled by $L2$   regularization and low-rank bias. In particular, we show that if two networks   are trained, such that the inputs in the training set of one network are   approximately orthogonal to the inputs in the training set of the other   network, the new network obtained by simply summing the weights of the two   networks will perform as well on both training sets as the respective   individual networks.  We demonstrate this for shallow ReLU neural networks   trained by gradient descent, as well as deep linear networks trained by gradient flow.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13743", "url": null, "sourceid": 211, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=EjOV1vJfGi", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11436, "modified": "2026-03-29T20:43:11.241910-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=EjOV1vJfGi", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "103", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13607, "uid": "051928341be67dcba03f0e04104d9047", "name": "Monotone and Conservative Policy Iteration Beyond the Tabular Case", "authors": [{"id": 20586, "fullname": "S R Eshwar", "url": "http://virtual.aistats.org/api/miniconf/users/20586?format=json", "institution": "Indian Institute of Science"}, {"id": 19890, "fullname": "Gugan Chandrashekhar Mallika Thoppe", "url": "http://virtual.aistats.org/api/miniconf/users/19890?format=json", "institution": "Indian Institute of Science"}, {"id": 22479, "fullname": "Ananyabrata Barua", "url": "http://virtual.aistats.org/api/miniconf/users/22479?format=json", "institution": "Indian Institute of Science"}, {"id": 882, "fullname": "Aditya Gopalan", "url": "http://virtual.aistats.org/api/miniconf/users/882?format=json", "institution": "Indian Institute of Science (IISc), Bangalore"}, {"id": 22480, "fullname": "Gal Dalal", "url": "http://virtual.aistats.org/api/miniconf/users/22480?format=json", "institution": "NVIDIA"}], "abstract": "We introduce Reliable Policy Iteration (RPI) and Conservative RPI (CRPI), variants of Policy Iteration (PI) and Conservative PI (CPI), that retain tabular guarantees under function approximation. RPI uses a novel Bellman-constrained optimization for policy evaluation. We show that RPI restores the textbook \\textit{monotonicity} of value estimates and that these estimates provably \\textit{lower-bound} the true return; moreover, their limit partially satisfies the \\textit{unprojected} Bellman equation. CRPI shares RPI's evaluation, but updates policies conservatively by  maximizing a new performance-difference \\textit{lower bound} that explicitly accounts for function-approximation-induced errors. CRPI inherits RPI's guarantees and, crucially, admits per-step improvement bounds. In initial simulations, RPI and CRPI outperform PI and its variants. Our work addresses a foundational gap in RL: popular algorithms such as TRPO and PPO derive from tabular CPI yet are deployed with function approximation, where CPI's guarantees often fail-leading to divergence, oscillations, or convergence to suboptimal policies. By restoring PI/CPI-style guarantees for \\textit{arbitrary} function classes, RPI and CRPI provide a principled basis for next-generation RL.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13607", "url": null, "sourceid": 2048, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=TPSX51x8oR", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11300, "modified": "2026-03-29T20:43:05.479943-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=TPSX51x8oR", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "103", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13895, "uid": "7250eb93b3c18cc9daa29cf58af7a004", "name": "Minimax-Optimal Two-Sample Test with Sliced Wasserstein", "authors": [{"id": 23073, "fullname": "Binh Thuan Tran", "url": "http://virtual.aistats.org/api/miniconf/users/23073?format=json", "institution": "Universit\u00e9 Gustave Eiffel"}, {"id": 19453, "fullname": "Nicolas Schreuder", "url": "http://virtual.aistats.org/api/miniconf/users/19453?format=json", "institution": "CNRS"}], "abstract": "We study the problem of nonparametric two-sample testing using the sliced Wasserstein (SW) distance. While prior theoretical and empirical work indicates that the SW distance offers a promising balance between strong statistical guarantees and computational efficiency, its theoretical foundations for hypothesis testing remain limited. We address this gap by proposing a permutation-based SW test and analyzing its performance. The test inherits finite-sample Type I error control from the permutation principle. Moreover, we establish non-asymptotic power bounds and show that the procedure achieves the minimax separation rate $n^{-1/2}$ over multinomial and bounded-support alternatives, matching the optimal guarantees of kernel-based tests while building on the geometric foundations of Wasserstein distances. Our analysis further quantifies the trade-off between the number of projections and statistical power. Finally, numerical experiments demonstrate that the test combines finite-sample validity with competitive power and scalability, and---unlike kernel-based tests, which require careful kernel tuning---it performs consistently well across all scenarios we consider.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13895", "url": null, "sourceid": 832, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=1x1kCzeMyx", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11588, "modified": "2026-03-29T20:43:17.654900-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=1x1kCzeMyx", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "104", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13718, "uid": "091d584fced301b442654dd8c23b3fc9", "name": "Lyapunov-Guided Self-Alignment: Test-Time Adaptation for Offline Safe Reinforcement Learning", "authors": [{"id": 20605, "fullname": "Seungyub Han", "url": "http://virtual.aistats.org/api/miniconf/users/20605?format=json", "institution": "Seoul National University"}, {"id": 22714, "fullname": "HyungJin Kim", "url": "http://virtual.aistats.org/api/miniconf/users/22714?format=json", "institution": "Seoul National University"}, {"id": 22715, "fullname": "Jungwoo Lee", "url": "http://virtual.aistats.org/api/miniconf/users/22715?format=json", "institution": "Seoul National University"}], "abstract": "Offline reinforcement learning (RL) agents often fail when deployed, as the gap between training datasets and real environments leads to unsafe behavior. To address this, we present SAS (Self-Alignment for Safety), a transformer-based framework that enables test-time adaptation in offline safe RL without retraining. In SAS, the main mechanism is self-alignment: at test time, the pretrained agent generates several imagined trajectories and selects those satisfying the Lyapunov condition. These feasible segments are then recycled as in-context prompts, allowing the agent to realign its behavior toward safety while avoiding parameter updates. In effect, SAS turns Lyapunov-guided imagination into control-invariant prompts, and its transformer architecture admits a hierarchical RL interpretation where prompting functions as Bayesian inference over latent skills. Across Safety Gymnasium and MuJoCo benchmarks, SAS consistently reduces cost and failure while maintaining or improving return.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13718", "url": null, "sourceid": 208, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=GPNwYOQECX", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11411, "modified": "2026-03-29T20:43:10.099824-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=GPNwYOQECX", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "104", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13652, "uid": "1e913e1b06ead0b66e30b6867bf63549", "name": "Multi-Agent Lipschitz Bandits", "authors": [{"id": 22562, "fullname": "Sourav Chakraborty", "url": "http://virtual.aistats.org/api/miniconf/users/22562?format=json", "institution": "University of Colorado at Boulder"}, {"id": 22563, "fullname": "Amit Rege", "url": "http://virtual.aistats.org/api/miniconf/users/22563?format=json", "institution": "University of Colorado, Boulder"}, {"id": 22564, "fullname": "Lijun Chen", "url": "http://virtual.aistats.org/api/miniconf/users/22564?format=json", "institution": "University of Colorado at Boulder"}, {"id": 17881, "fullname": "Claire Monteleoni", "url": "http://virtual.aistats.org/api/miniconf/users/17881?format=json", "institution": "University of Colorado, Boulder"}], "abstract": "We study the decentralized multi-player stochastic bandit problem over a continuous, Lipschitz-structured action space where hard collisions yield zero reward. Our objective is to design a communication-free policy that maximizes collective reward, with coordination costs that are independent of the time horizon $T$. We propose a modular protocol that first solves the multi-agent coordination problem\u2014identifying and seating players on distinct, high-value regions via a novel maxima-directed search\u2014and then decouples the problem into $N$ independent single-player Lipschitz bandits. We establish a near-optimal regret bound of $\\tilde{O}(T^{(d+1)/(d+2)})$ plus a $T$-independent coordination cost, matching the single-player rate. Our framework is, to our knowledge, the first to provide such guarantees and extends to general distance-threshold collision models.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13652", "url": null, "sourceid": 1933, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=NU614Ne4k3", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11345, "modified": "2026-03-29T20:43:07.387785-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=NU614Ne4k3", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "104", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13691, "uid": "e4bb4c5173c2ce17fd8fcd40041c068f", "name": "MDPs with a State Sensing Cost", "authors": [{"id": 22638, "fullname": "Vansh Kapoor", "url": "http://virtual.aistats.org/api/miniconf/users/22638?format=json", "institution": "School of Computer Science, Carnegie Mellon University"}, {"id": 486, "fullname": "Jayakrishnan Nair", "url": "http://virtual.aistats.org/api/miniconf/users/486?format=json", "institution": "&quot;Assist. Prof, EE, IIT Bombay&quot;"}], "abstract": "In many practical sequential decision-making problems, tracking the state of the environment incurs a sensing/computation cost. In these settings, the agent's interaction with its environment includes the additional component of deciding \\emph{when} to sense the state, in a manner that balances the value associated with optimal (state-specific) actions and the cost of sensing. We formulate this as an expected discounted cost Markov Decision Process (MDP), wherein the agent incurs an additional cost for sensing its next state, but has the option to take actions while remaining `blind' to the system state. We pose this problem as a classical discounted cost MDP with an expanded (countably infinite) state space. While computing the optimal policy for this MDP is intractable in general, we derive lower bounds on the optimal value function, which allow us to bound the suboptimality gap of any policy. We also propose a computationally efficient algorithm SPI, based on policy improvement, which in practice performs close to the optimal policy. Finally, we benchmark against the state-of-the-art via a numerical case study.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13691", "url": null, "sourceid": 695, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=J3VMBPRoeY", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11384, "modified": "2026-03-29T20:43:08.918850-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=J3VMBPRoeY", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "105", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13528, "uid": "4476b929e30dd0c4e8bdbcc82c6ba23a", "name": "Minimizing Human Intervention in Online Classification", "authors": [{"id": 19855, "fullname": "William R\u00e9veillard", "url": "http://virtual.aistats.org/api/miniconf/users/19855?format=json", "institution": "KTH Royal Institute of Technology"}, {"id": 22294, "fullname": "Vasileios Saketos", "url": "http://virtual.aistats.org/api/miniconf/users/22294?format=json", "institution": "School of Engineering and Applied Sciences, Harvard University"}, {"id": 4579, "fullname": "Alexandre Proutiere", "url": "http://virtual.aistats.org/api/miniconf/users/4579?format=json", "institution": "KTH Royal Institute of Technology"}, {"id": 22295, "fullname": "Richard Combes", "url": "http://virtual.aistats.org/api/miniconf/users/22295?format=json", "institution": "CentraleSupelec"}], "abstract": "Training or fine-tuning large language model (LLM)\u2013based systems often requires costly human feedback, yet there is limited understanding of how to minimize such intervention while maintaining strong error guarantees. We study this problem for LLM-based classification systems in an active learning framework: an agent sequentially labels $d$-dimensional query embeddings drawn i.i.d. from an unknown distribution by either calling a costly expert or guessing with no feedback, with the goal of minimizing regret relative to an oracle with free expert access. When the horizon $T$ is at least exponential in the embedding dimension $d$, the geometry of the class regions can be learned. In this regime, we propose the Conservative Hull-based Classifier (CHC), which maintains convex hulls of expert-labeled queries and calls the expert when a query lands outside all known hulls. CHC attains $\\mathcal{O}(\\log^d T)$ regret in $T$ and is minimax optimal for $d=1$. Otherwise, the geometry cannot be reliably learned in general. We show that for queries drawn from a subgaussian mixture and $T \\le e^d$, a Center-based Classifier (CC) achieves regret proportional to $N\\log{N}$ where $N$ is the number of labels. To bridge these regimes, we introduce the Generalized Hull-based Classifier (GHC), a practical extension of CHC that enables more aggressive guessing via a tunable parameter. Our approach is validated on real-world question-answering datasets using state-of-the-art text embedding models.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13528", "url": null, "sourceid": 1309, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=cWb7dddPyI", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11221, "modified": "2026-03-29T20:43:02.363667-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=cWb7dddPyI", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "105", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13359, "uid": "228499b55310264a8ea0e27b6e7c6ab6", "name": "Multi-Armed Sampling Problem and the End of Exploration", "authors": [{"id": 21931, "fullname": "Mohammad Pedramfar", "url": "http://virtual.aistats.org/api/miniconf/users/21931?format=json", "institution": "McGill University/Mila"}, {"id": 14764, "fullname": "Siamak Ravanbakhsh", "url": "http://virtual.aistats.org/api/miniconf/users/14764?format=json", "institution": "McGill - Mila"}], "abstract": "This paper introduces the framework of multi-armed sampling, which serves as the sampling counterpart to the optimization problem of multi-armed bandits. Our primary motivation is to rigorously examine the exploration-exploitation trade-off in the context of sampling. We systematically define plausible notions of regret for this framework and establish corresponding lower bounds. We then propose a simple algorithm that achieves near-optimal regret bounds. Our theoretical results suggest that, in contrast to optimization, sampling barely requires any exploration. To further connect our findings with those of multi-armed bandits, we define a continuous family of problems and associated regret measures that smoothly interpolate and unify multi-armed sampling and multi-armed bandit problems using a temperature parameter. We believe that the multi-armed sampling framework and our findings in this setting can play a foundational role in the study of sampling, including recent neural samplers, much like the role of multi-armed bandits in reinforcement learning. In particular, our work sheds light on the role of exploration (or lack thereof) and the convergence properties of algorithms for entropy-regularized reinforcement learning, fine-tuning of pretrained models and reinforcement learning with human feedback (RLHF).", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13359", "url": null, "sourceid": 1585, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=wVIdRkkbGe", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11052, "modified": "2026-03-29T20:42:55.721927-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=wVIdRkkbGe", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "105", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13742, "uid": "e1e32e235eee1f970470a3a6658dfdd5", "name": "Meet Me at the Arm: The Cooperative Multi Armed Bandits Problem with Shareable Arms", "authors": [{"id": 22752, "fullname": "Xinyi Hu", "url": "http://virtual.aistats.org/api/miniconf/users/22752?format=json", "institution": "Boston University"}, {"id": 17661, "fullname": "Aldo Pacchiano", "url": "http://virtual.aistats.org/api/miniconf/users/17661?format=json", "institution": "Boston University, Boston University"}], "abstract": "We study the decentralized multi-player multi-armed bandits (MMAB) problem under a no-sensing setting, where each player receives only their own reward and obtains no information about collisions. Each arm has an unknown capacity, and if the number of players pulling an arm exceeds its capacity, all players involved receive zero reward. This setting generalizes the classical unit-capacity model and introduces new challenges in coordination and capacity discovery under severe feedback limitations. We propose A-CAPELLA (Algorithm for Capacity-Aware Parallel Elimination for Learning and Allocation), a decentralized learning algorithm that achieves logarithmic regret in this generalized regime via protocol-driven coordination.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13742", "url": null, "sourceid": 483, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=ElXIobrn5n", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11435, "modified": "2026-03-29T20:43:11.207481-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=ElXIobrn5n", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "106", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13545, "uid": "0f3d014eead934bbdbacb62a01dc4831", "name": "Multiclass Local Calibration with the Jensen-Shannon Distance", "authors": [{"id": 22330, "fullname": "Cesare Barbera", "url": "http://virtual.aistats.org/api/miniconf/users/22330?format=json", "institution": "University of Pisa"}, {"id": 22331, "fullname": "Lorenzo Perini", "url": "http://virtual.aistats.org/api/miniconf/users/22331?format=json", "institution": "Meta"}, {"id": 22332, "fullname": "Giovanni De Toni", "url": "http://virtual.aistats.org/api/miniconf/users/22332?format=json", "institution": "Fondazione Bruno Kessler"}, {"id": 22333, "fullname": "Andrea Passerini", "url": "http://virtual.aistats.org/api/miniconf/users/22333?format=json", "institution": "University of Trento"}, {"id": 22334, "fullname": "Andrea Pugnana", "url": "http://virtual.aistats.org/api/miniconf/users/22334?format=json", "institution": "University of Trento"}], "abstract": "Developing trustworthy Machine Learning (ML) models requires their predicted probabilities to be well-calibrated, meaning they should reflect true-class frequencies.  Among calibration notions in multiclass classification, strong calibration is the most stringent, as it requires all predicted probabilities to be simultaneously calibrated across all classes. However, existing approaches to multiclass calibration lack a notion of distance among inputs, which makes them vulnerable to proximity bias: predictions in sparse regions of the feature space are systematically miscalibrated.  In this work, we address this main shortcoming by introducing a local perspective on multiclass calibration. First, we formally define multiclass local calibration and establish its relationship with strong calibration. Second, we theoretically analyze the pitfalls of existing evaluation metrics when applied to multiclass local calibration. Third, we propose a practical method to enhance local calibration in Neural Networks, which enforces alignment between predicted probabilities and local estimates of class frequencies using the Jensen-Shannon distance. Finally, we empirically validate our approach against existing multiclass calibration techniques.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13545", "url": null, "sourceid": 1512, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=atUXOpyCox", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11238, "modified": "2026-03-29T20:43:02.951936-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=atUXOpyCox", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "106", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13698, "uid": "5d6646aad9bcc0be55b2c82f69750387", "name": "Multiple Invertible and Partial-Equivariant Function for Latent Vector Transformation to Enhance Disentanglement in VAEs", "authors": [{"id": 20626, "fullname": "Hee-Jun Jung", "url": "http://virtual.aistats.org/api/miniconf/users/20626?format=json", "institution": "Gwangju Institute of Science and Technology"}, {"id": 22654, "fullname": "Jaehyoung Jeong", "url": "http://virtual.aistats.org/api/miniconf/users/22654?format=json", "institution": "Gwangju Institute of Science and Technology"}, {"id": 22655, "fullname": "Kangil Kim", "url": "http://virtual.aistats.org/api/miniconf/users/22655?format=json", "institution": "Gwangju Institute of Science and Technology"}], "abstract": "Disentanglement learning is central to understanding and reusing learned representations in variational autoencoders (VAEs).  Although equivariance has been explored in this context, effectively exploiting it for disentanglement remains challenging. In this paper, we propose a novel method, called \\textit{Multiple Invertible and Partial-Equivariant Transformation} (MIPE-Transformation), which integrates two main parts:  (1) \\textit{Invertible and Partial-Equivariant Transformation} (IPE-Transformation), guaranteeing an invertible latent-to\u2013transformed-latent mapping while preserving partial input-to-latent equivariance in the transformed latent space; and (2) \\textit{Exponential-Family Conversion} (EF-Conversion) to extend the standard Gaussian prior to an approximate exponential family via a learnable conversion. In experiments on the 3D Cars, 3D Shapes, and dSprites datasets, MIPE-Transformation improves the disentanglement performance of state-of-the-art VAEs.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13698", "url": null, "sourceid": 2141, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=IYZhiqul7Z", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11391, "modified": "2026-03-29T20:43:09.261292-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=IYZhiqul7Z", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "107", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13632, "uid": "1e48c4420b7073bc11916c6c1de226bb", "name": "Moonwalk: Inverse-Forward Differentiation", "authors": [{"id": 22524, "fullname": "Dmitrii Krylov", "url": "http://virtual.aistats.org/api/miniconf/users/22524?format=json", "institution": "University of California, Irvine"}, {"id": 22525, "fullname": "Armin Karamzade", "url": "http://virtual.aistats.org/api/miniconf/users/22525?format=json", "institution": "University of California, Irvine"}, {"id": 19913, "fullname": "Roy Fox", "url": "http://virtual.aistats.org/api/miniconf/users/19913?format=json", "institution": "UCI"}], "abstract": "Backpropagation is effective for gradient computation but requires large memory, limiting scalability. This work explores a novel inverse-forward automatic differentiation mode as an alternative for gradient computation in the relatively broad class of neural networks that have surjective differentials (called submersive networks), showing its potential to reduce the memory footprint without substantial drawbacks. We introduce a novel technique based on a vector\u2013inverse-Jacobian product that accelerates the forward computation of gradients compared to na\u00efve forward-mode methods while retaining their advantages of memory reduction and preserving the fidelity of true gradients. Our method, Moonwalk, significantly reduces the memory footprint of neural network training while achieving performance comparable to backpropagation, making it a compelling alternative for efficient model training.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13632", "url": null, "sourceid": 1010, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=PbtqGC9mIx", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11325, "modified": "2026-03-29T20:43:06.513910-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=PbtqGC9mIx", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "107", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13778, "uid": "f93882cbd8fc7fb794c1011d63be6fb6", "name": "Multilayer Correlation Clustering", "authors": [{"id": 23180, "fullname": "Atsushi Miyauchi", "url": "http://virtual.aistats.org/api/miniconf/users/23180?format=json", "institution": "Intesa Sanpaolo S.p.A."}, {"id": 22842, "fullname": "Florian Adriaens", "url": "http://virtual.aistats.org/api/miniconf/users/22842?format=json", "institution": "University of Helsinki"}, {"id": 22843, "fullname": "Francesco Bonchi", "url": "http://virtual.aistats.org/api/miniconf/users/22843?format=json", "institution": "Intesa Sanpaolo S.p.A."}, {"id": 22844, "fullname": "Nikolaj Tatti", "url": "http://virtual.aistats.org/api/miniconf/users/22844?format=json", "institution": "University of Helsinki"}], "abstract": "We establish Multilayer Correlation Clustering, a novel generalization of Correlation Clustering to the multilayer setting. In this model, we are given a series of inputs of Correlation Clustering (called layers) over the common set $V$ of $n$ elements. The goal is to find a clustering of $V$ that minimizes the $\\ell_p$-norm ($p\\geq 1$) of the multilayer-disagreements vector, which is defined as the vector (with dimension equal to the number of layers), each element of which represents the disagreements of the clustering on the corresponding layer. For this generalization, we first design an $O(L\\log n)$-approximation algorithm, where $L$ is the number of layers. We then study an important special case of our problem, namely the problem with the so-called probability constraint. For this case, we first give an $(\\alpha+2)$-approximation algorithm, where $\\alpha$ is any possible approximation ratio for the single-layer counterpart. Furthermore, we design a $4$-approximation algorithm, which improves the above approximation ratio of $\\alpha+2=4.5$ for the general probability-constraint case.  Computational experiments using real-world datasets support our theoretical findings and demonstrate the practical effectiveness of our proposed algorithms.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13778", "url": null, "sourceid": 1307, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=CDhk49SPuH", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11471, "modified": "2026-03-29T20:43:12.679719-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=CDhk49SPuH", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "109", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13540, "uid": "158f3069a435b314a80bdcb024f8e422", "name": "Mixture Proportion Estimation and Weakly-supervised Kernel Test for Conditional Independence", "authors": [{"id": 22318, "fullname": "Yushi Hirose", "url": "http://virtual.aistats.org/api/miniconf/users/22318?format=json", "institution": "Tokyo Institute of Technology, Tokyo Institute of Technology"}, {"id": 22319, "fullname": "Akito Narahara", "url": "http://virtual.aistats.org/api/miniconf/users/22319?format=json", "institution": "Institute of Science Tokyo"}, {"id": 20592, "fullname": "Takafumi Kanamori", "url": "http://virtual.aistats.org/api/miniconf/users/20592?format=json", "institution": "Institute of Science Tokyo/RIKEN AIP"}], "abstract": "Mixture proportion estimation (MPE) aims to estimate class priors from unlabeled data. This task is a critical component in weakly supervised learning such as PU learning, learning with label noise, and domain adaptation. Existing MPE methods rely on the _irreducibility_ assumption or its variant for identifiability. In this paper, we propose novel assumptions based on conditional independence (CI) given the class label, which ensure identifiability even when irreducibility does not hold. We develop method of moments estimators under these assumptions and analyze their asymptotic properties. Furthermore, we present weakly-supervised kernel tests to validate the CI assumptions, which are of independent interest in applications such as causal discovery and fairness evaluation. Empirically, we demonstrate the improved performance of our estimators compared with existing methods and that our tests successfully control both type I and type II errors.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13540", "url": null, "sourceid": 313, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=bS5M1ZF3Es", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11233, "modified": "2026-03-29T20:43:02.760874-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=bS5M1ZF3Es", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "109", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13382, "uid": "8a0e1141fd37fa5b98d5bb769ba1a7cc", "name": "MLorc: Momentum Low-rank Compression for Large Language Model Adaptation", "authors": [{"id": 12615, "fullname": "Wei Shen", "url": "http://virtual.aistats.org/api/miniconf/users/12615?format=json", "institution": "University of Virginia"}, {"id": 21975, "fullname": "Yaxiang Zhang", "url": "http://virtual.aistats.org/api/miniconf/users/21975?format=json", "institution": "National University of Singapore"}, {"id": 12616, "fullname": "Minhui Huang", "url": "http://virtual.aistats.org/api/miniconf/users/12616?format=json", "institution": "UC Davis"}, {"id": 18368, "fullname": "Mengfan Xu", "url": "http://virtual.aistats.org/api/miniconf/users/18368?format=json", "institution": "University of Massachusetts at Amherst"}, {"id": 21976, "fullname": "Jiawei Zhang", "url": "http://virtual.aistats.org/api/miniconf/users/21976?format=json", "institution": "University of Wisconsin - Madison"}, {"id": 182, "fullname": "Cong Shen", "url": "http://virtual.aistats.org/api/miniconf/users/182?format=json", "institution": "University of Virginia"}], "abstract": "With increasing size of large language models (LLMs), full-parameter fine-tuning imposes substantial memory demands. To alleviate this, we propose a novel memory-efficient training paradigm called Momentum Low-rank compression (MLorc). By directly compressing and reconstructing momentum rather than gradients, MLorc avoids imposing a fixed-rank constraint on weight update matrices and better preserves the training dynamics of full-parameter fine-tuning, in contrast to existing low-rank approaches such as LoRA and GaLore. Empirically, MLorc consistently outperforms other memory-efficient training methods, matches or even exceeds the performance of full fine-tuning with a small rank (e.g., $r=4$), and generalizes well across different optimizers -- all while not compromising time or memory efficiency. Furthermore, we provide a theoretical guarantee for its convergence under reasonable assumptions.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13382", "url": null, "sourceid": 697, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=sw7gHBmXls", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11075, "modified": "2026-03-29T20:42:56.623411-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=sw7gHBmXls", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "110", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13896, "uid": "362e80d4df43b03ae6d3f8540cd63626", "name": "Efficient Swap Regret Minimization in Combinatorial Bandits", "authors": [{"id": 23074, "fullname": "Andreas Kontogiannis", "url": "http://virtual.aistats.org/api/miniconf/users/23074?format=json", "institution": "National Technical University of Athens"}, {"id": 19904, "fullname": "Vasilis Pollatos", "url": "http://virtual.aistats.org/api/miniconf/users/19904?format=json", "institution": "Archimedes, Athena Research Center, Greece &amp; NKUA"}, {"id": 18546, "fullname": "Panayotis Mertikopoulos", "url": "http://virtual.aistats.org/api/miniconf/users/18546?format=json", "institution": "French National Center for Scientific Research"}, {"id": 3969, "fullname": "Ioannis Panageas", "url": "http://virtual.aistats.org/api/miniconf/users/3969?format=json", "institution": "UC Irvine"}], "abstract": "This paper addresses the problem of designing efficient no-swap regret algorithms for combinatorial bandits, where the number of actions $N$ is exponentially large in the dimensionality of the problem.  In this setting, designing efficient no-swap regret translates to sublinear -- in horizon $T$ -- swap regret with polylogarithmic dependence on $N$. In contrast to the weaker notion of external regret minimization -- a problem which is fairly well understood in the literature -- achieving no-swap regret with a polylogarithmic dependence on $N$ has remained elusive in combinatorial bandits. Our paper resolves this challenge, by introducing a no-swap-regret learning algorithm with regret that scales polylogarithmically in $N$ and is tight for the class of combinatorial bandits. To ground our results, we also demonstrate how to implement the proposed algorithm efficiently -- that is, with a per-iteration complexity that also scales polylogarithmically in $N$ -- across a wide range of well-studied applications.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13896", "url": null, "sourceid": 848, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=1ryyvVCWgD", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11589, "modified": "2026-03-29T20:43:17.692527-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=1ryyvVCWgD", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "111", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13681, "uid": "2a38a4a9316c49e5a833517c45d31070", "name": "Multi-Component VAE with Gaussian Markov Random Field", "authors": [{"id": 19749, "fullname": "Fouad Oubari", "url": "http://virtual.aistats.org/api/miniconf/users/19749?format=json", "institution": "ENS Paris Saclay"}, {"id": 20634, "fullname": "Mohamed El Baha", "url": "http://virtual.aistats.org/api/miniconf/users/20634?format=json", "institution": "Michelin"}, {"id": 22616, "fullname": "Rapha\u00ebl Meunier", "url": "http://virtual.aistats.org/api/miniconf/users/22616?format=json", "institution": "Michelin"}, {"id": 22617, "fullname": "Rodrigue D\u00e9catoire", "url": "http://virtual.aistats.org/api/miniconf/users/22617?format=json", "institution": "Michelin"}, {"id": 22618, "fullname": "Mathilde MOUGEOT", "url": "http://virtual.aistats.org/api/miniconf/users/22618?format=json", "institution": "Ecole Normale Superieure"}], "abstract": "Multi-component datasets with intricate dependencies challenge current generative modeling techniques. Existing Multi-component Variational AutoEncoders rely on simplified aggregation strategies that compromise structural coherence across generated components. We introduce the Gaussian Markov Random Field Multi-Component Variational AutoEncoder, embedding Gaussian Markov Random Fields into both prior and posterior distributions to explicitly model cross-component relationships. This enables richer representation and faithful reproduction of complex interactions. Empirically, our model achieves state-of-the-art performance on a synthetic Copula dataset designed for intricate component relationships, competitive results on PolyMNIST, and significantly enhanced structural coherence on the real-world BIKED dataset, demonstrating its suitability for applications demanding robust multi-component coherence.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13681", "url": null, "sourceid": 88, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=KGvXelrMHL", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11374, "modified": "2026-03-29T20:43:08.540966-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=KGvXelrMHL", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "111", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13615, "uid": "020bf2c45e7bb322f89a226bd2c5d41b", "name": "Non-Stationary Functional Bilevel Optimization", "authors": [{"id": 14183, "fullname": "Jason Bohne", "url": "http://virtual.aistats.org/api/miniconf/users/14183?format=json", "institution": ""}, {"id": 22495, "fullname": "Ieva Petrulionyt\u0117", "url": "http://virtual.aistats.org/api/miniconf/users/22495?format=json", "institution": "INRIA Rhone-Alpes"}, {"id": 4623, "fullname": "Michael Arbel", "url": "http://virtual.aistats.org/api/miniconf/users/4623?format=json", "institution": "UCL"}, {"id": 4098, "fullname": "Julien Mairal", "url": "http://virtual.aistats.org/api/miniconf/users/4098?format=json", "institution": "INRIA"}, {"id": 19596, "fullname": "Pawel Polak", "url": "http://virtual.aistats.org/api/miniconf/users/19596?format=json", "institution": "Stony Brook University"}], "abstract": "Functional bilevel optimization (FBO) provides a powerful framework for hierarchical learning in function spaces, yet current methods are limited to static offline settings and perform suboptimally in online, non-stationary scenarios. We propose **SmoothFBO**, the first algorithm for non-stationary FBO with both theoretical guarantees and practical scalability. SmoothFBO introduces a time-smoothed stochastic hypergradient estimator that reduces variance through a window parameter, enabling stable outer-loop updates with sublinear regret. Importantly, the classical parametric bilevel case is a special reduction of our framework, making SmoothFBO a natural extension to online, non-stationary settings. Empirically, SmoothFBO consistently outperforms existing FBO methods in non-stationary hyperparameter optimization and model-based reinforcement learning, demonstrating its practical effectiveness. Together, these results establish SmoothFBO as a general, theoretically grounded, and practically viable foundation for bilevel optimization in online, non-stationary scenarios.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13615", "url": null, "sourceid": 2419, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=SqSQKeduRH", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11308, "modified": "2026-03-29T20:43:05.746066-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=SqSQKeduRH", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "111", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13901, "uid": "32b30a250abd6331e03a2a1f16466346", "name": "NeST-BO: Fast Local Bayesian Optimization via Newton-Step Targeting of Gradient and Hessian Information", "authors": [{"id": 23081, "fullname": "Wei-Ting Tang", "url": "http://virtual.aistats.org/api/miniconf/users/23081?format=json", "institution": "University of Wisconsin - Madison"}, {"id": 23082, "fullname": "Akshay Kudva", "url": "http://virtual.aistats.org/api/miniconf/users/23082?format=json", "institution": "Ohio State University, Columbus"}, {"id": 23083, "fullname": "Joel Paulson", "url": "http://virtual.aistats.org/api/miniconf/users/23083?format=json", "institution": "University of Wisconsin - Madison"}], "abstract": "Bayesian optimization (BO) is effective for expensive black-box problems but remains challenging in high dimensions. We propose NeST-BO, a curvature-aware local BO method that targets a (modified) Newton step by jointly learning gradient and Hessian information with Gaussian process (GP) surrogates, and selecting evaluations via a one-step lookahead bound on the Newton-step error. We show that this bound contracts with batch size, so NeST-BO drives the step error to zero; in well-behaved neighborhoods it recovers the fast local convergence behavior of inexact/modified Newton methods, while standard safeguards support global convergence to stationary points. To improve scaling with problem dimension, we optimize the acquisition in low-dimensional embedded subspaces (random or learned), reducing the dominant cost of learning curvature from $O(d^2)$ to $O(m^2)$ with $m \\ll d$ while preserving step targeting. Across high-dimensional synthetic and real-world problems, including cases with thousands of variables and unknown active subspaces, NeST-BO consistently yields faster convergence and better final values than state-of-the-art local and high-dimensional BO baselines.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13901", "url": null, "sourceid": 809, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=1MHLQM0MKM", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11594, "modified": "2026-03-29T20:43:17.873959-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=1MHLQM0MKM", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "112", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13665, "uid": "b5b41fac0361d157d9673ecb926af5ae", "name": "Narrowing Action Choices with AI Improves Human Sequential Decisions", "authors": [{"id": 19839, "fullname": "Eleni Straitouri", "url": "http://virtual.aistats.org/api/miniconf/users/19839?format=json", "institution": "Max Planck Institute for Software Systems"}, {"id": 22160, "fullname": "Stratis Tsirtsis", "url": "http://virtual.aistats.org/api/miniconf/users/22160?format=json", "institution": "Hasso Plattner Institute"}, {"id": 22162, "fullname": "Ander Artola Velasco", "url": "http://virtual.aistats.org/api/miniconf/users/22162?format=json", "institution": "Max Planck Institute for Software Systems"}, {"id": 22164, "fullname": "Manuel Gomez Rodriguez", "url": "http://virtual.aistats.org/api/miniconf/users/22164?format=json", "institution": "MPI-SWS"}], "abstract": "Recent work has shown that, in classification tasks, it is possible to design decision support systems that do not require human experts to understand when to cede agency to a classifier or when to exercise their own agency to achieve complementarity---experts using these systems make more accurate predictions than those made by the experts or the classifier alone. The key principle underpinning these systems reduces to adaptively controlling the level of human agency, by design. Can we use the same principle to achieve complementarity in sequential decision making tasks? In this paper, we answer this question affirmatively. We develop a decision support system that uses a pre-trained AI agent to narrow down the set of actions a human can take to a subset, and then asks the human to take an action from the action set. Along the way, we also introduce a bandit algorithm that leverages the smoothness properties of the action sets provided by our system to efficiently optimize the level of human agency. To evaluate our decision support system, we conduct a large-scale human subject study ($n = 1{,}600$) where participants play a wildfire mitigation game. We find that participants who play the game supported by our system outperform those who play on their own by $\\sim$$30$\\% and the AI agent used by our system by $>$$2$\\%, even though the AI agent largely outperforms participants playing without support.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13665", "url": null, "sourceid": 502, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=MW1Pb6dqrK", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11358, "modified": "2026-03-29T20:43:07.875407-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=MW1Pb6dqrK", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "112", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13533, "uid": "074177d3eb6371e32c16c55a3b8f706b", "name": "Numerical Fragility in Transformers: A Layer-wise Theory for Explaining, Forecasting, and Mitigating Instability", "authors": [{"id": 23100, "fullname": "Jinwoo Baek", "url": "http://virtual.aistats.org/api/miniconf/users/23100?format=json", "institution": "Oregon State University"}], "abstract": "We study numerical instability in Transformers trained in low precision and give a first-order, module-wise theory that predicts when and where forward errors amplify. For self-attention we derive a layer-wise bound that factors into three interpretable diagnostics: a score-scale ratio $\\kappa_{\\mathrm{score}}$, a row-wise softmax sensitivity $\\kappa_{\\mathrm{softmax}}$, and the value conditioning $\\kappa(V)$. We prove a residual relaxation inequality showing that residual blocks attenuate depth-wise accumulation, and we give a precision- and width-aware LayerNorm indicator $\\rho_{\\mathrm{LN}}$ with a corresponding first-order error bound in the $\\varepsilon$-dominated regime. These pieces combine into a unified forward-stability bound whose per-layer right-hand side is directly estimable during training.  On Tiny-ViT/CIFAR-10, we evaluate the bound and its components in three studies. (Exp-1) A decisive scatter shows that the combined predictor $ \\kappa_{\\mathrm{softmax}} \\cdot (1 + \\kappa_{\\mathrm{score}}) \\cdot \\kappa(V) \\cdot \\lVert W_O \\rVert_2 + \\kappa_{\\mathrm{eff}} + C_{\\mathrm{LN}} $ tracks the observed FP32 $ \\leftrightarrow $ LP forward mismatch across seeds, widths, and precisions; scaling by $ \\epsilon_{\\mathrm{mach}} $ collapses mixed-precision points. (Exp-2) The time series of the maximal $ \\kappa_{\\mathrm{softmax}} $ acts as an early-warning signal, consistently leading increases in forward error by 16--24 steps with correlations of 0.65--0.82 (permutation-test $ p \\approx 10^{-3} $) and a Precision@K of 0.89--1.00. (Exp-3) Guided by $ \\rho_{\\mathrm{LN}} $, a simple LayerNorm-$ \\varepsilon $ intervention that targets $ \\rho_* $ yields small but consistent stabilization (e.g., mean tail-loss reduction $ \\approx 0.010 $ at $ \\rho_* = 0.6 $, cap $ = 10^{-2} $), with no architectural changes and negligible overhead.  Overall, our theory provides actionable, unitless diagnostics that (i) explain when self-attention is numerically fragile, (ii) forecast instability in advance, and (iii) motivate a minimally invasive mitigation that trends beneficial in practice.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13533", "url": null, "sourceid": 2273, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=c2dW3ivn7Y", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11226, "modified": "2026-03-29T20:43:02.529927-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=c2dW3ivn7Y", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "112", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13831, "uid": "c5cc17e395d3049b03e0f1ccebb02b4d", "name": "Neural Doubly Robust Proximal Causal Estimation", "authors": [{"id": 22955, "fullname": "Ruolin Meng", "url": "http://virtual.aistats.org/api/miniconf/users/22955?format=json", "institution": "Duke University"}, {"id": 22956, "fullname": "Dhanajit Brahma", "url": "http://virtual.aistats.org/api/miniconf/users/22956?format=json", "institution": "Duke University"}, {"id": 914, "fullname": "Ricardo Henao", "url": "http://virtual.aistats.org/api/miniconf/users/914?format=json", "institution": "Duke University"}, {"id": 317, "fullname": "Lawrence Carin Duke", "url": "http://virtual.aistats.org/api/miniconf/users/317?format=json", "institution": "CS"}], "abstract": "We consider the challenging task of estimating treatment effects from observational data under the assumption that there are unobserved confounders. We employ the proximal causal estimation framework, that assumes access to control (proxy) measurements that contain information about unobserved confounders. We consider outcome and treatment bridges, which provide two distinct ways of estimating causal effects. We also consider a doubly-robust approach, based on combining the outcome and treatment bridges, which is robust in expectation to either (but not both) of the two bridge functions being misspecified. We present a new theoretical bound on the estimation accuracy of the treatment bridge, and we analyze the variance of the doubly-robust estimator. We investigate the impact of autoencoder-based regularization through an ablation study, finding that simpler models sometimes outperform more complex variants. Comparisons with state-of-the-art methods on synthetic and real-world data demonstrate the advantages of our approach.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13831", "url": null, "sourceid": 1450, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=7vNGiz4CoH", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11524, "modified": "2026-03-29T20:43:14.870485-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=7vNGiz4CoH", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "113", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13551, "uid": "635440afdfc39fe37995fed127d7df4f", "name": "On Global Convergence Rates for Federated Softmax Policy Gradient under Heterogeneous Environments", "authors": [{"id": 14780, "fullname": "Safwan Labbi", "url": "http://virtual.aistats.org/api/miniconf/users/14780?format=json", "institution": "Ecole Polytechnique"}, {"id": 13050, "fullname": "Paul Mangold", "url": "http://virtual.aistats.org/api/miniconf/users/13050?format=json", "institution": "\u00c9cole polytechnique, France"}, {"id": 23265, "fullname": "Daniil Tiapkin", "url": "http://virtual.aistats.org/api/miniconf/users/23265?format=json", "institution": "Google Deepmind"}, {"id": 22351, "fullname": "Eric Moulines", "url": "http://virtual.aistats.org/api/miniconf/users/22351?format=json", "institution": "Mohamed bin Zayed University of Artificial Intelligence"}], "abstract": "We provide global convergence rates for vanilla and entropy-regularized federated softmax stochastic policy gradient ($\\texttt{FedPG}$) with local training.  We show that $\\texttt{FedPG}$ converges to a near-optimal policy in terms of the average agent value, with a gap controlled by the level of heterogeneity. Remarkably, we obtain the first convergence rates for entropy-regularized policy gradient *with explicit constants*, leveraging a projection-like operator. Our results build upon a new analysis of federated averaging for non-convex objectives, based on the observation that the \u0141ojasiewicz-type inequalities from the single-agent setting  (Mei et al., 2020) do not hold for the federated objective.  This uncovers a fundamental difference between single-agent and federated reinforcement learning: while single-agent optimal policies can be deterministic, federated objectives may inherently require stochastic policies.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13551", "url": null, "sourceid": 2252, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=aPAq9U7MFm", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11244, "modified": "2026-03-29T20:43:03.180587-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=aPAq9U7MFm", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "114", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13365, "uid": "1579779b98ce9edb98dd85606f2c119d", "name": "Neuron Block Dynamics for XOR Classification with Zero-Margin", "authors": [{"id": 12427, "fullname": "Guillaume Braun", "url": "http://virtual.aistats.org/api/miniconf/users/12427?format=json", "institution": "RIKEN AIP"}, {"id": 10230, "fullname": "Masaaki Imaizumi", "url": "http://virtual.aistats.org/api/miniconf/users/10230?format=json", "institution": "The University of Tokyo / RIKEN AIP"}], "abstract": "The ability of neural networks to learn useful features through stochastic gradient descent (SGD) is a cornerstone of their success. Most theoretical analyses focus on regression or on classification tasks with a positive margin, where worst-case gradient bounds suffice. In contrast, we study zero-margin nonlinear classification by analyzing the Gaussian XOR problem, where inputs are Gaussian and the XOR decision boundary determines labels. In this setting, a non-negligible fraction of data lies arbitrarily close to the boundary, breaking standard margin-based arguments. Building on Glasgow\u2019s (2024) analysis, we extend the study of training dynamics from discrete to Gaussian inputs and develop a framework for the dynamics of neuron blocks. We show that neurons cluster into four directions and that block-level signals evolve coherently, a phenomenon essential in the Gaussian setting where individual neuron signals vary significantly. Leveraging this block perspective, we analyze generalization without relying on margin assumptions, adopting an average-case view that distinguishes regions of reliable prediction from regions of persistent error. Numerical experiments confirm the predicted two-phase block dynamics and demonstrate their robustness beyond the Gaussian setting.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13365", "url": null, "sourceid": 1046, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=vtMOjOW6Gp", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11058, "modified": "2026-03-29T20:42:55.928136-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=vtMOjOW6Gp", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "114", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13641, "uid": "8698ff92115213ab187d31d4ee5da8ea", "name": "On propagation of chaos for the Fisher-Rao gradient flow in entropic mean-field optimization", "authors": [{"id": 22539, "fullname": "Petra Lazi\u0107", "url": "http://virtual.aistats.org/api/miniconf/users/22539?format=json", "institution": "University of Ljubljana"}, {"id": 22540, "fullname": "Linshan Liu", "url": "http://virtual.aistats.org/api/miniconf/users/22540?format=json", "institution": "Heriot-Watt University"}, {"id": 22541, "fullname": "Mateusz Majka", "url": "http://virtual.aistats.org/api/miniconf/users/22541?format=json", "institution": "Heriot-Watt University"}], "abstract": "We consider a class of optimization problems on the space of probability measures motivated by the mean-field approach to studying neural networks. Such problems can be solved by constructing continuous-time gradient flows that converge to the minimizer of the energy function under consideration, and then implementing discrete-time algorithms that approximate the flow. In this work, we focus on the Fisher-Rao gradient flow and we construct an interacting particle system that approximates the flow as its mean-field limit. We discuss the connection between the energy function, the gradient flow and the particle system and explain different approaches to smoothing out the energy function with an appropriate kernel in a way that allows for the particle system to be well-defined. We provide a rigorous proof of the existence and uniqueness of thus obtained kernelized flows, as well as a propagation of chaos result that provides a theoretical justification for using the corresponding kernelized particle systems as approximation algorithms in entropic mean-field optimization.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13641", "url": null, "sourceid": 1765, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=OYBQs5zwO4", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11334, "modified": "2026-03-29T20:43:06.848315-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=OYBQs5zwO4", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "115", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13763, "uid": "e96ed478dab8595a7dbda4cbcbee168f", "name": "Noise-Free Dynamic Rank-Adaptation via Riemannian Methods in Federated Fine-Tuning", "authors": [{"id": 22793, "fullname": "Zihan Zhou", "url": "http://virtual.aistats.org/api/miniconf/users/22793?format=json", "institution": "Auburn University"}, {"id": 22794, "fullname": "Yang Zhou", "url": "http://virtual.aistats.org/api/miniconf/users/22794?format=json", "institution": "Auburn University"}, {"id": 22795, "fullname": "Tianshi Che", "url": "http://virtual.aistats.org/api/miniconf/users/22795?format=json", "institution": "Auburn University"}, {"id": 22796, "fullname": "Zeru Zhang", "url": "http://virtual.aistats.org/api/miniconf/users/22796?format=json", "institution": "Auburn University"}, {"id": 22797, "fullname": "Jiaxiang Ren", "url": "http://virtual.aistats.org/api/miniconf/users/22797?format=json", "institution": "Aicadium"}, {"id": 22798, "fullname": "Da Yan", "url": "http://virtual.aistats.org/api/miniconf/users/22798?format=json", "institution": "Indiana University Bloomington"}, {"id": 22799, "fullname": "Zhe Jiang", "url": "http://virtual.aistats.org/api/miniconf/users/22799?format=json", "institution": "University of Florida"}, {"id": 22800, "fullname": "yelong shen", "url": "http://virtual.aistats.org/api/miniconf/users/22800?format=json", "institution": "Microsoft"}, {"id": 22801, "fullname": "Ruoming Jin", "url": "http://virtual.aistats.org/api/miniconf/users/22801?format=json", "institution": "Kent State University"}, {"id": 22802, "fullname": "Jianfeng Gao", "url": "http://virtual.aistats.org/api/miniconf/users/22802?format=json", "institution": "Microsoft Research"}], "abstract": "Rank-adaptive low-rank adaptation (LoRA), a parameter-efficient fine-tuning (PEFT) technology, has achieved state-of-the-art performance in fine-tuning foundation models (FM). Directly transplanting the rank-adaptive LoRA methods from centralized learning to federated learning raises two critical issues: aggregation noise and rank drift. We presents Riemannian LoRA algorithm with adaptive rank for federated fine-tuning of foundation models (FFT-FM), RAFFT, which resolves both issues and significantly improves the computational cost. First, by utilizing Riemannian Procrustes analysis, we propose a Riemannian parameter matching method to avoid aggregation noise and ensure effective FFT-FM with rank-adaptive LoRA while cutting SVD cost by decomposing only low-dimensional $r \\times r$ matrices, where $r$ is the rank parameter in the LoRA. We theoretically derive the equivalence between our RAFFT algorithm with rank-adaptive LoRA for the FFT-FM and the standard FFT-FM on the full parameter matrices based on FedAvg and verify the bounded error introduced by approximation. Second, by leveraging Riemannian manifold theory, we develop a Riemannian gradient descent (RGD) method to guarantee the local full parameter matrices on clients in the form of low-rank ones with fixed rank optimized by the server in each FFT-FM round, for alleviating the rank-drift issue to speed up the convergence of RAFFT. We theoretically demonstrate that the RGD optimization on the Riemannian manifold ensures the rank invariance during the local update process and the RGD optimization can converge in the FFT-FM context.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13763", "url": null, "sourceid": 218, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=DT6asUtXRl", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11456, "modified": "2026-03-29T20:43:12.070452-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=DT6asUtXRl", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "116", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13370, "uid": "d095a94d20dcaf7aa07301948549bede", "name": "On the bias of variational resampling", "authors": [{"id": 19593, "fullname": "Axel Finke", "url": "http://virtual.aistats.org/api/miniconf/users/19593?format=json", "institution": "Newcastle University"}, {"id": 4625, "fullname": "Oskar Kviman", "url": "http://virtual.aistats.org/api/miniconf/users/4625?format=json", "institution": "KTH, Sweden"}, {"id": 23260, "fullname": "Nicola Branchini", "url": "http://virtual.aistats.org/api/miniconf/users/23260?format=json", "institution": "The University of Warwick"}, {"id": 4628, "fullname": "Victor Elvira", "url": "http://virtual.aistats.org/api/miniconf/users/4628?format=json", "institution": "University of Edinburgh"}], "abstract": "Variational resampling (VR) is a method for deterministically resampling the $N$ particles in sequential Monte Carlo (SMC) algorithms (also known as particle filters), by minimising the Kullback--Leibler divergence from the empirical measure of the $N$ weighted original particles to the empirical measure of $M$ unweighted resampled particles. The combination of VR with a weight transformation (called smoothing weights) has shown to often yield a smaller mean-square error (MSE) than standard resampling schemes in the literature. However, its bias has never been investigated. In this paper, we first show that VR incurs a weighting bias and a truncation bias. We then propose a mechanism to alleviate the weighting bias through an uneven weighting of the resampled particles. We also show that the truncation bias implies that the particle approximation of the target distribution is restricted to a region in which the unnormalised weights are larger than some threshold with high probability. We prove that this probability approaches $1$ if $M = \\mathrm{O}(N)$ as $N \\to \\infty$. Finally, we empirically illustrate that the smaller MSE of VR observed in the literature can be attributed to an underestimation of uncertainty caused by the use of the smoothing weights.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13370", "url": null, "sourceid": 2357, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=uLtVt5lnAF", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11063, "modified": "2026-03-29T20:42:56.105572-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=uLtVt5lnAF", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "116", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13327, "uid": "359f38463d487e9e29bd20e24f0c050a", "name": "On the Convergence and Stability of Distributed Sub-model Training", "authors": [{"id": 21873, "fullname": "Yuyang Deng", "url": "http://virtual.aistats.org/api/miniconf/users/21873?format=json", "institution": "Columbia University"}, {"id": 14931, "fullname": "Fuli Qiao", "url": "http://virtual.aistats.org/api/miniconf/users/14931?format=json", "institution": "Pennsylvania State University"}, {"id": 724, "fullname": "Mehrdad Mahdavi", "url": "http://virtual.aistats.org/api/miniconf/users/724?format=json", "institution": "Penn State"}], "abstract": "As learning models continue to grow in size, enabling on-device local training of these models has emerged as a critical challenge in federated learning. A popular solution is sub-model training, where the server only distributes randomly sampled sub-models to the edge clients, and clients only update these small models. However, those random sampling of sub-models may not give satisfying convergence performance. In this paper, observing the success of SGD with shuffling, we propose a distributed shuffled sub-model training, where the full model is partitioned into several sub-models in advance, and the server shuffles those sub-models, sends each of them to clients at each round, and by the end of local updating period, clients send back the updated sub-models, and server averages them. We establish the convergence rate of this algorithm. We also study the generalization of distributed sub-model training via stability analysis, and find that the sub-model training can improve the generalization via amplifying the stability of training process. The extensive experiments also validate our theoretical findings.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13327", "url": null, "sourceid": 1390, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=zacHFjKPyB", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11020, "modified": "2026-03-29T20:42:54.273804-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=zacHFjKPyB", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "116", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13541, "uid": "3a066bda8c96b9478bb0512f0a43028c", "name": "Non-Asymptotic Generalization and Optimization Bounds for Stochastic Gauss-Newton in Deep Neural Networks", "authors": [{"id": 22320, "fullname": "Semih Cayci", "url": "http://virtual.aistats.org/api/miniconf/users/22320?format=json", "institution": "Rheinisch Westf\u00e4lische Technische Hochschule Aachen"}], "abstract": "An important question in deep learning is how higher-order optimization methods affect generalization. In this work, we analyze a stochastic Gauss-Newton (SGN) method with Levenberg\u2013Marquardt damping and mini-batch sampling for training overparameterized deep neural networks with smooth activations in a regression setting. Our theoretical contributions are twofold. First, we establish finite-time convergence bounds via a variable-metric analysis in parameter space, with explicit dependencies on the batch size, network width and depth. Second, we derive non-asymptotic generalization bounds for SGN using uniform stability in the overparameterized regime, characterizing the impact of curvature, batch size, and overparameterization on generalization performance. Our theoretical results identify a favorable generalization regime for SGN in which a larger minimum eigenvalue of the Gauss\u2013Newton matrix along the optimization path, together with smaller batch sizes, yields tighter stability bounds.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13541", "url": null, "sourceid": 661, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=bQcZwpHJC8", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11234, "modified": "2026-03-29T20:43:02.793711-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=bQcZwpHJC8", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "117", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13329, "uid": "85d8ce590ad8981ca2c8286f79f59954", "name": "Off-policy Distributional Q($\\lambda$): Distributional RL without Importance Sampling", "authors": [{"id": 148, "fullname": "Yunhao Tang", "url": "http://virtual.aistats.org/api/miniconf/users/148?format=json", "institution": "DeepMind"}, {"id": 107, "fullname": "Mark Rowland", "url": "http://virtual.aistats.org/api/miniconf/users/107?format=json", "institution": "DeepMind"}, {"id": 21878, "fullname": "R\u00e9mi Munos", "url": "http://virtual.aistats.org/api/miniconf/users/21878?format=json", "institution": "Meta"}, {"id": 18575, "fullname": "Bernardo Avila Pires", "url": "http://virtual.aistats.org/api/miniconf/users/18575?format=json", "institution": "Google DeepMind"}, {"id": 289, "fullname": "Will Dabney", "url": "http://virtual.aistats.org/api/miniconf/users/289?format=json", "institution": "DeepMind"}], "abstract": "We introduce off-policy distributional Q($\\lambda$), a new addition to the family of off-policy distributional evaluation algorithms. Off-policy distributional Q($\\lambda$) does not apply importance sampling for off-policy learning, which introduces intriguing interactions with signed measures. Such unique properties distributional Q($\\lambda$) from other existing alternatives such as distributional Retrace. We characterize the algorithmic properties of distributional Q($\\lambda$) and validate theoretical insights with tabular experiments. We show how distributional Q($\\lambda$)-C51, a combination of Q($\\lambda$) with the C51 agent, exhibits promising results on deep RL benchmarks.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13329", "url": null, "sourceid": 197, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=zQT82QJh8c", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11022, "modified": "2026-03-29T20:42:54.357845-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=zQT82QJh8c", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "118", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13897, "uid": "fb2fcd534b0ff3bbed73cc51df620323", "name": "Online Learning-to-Defer with Varying Experts", "authors": [{"id": 22148, "fullname": "Yannis Montreuil", "url": "http://virtual.aistats.org/api/miniconf/users/22148?format=json", "institution": "National University of Singapore"}, {"id": 23075, "fullname": "Hoang Dang", "url": "http://virtual.aistats.org/api/miniconf/users/23075?format=json", "institution": "National University of Singapore"}, {"id": 23076, "fullname": "Maxime Meyer", "url": "http://virtual.aistats.org/api/miniconf/users/23076?format=json", "institution": "National University of Singapore"}, {"id": 22150, "fullname": "Lai Xing Ng", "url": "http://virtual.aistats.org/api/miniconf/users/22150?format=json", "institution": "Institute for Infocomm Research (I2R), A*STAR"}, {"id": 22151, "fullname": "Axel Carlier", "url": "http://virtual.aistats.org/api/miniconf/users/22151?format=json", "institution": "Institut Sup\u00e9rieur de l&#x27;A\u00e9ronautique et de l&#x27;Espace"}, {"id": 22152, "fullname": "Wei Ooi", "url": "http://virtual.aistats.org/api/miniconf/users/22152?format=json", "institution": "National University of Singapore"}], "abstract": "Learning-to-Defer (L2D) methods route each query either to a predictive model or to external experts. While existing work studies this problem in batch settings, real-world deployments require handling streaming data,  changing expert availability, and shifting expert distribution. We introduce the first online L2D algorithm for multiclass classification with bandit feedback and a dynamically varying pool of experts. Our method achieves regret guarantees of $O((n+n_e)T^{2/3})$ in general and $O((n+n_e)\\sqrt{T})$ under a low-noise condition, where $T$ is the time horizon, $n$ the number of labels, and $n_e$ the number of distinct experts observed across rounds. The analysis builds on novel $\\mathcal{H}$-consistency bounds for the online framework, combined with first-order methods for online convex optimization. Experiments on synthetic and real-world datasets demonstrate that our approach effectively extends standard Learning-to-Defer to settings with varying expert availability and reliability.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13897", "url": null, "sourceid": 1204, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=1lix8ppUJ7", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11590, "modified": "2026-03-29T20:43:17.728360-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=1lix8ppUJ7", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "120", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13571, "uid": "abdbeb4d8dbe30df8430a8394b7218ef", "name": "On the Hardness of Auditing Model Properties Under Updates: Complexity of Property-Preserving Updates", "authors": [{"id": 22398, "fullname": "Ayoub Ajarra", "url": "http://virtual.aistats.org/api/miniconf/users/22398?format=json", "institution": "INRIA"}, {"id": 12340, "fullname": "Debabrota Basu", "url": "http://virtual.aistats.org/api/miniconf/users/12340?format=json", "institution": "Inria, CNRS, Univ. Lille"}], "abstract": "As machine learning becomes deeply embedded in societal infrastructure, assessing the risks posed by these models has grown increasingly critical. Real-world deployment further complicates this assessment: model owners may apply strategic updates in response to dynamic environments (e.g., financial markets), potentially undermining key guarantees.  We formalize this setting and address two goals: (i) accurately estimating a target auditing property-- such as group fairness-- using a minimal number of labeled samples; and (ii) characterizing the complexity of strategic updates by identifying the subset of admissible updates that preserve the property.  To this end, we propose a generic algorithmic framework for efficient PAC auditing, powered by an Empirical Property Optimization (EPO) oracle. For statistical parity, we establish distribution-free audit bounds characterized by the SP dimension, a new combinatorial measure that captures the complexity of admissible strategic updates. Finally, we show that our framework naturally extends to other properties, including prediction error and robust risk.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13571", "url": null, "sourceid": 2122, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=XuptGYEoXO", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11264, "modified": "2026-03-29T20:43:04.082784-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=XuptGYEoXO", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "120", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13363, "uid": "5b8e4fd39d9786228649a8a8bec4e008", "name": "On the Identifiability of Tensor Ranks via Prior Predictive Matching", "authors": [{"id": 21942, "fullname": "Eliezer de Souza da Silva", "url": "http://virtual.aistats.org/api/miniconf/users/21942?format=json", "institution": "University of Coimbra, DEI, CISUC/LASI"}, {"id": 4019, "fullname": "Arto Klami", "url": "http://virtual.aistats.org/api/miniconf/users/4019?format=json", "institution": "University of Helsinki"}, {"id": 10049, "fullname": "Diego Mesquita", "url": "http://virtual.aistats.org/api/miniconf/users/10049?format=json", "institution": "Getulio Vargas Foundation (FGV EMAp)"}, {"id": 21943, "fullname": "I\u00f1igo Urteaga", "url": "http://virtual.aistats.org/api/miniconf/users/21943?format=json", "institution": "Basque Center for Applied Mathematics"}], "abstract": "Selecting the latent dimensions (ranks) in tensor factorization is a central challenge that often relies on heuristic methods. This paper introduces a rigorous approach to determine rank identifiability in probabilistic tensor models, based on prior predictive moment matching. We transform a set of moment matching conditions into a log-linear system of equations in terms of marginal moments, prior hyperparameters, and ranks; establishing an equivalence between rank identifiability and the solvability of such system. We apply this framework to four foundational tensor-models, demonstrating that the linear structure of the PARAFAC/CP model, the chain structure of the Tensor Train model, and the closed-loop structure of the Tensor Ring model yield solvable systems, making their ranks identifiable. In contrast, we prove that the symmetric topology of the Tucker model leads to an underdetermined system, rendering the ranks unidentifiable by this method. For the identifiable models, we derive explicit closed-form rank estimators based on the moments of observed data only. We empirically validate these estimators and evaluate the robustness of the proposal.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13363", "url": null, "sourceid": 2287, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=wGHcWy5Te1", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11056, "modified": "2026-03-29T20:42:55.859716-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=wGHcWy5Te1", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "121", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13656, "uid": "1595af6435015c77a7149e92a551338e", "name": "On Kernel based Variational Autoencoders", "authors": [{"id": 22567, "fullname": "Tian Qin", "url": "http://virtual.aistats.org/api/miniconf/users/22567?format=json", "institution": "Lehigh University"}, {"id": 22568, "fullname": "Wei-Min Huang", "url": "http://virtual.aistats.org/api/miniconf/users/22568?format=json", "institution": "Lehigh University"}], "abstract": "In this paper, we bridge Variational Autoencoders (VAEs)  and  kernel density estimations (KDEs)  by approximating the posterior by the expectation of kernel density estimator and deriving a new lower bound of empirical log likelihood. The flexibility of KDEs provides a new perspective of controlling the KL-divergence term in original evidence lower bound (ELBO) which enriches the choice of the posterior and prior pairs in VAE.  We show that  the Epanechnikov kernel gives the tightest upper bound in controlling the KL-divergence under appropriate conditions in theory and develop a kernel-based VAE called Epanechnikov Variational Autoenocoder (EVAE). The implementation of EVAE is straightforward  as Epanechnikov kernel lies in the ``location-scale'' family of distributions where reparametrization tricks can be applied directly. Compared with Gaussian kernel, Epanechnikov kernel has compact support which should make the generated sample less blurry. The flexibility of new lower bound of ELBO also enables us to employ a two-stage training strategy to treat reconstruction and generation separately, which is an analogue of the idea in VQ-VAE. Extensive experiments  illustrate the potential of EVAE in image generation and the superiority of EVAE over vanilla VAE and other baseline models in the quality of reconstructed images, as measured by the FID score and Sharpness. We also carried out additional experiments about the application of EVAE in downstream classification tasks.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13656", "url": null, "sourceid": 681, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=N59DW4lA4X", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11349, "modified": "2026-03-29T20:43:07.519639-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=N59DW4lA4X", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "122", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13474, "uid": "2b45c629e577731c4df84fc34f936a89", "name": "On the Neural Feature Ansatz for Deep Neural Networks", "authors": [{"id": 22182, "fullname": "Edward Tansley", "url": "http://virtual.aistats.org/api/miniconf/users/22182?format=json", "institution": "University of Oxford"}, {"id": 22183, "fullname": "Estelle Massart", "url": "http://virtual.aistats.org/api/miniconf/users/22183?format=json", "institution": "UCL"}, {"id": 22184, "fullname": "Coralia Cartis", "url": "http://virtual.aistats.org/api/miniconf/users/22184?format=json", "institution": "University of Oxford"}], "abstract": "Understanding feature learning is an important open question in establishing a mathematical foundation for deep neural networks.   The Neural Feature Ansatz (NFA) states that after training, the Gram matrix of the first-layer weights of a deep neural network is proportional to some power $\\alpha>0$ of the average gradient outer product (AGOP) of this network with respect to its inputs. Assuming gradient flow dynamics with balanced weight  initialization, the NFA was proven to hold throughout training for two-layer linear networks with exponent $\\alpha = 1/2$ (Radhakrishnan et al., 2024).   We extend this result to networks with $L \\geq 2$ layers, showing that the NFA holds with exponent $\\alpha = 1/L$, thus demonstrating a depth dependency of the NFA.   Furthermore, we prove that for unbalanced initialization, the NFA holds asymptotically through training if weight decay is applied.   We also provide counterexamples showing that the NFA does not hold for some network architectures with nonlinear activations, even when these networks fit arbitrarily well the training data.    We thoroughly validate our theoretical results through numerical experiments across a variety of optimization algorithms, weight decay rates and initialization schemes.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13474", "url": null, "sourceid": 2301, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=hvTHKUuVUG", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11167, "modified": "2026-03-29T20:43:00.320692-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=hvTHKUuVUG", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "123", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13578, "uid": "9996535e07258a7bbfd8b132435c5962", "name": "Partial Monotonicity for Submodular Maximization with a Knapsack Constraint", "authors": [{"id": 19510, "fullname": "Tong Cheng", "url": "http://virtual.aistats.org/api/miniconf/users/19510?format=json", "institution": "Nanyang Technological University"}, {"id": 22411, "fullname": "Xueyan Tang", "url": "http://virtual.aistats.org/api/miniconf/users/22411?format=json", "institution": "Nanyang Technological University"}], "abstract": "Submodular maximization has become increasingly important in the fields of machine learning and data mining. For general submodular maximization without monotonicity, many previous analyses provide poor approximation guarantees, especially for submodular functions that are approximately monotone. To address this issue, the research community has proposed a continuous metric called the monotonicity ratio for submodular functions. The monotonicity ratio has been studied for submodular maximization under no constraint, a cardinality constraint, and a matroid constraint. However, the implications of using the monotonicity ratio for submodular maximization with a knapsack constraint remain unclear. Although a knapsack constraint can be regarded as a continuous extension of the cardinality constraint with non-uniform costs, the gap in analysis between these two constraints is substantial. In this paper, we analytically show that many previously proposed algorithms for monotone submodular maximization with a knapsack constraint can achieve improved approximation guarantees under partial monotonicity with a simple modification: enforcing positive marginal gain. In addition, we evaluate our proposed algorithms for two machine learning applications of movie recommendation and influence-and-exploit marketing, showing that our algorithms could achieve better empirical performance than state-of-the-art algorithms under partial monotonicity.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13578", "url": null, "sourceid": 1235, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=XGyjEEVUKD", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11271, "modified": "2026-03-29T20:43:04.303248-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=XGyjEEVUKD", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "125", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13523, "uid": "6f4922f45568161a8cdf4ad2299f6d23", "name": "On the hardness of Reinforcement Learning with Transition Lookahead", "authors": [{"id": 14801, "fullname": "Corentin Pla", "url": "http://virtual.aistats.org/api/miniconf/users/14801?format=json", "institution": "Criteo, CREST-ENSAE"}, {"id": 10954, "fullname": "Hugo Richard", "url": "http://virtual.aistats.org/api/miniconf/users/10954?format=json", "institution": "Criteo"}, {"id": 1431, "fullname": "Marc Abeille", "url": "http://virtual.aistats.org/api/miniconf/users/1431?format=json", "institution": "Criteo AI Lab"}, {"id": 12593, "fullname": "Nadav Merlis", "url": "http://virtual.aistats.org/api/miniconf/users/12593?format=json", "institution": "ENSAE Paris"}, {"id": 9406, "fullname": "Vianney Perchet", "url": "http://virtual.aistats.org/api/miniconf/users/9406?format=json", "institution": "ENSAE &amp; Criteo AI Lab"}], "abstract": "We study reinforcement learning (RL) with transition look-ahead, where the agent may observe which states would be visited upon playing any sequence of $\\ell$ actions before deciding its course of action. While such predictive information can drastically improve the achievable performance, we show that using this information optimally comes at a potentially prohibitive computational cost. Specifically, we prove that optimal planning with one-step look-ahead ($\\ell=1$) can be solved in polynomial time through a novel linear programming formulation. In contrast, for $\\ell \\geq 2$, the problem becomes NP-hard.  Our results delineate a precise boundary between tractable and intractable cases for the problem of planning with transition look-ahead in reinforcement learning.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13523", "url": null, "sourceid": 18, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=clyOoEL3pS", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11216, "modified": "2026-03-29T20:43:02.138236-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=clyOoEL3pS", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "125", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13815, "uid": "da0d1111d2dc5d489242e60ebcbaf988", "name": "Partial VOROS: A Cost-aware Performance Metric for Binary Classifiers with Precision and Capacity Constraints", "authors": [{"id": 22922, "fullname": "Christopher Ratigan", "url": "http://virtual.aistats.org/api/miniconf/users/22922?format=json", "institution": "Tufts University"}, {"id": 20610, "fullname": "Kyle Heuton", "url": "http://virtual.aistats.org/api/miniconf/users/20610?format=json", "institution": "Tufts University"}, {"id": 22923, "fullname": "Carissa Wang", "url": "http://virtual.aistats.org/api/miniconf/users/22923?format=json", "institution": "Tufts University"}, {"id": 22924, "fullname": "Lenore Cowen", "url": "http://virtual.aistats.org/api/miniconf/users/22924?format=json", "institution": "Tufts University"}, {"id": 4807, "fullname": "Michael Hughes", "url": "http://virtual.aistats.org/api/miniconf/users/4807?format=json", "institution": "Tufts University"}], "abstract": "The ROC curve is widely used to assess binary classifiers. Yet for some applications, such as alert systems for monitoring hospitalized patients, conventional ROC analysis cannot meet two key deployment needs:  enforcing a constraint on precision to avoid false alarm fatigue and imposing an upper bound on the number of predicted positives to represent the capacity of hospital staff. The usual area under the curve metric also does not reflect asymmetric costs for false positives and false negatives. In this paper we address all three of these issues. First, we show how the subset of classifiers that meet precision and capacity constraints occupy a feasible region in ROC space. We establish the polygon-shaped geometry of this region. We then define the partial area of lesser classifiers, a performance metric that is monotonic with cost and only accounts for the feasible region. Averaging this area over a desired distribution for cost parameters results in the partial volume over the ROC surface, or partial VOROS. In experiments predicting mortality risk from vital sign history on several datasets, we show this cost-aware metric can outperform alternatives at ranking classifiers for in-hospital alerts.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13815", "url": null, "sourceid": 917, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=9J9Ly2xUxL", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11508, "modified": "2026-03-29T20:43:14.155541-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=9J9Ly2xUxL", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "126", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13751, "uid": "1ee3dfcd8a0645a25a35977997223d22", "name": "Personalized Incentive Alignment: Correcting Utility-Driven Selection Bias in A/B Tests", "authors": [{"id": 22765, "fullname": "Jiachun Li", "url": "http://virtual.aistats.org/api/miniconf/users/22765?format=json", "institution": "Massachusetts Institute of Technology"}, {"id": 19886, "fullname": "yang meng", "url": "http://virtual.aistats.org/api/miniconf/users/19886?format=json", "institution": "university of chicago"}, {"id": 710, "fullname": "David Simchi-Levi", "url": "http://virtual.aistats.org/api/miniconf/users/710?format=json", "institution": "MIT"}, {"id": 22766, "fullname": "Chonghuan Wang", "url": "http://virtual.aistats.org/api/miniconf/users/22766?format=json", "institution": "University of Texas at Dallas"}], "abstract": "Although A/B testing is a powerful tool for estimating the average treatment effect (ATE), it often proves impractical in social or commercial settings because ethical and business constraints induce participant non-compliance. For example, patients may refuse assignment to less promising therapies, and users may choose whether to adopt a newly released feature based on personal preferences. In this work, we posit that participants act to maximize individual incentives. To capture this behavior, we adopt a utility-based random choice model that explicitly characterizes the identification bias introduced by self-selection and the estimation instability caused by feature imbalance. We then demonstrate how heterogeneous incentives generate both selection bias and inflated variance. Building on these insights, we design an optimal incentive mechanism that equalizes preference distributions across treatment arms, thereby achieving a more balanced covariate profile, lower variance, and a sharper identified set with minimal bias. Finally, we propose an online learning framework that adaptively identifies the optimal incentive scheme during the experiment and produces valid treatment-effect estimates. We validate our theoretical results through both simulation studies and field experiments.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13751", "url": null, "sourceid": 1319, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=ESmbflvAcA", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11444, "modified": "2026-03-29T20:43:11.571070-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=ESmbflvAcA", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "127", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13575, "uid": "97d0145823aeb8ed80617be62e08bdcc", "name": "Optimal Transport Guarantees to Nonparametric Regression for Locally Stationary Time Series", "authors": [{"id": 19504, "fullname": "Jan Nino Tinio", "url": "http://virtual.aistats.org/api/miniconf/users/19504?format=json", "institution": "Caraga State University"}, {"id": 11035, "fullname": "Mokhtar Alaya", "url": "http://virtual.aistats.org/api/miniconf/users/11035?format=json", "institution": "LMAC - Universit\u00e9 de Technologie de Compi\u00e8gne"}, {"id": 22406, "fullname": "Salim Bouzebda", "url": "http://virtual.aistats.org/api/miniconf/users/22406?format=json", "institution": "Universit\u00e9 de Technologie de Compi\u00e8gne"}], "abstract": "Locally stationary time series (LSTS) represent an essential modeling paradigm for capturing the nuanced dynamics inherent in time series data, whose statistical characteristics, including mean and variance, evolve smoothly over time. In this paper, we propose a conditional probability distribution estimator for LSTS through Nadaraya\u2013Watson (NW) kernel smoothing. NW estimator leverages local kernel smoothing to approximate the conditional distribution of a response variable given its covariates. Under mild conditions, we establish optimal transport convergence guarantees to the proposed NW-based conditional probability estimator. These guarantees are initially proven in the univariate setting using the Wasserstein distance, and subsequently in a multivariate setting employing the sliced Wasserstein distance. To corroborate our theoretical findings, we conduct a wide range of numerical experiments to assess the convergence rates and showcase the practical relevance of the estimator in capturing intricate temporal dependencies in complex nonstationary phenomena.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13575", "url": null, "sourceid": 1627, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=Xaf7Y78tXe", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11268, "modified": "2026-03-29T20:43:04.205910-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=Xaf7Y78tXe", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "129", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13336, "uid": "c06d06da9666a219db15cf575aff2824", "name": "Optimal Arm Elimination Algorithms for Combinatorial Bandits", "authors": [{"id": 21888, "fullname": "Yuxiao Wen", "url": "http://virtual.aistats.org/api/miniconf/users/21888?format=json", "institution": "New York University"}, {"id": 21889, "fullname": "Yanjun Han", "url": "http://virtual.aistats.org/api/miniconf/users/21889?format=json", "institution": "New York University"}, {"id": 55, "fullname": "Zhengyuan Zhou", "url": "http://virtual.aistats.org/api/miniconf/users/55?format=json", "institution": "New York University"}], "abstract": "Combinatorial bandits extend the classical bandit framework to settings where the learner selects multiple arms in each round, motivated by applications such as online recommendation and assortment optimization. While extensions of upper confidence bound (UCB) algorithms arise naturally in this context, adapting arm elimination methods has proved more challenging. We introduce a novel elimination scheme that partitions arms into three categories (confirmed, active, and eliminated), and incorporates explicit exploration to update these sets. We demonstrate the efficacy of our algorithm in two settings: the combinatorial multi-armed bandit with general graph feedback, and the combinatorial linear contextual bandit. Matching lower bounds are also provided. In both cases, our approach achieves near-optimal regret, whereas UCB-based methods can provably fail due to insufficient explicit exploration.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13336", "url": null, "sourceid": 690, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=yjn4MENlun", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11029, "modified": "2026-03-29T20:42:54.686232-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=yjn4MENlun", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "129", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13825, "uid": "b5a1fc2085986034e448d2ccc5bb9703", "name": "Optimistic Reinforcement Learning with Quantile Objectives", "authors": [{"id": 19423, "fullname": "Mohammad Alipour-Vaezi", "url": "http://virtual.aistats.org/api/miniconf/users/19423?format=json", "institution": "Virginia Tech"}, {"id": 22941, "fullname": "Huaiyang Zhong", "url": "http://virtual.aistats.org/api/miniconf/users/22941?format=json", "institution": "Virginia Polytechnic Institute and State University"}, {"id": 22942, "fullname": "Kwok-leung Tsui", "url": "http://virtual.aistats.org/api/miniconf/users/22942?format=json", "institution": "University of Texas at Arlington"}, {"id": 22943, "fullname": "Sajad Khodadadian", "url": "http://virtual.aistats.org/api/miniconf/users/22943?format=json", "institution": "Virginia Polytechnic Institute and State University"}], "abstract": "Reinforcement Learning (RL) has achieved tremendous success in recent years. However, the classical foundations of RL do not account for the risk sensitivity of the objective function, which is critical in various fields, including healthcare, finance, etc. A popular approach to incorporate risk sensitivity is to optimize a specific quantile of the cumulative reward distribution. In this paper, we develop UCB-QRL, an optimistic learning algorithm for the $\\tau$-quantile objective in finite-horizon Markov decision processes (MDPs). UCB-QRL is an iterative algorithm in which, at each iteration, we first estimate the underlying transition probability and then optimize the quantile value function over a confidence ball around this estimate. Here, we show that UCB-QRL yields high-probability regret bounds $\\mathcal O\\left((2/\\kappa)^HH\\sqrt{SATH\\log(2SATH/\\delta)}\\right)$ in the episodic setting with $S$ states, $A$ actions, $T$ episodes, and $H$ horizons. Here, $\\kappa>0$ is a problem-dependent constant that captures the sensitivity of the underlying MDP's quantile value.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13825", "url": null, "sourceid": 1536, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=8MuJDoP8XK", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11518, "modified": "2026-03-29T20:43:14.629444-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=8MuJDoP8XK", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "130", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13759, "uid": "bac9162b47c56fc8a4d2a519803d51b3", "name": "Optimal Posterior Sampling for Policy Identification in Tabular Markov Decision Processes", "authors": [{"id": 11117, "fullname": "Cyrille Kone", "url": "http://virtual.aistats.org/api/miniconf/users/11117?format=json", "institution": "UZH"}, {"id": 1244, "fullname": "Kevin Jamieson", "url": "http://virtual.aistats.org/api/miniconf/users/1244?format=json", "institution": "University of Washington"}], "abstract": "We study the $(\\varepsilon, \\delta)$-PAC policy identification problem in finite-horizon episodic Markov Decision Processes. Existing approaches provide finite-time guarantees for approximate settings ($\\varepsilon>0$) but suffer from high computational cost, rendering them hard to implement, and also suffer from suboptimal dependence on $\\log(1/\\delta)$. We propose a randomized and computationally efficient algorithm for best policy identification that combines posterior sampling with an online learning algorithm to guide the exploration. Our method achieves asymptotic optimality in sample complexity, also in terms of posterior contraction rate, and runs in $O(S^2AH)$ per episode, matching standard model-based approaches. Unlike prior algorithms such as MOCA and PEDEL, our guarantees remain meaningful in the asymptotic regime and avoid sub-optimal polynomial dependence on $\\log(1/\\delta)$. Our results provide both theoretical insights and practical tools for efficient policy identification in tabular MDPs.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13759", "url": null, "sourceid": 364, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=DmBeNcPi7n", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11452, "modified": "2026-03-29T20:43:11.931794-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=DmBeNcPi7n", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "130", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13605, "uid": "c60d060b946d6dd6145dcbad5c4ccf6f", "name": "Process-Tensor Tomography of SGD: Measuring Non-Markovian Memory via Back-Flow of Distinguishability", "authors": [{"id": 19742, "fullname": "Vasileios Sevetlidis", "url": "http://virtual.aistats.org/api/miniconf/users/19742?format=json", "institution": "Athena RC"}, {"id": 22477, "fullname": "George Pavlidis", "url": "http://virtual.aistats.org/api/miniconf/users/22477?format=json", "institution": "Athena Research Center"}], "abstract": "We model neural training as a classical multi-time map from controllable interventions---batch choices, augmentations, and optimizer micro-steps---to model predictions on a fixed probe set. On this basis, we introduce a simple, model-agnostic witness of training memory based on back-flow of distinguishability. In a controlled two-step protocol, we compare predictive distributions after one intervention versus two; a positive increase $\\Delta_{\\mathrm{BF}} = D_2 - D_1 > 0$, with $D\\in\\{\\mathrm{TV}, \\mathrm{JS}, \\mathrm{H}\\}$, certifies observable-level non-Markovianity. Across controlled SGD experiments, we observe consistent positive back-flow with tight bootstrap confidence intervals, stronger effects under higher momentum, larger batch overlap, and more micro-steps, and marked collapse under a \\emph{causal break} that resets optimizer state. The witness is inexpensive, requires no architectural changes, and is robust across TV/JS/Hellinger. We position this as a measurement contribution: a practical diagnostic, and empirical evidence, that real training often deviates from the Markov idealization in ways that matter for optimizer behavior, data order, and schedule design.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13605", "url": null, "sourceid": 1118, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=TonOzlbE3k", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11298, "modified": "2026-03-29T20:43:05.409747-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=TonOzlbE3k", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "130", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13516, "uid": "149815eb972b3c370dee3b89d645ae14", "name": "Optimized projection-free algorithms for online learning: construction and worst-case analysis", "authors": [{"id": 20601, "fullname": "Julien Weibel", "url": "http://virtual.aistats.org/api/miniconf/users/20601?format=json", "institution": "Inria Paris"}, {"id": 4100, "fullname": "Pierre Gaillard", "url": "http://virtual.aistats.org/api/miniconf/users/4100?format=json", "institution": "Inria"}, {"id": 13101, "fullname": "Wouter Koolen", "url": "http://virtual.aistats.org/api/miniconf/users/13101?format=json", "institution": "CWI Amsterdam"}, {"id": 3810, "fullname": "Adrien Taylor", "url": "http://virtual.aistats.org/api/miniconf/users/3810?format=json", "institution": "INRIA/ENS"}], "abstract": "This work studies and develop projection-free algorithms for online learning with linear optimization oracles (a.k.a. Frank--Wolfe) for handling the constraint set. More precisely, this work   (i) shows how to exploit semidefinite programming to jointly design and analyze online Frank--Wolfe-type algorithms numerically in a variety of settings, (ii) leverage those design techniques to propose an improved (optimized) variant of an online Frank--Wolfe algorithm along with its conceptually simple potential-based proof, and (iii) its anytime version which benefits from similar $O(T^{3/4})$ regret rate without requiring to know the time horizon $T$ in advance. We are not aware of other direct regret guarantees for anytime version of online Frank--Wolfe  without using the classical doubling trick.    Based on the semidefinite technique, we conclude with strong numerical evidence suggesting that no pure online Frank--Wolfe algorithm within our model class can have a regret guarantee better than $O(T^{3/4})$  without additional assumptions, that the current algorithms do not have optimal constants, and that multiple linear optimization rounds do not generally help to obtain better regret bounds.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13516", "url": null, "sourceid": 2323, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=daGYvxmQDC", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11209, "modified": "2026-03-29T20:43:01.895424-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=daGYvxmQDC", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "131", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13613, "uid": "27ed0fb950b856b06e1273989422e7d3", "name": "Provably Efficient and Agile Randomized Q-Learning", "authors": [{"id": 22491, "fullname": "He Wang", "url": "http://virtual.aistats.org/api/miniconf/users/22491?format=json", "institution": "Carnegie Mellon University"}, {"id": 22492, "fullname": "Xingyu Xu", "url": "http://virtual.aistats.org/api/miniconf/users/22492?format=json", "institution": "CMU, Carnegie Mellon University"}, {"id": 19827, "fullname": "Yuejie Chi", "url": "http://virtual.aistats.org/api/miniconf/users/19827?format=json", "institution": "Yale University"}], "abstract": "While Bayesian-based exploration often demonstrates superior empirical performance compared to bonus-based methods in model-based reinforcement learning (RL), its theoretical understanding remains limited for model-free settings. Existing provable algorithms either suffer from computational intractability or rely on stage-wise policy updates which reduce responsiveness and slow down the learning process. In this paper, we propose a novel variant of Q-learning algorithm, referred to as RandomizedQ, which integrates sampling-based exploration with agile, step-wise, policy updates, for episodic tabular RL. We establish a sublinear regret bound $\\widetilde{O}(\\sqrt{H^5SAT})$, where $S$ is the number of states, $A$ is the number of actions, $H$ is the episode length, and $T$ is the total number of episodes. In addition, we present a logarithmic regret bound $ O\\left(\\frac{H^6SA}{\\Delta_{\\min}}\\log^5(SAHT)\\right)$ when the optimal Q-function has a positive sub-optimality $\\Delta_{\\min}$. Empirically, RandomizedQ exhibits outstanding performance compared to existing Q-learning variants with both bonus-based and Bayesian-based exploration on standard benchmarks.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13613", "url": null, "sourceid": 1039, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=SsaipDZ2xI", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11306, "modified": "2026-03-29T20:43:05.685275-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=SsaipDZ2xI", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "132", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13453, "uid": "8c19f571e251e61cb8dd3612f26d5ecf", "name": "Optimal Variance and Covariance Estimation under Differential Privacy in the Add-Remove Model and Beyond", "authors": [{"id": 20615, "fullname": "Shokichi Takakura", "url": "http://virtual.aistats.org/api/miniconf/users/20615?format=json", "institution": "LY Corporation"}, {"id": 22137, "fullname": "Seng Pei Liew", "url": "http://virtual.aistats.org/api/miniconf/users/22137?format=json", "institution": "SB Intuitions"}, {"id": 22138, "fullname": "Satoshi Hasegawa", "url": "http://virtual.aistats.org/api/miniconf/users/22138?format=json", "institution": "LY Corporation"}], "abstract": "In this paper, we study the problem of estimating the variance and covariance of datasets under differential privacy in the add-remove model.   While estimation in the swap model has been extensively studied in the literature, the add-remove model remains less explored and more challenging, as the dataset size must also be kept private.   To address this issue, we develop efficient mechanisms for variance and covariance estimation based on the Bezier mechanism, a novel moment-release framework that leverages Bernstein bases.   We prove that our proposed mechanisms are minimax optimal in the high-privacy regime by establishing new minimax lower bounds.   Moreover, beyond worst-case scenarios, we analyze instance-wise utility and show that the B\u00e9zier-based estimator consistently achieves better utility compared to alternative mechanisms.   Finally, we demonstrate the effectiveness of the B\u00e9zier mechanism beyond variance and covariance estimation, showcasing its applicability to other statistical tasks.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13453", "url": null, "sourceid": 263, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=kyi3n1DvFl", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11146, "modified": "2026-03-29T20:42:59.544176-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=kyi3n1DvFl", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "132", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13355, "uid": "980ecd059122ce2e50136bda65c25e07", "name": "Provably Efficient Reinforcement Learning for Sparse Dynamical Systems with Non-Gaussian Noise", "authors": [{"id": 21923, "fullname": "Davide Maran", "url": "http://virtual.aistats.org/api/miniconf/users/21923?format=json", "institution": "Politecnico di Milano"}, {"id": 21924, "fullname": "Gianmarco Tedeschi", "url": "http://virtual.aistats.org/api/miniconf/users/21924?format=json", "institution": "Polytechnic Institute of Milan"}, {"id": 21925, "fullname": "Enea Gusmeroli", "url": "http://virtual.aistats.org/api/miniconf/users/21925?format=json", "institution": "Polytechnic Institute of Milan"}, {"id": 3833, "fullname": "Marcello Restelli", "url": "http://virtual.aistats.org/api/miniconf/users/3833?format=json", "institution": "Politecnico di Milano"}], "abstract": "The recent development of sparse methods for identifying nonlinear dynamical systems has opened new avenues for efficient and interpretable model-based reinforcement learning (RL). In this work, we study online RL in environments where the system dynamics, modeled as $s'=f(s,a)+$ noise, is assumed to be sparse with respect to a big feature map, a structural idea inspired by the SINDy framework. We introduce an optimistic algorithm that combines online sparse regression with confidence set construction to guide exploration and planning. On the theoretical side, we provide the first regret bounds for sparse nonlinear dynamics, showing that regret scales with the sparsity level $d_0$. This result holds even when relaxing standard Gaussian noise assumptions by allowing a much more general, non-parametric, family of densities and when the model is misspecified. The algorithm achieving the regret bound is not computationally efficient, as it relies on a very computationally intensive online regression method. To bridge this gap, we propose a practical variant that draws inspiration from theoretical principles but incorporates more scalable components. We adopt SINDy for sparse system identification algorithm and couple it with SAC in a Dyna-style planning framework. Empirical results on classic continuous control tasks demonstrate the practical viability and robustness of our approach.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13355", "url": null, "sourceid": 1356, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=wy98vOBhsM", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11048, "modified": "2026-03-29T20:42:55.537076-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=wy98vOBhsM", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "133", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13345, "uid": "19de10adbaa1b2ee13f77f679fa1483a", "name": "Out-of-Distribution Generalization of In-Context Learning: A Low-Dimensional Subspace Perspective", "authors": [{"id": 12579, "fullname": "Soo Min Kwon", "url": "http://virtual.aistats.org/api/miniconf/users/12579?format=json", "institution": "University of Michigan"}, {"id": 19766, "fullname": "Alec Xu", "url": "http://virtual.aistats.org/api/miniconf/users/19766?format=json", "institution": "University of Michigan"}, {"id": 21906, "fullname": "Can Yaras", "url": "http://virtual.aistats.org/api/miniconf/users/21906?format=json", "institution": "University of Michigan - Ann Arbor"}, {"id": 4251, "fullname": "Laura Balzano", "url": "http://virtual.aistats.org/api/miniconf/users/4251?format=json", "institution": "University of Michigan"}, {"id": 12582, "fullname": "Qing Qu", "url": "http://virtual.aistats.org/api/miniconf/users/12582?format=json", "institution": "University of Michigan"}], "abstract": "The transformer's emergent ability to perform in-context learning (ICL) has sparked a wide range of studies designed to understand its strengths and limitations. However, a theoretical understanding of when ICL can and cannot generalize beyond its pre-training data still remains unclear. This paper puts forth a minimal mathematical model that provably identifies when ICL can generalize out-of-distribution (OOD). By studying linear regression tasks parameterized with low-rank covariance matrices, we model distribution shifts as varying angles between subspaces and derive conditions under which a single-layer linear attention model interpolates across all angles. We show that if pre-training task vectors are drawn from a union of subspaces, transformers can generalize to all angle shifts\u2014enabling ICL even in regions with zero probability mass in the training distribution. On the other hand, if the pre-training tasks are drawn from a single Gaussian, the test risk shows a non-negligible dependence on the angle, implying that ICL cannot generalize OOD. We empirically show that our results also hold for models such as GPT-2, and present experiments on how our observations extend to nonlinear function classes.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13345", "url": null, "sourceid": 1866, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=xrmPHv8SNT", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11038, "modified": "2026-03-29T20:42:55.122321-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=xrmPHv8SNT", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "133", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13478, "uid": "8f14e45fceea167a5a36dedd4bea2543", "name": "Partially Lazy Gradient Descent for Smoothed Online Learning", "authors": [{"id": 20637, "fullname": "Naram Mhaisen", "url": "http://virtual.aistats.org/api/miniconf/users/20637?format=json", "institution": "TU Delft"}, {"id": 22194, "fullname": "George Iosifidis", "url": "http://virtual.aistats.org/api/miniconf/users/22194?format=json", "institution": "Delft University of Technology"}], "abstract": "We introduce \\textsc{$k$-lazyGD}, an online learning algorithm that bridges the gap between greedy Online Gradient Descent (OGD, for $k=1$) and lazy GD/dual-averaging (for $k=T$), creating a spectrum between reactive and stable updates. We analyze this spectrum in Smoothed Online Convex Optimization (SOCO), where the learner incurs both hitting and movement costs. Our main contribution is establishing that laziness is possible without sacrificing hitting performance: we prove that \\textsc{$k$-lazyGD} achieves the optimal dynamic regret $\\mathcal{O}(\\sqrt{(P_T+1)T})$ for any laziness slack $k$ up to $\\Theta(\\sqrt{T/P_T})$, where $P_T$ is the comparator path length. This result formally connects the allowable laziness to the comparator's shifts, showing that \\textsc{$k$-lazyGD} can retain the inherently small movements of lazy methods without compromising tracking ability. We base our analysis on the Follow the Regularized Leader (FTRL) framework, and derive a matching lower bound. Since the slack depends on $P_T$, an ensemble of learners with various slacks is used, yielding a method that is provably stable when it can be, and agile when it must be.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13478", "url": null, "sourceid": 7, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=hIAkL2BYCI", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11171, "modified": "2026-03-29T20:43:00.468083-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=hIAkL2BYCI", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "135", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13496, "uid": "86df7dcfd896fcaf2674f757a2463eba", "name": "Rate optimal learning of equilibria from data", "authors": [{"id": 22238, "fullname": "Till Freihaut", "url": "http://virtual.aistats.org/api/miniconf/users/22238?format=json", "institution": "University of Zurich"}, {"id": 22239, "fullname": "Luca Viano", "url": "http://virtual.aistats.org/api/miniconf/users/22239?format=json", "institution": "EPFL - EPF Lausanne"}, {"id": 22240, "fullname": "Emanuele Nevali", "url": "http://virtual.aistats.org/api/miniconf/users/22240?format=json", "institution": "EPFL - EPF Lausanne"}, {"id": 4411, "fullname": "Volkan Cevher", "url": "http://virtual.aistats.org/api/miniconf/users/4411?format=json", "institution": "EPFL"}, {"id": 22241, "fullname": "Matthieu Geist", "url": "http://virtual.aistats.org/api/miniconf/users/22241?format=json", "institution": "Earth Species Project"}, {"id": 22242, "fullname": "Giorgia Ramponi", "url": "http://virtual.aistats.org/api/miniconf/users/22242?format=json", "institution": "Department of Informatics, University of Zurich, University of Zurich"}], "abstract": "We close open theoretical gaps in Multi-Agent Imitation Learning (MAIL) by characterizing the limits of non-interactive MAIL and presenting the first interactive algorithm with near-optimal sample complexity. In the non-interactive setting, we prove a statistical lower bound that identifies the \\emph{all-policy deviation concentrability coefficient} as the fundamental complexity measure, and we show that Behavior Cloning (BC) is rate-optimal. For the interactive setting, we introduce a framework that combines reward-free reinforcement learning with interactive MAIL and instantiate it with an algorithm, \\emph{\\ours}. It improves the best previously known sample complexity from $\\mathcal{O}(\\varepsilon^{-8})$ to $\\mathcal{O}(\\varepsilon^{-2}),$ matching the dependence on $\\varepsilon$ implied by our lower bound. Finally, we provide numerical results that support our theory and illustrate, in environments such as grid worlds, cases where Behavior Cloning fails to learn.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13496", "url": null, "sourceid": 1486, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=g6wdvJkhaY", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11189, "modified": "2026-03-29T20:43:01.159108-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=g6wdvJkhaY", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "137", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13404, "uid": "98c7242894844ecd6ec94af67ac8247d", "name": "Policy Learning with Abstention", "authors": [{"id": 22016, "fullname": "Ayush Sawarni", "url": "http://virtual.aistats.org/api/miniconf/users/22016?format=json", "institution": "Stanford"}, {"id": 22017, "fullname": "Jikai Jin", "url": "http://virtual.aistats.org/api/miniconf/users/22017?format=json", "institution": "Stanford University"}, {"id": 22018, "fullname": "Justin Whitehouse", "url": "http://virtual.aistats.org/api/miniconf/users/22018?format=json", "institution": "Stanford University"}, {"id": 12717, "fullname": "Vasilis Syrgkanis", "url": "http://virtual.aistats.org/api/miniconf/users/12717?format=json", "institution": "Stanford University"}], "abstract": "Policy learning algorithms are regularly leveraged in domains such as personalized medicine and advertising to develop individualized treatment regimes. One deficit of existing policy learning algorithms is that they do not adjust their decisions based on uncertainty in their predictions\u2014that is, they fail to \\textit{abstain}. To remedy this, we introduce a framework for \\textit{policy learning with abstention}, in which policies that choose not to assign a treatment to some customers/patients receive a small additive reward on top of the value of a random guess. Building on empirical welfare maximization, we propose a two-stage learner that first identifies a set of near-optimal policies and then constructs an abstention class based on disagreements between the policies. We establish fast $O(1/n)$\u2013type regret guarantees for the learned policy from offline data when the treatment propensity in the offline data is known, and we show how to extend these guarantees to the unknown\u2013propensity case via a doubly robust (DR) objective. Further, we use our algorithm as a black box to obtain improved guarantees under margin conditions that go beyond realizability, which has been a standard assumption in prior work on policy learning with a margin. We also study links to distributionally robust policy learning\u2014where abstention acts as a hedge against small shifts\u2014and to safe policy improvement, where the objective is to improve upon a given baseline policy with high probability. We validate our theoretical findings through extensive synthetic experiments.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13404", "url": null, "sourceid": 1968, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=ppWhBLJNbD", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11097, "modified": "2026-03-29T20:42:57.549894-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=ppWhBLJNbD", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "137", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13585, "uid": "8cbd005a556ccd4211ce43f309bc0eac", "name": "PowerSoftmax: Towards Secure LLM Inference Over Encrypted Data", "authors": [{"id": 12519, "fullname": "Itamar Zimerman", "url": "http://virtual.aistats.org/api/miniconf/users/12519?format=json", "institution": "Tel-Aviv University"}, {"id": 22427, "fullname": "Allon Adir", "url": "http://virtual.aistats.org/api/miniconf/users/22427?format=json", "institution": "International Business Machines"}, {"id": 22428, "fullname": "Ehud Aharoni", "url": "http://virtual.aistats.org/api/miniconf/users/22428?format=json", "institution": "International Business Machines"}, {"id": 22429, "fullname": "Matan Avitan", "url": "http://virtual.aistats.org/api/miniconf/users/22429?format=json", "institution": "Bar-Ilan University"}, {"id": 22430, "fullname": "Moran Baruch", "url": "http://virtual.aistats.org/api/miniconf/users/22430?format=json", "institution": "International Business Machines"}, {"id": 22431, "fullname": "Nir Drucker", "url": "http://virtual.aistats.org/api/miniconf/users/22431?format=json", "institution": "International Business Machines"}, {"id": 22432, "fullname": "Jenny Lerner", "url": "http://virtual.aistats.org/api/miniconf/users/22432?format=json", "institution": "International Business Machines"}, {"id": 22433, "fullname": "Ramy Masalha", "url": "http://virtual.aistats.org/api/miniconf/users/22433?format=json", "institution": "University of Haifa"}, {"id": 22434, "fullname": "Reut Moshe", "url": "http://virtual.aistats.org/api/miniconf/users/22434?format=json", "institution": "Ben Gurion University of the Negev"}, {"id": 22435, "fullname": "Omri Soceanu", "url": "http://virtual.aistats.org/api/miniconf/users/22435?format=json", "institution": "International Business Machines"}], "abstract": "Modern cryptographic methods for implementing privacy-preserving LLMs such as HE require the LLMs to have a polynomial form. Forming such a representation is challenging because transformers include non-polynomial components, such as Softmax and layer normalization. Previous approaches have either directly approximated pre-trained models with large-degree polynomials, which are less efficient over HE, or replaced non-polynomial components with easier-to-approximate primitives before training, e.g., Softmax with pointwise attention. The latter approach might introduce scalability challenges. We present a new HE-friendly variant of self-attention that offers a stable form for training and is easy to approximate with polynomials for secure inference. Our work introduces the first polynomial LLMs over a billion parameters, exceeding the size of previous models by more than tenfold. The resulting models demonstrate reasoning and in-context learning (ICL) capabilities comparable to standard transformers of the same size, representing a breakthrough in the field. Finally, we provide a detailed latency breakdown for each computation over encrypted data, paving the way for further optimization, and explore the differences in inductive bias between models relying on our HE-friendly variant and standard transformers. Our code is attached as a supplement.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13585", "url": null, "sourceid": 2374, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=WeIUIh1xwt", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11278, "modified": "2026-03-29T20:43:04.520847-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=WeIUIh1xwt", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "138", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13617, "uid": "389bc7bb1e1c2a5e7e147703232a88f6", "name": "Preconditioned Attention: Enhancing Efficiency in Transformer Blocks", "authors": [{"id": 22499, "fullname": "Hemanth Saratchandran", "url": "http://virtual.aistats.org/api/miniconf/users/22499?format=json", "institution": "University of Adelaide/Australian Institute of Machine Learning"}], "abstract": "Central to the success of Transformers is the attention block, which effectively models global dependencies among input tokens associated to a dataset. However, we theoretically demonstrate that standard attention mechanisms in transformers often produce ill-conditioned matrices with large condition numbers. This ill-conditioning is a well-known obstacle for gradient-based optimizers, leading to inefficient training. To address this issue, we introduce preconditioned attention, a novel approach that incorporates a conditioning matrix into each attention head. Our theoretical analysis shows that this method significantly reduces the condition number of attention matrices, resulting in better-conditioned matrices that improve optimization. Conditioned attention serves as a simple drop-in replacement for a wide variety of attention mechanisms in the literature. We validate the effectiveness of preconditioned attention across a diverse set of transformer applications, including image classification, object detection, instance segmentation, long sequence modeling and language modeling.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13617", "url": null, "sourceid": 508, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=SoFvbfAbRs", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11310, "modified": "2026-03-29T20:43:05.812572-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=SoFvbfAbRs", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "139", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13452, "uid": "98d6f58ab0dafbb86b083a001561bb34", "name": "Reconciling Communication Compression and Byzantine-Robustness in Distributed Learning", "authors": [{"id": 22133, "fullname": "Diksha Gupta", "url": "http://virtual.aistats.org/api/miniconf/users/22133?format=json", "institution": "University of Virginia, Charlottesville"}, {"id": 22134, "fullname": "Antonio Honsell", "url": "http://virtual.aistats.org/api/miniconf/users/22134?format=json", "institution": "Bocconi University"}, {"id": 22135, "fullname": "Chuan Xu", "url": "http://virtual.aistats.org/api/miniconf/users/22135?format=json", "institution": "Universit\u00e9 C\u00f4te d&#x27;Azur,INRIA,i3S"}, {"id": 22136, "fullname": "Nirupam Gupta", "url": "http://virtual.aistats.org/api/miniconf/users/22136?format=json", "institution": "University of Copenhagen"}, {"id": 9965, "fullname": "Giovanni Neglia", "url": "http://virtual.aistats.org/api/miniconf/users/9965?format=json", "institution": "Inria"}], "abstract": "Distributed learning enables scalable model training over decentralized data, but remains hindered by Byzantine faults and high communication costs. While both challenges have been studied extensively in isolation, their interplay has received limited attention. Prior work has shown that naively combining communication compression with Byzantine-robust aggregation can severely weaken resilience to faulty nodes. The current state-of-the-art, Byz-DASHA-PAGE, leverages a momentum-based variance reduction scheme to counteract the negative effect of compression noise on Byzantine robustness. In this work, we introduce RoSDHB, a new algorithm that integrates classical Polyak momentum with a coordinated compression strategy. Theoretically, RoSDHB matches the convergence guarantee of Byz-DASHA-PAGE under the standard $(G, B)$-gradient dissimilarity model, but relies on milder assumptions. Empirically, RoSDHB demonstrates stronger robustness while achieving substantial communication savings compared to Byz-DASHA-PAGE.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13452", "url": null, "sourceid": 873, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=l5ofGnDuzC", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11145, "modified": "2026-03-29T20:42:59.508511-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=l5ofGnDuzC", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "139", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13544, "uid": "86d7c8a08b4aaa1bc7c599473f5dddda", "name": "Private, Efficient, and Robust Federated Statistical Learning", "authors": [{"id": 22326, "fullname": "Jaemu Heo", "url": "http://virtual.aistats.org/api/miniconf/users/22326?format=json", "institution": "Ulsan National Institute of Science and Technology"}, {"id": 22327, "fullname": "Xiwen Feng", "url": "http://virtual.aistats.org/api/miniconf/users/22327?format=json", "institution": "Indiana University"}, {"id": 22328, "fullname": "Jeonghun Kang", "url": "http://virtual.aistats.org/api/miniconf/users/22328?format=json", "institution": "TVING"}, {"id": 22329, "fullname": "Taehwan Kim", "url": "http://virtual.aistats.org/api/miniconf/users/22329?format=json", "institution": "Ulsan National Institute of Science and Technology"}, {"id": 20625, "fullname": "Changgee Chang", "url": "http://virtual.aistats.org/api/miniconf/users/20625?format=json", "institution": "Indiana University School of Medicine"}], "abstract": "Federated Learning (FL) enables collaborative model training across multiple sites while preserving data privacy and differential privacy (DP) provides a probabilistic framework to safeguard sensitive information when sharing output derived from data. While numerous DP-FL methods exist for large-scale prediction models, achieving DP, efficiency, and robustness in federated statistical learning remains a significant challenge. In this work, we propose a novel federated statistical learning framework that ensures efficient, robust, and privacy-preserving estimation. We introduce a new noising mechanism that encodes Fisher information along with the maximum likelihood estimate (MLE) by leveraging multiple noisy copies of the MLE. To calibrate noise effectively, we extend the smooth sensitivity to account for data-dependent correlations, ensuring strong DP guarantees while maintaining utility. Additionally, we develop INFEMBLER, an information-assembling algorithm that efficiently de-noises multiple noisy MLE copies using a hierarchical Bayesian model and an expectation-maximization (EM) algorithm. INFEMBLER significantly enhances estimation efficiency over existing methods and is inherently robust, providing estimates at least as reliable as those derived from local data alone, thereby preserving the benefits of FL. We establish its asymptotic properties and validate its effectiveness through experiments on both simulated and real datasets, demonstrating its superior statistical efficiency and robustness.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13544", "url": null, "sourceid": 1666, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=azYahrqBbC", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11237, "modified": "2026-03-29T20:43:02.918912-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=azYahrqBbC", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "140", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13503, "uid": "f899139df5e1059396431415e770c6dd", "name": "Prior Knowledge Makes It Possible: From Sublinear Graph Algorithms to LLM Test-Time Methods", "authors": [{"id": 776, "fullname": "Avrim Blum", "url": "http://virtual.aistats.org/api/miniconf/users/776?format=json", "institution": "Toyota Technological Institute at Chicago"}, {"id": 17670, "fullname": "Daniel Hsu", "url": "http://virtual.aistats.org/api/miniconf/users/17670?format=json", "institution": "Columbia University"}, {"id": 22254, "fullname": "Cyrus Rashtchian", "url": "http://virtual.aistats.org/api/miniconf/users/22254?format=json", "institution": "Google Research"}, {"id": 22255, "fullname": "Donya Saless", "url": "http://virtual.aistats.org/api/miniconf/users/22255?format=json", "institution": "University of California, Berkeley"}], "abstract": "Test-time augmentation, such as Retrieval-Augmented Generation (RAG) or tool use, critically depends on an interplay between a model's parametric knowledge and externally retrieved information. However, the theoretical underpinnings of this relationship remain poorly understood. Specifically, it is not clear how much pre-training knowledge is required to answer queries with a small number of augmentation steps, which is a desirable property in practice. To address this question, we formulate multi-step reasoning as an $s$-$t$ connectivity problem on a knowledge graph. We represent a model's pre-training parametric knowledge as a partial, potentially noisy subgraph.  We view augmentation as querying an oracle for true edges that augment the model's knowledge. From a technical point of view, we are the first to study graph query complexity when given partial prior knowledge. Then, we characterize the necessary and sufficient number of augmentation steps for the model to generate an accurate answer. One key result shows a phase transition: if the prior knowledge graph over $n$ vertices is disconnected into small components, then finding a path via augmentation is inefficient and requires $\\Omega(\\sqrt{n})$ queries. On the other hand, once the density of correct knowledge surpasses a threshold, forming a giant component, we can find paths with an expected constant number of queries. We also extend our analysis to verifier-based approaches, quantifying how the cost of test-time verification scales with the unreliability of the pre-trained knowledge.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13503", "url": null, "sourceid": 100, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=fgUjxaumpO", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11196, "modified": "2026-03-29T20:43:01.418592-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=fgUjxaumpO", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "141", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13875, "uid": "92c8c96e4c37100777c7190b76d28233", "name": "Prior shift estimation for positive unlabeled data through the lens of kernel embedding", "authors": [{"id": 23029, "fullname": "Jan Mielniczuk", "url": "http://virtual.aistats.org/api/miniconf/users/23029?format=json", "institution": "Institute of Computer Science Polish Academy of Sciences  NIP 525-000-94-01"}, {"id": 23030, "fullname": "Wojciech Rejchel", "url": "http://virtual.aistats.org/api/miniconf/users/23030?format=json", "institution": "Nicolaus Copernicus University"}, {"id": 18053, "fullname": "Pawe\u0142 Teisseyre", "url": "http://virtual.aistats.org/api/miniconf/users/18053?format=json", "institution": "Warsaw University of Technology"}], "abstract": "We study estimation  of a class prior  for unlabeled target samples which  possibly differs from that of source population. Moreover, it is assumed that the source data is partially observable: only samples from the positive class and from the whole population are available (PU learning scenario). We introduce a novel direct estimator of the class prior which avoids estimation of posterior probabilities in both populations and has a simple geometric interpretation. It is based on a distribution matching technique together with kernel embedding in Reproducing Kernel Hilbert  Space and is obtained  as an explicit solution to an optimisation task. We establish its asymptotic consistency as well as an explicit non-asymptotic    bound  on its deviation from the unknown prior, which is calculable in practice. We study  finite sample behaviour for synthetic and real data and show that the proposal works consistently on par or better than its competitors.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13875", "url": null, "sourceid": 280, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=40aNv2nho5", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11568, "modified": "2026-03-29T20:43:16.883321-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=40aNv2nho5", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "142", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13469, "uid": "c73dfe6c630edb4c1692db67c510f65c", "name": "Representative, Informative, and De-Amplifying: Requirements for Robust Bayesian Active Learning under Model Misspecification", "authors": [{"id": 22173, "fullname": "Roubing Tang", "url": "http://virtual.aistats.org/api/miniconf/users/22173?format=json", "institution": "University of Manchester"}, {"id": 22174, "fullname": "Sabina Sloman", "url": "http://virtual.aistats.org/api/miniconf/users/22174?format=json", "institution": "University of Manchester"}, {"id": 3700, "fullname": "Samuel Kaski", "url": "http://virtual.aistats.org/api/miniconf/users/3700?format=json", "institution": "Aalto University and University of Manchester"}], "abstract": "In many settings in science and industry, such as drug discovery and clinical trials, a central challenge is designing experiments under time and budget constraints.     *Bayesian Optimal Experimental Design (BOED)* is a paradigm to pick maximally informative designs that has been increasingly applied to such problems. During training, BOED selects inputs according to a pre-determined acquisition criterion to target *informativeness*. During testing, the model learned during training encounters a naturally occurring distribution of test samples. This leads to an instance of covariate shift, where the train and test samples are drawn from different distributions (the training samples are not *representative* of the test distribution).      Prior work has shown that in the presence of model misspecification, covariate shift amplifies generalization error.     Our first contribution is to provide a mathematical analysis of generalization error that reveals key contributors to generalization error in the presence of model misspecification. We show that generalization error under misspecification is the result of, in addition to covariate shift, a phenomenon we term *error (de-)amplification* which has not been identified or studied in prior work.     We then develop a new acquisition function that mitigates the effects of model misspecification by including terms for representativeness, informativeness, and de-amplification (R-IDeA).     Our experimental results demonstrate that the proposed method performs better than methods that target either only informativeness, representativeness, or both.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13469", "url": null, "sourceid": 1348, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=iyKz5REmeQ", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11162, "modified": "2026-03-29T20:43:00.141804-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=iyKz5REmeQ", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "142", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13612, "uid": "9bf31c7ff062936a96d3c8bd1f8f2ff3", "name": "Probabilistic multi-dimensional classification with incomplete data at the prediction time", "authors": [{"id": 19738, "fullname": "Thu Ha DO", "url": "http://virtual.aistats.org/api/miniconf/users/19738?format=json", "institution": "HEUDIASYC - UMR CNRS 7253, Universit\u00e9 de Technologie de Compi\u00e8gne"}, {"id": 22489, "fullname": "Vu-Linh Nguyen", "url": "http://virtual.aistats.org/api/miniconf/users/22489?format=json", "institution": "Universit\u00e9 de Technologie de Compi\u00e8gne"}, {"id": 22490, "fullname": "Yves Grandvalet", "url": "http://virtual.aistats.org/api/miniconf/users/22490?format=json", "institution": "Centre National de la Recherche Scientifique  Universit\u00e9 de technologie de Compi\u00e8gne"}], "abstract": "Multi-dimensional classification (MDC) extends multi-class and multi-label learning by predicting several class variables per instance.  We revisit probabilistic MDC methods with mixed features (discrete and continuous), focusing on their strengths and limits for handling incomplete data at prediction time.  We present theoretical results leading to a new probabilistic approach with efficient learning and prediction algorithms that address scalability and robustness issues.  Experiments demonstrate its benefits in different missingness scenarios.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13612", "url": null, "sourceid": 15, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=Sv2frNivLO", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11305, "modified": "2026-03-29T20:43:05.647557-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=Sv2frNivLO", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "143", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13717, "uid": "9fdb62f932adf55af2c0e09e55861964", "name": "ProxRouter: Proximity-Weighted LLM Query Routing for Improved Robustness to Outliers", "authors": [{"id": 22712, "fullname": "Shivam Patel", "url": "http://virtual.aistats.org/api/miniconf/users/22712?format=json", "institution": "CMU, Carnegie Mellon University"}, {"id": 12703, "fullname": "Neharika Jali", "url": "http://virtual.aistats.org/api/miniconf/users/12703?format=json", "institution": "Carnegie Mellon University"}, {"id": 22713, "fullname": "Ankur Mallick", "url": "http://virtual.aistats.org/api/miniconf/users/22713?format=json", "institution": "Microsoft"}, {"id": 4585, "fullname": "Gauri Joshi", "url": "http://virtual.aistats.org/api/miniconf/users/4585?format=json", "institution": "Carnegie Mellon University"}], "abstract": "Large language model (LLM) query routers are critical to modern AI platforms as they seek to improve efficiency by assigning inference queries to accurate, yet low-cost models. Parametric routers typically use trained neural networks for LLM selection but suffer from retraining and maintenance overheads. Nonparametric routers are training-free, instead estimating LLM accuracy and cost via similarity between encodings of the input query and training set queries. However, like their parametric counterparts, nonparametric routers struggle to generalize to outlier queries, an issue exacerbated by limited diversity in training sets which are costly to expand and difficult to keep current with ever-evolving use cases. We propose ProxRouter, which applies an exponentially tilted aggregation mechanism to balance bias and variance in nonparametric routers, improving their robustness to outliers. Experiments show ProxRouter enhances outlier routing while preserving inlier performance with minimal overhead.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13717", "url": null, "sourceid": 1900, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=GYWls3tyqM", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11410, "modified": "2026-03-29T20:43:10.055805-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=GYWls3tyqM", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "144", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13692, "uid": "0537fb40a68c18da59a35c2bfe1ca554", "name": "Robust estimation of heterogeneous treatment effects in randomized trials leveraging external data", "authors": [{"id": 22639, "fullname": "Rickard Karlsson", "url": "http://virtual.aistats.org/api/miniconf/users/22639?format=json", "institution": "Delft University of Technology"}, {"id": 13247, "fullname": "Piersilvio De Bartolomeis", "url": "http://virtual.aistats.org/api/miniconf/users/13247?format=json", "institution": "ETH Z\u00fcrich"}, {"id": 22640, "fullname": "Issa Dahabreh", "url": "http://virtual.aistats.org/api/miniconf/users/22640?format=json", "institution": "Harvard University"}, {"id": 22641, "fullname": "Jesse Krijthe", "url": "http://virtual.aistats.org/api/miniconf/users/22641?format=json", "institution": "Delft University of Technology"}], "abstract": "Randomized trials are typically designed to detect average treatment effects but often lack the statistical power to uncover individual-level treatment effect heterogeneity, limiting their value for personalized decision-making. To address this, we propose the QR-learner, a model-agnostic learner that estimates conditional average treatment effects (CATE) within the trial population by leveraging external data from other trials or observational studies. The proposed method is robust: it can reduce the mean squared error relative to a trial-only CATE learner, and is guaranteed to recover the true CATE even when the external data are not aligned with the trial. Moreover, we introduce a procedure that combines the QR-learner with a trial-only CATE learner and show that it asymptotically matches or exceeds both component learners in terms of mean squared error. We examine the performance of our approach in simulation studies and apply the methods to a real-world dataset, demonstrating improvements in both CATE estimation and statistical power for detecting heterogeneous effects.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13692", "url": null, "sourceid": 744, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=J1EPu0h4hW", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11385, "modified": "2026-03-29T20:43:08.956709-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=J1EPu0h4hW", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "144", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13750, "uid": "41a60377ba920919939d83326ebee5a1", "name": "Scalable Model-Based Clustering with Sequential Monte Carlo", "authors": [{"id": 22761, "fullname": "Connie Trojan", "url": "http://virtual.aistats.org/api/miniconf/users/22761?format=json", "institution": "Lancaster University"}, {"id": 22762, "fullname": "Pavel Myshkov", "url": "http://virtual.aistats.org/api/miniconf/users/22762?format=json", "institution": "Research, Microsoft"}, {"id": 4569, "fullname": "Paul Fearnhead", "url": "http://virtual.aistats.org/api/miniconf/users/4569?format=json", "institution": ""}, {"id": 22763, "fullname": "James Hensman", "url": "http://virtual.aistats.org/api/miniconf/users/22763?format=json", "institution": "Microsoft Research"}, {"id": 22764, "fullname": "Tom Minka", "url": "http://virtual.aistats.org/api/miniconf/users/22764?format=json", "institution": "Microsoft"}, {"id": 9292, "fullname": "Christopher Nemeth", "url": "http://virtual.aistats.org/api/miniconf/users/9292?format=json", "institution": "University of Lancaster"}], "abstract": "In online clustering problems, there is often a large amount of uncertainty over possible cluster assignments that cannot be resolved until more data are observed. This difficulty is compounded when clusters follow complex distributions, as is the case with text data. Sequential Monte Carlo (SMC) methods give a natural way of representing and updating this uncertainty over time, but have prohibitive memory requirements for large-scale problems. We propose a novel SMC algorithm that decomposes clustering problems into approximately independent subproblems, allowing a more compact representation of the algorithm state. Our approach is motivated by the knowledge base construction problem, and we show that our method is able to accurately and efficiently solve clustering problems in this setting and others where traditional SMC struggles.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13750", "url": null, "sourceid": 1510, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=EVDivDL9jD", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11443, "modified": "2026-03-29T20:43:11.519300-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=EVDivDL9jD", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "145", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13689, "uid": "f18a6d1cde4b205199de8729a6637b42", "name": "Q-Learning with Shift-Aware Upper Confidence Bound in Non-Stationary Reinforcement Learning", "authors": [{"id": 12864, "fullname": "Ha Manh Bui", "url": "http://virtual.aistats.org/api/miniconf/users/12864?format=json", "institution": "Johns Hopkins University"}, {"id": 22634, "fullname": "Felix Parker", "url": "http://virtual.aistats.org/api/miniconf/users/22634?format=json", "institution": "Johns Hopkins University"}, {"id": 22635, "fullname": "Kimia Ghobadi", "url": "http://virtual.aistats.org/api/miniconf/users/22635?format=json", "institution": "Johns Hopkins University"}, {"id": 9722, "fullname": "Anqi Liu", "url": "http://virtual.aistats.org/api/miniconf/users/9722?format=json", "institution": "Johns Hopkins University"}], "abstract": "We study the Non-Stationary Reinforcement Learning (RL) under distribution shifts in both finite-horizon episodic and infinite-horizon discounted Markov Decision Processes (MDPs). In the finite-horizon case, the transition functions may suddenly change at a particular episode. In the infinite-horizon setting, such changes can occur at an arbitrary time step during the agent's interaction with the environment. While the Q-learning Upper Confidence Bound algorithm (QUCB) can discover a proper policy during learning, due to the distribution shifts, this policy can exploit sub-optimal rewards after the shift happens. To address this issue, we propose Density-QUCB (DQUCB), a shift-aware Q-learning UCB algorithm, which uses a transition density function to detect distribution shifts, then leverages its likelihood to enhance the uncertainty estimation quality of Q-learning UCB, resulting in a balance between exploration and exploitation. Theoretically, we prove that our oracle DQUCB achieves a better regret guarantee than QUCB. Empirically, our DQUCB enjoys the computational efficiency of model-free RL and outperforms QUCB baselines by having a lower regret across RL tasks, as well as a COVID-19 patient hospital allocation task using a Deep-Q-learning architecture.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13689", "url": null, "sourceid": 1605, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=Jb5dJwWGF9", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11382, "modified": "2026-03-29T20:43:08.852867-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=Jb5dJwWGF9", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "145", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13784, "uid": "e60e81c4cbe5171cd654662d9887aec2", "name": "Refining Covariance Matrix Estimation in Stochastic Gradient Descent through Bias Reduction", "authors": [{"id": 22863, "fullname": "Ziyang Wei", "url": "http://virtual.aistats.org/api/miniconf/users/22863?format=json", "institution": "University of Chicago"}, {"id": 22864, "fullname": "Wanrong Zhu", "url": "http://virtual.aistats.org/api/miniconf/users/22864?format=json", "institution": "University of California, Irvine"}, {"id": 22865, "fullname": "Jingyang Lyu", "url": "http://virtual.aistats.org/api/miniconf/users/22865?format=json", "institution": "University of Wisconsin - Madison"}, {"id": 22866, "fullname": "Wei Biao Wu", "url": "http://virtual.aistats.org/api/miniconf/users/22866?format=json", "institution": "University of Chicago"}], "abstract": "We study online inference and asymptotic covariance estimation for the stochastic gradient descent (SGD) algorithm. While classical methods\u2014such as plug-in and batch-means estimators\u2014are available, they either require inaccessible second-order (Hessian) information or suffer from slow convergence. To address these challenges, we propose a novel, fully online de-biased covariance estimator that eliminates the need for second-order derivatives while significantly improving estimation accuracy. Our method employs a bias-reduction technique to achieve a convergence rate of $n^{(\\alpha-1)/2}\\sqrt{\\log n}$, outperforming existing Hessian-free alternatives. We provide theoretical guarantees for consistency and validate the estimator\u2019s superior finite-sample performance through extensive simulations.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13784", "url": null, "sourceid": 1526, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=Bttn4JfaZH", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11477, "modified": "2026-03-29T20:43:12.888795-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=Bttn4JfaZH", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "152", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13796, "uid": "c8c41c4a18675a74e01c8a20e8a0f662", "name": "Semi-Random Noisy and One-Bit Matrix Completion With Nonconvex Primal-Dual Framework", "authors": [{"id": 22887, "fullname": "Xing Gao", "url": "http://virtual.aistats.org/api/miniconf/users/22887?format=json", "institution": "University of Illinois, Chicago"}, {"id": 22888, "fullname": "Binhao Chen", "url": "http://virtual.aistats.org/api/miniconf/users/22888?format=json", "institution": "Department of Computer Science, Brown University"}, {"id": 22889, "fullname": "Yu Cheng", "url": "http://virtual.aistats.org/api/miniconf/users/22889?format=json", "institution": "Brown University"}], "abstract": "We study low-rank matrix completion in the semi-random model, where each entry is revealed with some unknown probability at least $p$, in contrast to the standard setting with a uniform probability $p$. While prior work [CG18] has shown that the nonconvex approach succeeds in the semi-random model for quadratic loss, it remains unclear whether similar guarantees extend to more general settings, such as noisy or one-bit matrix completion. We give a nearly-linear time algorithmic framework for semi-random matrix completion with a broad family of loss functions. Our approach builds on a preprocessing step from [CG18] to restore regularity conditions failing under the semi-random setting, as well as a general primal-dual scheme of [ZWYG18], achieving nearly-linear runtime and recovery guarantees that are optimal up to logarithmic factors. As concrete corollaries, this yields nearly-linear time solutions for noisy and one-bit matrix completion under the semi-random model.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13796", "url": null, "sourceid": 946, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=B2u4zRSt2d", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11489, "modified": "2026-03-29T20:43:13.376372-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=B2u4zRSt2d", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "146", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13887, "uid": "0d4f4805c36dc6853edfa4c7e1638b48", "name": "Quantifying Epistemic Uncertainty in Diffusion Models", "authors": [{"id": 23055, "fullname": "Aditi Gupta", "url": "http://virtual.aistats.org/api/miniconf/users/23055?format=json", "institution": "Lawrence Berkeley National Lab"}, {"id": 23056, "fullname": "Raphael Meyer", "url": "http://virtual.aistats.org/api/miniconf/users/23056?format=json", "institution": "University of California, Berkeley"}, {"id": 23057, "fullname": "Yotam Yaniv", "url": "http://virtual.aistats.org/api/miniconf/users/23057?format=json", "institution": "Lawrence Berkeley National Lab"}, {"id": 12836, "fullname": "Elynn Chen", "url": "http://virtual.aistats.org/api/miniconf/users/12836?format=json", "institution": "New York University"}, {"id": 12654, "fullname": "N. Benjamin Erichson", "url": "http://virtual.aistats.org/api/miniconf/users/12654?format=json", "institution": "Berkeley Lab and ICSI"}], "abstract": "To ensure high quality outputs, it is important to quantify the epistemic uncertainty of diffusion models. Existing methods are often unreliable because they mix epistemic and aleatoric uncertainty. We introduce a method based on Fisher information that explicitly isolates epistemic variance, producing more reliable plausibility scores for generated data. To make this approach scalable, we propose FLARE (Fisher-Laplace Randomized Estimator), which approximates the Fisher information using a uniformly random subset of model parameters. Empirically, FLARE improves uncertainty estimation in synthetic time-series generation tasks, achieving more accurate and reliable filtering than other methods. Theoretically, we bound the convergence rate of our randomized approximation and provide analytic and empirical evidence that last-layer Laplace approximations are insufficient for this task.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13887", "url": null, "sourceid": 1574, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=2mAIPSwn1C", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11580, "modified": "2026-03-29T20:43:17.319462-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=2mAIPSwn1C", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "147", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13611, "uid": "1abb1e1ea5f481b589da52303b091cbb", "name": "Set to Be Fair: Demographic Parity Constraints for Set-Valued Classification", "authors": [{"id": 22487, "fullname": "Eyal Cohen", "url": "http://virtual.aistats.org/api/miniconf/users/22487?format=json", "institution": "Sorbonne Universit\u00e9 - Facult\u00e9 des Sciences (Paris VI)"}, {"id": 22488, "fullname": "Christophe Denis", "url": "http://virtual.aistats.org/api/miniconf/users/22488?format=json", "institution": "Universit\u00e9 Panth\u00e9on-Sorbonne (Paris I)"}, {"id": 20639, "fullname": "Mohamed Hebiri", "url": "http://virtual.aistats.org/api/miniconf/users/20639?format=json", "institution": "Universit\u00e9 Gustave Eiffel"}], "abstract": "Set-valued classification is used in multiclass settings where confusion between classes can occur and lead to misleading predictions. However, its application may amplify discriminatory bias motivating the development of set-valued approaches under fairness constraints.  In this paper, we address the problem of set-valued classification under demographic parity and expected size constraints. We propose two complementary strategies: an oracle-based method that minimizes classification risk while satisfying both constraints, and a computationally efficient proxy that prioritizes constraint satisfaction. For both strategies, we derive closed-form expressions for the (optimal) fair set-valued classifiers and use these to build plug-in, data-driven procedures for empirical predictions. We establish distribution-free convergence rates for violations of the size and fairness constraints for both methods, and under mild assumptions we also provide excess-risk bounds for the oracle-based approach. Empirical results demonstrate the effectiveness of both strategies and highlight the efficiency of our proxy method.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13611", "url": null, "sourceid": 1459, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=T0mIZZEhJ6", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11304, "modified": "2026-03-29T20:43:05.606497-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=T0mIZZEhJ6", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "148", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13522, "uid": "cb8acb1dc9821bf74e6ca9068032d623", "name": "Randomized HyperSteiner: A Stochastic Delaunay Triangulation Heuristic for the Hyperbolic Steiner Minimal Tree", "authors": [{"id": 19825, "fullname": "Aniss Aiman Medbouhi", "url": "http://virtual.aistats.org/api/miniconf/users/19825?format=json", "institution": "KTH Royal Institute of Technology"}, {"id": 22291, "fullname": "Alejandro Garc\u00eda-Castellanos", "url": "http://virtual.aistats.org/api/miniconf/users/22291?format=json", "institution": "University of Amsterdam"}, {"id": 5613, "fullname": "Giovanni Luca Marchetti", "url": "http://virtual.aistats.org/api/miniconf/users/5613?format=json", "institution": "KTH"}, {"id": 22292, "fullname": "Daniel Pelt", "url": "http://virtual.aistats.org/api/miniconf/users/22292?format=json", "institution": "Leiden University"}, {"id": 22293, "fullname": "Erik Bekkers", "url": "http://virtual.aistats.org/api/miniconf/users/22293?format=json", "institution": "University of Amsterdam"}, {"id": 9273, "fullname": "Danica Kragic", "url": "http://virtual.aistats.org/api/miniconf/users/9273?format=json", "institution": "KTH Royal Institute of Technology"}], "abstract": "We study the problem of constructing Steiner Minimal Trees (SMTs) in hyperbolic space. Exact SMT computation is NP-hard, and existing hyperbolic heuristics such as HyperSteiner are deterministic and often get trapped in locally suboptimal configurations. We introduce Randomized HyperSteiner (RHS), a stochastic Delaunay triangulation heuristic that incorporates randomness into the expansion process and refines candidate trees via Riemannian gradient descent optimization. Experiments on synthetic data sets and a real-world single-cell transcriptomic data show that RHS outperforms Minimum Spanning Tree (MST), Neighbour Joining, and vanilla HyperSteiner (HS). In near-boundary configurations, RHS can achieve a 32\\% reduction in total length over HS, demonstrating its effectiveness and robustness in diverse data regimes.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13522", "url": null, "sourceid": 1530, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=cq2lc89TqM", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11215, "modified": "2026-03-29T20:43:02.099361-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=cq2lc89TqM", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "149", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13337, "uid": "9b698eb3105bd82528f23d0c92dedfc0", "name": "Random Features for Operator-Valued Kernels: Bridging Kernel Methods and Neural Operators", "authors": [{"id": 21890, "fullname": "Mike Nguyen", "url": "http://virtual.aistats.org/api/miniconf/users/21890?format=json", "institution": "Technische Universit\u00e4t Braunschweig"}, {"id": 114, "fullname": "Nicole M\u00fccke", "url": "http://virtual.aistats.org/api/miniconf/users/114?format=json", "institution": "TU Braunschweig"}], "abstract": "In this work, we investigate the generalization properties of random feature methods. Our analysis extends prior  results for Tikhonov regularization to a broad class of spectral regularization techniques and further generalizes  the setting to operator-valued kernels. This unified framework enables, for the first time, a rigorous theoretical  analysis of neural operators and neural networks through the lens of the Neural Tangent Kernel (NTK). In  particular, it allows us to establish optimal learning rates and provides a good understanding of how many  neurons are required to achieve a given accuracy. Furthermore, we establish minimax rates in the well-specified  case and also in the misspecified case, where the target is not contained in the reproducing kernel Hilbert space.  These results sharpen and complete earlier findings for specific kernel algorithms.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13337", "url": null, "sourceid": 643, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=yX21eki5p1", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11030, "modified": "2026-03-29T20:42:54.723043-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=yX21eki5p1", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "149", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13848, "uid": "24b16fede9a67c9251d3e7c7161c83ac", "name": "Regret Guarantees for Linear Contextual Stochastic Shortest Path", "authors": [{"id": 20589, "fullname": "Dor Polikar", "url": "http://virtual.aistats.org/api/miniconf/users/20589?format=json", "institution": "Tel Aviv University"}, {"id": 18626, "fullname": "Alon Cohen", "url": "http://virtual.aistats.org/api/miniconf/users/18626?format=json", "institution": "Tel Aviv University, Tel Aviv University"}], "abstract": "We define the problem of linear Contextual Stochastic Shortest Path (CSSP), where at the beginning of each episode, the learner observes an adversarially chosen context that determines the MDP through a fixed but unknown linear function. The learner's objective is to reach a designated goal state with minimal expected cumulative loss, despite having no prior knowledge of the transition dynamics, loss functions, or the mapping from context to MDP. In this work, we propose LR-CSSP, an algorithm that achieves a regret bound of $\\widetilde{O}(K^{2/3} d^{2/3} |S| |A|^{1/3} B_\\star^2 T_\\star \\log (1/ \\delta))$, where $K$ is the number of episodes, $d$ is the context dimension, $S$ and $A$ are the sets of states and actions respectively, $B_\\star$ bounds the optimal cumulative loss and $T_\\star$, unknown to the learner, bounds the expected time for the optimal policy to reach the goal. In the case where all costs exceed $\\ell_{\\min}$, LR-CSSP attains a regret of $\\widetilde O(\\sqrt{K \\cdot d^2 |S|^3 |A| B_\\star^3 \\log(1/\\delta)/\\ell_{\\min}})$. Unlike in contextual finite-horizon MDPs, where limited knowledge primarily leads to higher losses and regret, in the CSSP setting, insufficient knowledge can also prolong episodes and may even lead to non-terminating episodes. Our analysis reveals that LR-CSSP effectively handles continuous context spaces, while ensuring all episodes terminate within a reasonable number of time steps.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13848", "url": null, "sourceid": 372, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=6UCzhaqxYy", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11541, "modified": "2026-03-29T20:43:15.576782-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=6UCzhaqxYy", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "151", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13811, "uid": "2bd7f907b7f5b6bbd91822c0c7b835f6", "name": "Statistical-computational gap in multiple Gaussian graph alignment", "authors": [{"id": 22916, "fullname": "Bertrand Even", "url": "http://virtual.aistats.org/api/miniconf/users/22916?format=json", "institution": "Universit\u00e9 Paris-Saclay"}, {"id": 22917, "fullname": "Luca Ganassali", "url": "http://virtual.aistats.org/api/miniconf/users/22917?format=json", "institution": "Universit\u00e9 Paris-Saclay"}], "abstract": "We investigate the existence of a statistical-computational gap in multiple Gaussian graph alignment. We first generalize a previously established informational threshold from Vassaux and Massouli\u00e9 (2025) to regimes where the number of observed graphs $p$ may also grow with the number of nodes $n$: when $p \\leq O(n/\\log(n))$, we recover the results from  Vassaux and Massouli\u00e9 (2025), and $p \\geq \\Omega(n/\\log(n))$ corresponds to a regime where the problem is as difficult as aligning one single graph with some unknown \"signal\" graph. Moreover, when $p = \\omega(n)$, the informational thresholds for partial and exact recovery no longer coincide, in contrast to the all-or-nothing phenomenon observed when $p=O(n)$. Then, we provide the first computational barrier in the low-degree framework for (multiple) Gaussian graph alignment.  We prove that when the correlation $\\rho$ is less than $1$, up to logarithmic terms, low degree non-trivial estimation fails. Our results suggest that the task of aligning $p$ graphs in polynomial time is as hard as the problem of aligning two graphs in polynomial time, up to logarithmic factors. These results characterize the existence of a statistical-computational gap and provide another example in which polynomial-time algorithms cannot handle complex combinatorial bi-dimensional structures.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13811", "url": null, "sourceid": 1389, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=9umc4n7Qek", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11504, "modified": "2026-03-29T20:43:13.999718-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=9umc4n7Qek", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "153", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13658, "uid": "e8fd4a8a5bab2b3785d794ab51fef55c", "name": "Regression Descent: A Statistical Framework for Neural Network Optimization", "authors": [{"id": 22569, "fullname": "Kamaljeet Singh", "url": "http://virtual.aistats.org/api/miniconf/users/22569?format=json", "institution": "University of Arizona"}, {"id": 22570, "fullname": "Nicolas Hengartner", "url": "http://virtual.aistats.org/api/miniconf/users/22570?format=json", "institution": "Los Alamos National Laboratory"}, {"id": 23271, "fullname": "Hao Zhang", "url": "http://virtual.aistats.org/api/miniconf/users/23271?format=json", "institution": "University of Arizona"}, {"id": 22572, "fullname": "Brian Bell", "url": "http://virtual.aistats.org/api/miniconf/users/22572?format=json", "institution": "Los Alamos National Laboratory"}, {"id": 22573, "fullname": "James Hyman", "url": "http://virtual.aistats.org/api/miniconf/users/22573?format=json", "institution": "Tulane University"}], "abstract": "We present Regression Descent (RD), a novel optimization algorithm for training deep neural networks that reformulates each gradient step as a regression problem in the span of the Jacobian. By leveraging the implicit function theorem in over-parameterized settings where the number of parameters exceed observations $(p > n)$, we project the optimization onto an $n$-dimensional subspace, enabling the use of statistical techniques and potentially improved conditioning. Our key insight is that in the over-parameterized regime, meaningful parameter updates lie in the row space of the Jacobian matrix, allowing us to solve a lower-dimensional regression problem with explicit regularization control. We establish convergence guarantees for RD under standard smoothness assumptions, showing that it achieves a convergence rate of $O(1/k)$ for smooth non-convex objectives. Furthermore, we prove that RD exhibits local linear convergence in neighborhoods of strict local minima, with the convergence rate dependent on the condition number of the regularized Gram matrix. The algorithm naturally handles the ill-conditioning common in neural network optimization through adaptive regularization and extends seamlessly to multi-output problems and mini-batch settings. Experimental results on Lorenz96, MNIST, and FMNIST datasets demonstrate that RD achieves up to 40\\% faster convergence compared to SGD and Adam in terms of wall-clock time, with strong performance in the presence of activation function saturation. The computational overhead of solving $m \\times m$ linear systems (where $m$ is the batch size) is offset by improved convergence properties and GPU-efficient operations. Our work opens new avenues for understanding neural network optimization through the lens of statistical regression, providing a practical algorithm for scenarios where standard gradient methods struggle.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13658", "url": null, "sourceid": 1767, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=MxWtUEoLSV", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11351, "modified": "2026-03-29T20:43:07.583241-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=MxWtUEoLSV", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "153", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13422, "uid": "41f1f19176d383480afa65d325c06ed0", "name": "Regularizing Extrapolation in Causal Inference", "authors": [{"id": 1783, "fullname": "David Arbour", "url": "http://virtual.aistats.org/api/miniconf/users/1783?format=json", "institution": "Adobe Research"}, {"id": 22054, "fullname": "Harsh Parikh", "url": "http://virtual.aistats.org/api/miniconf/users/22054?format=json", "institution": "Yale University"}, {"id": 22055, "fullname": "Bijan Niknam", "url": "http://virtual.aistats.org/api/miniconf/users/22055?format=json", "institution": "Harvard University"}, {"id": 22056, "fullname": "Elizabeth Stuart", "url": "http://virtual.aistats.org/api/miniconf/users/22056?format=json", "institution": "Johns Hopkins University"}, {"id": 22057, "fullname": "Kara Rudolph", "url": "http://virtual.aistats.org/api/miniconf/users/22057?format=json", "institution": "Columbia University"}, {"id": 4077, "fullname": "Avi Feller", "url": "http://virtual.aistats.org/api/miniconf/users/4077?format=json", "institution": "UC Berkeley"}], "abstract": "Many common estimators in machine learning and causal inference are linear smoothers, where the prediction is a weighted average of the training outcomes. Some estimators, such as ordinary least squares and kernel ridge regression, allow for arbitrarily negative weights, which improve feature imbalance but often at the cost of increased dependence on parametric modeling assumptions and higher variance. By contrast, estimators like importance weighting and random forests (sometimes implicitly) restrict weights to be non-negative, reducing dependence on parametric modeling and variance at the cost of worse imbalance. In this paper, we propose a unified framework that directly penalizes the level of extrapolation, replacing the current practice of a hard non-negativity constraint with a soft constraint and corresponding hyperparameter. We derive a worst-case extrapolation error bound and introduce a novel ``bias-bias-variance'' tradeoff, encompassing biases due to feature imbalance, model misspecification, and estimator variance; this tradeoff is especially pronounced in high dimensions, when positivity is poor. We then develop an optimization procedure that regularizes this bound while minimizing imbalance and outline how to use this approach as a sensitivity analysis for dependence on parametric modeling assumptions. We demonstrate the effectiveness of our approach through synthetic experiments and a real-world application, involving the generalization of randomized controlled trial estimates to a target population of interest.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13422", "url": null, "sourceid": 371, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=oMI8LbL9cG", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11115, "modified": "2026-03-29T20:42:58.243586-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=oMI8LbL9cG", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "153", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13683, "uid": "cd63a3eec3319fd9c84c942a08316e00", "name": "Regularized $f-$Divergence Kernel Tests", "authors": [{"id": 22620, "fullname": "M\u00f3nica Ribero", "url": "http://virtual.aistats.org/api/miniconf/users/22620?format=json", "institution": "Google"}, {"id": 22621, "fullname": "Antonin Schrab", "url": "http://virtual.aistats.org/api/miniconf/users/22621?format=json", "institution": "University of Cambridge"}, {"id": 3537, "fullname": "Arthur Gretton", "url": "http://virtual.aistats.org/api/miniconf/users/3537?format=json", "institution": "Gatsby Computational Neuroscience Unit"}], "abstract": "We propose a framework to construct practical kernel-based two-sample tests from the family of $f$-divergences. The test statistic is computed from the witness function of a regularized variational representation of the divergence, which we estimate using kernel methods. Aggregation is used to adapt the test over hyperparameters such as the kernel bandwidth and the regularization parameter. While our test covers a variety of $f$-divergences, we bring particular focus to the hockey-stick divergence, motivated by its applications to differential privacy auditing and machine unlearning evaluation. We provide theoretical guarantees for statistical test power across our family of $f-$divergence estimates. For two-sample testing, experiments demonstrate that different $f$-divergences are sensitive to different localized differences, illustrating the importance of leveraging diverse statistics. For machine unlearning, we propose a relative test that distinguishes true unlearning failures from safe distributional variations.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13683", "url": null, "sourceid": 1849, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=K7Ds0sjbEE", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11376, "modified": "2026-03-29T20:43:08.634585-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=K7Ds0sjbEE", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "154", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13543, "uid": "f52378e14237225a6f6c7d802dc6abbd", "name": "Stick-Breaking Embedded Topic Model with Continuous Optimal Transport for Online Analysis of Document Streams", "authors": [{"id": 19789, "fullname": "Federica Granese", "url": "http://virtual.aistats.org/api/miniconf/users/19789?format=json", "institution": "INRIA"}, {"id": 22324, "fullname": "Serena Villata", "url": "http://virtual.aistats.org/api/miniconf/users/22324?format=json", "institution": "CNRS"}, {"id": 22325, "fullname": "Charles Bouveyron", "url": "http://virtual.aistats.org/api/miniconf/users/22325?format=json", "institution": "Universit\u00e9 C\u00f4te d&#x27;Azur"}], "abstract": "Online topic models are unsupervised algorithms to identify latent topics in data streams that continuously evolve over time. Although these methods naturally align with real-world scenarios, they have received considerably less attention from the community compared to their offline counterparts, due to specific additional challenges. To tackle these issues, we present SB-SETM, an innovative model extending the Embedded Topic Model (ETM) to process data streams by merging models formed on successive partial document batches. To this end, SB-SETM (i) leverages a truncated stick-breaking construction for the topic\u2013per-document distribution, enabling the model to automatically infer from the data the appropriate number of active topics at each timestep; and (ii) introduces a merging strategy for topic embeddings based on a continuous formulation of optimal transport adapted to the high dimensionality of the latent topic space. Numerical experiments show SB-SETM outperforming baselines on simulated scenarios. We extensively test it on a real-world corpus of news articles covering the Russian\u2013Ukrainian war throughout 2022\u20132023.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13543", "url": null, "sourceid": 1377, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=bDWSCHWPPl", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11236, "modified": "2026-03-29T20:43:02.879993-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=bDWSCHWPPl", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "154", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13766, "uid": "f29b38f160f87ae86df31cee1982066f", "name": "Structured Matching via Cost-Regularized Unbalanced Optimal Transport", "authors": [{"id": 22809, "fullname": "Emanuele Pardini", "url": "http://virtual.aistats.org/api/miniconf/users/22809?format=json", "institution": "University of Pisa"}, {"id": 19407, "fullname": "Katerina Papagiannouli", "url": "http://virtual.aistats.org/api/miniconf/users/19407?format=json", "institution": null}], "abstract": "Unbalanced optimal transport (UOT) provides a flexible way to match or compare nonnegative finite Radon measures. However, UOT requires a predefined ground transport cost, which may misrepresent the data\u2019s underlying geometry. Choosing such a cost is particularly challenging when datasets live in heterogeneous spaces, often motivating practitioners to adopt Gromov\u2013Wasserstein formulations. To address this challenge, we introduce cost-regularized unbalanced optimal transport (CR-UOT), a framework that allows the ground cost to vary while allowing mass creation and removal. We show that CR-UOT incorporates unbalanced Gromov\u2013Wasserstein\u2013type problems through families of inner-product costs parameterized by linear transformations, enabling the matching of measures (or point clouds) across Euclidean spaces. We develop algorithms for such CR-UOT problems using entropic regularization and demonstrate that this approach improves the alignment of heterogeneous single-cell omics profiles, especially when many cells lack direct matches.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13766", "url": null, "sourceid": 1312, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=D0FQJ53iwY", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11459, "modified": "2026-03-29T20:43:12.172885-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=D0FQJ53iwY", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "155", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13461, "uid": "a87ff679a2f3e71d9181a67b7542122c", "name": "Rethinking Intrinsic Dimension Estimation in Neural Representations", "authors": [{"id": 14366, "fullname": "Rickmer Schulte", "url": "http://virtual.aistats.org/api/miniconf/users/14366?format=json", "institution": "LMU Munich, MCML"}, {"id": 14293, "fullname": "David R\u00fcgamer", "url": "http://virtual.aistats.org/api/miniconf/users/14293?format=json", "institution": "LMU Munich, MCML"}], "abstract": "The analysis of neural representation has become an integral part of research aiming to better understand the inner workings of neural networks. While there are many different approaches to investigate neural representations, an important line of research has focused on doing so through the lens of intrinsic dimensions (IDs). Although this perspective has provided valuable insights and stimulated substantial follow-up research, important limitations of this approach have remained largely unaddressed. In this paper, we highlight a crucial discrepancy between theory and practice of IDs in neural representations, theoretically and empirically showing that common ID estimators are, in fact, not tracking the true underlying ID of the representation. We contrast this negative result with an investigation of the underlying factors that may drive commonly reported ID-related results on neural representation in the literature. Building on these insights, we offer a new perspective on ID estimation in neural representations.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13461", "url": null, "sourceid": 4, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=kH1gPRbYqh", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11154, "modified": "2026-03-29T20:42:59.813359-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=kH1gPRbYqh", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "155", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13526, "uid": "285f89b802bcb2651801455c86d78f2a", "name": "Structured Temporal Inference in State-Space Models", "authors": [{"id": 19758, "fullname": "hamidreza hashemp", "url": "http://virtual.aistats.org/api/miniconf/users/19758?format=json", "institution": ""}], "abstract": "We propose a framework for structured temporal inference in nonlinear state-space    models (SSMs) with hybrid latent dynamics that mix discrete and continuous variables.     Our method follows a two-stage inference: continuous states are estimated via      Kalman inspired updates, while discrete variables are sampled by a neural model      conditioned on these states, avoiding explicit Markov assumptions. To handle      instabilities arising from recurrent dynamics, we introduce stabilization      approach, and train all components jointly using surrogate gradient estimators      that support REINFORCE-style updates.     This design achieves SOTA results across synthetic and real-world datasets,      in state estimation, regime detection, and imputation under noise and      partial observability.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13526", "url": null, "sourceid": 1155, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=ccV8Efl4yb", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11219, "modified": "2026-03-29T20:43:02.261560-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=ccV8Efl4yb", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "156", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13761, "uid": "4dcf435435894a4d0972046fc566af76", "name": "T$_k$CP: Context-Aware Pooling via Top-k% Activation Selection", "authors": [{"id": 22789, "fullname": "Seo-Yeon Choi", "url": "http://virtual.aistats.org/api/miniconf/users/22789?format=json", "institution": "Jeonbuk National University"}, {"id": 22790, "fullname": "Kyungsu Lee", "url": "http://virtual.aistats.org/api/miniconf/users/22790?format=json", "institution": "Jeonbuk National University"}], "abstract": "Pooling is a core operation in convolutional neural networks (CNNs), enabling spatial reduction and hierarchical abstraction. However, standard methods such as max or average pooling operate locally and often fail to capture global context, leading to under- or over-estimation of features. This limits performance on tasks requiring both fine localization and holistic understanding. To address this, we propose Top-$k$\\% Contextual Pooling (TkCP), a framework that preserves informative activations based on contextual importance. TkCP includes two variants: (1) Sparse Contextual Pooling, selecting top-$k$\\% activations within local windows, and (2) Global Contextual Pooling, selecting top-$k$\\% across the entire feature map. Given a kernel size and target resolution, TkCP deterministically sets the stride and reconstructs outputs without additional parameters. Experiments across classification, detection, tracking, segmentation, and generation show consistent improvements in accuracy and robustness. Additionally, TkCP enhances interpretability by tracing high-activation regions across layers.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13761", "url": null, "sourceid": 1507, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=DZVbzYrUFN", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11454, "modified": "2026-03-29T20:43:11.998638-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=DZVbzYrUFN", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "157", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13579, "uid": "4dcae38ee11d3a6606cc6cd636a3628b", "name": "Tensor Gaussian Processes: Efficient Solvers for Nonlinear PDEs", "authors": [{"id": 18372, "fullname": "Qiwei Yuan", "url": "http://virtual.aistats.org/api/miniconf/users/18372?format=json", "institution": "University of Utah"}, {"id": 18387, "fullname": "Zhitong Xu", "url": "http://virtual.aistats.org/api/miniconf/users/18387?format=json", "institution": "University of Utah"}, {"id": 22412, "fullname": "Yinghao Chen", "url": "http://virtual.aistats.org/api/miniconf/users/22412?format=json", "institution": "University of Utah"}, {"id": 22413, "fullname": "Yiming Xu", "url": "http://virtual.aistats.org/api/miniconf/users/22413?format=json", "institution": "University of Kentucky"}, {"id": 22414, "fullname": "Houman Owhadi", "url": "http://virtual.aistats.org/api/miniconf/users/22414?format=json", "institution": "California Institute of Technology"}, {"id": 552, "fullname": "Shandian Zhe", "url": "http://virtual.aistats.org/api/miniconf/users/552?format=json", "institution": "University of Utah"}], "abstract": "Machine learning solvers for partial differential equations (PDEs) have attracted growing interest. However, most existing approaches, such as neural network solvers, rely on stochastic training, which is inefficient and typically requires a great many training epochs.  Gaussian process (GP)/kernel-based solvers, while mathematical principled, suffer from scalability issues when handling large numbers of collocation points often needed for challenging or higher-dimensional  PDEs.  To overcome these limitations, we propose TGPS, a tensor-GP-based solver  that introduces factor functions along each input dimension using one-dimensional GPs and combines them via tensor decomposition to approximate the full solution. This design reduces the task to learning a collection of one-dimensional GPs, substantially lowering computational complexity, and enabling scalability to massive collocation sets.  For efficient nonlinear PDE solving, we use a partial freezing strategy and Newton's method to linerize the nonlinear terms. We then develop an alternating least squares (ALS) approach that admits closed-form  updates, thereby substantially enhancing the training efficiency. We establish theoretical guarantees on the expressivity of our model, together with convergence proof and error analysis  under standard regularity assumptions.  Experiments on several benchmark PDEs demonstrate that our method achieves superior accuracy and efficiency compared to existing approaches. The code  is released at \\url{https://github.com/BayesianAIGroup/TGPSolve-NonLinear-PDEs}", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13579", "url": null, "sourceid": 1458, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=XCudRYbJ1I", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11272, "modified": "2026-03-29T20:43:04.331416-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=XCudRYbJ1I", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "158", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13475, "uid": "dca5672ff3444c7e997aa9a2c4eb2094", "name": "Robust Federated Clustering under Heterogeneity and Adversaries", "authors": [{"id": 22185, "fullname": "Mart\u00edn Bravo", "url": "http://virtual.aistats.org/api/miniconf/users/22185?format=json", "institution": "Mohamed bin Zayed University of Artificial Intelligence"}, {"id": 22186, "fullname": "Sebastian Dalleiger", "url": "http://virtual.aistats.org/api/miniconf/users/22186?format=json", "institution": "KTH Royal Institute of Technology"}], "abstract": "Clustering distributed and private data is an increasingly important task across domains that handle sensitive information, such as life sciences and clinical research. In federated settings, clustering faces three challenges: heterogeneous client data distributions, adversarial behavior, and strict privacy requirements. Existing approaches often exhibit significant performance degradation under these conditions and fail to return accurate solutions. To overcome these limitations, we introduce a novel federated clustering algorithm that combines client-level differential privacy with Byzantine-robust aggregation at the server, based on a novel efficient and robust clustering procedure. Our method comes with theoretical robustness guarantees, and through extensive experiments on synthetic and real-world data, we demonstrate that it produces high-quality clusters in just a few communication rounds, even in scenarios where state-of-the-art methods fail.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13475", "url": null, "sourceid": 1589, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=hlw2laVrmW", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11168, "modified": "2026-03-29T20:43:00.354552-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=hlw2laVrmW", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "158", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13787, "uid": "e4873aa9a05cc5ed839561d121516766", "name": "Robustness and Generalization in Uncertainty-Aware Message Passing Neural Networks", "authors": [{"id": 22872, "fullname": "Alesia Chernikova", "url": "http://virtual.aistats.org/api/miniconf/users/22872?format=json", "institution": "Northeastern University"}, {"id": 19841, "fullname": "Moritz Laber", "url": "http://virtual.aistats.org/api/miniconf/users/19841?format=json", "institution": "Northeastern University"}, {"id": 19922, "fullname": "Narayan G. Sabhahit", "url": "http://virtual.aistats.org/api/miniconf/users/19922?format=json", "institution": "Northeastern University, Boston, USA"}, {"id": 22873, "fullname": "Tina Eliassi-Rad", "url": "http://virtual.aistats.org/api/miniconf/users/22873?format=json", "institution": "Northeastern University"}], "abstract": "Existing theoretical guarantees for message passing neural networks (MPNNs) assume deterministic node features. We address a more realistic scenario where noise or finite measurement precision introduces uncertainties in node feature values. First, we quantify uncertainty by propagating the moments of node-feature distributions through the MPNN architecture. To propagate moments through activation functions, we use the Taylor expansion and the pseudo-Taylor polynomial expansion. We then use the resulting node embedding distributions to analytically derive probabilistic adversarial robustness certificates for node classification tasks against L2-bounded perturbations of node features. Second, we model node features as multivariate random variables and introduce Feature Convolution Distance $FCD_p$, a pseudometric based on the Wasserstein distance. $FCD_p$ corresponds to the discriminative power of MPNNs at the node level. We show that MPNNs are globally Lipschitz continuous functions with respect to the pseudometric $FCD_p$. Using the covering number of the resulting pseudometric space, which is a subset of the Wasserstein space, we derive generalization bounds for MPNNs with uncertainties in node features. Together, these two complementary approaches---moment propagation for adversarial robustness and $FCD_p$ on the subset of the Wasserstein space for generalization---establish a unified theoretical framework that comprehensively addresses MPNN reliability under node feature uncertainty.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13787", "url": null, "sourceid": 1646, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=BcqtGTw9OZ", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11480, "modified": "2026-03-29T20:43:13.022080-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=BcqtGTw9OZ", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "160", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13745, "uid": "eae27d77ca20db309e056e3d2dcd7d69", "name": "Robust Generalization with Adaptive Optimal Transport Priors for Decision-Focused Learning", "authors": [{"id": 19799, "fullname": "Haixiang Sun", "url": "http://virtual.aistats.org/api/miniconf/users/19799?format=json", "institution": "Purdue University"}, {"id": 22755, "fullname": "Andrew Liu", "url": "http://virtual.aistats.org/api/miniconf/users/22755?format=json", "institution": "Purdue University"}], "abstract": "Few-shot learning requires models to generalize under limited supervision while remaining robust to distribution shifts. Existing Sinkhorn Distributionally Robust Optimization (DRO) methods provide theoretical guarantees but rely on a fixed reference distribution, which limits their adaptability. We propose a Prototype-Guided Distributionally Robust Optimization (PG-DRO) framework that learns class-adaptive priors from abundant base data via hierarchical optimal transport and embeds them into the Sinkhorn DRO formulation. This design enables few-shot information to be organically integrated into producing class-specific robust decisions that are both theoretically grounded and efficient, and further aligns the uncertainty set with transferable structural knowledge. Experiments show that PG-DRO achieves stronger robust generalization in few-shot scenarios, outperforming both standard learners and DRO baselines.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13745", "url": null, "sourceid": 205, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=EXv1wBpzaY", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11438, "modified": "2026-03-29T20:43:11.310917-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=EXv1wBpzaY", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "160", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13350, "uid": "414e773d5b7e5c06d564f594bf6384d0", "name": "Scalable Policy Maximization Under Network Interference", "authors": [{"id": 21914, "fullname": "Aidan Gleich", "url": "http://virtual.aistats.org/api/miniconf/users/21914?format=json", "institution": "Duke University"}, {"id": 21915, "fullname": "Eric Laber", "url": "http://virtual.aistats.org/api/miniconf/users/21915?format=json", "institution": "Duke University"}, {"id": 12626, "fullname": "Alexander Volfovsky", "url": "http://virtual.aistats.org/api/miniconf/users/12626?format=json", "institution": "Duke University"}], "abstract": "Many interventions, such as vaccines in clinical trials or coupons in online marketplaces, must be assigned sequentially without full knowledge of their effects. Multi-armed bandit algorithms have proven successful in such settings. However, standard independence assumptions fail when the treatment status of one individual impacts the outcomes of others, a phenomenon known as interference. We study optimal-policy learning under interference on large networks. Existing approaches to this problem require repeated observations of the same fixed network and struggle to scale in sample size beyond as few as fifteen connected units --- both limit applications. We show that common assumptions on the structure of interference enable a parsimonious linear parameterization of the reward function. We develop a scalable Thompson sampling algorithm that maximizes cumulative rewards on a $n$-node network while allowing for both nodes and edges to be sampled at each time period. We prove upper and lower bounds on Bayesian regret that imply near-optimality.  Simulation experiments show that our algorithm learns quickly and outperforms existing methods. The results close a key scalability gap between causal inference methods for interference and practical bandit algorithms, enabling policy optimization in large-scale networked systems.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13350", "url": null, "sourceid": 1923, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=x6pYSOVcNC", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11043, "modified": "2026-03-29T20:42:55.356410-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=x6pYSOVcNC", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "161", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13341, "uid": "39059724f73a9969845dfe4146c5660e", "name": "Root Cause Analysis of Outliers in Unknown Cyclic Graphs", "authors": [{"id": 21897, "fullname": "Daniela Schkoda", "url": "http://virtual.aistats.org/api/miniconf/users/21897?format=json", "institution": "Amazon, Technical University of Munich"}, {"id": 815, "fullname": "Dominik Janzing", "url": "http://virtual.aistats.org/api/miniconf/users/815?format=json", "institution": "Amazon"}], "abstract": "We study the propagation of outliers in cyclic causal graphs with linear structural equations, tracing them back to one or several \"root cause\" nodes. We show that it is possible to identify a short list of potential root causes provided that the perturbation is sufficiently strong and propagates according to the same structural equations as in the normal mode. This shortlist consists of the true root causes together with those of its parents lying on a cycle with the root cause. Notably, our method does not require prior knowledge of the causal graph and yields encouraging results on simulated data and real data from biology and cloud computing.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13341", "url": null, "sourceid": 270, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=yMy87GlRfE", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11034, "modified": "2026-03-29T20:42:54.949902-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=yMy87GlRfE", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "161", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13907, "uid": "41ae36ecb9b3eee609d05b90c14222fb", "name": "Scalable Learning of Multivariate Distributions via Coresets", "authors": [{"id": 19921, "fullname": "Zeyu Ding", "url": "http://virtual.aistats.org/api/miniconf/users/19921?format=json", "institution": "Lamarr Institute for Machine Learning and Artificial Intelligence"}, {"id": 23089, "fullname": "Katja Ickstadt", "url": "http://virtual.aistats.org/api/miniconf/users/23089?format=json", "institution": "Technische Universit\u00e4t Dortmund"}, {"id": 23090, "fullname": "Nadja Klein", "url": "http://virtual.aistats.org/api/miniconf/users/23090?format=json", "institution": "Karlsruhe Institute of Technology"}, {"id": 3725, "fullname": "Alexander Munteanu", "url": "http://virtual.aistats.org/api/miniconf/users/3725?format=json", "institution": "TU Dortmund"}, {"id": 23091, "fullname": "Simon Omlor", "url": "http://virtual.aistats.org/api/miniconf/users/23091?format=json", "institution": "TU Dortmund"}], "abstract": "Efficient and scalable non-parametric or semi-parametric regression analysis and density estimation are of crucial importance to the fields of statistics and machine learning. However, available methods are limited in their ability to handle large-scale data. We address this issue by developing a novel coreset construction for multivariate conditional transformation models (MCTMs) to enhance their scalability and training efficiency. To the best of our knowledge, these are the first coresets for semi-parametric distributional models. Our approach yields substantial data reduction via importance sampling. It ensures with high probability that the log-likelihood remains within multiplicative error bounds of $(1\\pm\\varepsilon)$ and thereby maintains statistical model accuracy. Compared to conventional full-parametric models, where coresets have been incorporated before, our semi-parametric approach exhibits enhanced adaptability, particularly in scenarios where complex distributions and non-linear relationships are present, but not fully understood. To address numerical problems associated with normalizing logarithmic terms, we follow a geometric approximation based on the convex hull of input data. This ensures feasible, stable, and accurate inference in scenarios involving large amounts of data. Numerical experiments demonstrate substantially improved computational efficiency when handling large and complex datasets, thus laying the foundation for a broad range of applications within the statistics and machine learning communities.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13907", "url": null, "sourceid": 417, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=0jJQy8Ofdg", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11600, "modified": "2026-03-29T20:43:18.113194-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=0jJQy8Ofdg", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "164", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13647, "uid": "7bccfde7714a1ebadf06c5f4cea752c1", "name": "FairSHAP: Preprocessing for Fairness Through Attribution-Based Data Augmentation", "authors": [{"id": 23270, "fullname": "Lin Zhu", "url": "http://virtual.aistats.org/api/miniconf/users/23270?format=json", "institution": "Aalto University"}, {"id": 19821, "fullname": "Yijun Bian", "url": "http://virtual.aistats.org/api/miniconf/users/19821?format=json", "institution": "University of Copenhagen"}, {"id": 14537, "fullname": "Lei You", "url": "http://virtual.aistats.org/api/miniconf/users/14537?format=json", "institution": "Technical University of Denmark"}], "abstract": "Ensuring fairness in machine learning models is critical, particularly in high-stakes domains where biased decisions can lead to serious societal consequences. However, existing preprocessing approaches generally lack transparent mechanisms for identifying which features are responsible for unfairness. This obscures the rationale behind data modifications. We introduce FairSHAP, a novel preprocessing framework that leverages Shapley value attribution to improve both individual and group fairness. FairSHAP identifies fairness-critical features in the training data using an interpretable measure of feature importance, and systematically modifies them through instance-level matching across sensitive groups. Our method effectively reduces discriminative risk (DR) with an instance-wise guarantee up to an interaction residual term, which is bounded under local matching, while simultaneously bounding the upper limit of demographic parity (DP), which in practice leads to its reduction. Experiments on multiple tabular datasets show that we achieve state-of-the-art or comparable performance across DR, DP, and equality of opportunity (EO) with minimal modifications, thereby preserving data fidelity. As a model-agnostic and transparent method, FairSHAP integrates seamlessly into existing machine learning pipelines and provides actionable insights into the sources of bias. Our code is available on https://github.com/ZhuMuMu0216/FairSHAP.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13647", "url": null, "sourceid": 1236, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=Ns2v9TPSMc", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11340, "modified": "2026-03-29T20:43:07.106179-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=Ns2v9TPSMc", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "55", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13755, "uid": "24f0d2c90473b2bc949ad962e61d9bcb", "name": "Shift is Good: Mismatched Data Mixing Improves Test Performance", "authors": [{"id": 22777, "fullname": "Marko Medvedev", "url": "http://virtual.aistats.org/api/miniconf/users/22777?format=json", "institution": "University of Chicago"}, {"id": 22778, "fullname": "Kaifeng Lyu", "url": "http://virtual.aistats.org/api/miniconf/users/22778?format=json", "institution": "Tsinghua University"}, {"id": 22779, "fullname": "Zhiyuan Li", "url": "http://virtual.aistats.org/api/miniconf/users/22779?format=json", "institution": "Toyota Technological Institute at Chicago"}, {"id": 1015, "fullname": "Nathan Srebro", "url": "http://virtual.aistats.org/api/miniconf/users/1015?format=json", "institution": "Toyota Technical Institute of Chicago"}], "abstract": "We consider training and testing on mixture distributions with different training and test proportions.  We show that in many settings, and in some sense generically, distribution shift can be beneficial, and test performance can improve due to mismatched training proportions.  In a variety of scenarios, we identify the optimal training proportions and the extent to which such a distribution shift can be beneficial.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13755", "url": null, "sourceid": 2105, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=E3BHEqHhmO", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11448, "modified": "2026-03-29T20:43:11.768725-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=E3BHEqHhmO", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "165", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13467, "uid": "38af86134b65d0f10fe33d30dd76442e", "name": "Secretary Problem with Predictions and Ordering", "authors": [{"id": 20614, "fullname": "Yiming Kang", "url": "http://virtual.aistats.org/api/miniconf/users/20614?format=json", "institution": "Independent"}], "abstract": "The classic secretary problem, a cornerstone of optimal stopping theory, assumes a random, immutable arrival order of candidates. While recent work has integrated machine-learned predictions to improve selection, the power to set the arrival order based on these predictions remains largely untapped. This paper introduces a novel framework for the secretary problem that leverages predictions for both valuation and strategic scheduling. We propose an algorithm that strategically controls the arrival time of the top-predicted candidate and dynamically adapts its hiring policy based on observed prediction accuracy. Our analysis shows that this approach achieves a worst-case competitive ratio of 0.229, surpassing the 0.215 bound of state-of-the-art algorithms that do not control ordering, bringing it closer to the upper bound of $1/e \\approx 0.368$ while maintaining consistency guarantees. Furthermore, we demonstrate that our ordering framework can be adapted to improve fairness guarantees, doubling the success probability in a known fair algorithm from 1/16 to 1/8. Our results highlight that controlling the sequence is a powerful tool for building more robust and fair learning-augmented online algorithms.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13467", "url": null, "sourceid": 176, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=jI1xC44cTn", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11160, "modified": "2026-03-29T20:43:00.077630-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=jI1xC44cTn", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "166", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13699, "uid": "16ba72172e6a4f1de54d11ab6967e371", "name": "Slithering through Gaps: Capturing Discrete Isolated Modes via Logistic Bridging", "authors": [{"id": 22656, "fullname": "Pinaki Mohanty", "url": "http://virtual.aistats.org/api/miniconf/users/22656?format=json", "institution": "Purdue University"}, {"id": 10220, "fullname": "Ruqi Zhang", "url": "http://virtual.aistats.org/api/miniconf/users/10220?format=json", "institution": "Purdue University"}], "abstract": "High-dimensional and complex discrete distributions often exhibit multimodal behavior due to inherent discontinuities, posing significant challenges for sampling. Gradient-based discrete samplers, while effective, frequently become trapped in local modes when confronted with rugged or disconnected energy landscapes. This limitation makes it difficult for sampling methods to achieve adequate mixing and convergence in high-dimensional multimodal discrete spaces. To address these challenges, we propose Hyperbolic Secant-squared Gibbs-Sampling (HiSS), a novel family of sampling algorithms that integrates a Metropolis-within-Gibbs framework to enhance mixing efficiency. HiSS leverages a logistic convolution kernel to couple the discrete sampling variable with the continuous auxiliary variable in a joint distribution. This design ensures that the auxiliary variable encapsulates the true target distribution while facilitating easy transitions between distant and disconnected modes. We provide theoretical guarantees of convergence and demonstrate empirically that HiSS outperforms many popular alternatives on a wide variety of tasks, including Ising models, binary neural networks, and combinatorial optimization.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13699", "url": null, "sourceid": 2063, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=IPVCtuZL8m", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11392, "modified": "2026-03-29T20:43:09.295606-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=IPVCtuZL8m", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "167", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13609, "uid": "698d51a19d8a121ce581499d7b701668", "name": "Sequential 1-bit Mean Estimation with Near-Optimal Sample Complexity", "authors": [{"id": 22484, "fullname": "Ivan Lau", "url": "http://virtual.aistats.org/api/miniconf/users/22484?format=json", "institution": "National University of Singapore"}, {"id": 596, "fullname": "Jonathan Scarlett", "url": "http://virtual.aistats.org/api/miniconf/users/596?format=json", "institution": "National University of Singapore"}], "abstract": "In this paper, we study the problem of distributed mean estimation with 1-bit communication constraints. We propose a mean estimator that is based on (randomized and sequentially-chosen) interval queries, whose 1-bit outcome indicates whether the given sample lies in the specified interval. Our estimator is $(\\epsilon, \\delta)$-PAC for all distributions with bounded mean ($-\\lambda \\le \\mathbb{E}(X) \\le \\lambda $) and variance ($\\mathrm{Var}(X) \\le \\sigma^2$) for some known parameters $\\lambda$ and $\\sigma$.  We derive a sample complexity bound $\\widetilde{O}\\big( \\frac{\\sigma^2}{\\epsilon^2}\\log\\frac{1}{\\delta} + \\log\\frac{\\lambda}{\\sigma}\\big)$, which matches the minimax lower bound for the unquantized setting up to logarithmic factors and the additional $\\log\\frac{\\lambda}{\\sigma}$ term that we show to be unavoidable.  We also establish an adaptivity gap for interval-query based estimators: the best non-adaptive mean estimator is considerably worse than our adaptive mean estimator for large $\\frac{\\lambda}{\\sigma}$.  Finally, we give tightened sample complexity bounds for distributions with stronger tail decay, and present additional variants that (i) handle an unknown sampling budget (ii) adapt to the unknown true variance given (possibly loose) upper and lower bounds on the variance, and (iii) use only two stages of adaptivity at the expense of more complicated (non-interval) queries.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13609", "url": null, "sourceid": 111, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=T8nCp7UYSv", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11302, "modified": "2026-03-29T20:43:05.537790-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=T8nCp7UYSv", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "168", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13364, "uid": "e8dfff4676a47048d6f0c4ef899593dd", "name": "Spectral Text Fusion: A Frequency-Aware Approach to Multimodal Time-Series Forecasting", "authors": [{"id": 19811, "fullname": "Huu Hiep Nguyen", "url": "http://virtual.aistats.org/api/miniconf/users/19811?format=json", "institution": "Applied Artificial Intelligence Initiative"}, {"id": 21944, "fullname": "Minh Nguyen", "url": "http://virtual.aistats.org/api/miniconf/users/21944?format=json", "institution": "Deakin University"}, {"id": 21945, "fullname": "Dung Nguyen", "url": "http://virtual.aistats.org/api/miniconf/users/21945?format=json", "institution": "Deakin University"}, {"id": 21946, "fullname": "Hung Le", "url": "http://virtual.aistats.org/api/miniconf/users/21946?format=json", "institution": "Deakin University"}], "abstract": "Multimodal time series forecasting is crucial in real-world applications, where decisions depend on both numerical data and contextual signals. The core challenge is to effectively combine temporal numerical patterns with the context embedded in other modalities, such as text. While most existing methods align textual features with time-series patterns one step at a time, they neglect the multiscale temporal influences of contextual information such as time-series cycles and dynamic shifts. This mismatch between local alignment and global textual context can be addressed by spectral decomposition, which separates time series into frequency components capturing both short-term changes and long-term trends. In this paper, we propose SpecTF, a simple yet effective framework that integrates the effect of textual data on time series in the frequency domain. Our method extracts textual embeddings, projects them into the frequency domain, and fuses them with the time series' spectral components using a lightweight cross-attention mechanism. This adaptively reweights frequency bands based on textual relevance before mapping the results back to the temporal domain for predictions. Experimental results demonstrate that SpecTF significantly outperforms state-of-the-art models across diverse multi-modal time series datasets while utilizing considerably fewer parameters. Code is available at \\url{https://github.com/hiepnh137/SpecTF}.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13364", "url": null, "sourceid": 1841, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=w4yjJc06zt", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11057, "modified": "2026-03-29T20:42:55.892189-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=w4yjJc06zt", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "168", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13757, "uid": "0f46c64b74a6c964c674853a89796c8e", "name": "SQuaT: Self-Supervised Knowledge Distillation via Student-Aware Quantized Teacher Features", "authors": [{"id": 19865, "fullname": "Hyeon Jun Lee", "url": "http://virtual.aistats.org/api/miniconf/users/19865?format=json", "institution": "Kookmin University"}, {"id": 22783, "fullname": "Hyeonsik Jo", "url": "http://virtual.aistats.org/api/miniconf/users/22783?format=json", "institution": "Kookmin University"}, {"id": 19373, "fullname": "Jinwoo Chung", "url": "http://virtual.aistats.org/api/miniconf/users/19373?format=json", "institution": ""}, {"id": 22784, "fullname": "Jangho Kim", "url": "http://virtual.aistats.org/api/miniconf/users/22784?format=json", "institution": "Kookmin University"}], "abstract": "Quantization-Aware Training (QAT) enables the deployment of quantized models with minimal accuracy degradation. However, in practical scenarios, training labels are often unavailable due to privacy, copyright, or cost constraints. Knowledge Distillation (KD) is a common approach to address this challenge, but we observe that prior work combining QAT with KD suffers from a fundamental limitation: during distillation, the range mismatch between the teacher and the quantized student model induces an unattainable residual, resulting in an irreducible lower bound on the distillation loss. Motivated by this observation, we propose SQuaT (Student-Aware Quantized Teacher Features), a label-free QAT framework with KD that theoretically eliminates this lower bound by applying the student\u2019s quantization parameters to quantize the teacher\u2019s features during distillation. Through comprehensive experiments across diverse settings, we demonstrate that SQuaT consistently outperforms strong baselines, with particularly pronounced gains in extreme low-bit (e.g., 1- and 2-bit) settings. Furthermore, extensive evaluations across various model design choices show that our approach does not rely on specific architectural assumptions, making it broadly applicable across diverse architectures and quantization settings. The source code is available at https://github.com/lcdbsa522/SQuaT.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13757", "url": null, "sourceid": 2281, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=E1rZnlgy6O", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11450, "modified": "2026-03-29T20:43:11.848044-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=E1rZnlgy6O", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "170", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13873, "uid": "facf9f743b083008a894eee7baa16469", "name": "TabTreeFormer: Tabular Data Generation Using Hybrid Tree-Transformer", "authors": [{"id": 23024, "fullname": "Jiayu Li", "url": "http://virtual.aistats.org/api/miniconf/users/23024?format=json", "institution": "University of Illinois at Urbana-Champaign"}, {"id": 23025, "fullname": "Bingyin Zhao", "url": "http://virtual.aistats.org/api/miniconf/users/23025?format=json", "institution": "Pixocial Technology"}, {"id": 23026, "fullname": "Zilong Zhao", "url": "http://virtual.aistats.org/api/miniconf/users/23026?format=json", "institution": "National University of Singapore"}, {"id": 23027, "fullname": "Uzair Javaid", "url": "http://virtual.aistats.org/api/miniconf/users/23027?format=json", "institution": "Betterdata"}, {"id": 23028, "fullname": "Biplab Sikdar", "url": "http://virtual.aistats.org/api/miniconf/users/23028?format=json", "institution": "National University of Singapore"}], "abstract": "Transformers have shown impressive results in tabular data generation. However, they lack domain-specific inductive biases which are critical for preserving the intrinsic characteristics of tabular data. They also suffer from poor scalability and efficiency due to quadratic computational complexity. In this paper, we propose TabTreeFormer, a hybrid transformer architecture that integrates inductive biases of tree-based models (e.g., non-smoothness and non-rotational invariance) to effectively handle the discrete and weakly correlated features in tabular datasets. To improve numerical fidelity and capture multimodal distributions, we introduce a novel tokenizer that learns token sequences based on the complexity of tabular values. This reduces vocabulary size and sequence length, yielding more compact and efficient representations without sacrificing performance. We evaluate TabTreeFormer on nine diverse datasets, benchmarking against eight generative models. We show that TabTreeFormer consistently outperforms baselines in utility, fidelity, and privacy metrics with competitive efficiency. Notably, in scenarios prioritizing data utility over privacy and efficiency, the best variant of TabTreeFormer delivers a 44% performance gain relative to its baseline variant. Our code is available at: https://github.com/li-jiayu-ljy/tabtreeformer.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13873", "url": null, "sourceid": 1564, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=4Kj1Xoqi23", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11566, "modified": "2026-03-29T20:43:16.819830-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=4Kj1Xoqi23", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "171", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13732, "uid": "a532400ed62e772b9dc0b86f46e583ff", "name": "Sparse Linear Bandits with Fixed Sparsity Support: Adversarial and Stochastic Regimes", "authors": [{"id": 5305, "fullname": "Kyoungseok Jang", "url": "http://virtual.aistats.org/api/miniconf/users/5305?format=json", "institution": "Chung-Ang University"}, {"id": 22114, "fullname": "Nam Tran", "url": "http://virtual.aistats.org/api/miniconf/users/22114?format=json", "institution": "University of Warwick"}, {"id": 3610, "fullname": "Nicol\u00f2 Cesa-Bianchi", "url": "http://virtual.aistats.org/api/miniconf/users/3610?format=json", "institution": "University of Milan"}], "abstract": "We study sparse linear bandits in both adversarial and stochastic settings. While existing literature has extensively explored sparse linear bandits in the stochastic regime, the adversarial setting, particularly for general $l_p$-ball action sets $(p>1)$, remains poorly understood. Our work addresses this gap by showing that the curse of dimensionality in adversarial linear bandits can be broken under a natural fixed sparsity support assumption. Specifically, we design algorithms for the $l_\\infty$- and $l_2$-balls that integrate sparsity support identification with the OSMD algorithm, achieving regret bounds $O(s\\sqrt{T}\\log T )$ and $O(\\sqrt{sT}\\log T )$, respectively. These results nearly match the optimal results when the sparsity support is known, and significantly improve upon the $ O(d\\sqrt{T}) $ regret of algorithms ignoring sparsity. Furthermore, in the stochastic setting, we show how the geometry of the $l_p$-ball action set influences both exploration and regret. Our work highlights fundamental contrasts between adversarial and stochastic regimes, and establishes the first regret guarantees for sparse adversarial linear bandits beyond the $l_1$-ball action set.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13732", "url": null, "sourceid": 637, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=FUiZHL1WNu", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11425, "modified": "2026-03-29T20:43:10.611937-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=FUiZHL1WNu", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "171", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13463, "uid": "5eac43aceba42c8757b54003a58277b5", "name": "Training Latent Diffusion Models with Interacting Particle Algorithms", "authors": [{"id": 19835, "fullname": "Tim Y. J. Wang", "url": "http://virtual.aistats.org/api/miniconf/users/19835?format=json", "institution": "Imperial College London"}, {"id": 22153, "fullname": "Juan Kuntz", "url": "http://virtual.aistats.org/api/miniconf/users/22153?format=json", "institution": "Relation Therapeutics"}, {"id": 22154, "fullname": "O. Deniz Akyildiz", "url": "http://virtual.aistats.org/api/miniconf/users/22154?format=json", "institution": "Imperial College London"}], "abstract": "We introduce a novel particle-based algorithm for end-to-end training of latent diffusion models. We reformulate the training task as minimizing a free energy functional and obtain a gradient flow that does so. By approximating the latter with a system of interacting particles, we obtain the algorithm, which we underpin theoretically by providing error guarantees. The novel algorithm compares favorably in experiments with previous particle-based methods and variational inference analogues.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13463", "url": null, "sourceid": 1245, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=jtigBUpHeq", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11156, "modified": "2026-03-29T20:42:59.896375-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=jtigBUpHeq", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "171", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13772, "uid": "9978b7063e297d84bb2ac8e46c1c845f", "name": "TENDE: Transfer Entropy Neural Diffusion Estimation", "authors": [{"id": 22824, "fullname": "Simon Pedro Galeano Munoz", "url": "http://virtual.aistats.org/api/miniconf/users/22824?format=json", "institution": "King Abdullah University of Science and Technology"}, {"id": 18275, "fullname": "Maurizio Filippone", "url": "http://virtual.aistats.org/api/miniconf/users/18275?format=json", "institution": "King Abdullah University of Science and Technology"}, {"id": 22825, "fullname": "Giulio Franzese", "url": "http://virtual.aistats.org/api/miniconf/users/22825?format=json", "institution": "Eurecom"}, {"id": 22826, "fullname": "Mustapha Bounoua", "url": "http://virtual.aistats.org/api/miniconf/users/22826?format=json", "institution": "Eurecom"}, {"id": 22827, "fullname": "Pietro Michiardi", "url": "http://virtual.aistats.org/api/miniconf/users/22827?format=json", "institution": "EURECOM"}], "abstract": "Transfer entropy is a fundamental measure for quantifying directed information flow in time series, with applications spanning neuroscience, finance, and complex systems analysis. However, existing estimation methods suffer from the curse of dimensionality, require restrictive distributional assumptions, or need exponentially large datasets for reliable convergence. We address these limitations in the literature by proposing TENDE (Transfer Entropy Neural Diffusion Estimation), a novel approach that leverages score-based diffusion models to estimate transfer entropy through conditional mutual information. By learning score functions of the relevant conditional distributions, TENDE provides flexible, scalable estimation while making minimal assumptions about the underlying data-generating process. We demonstrate superior accuracy and robustness compared to existing neural estimators and other state-of-the-art approaches across synthetic benchmarks and real data.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13772", "url": null, "sourceid": 2172, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=Cje2pdPusB", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11465, "modified": "2026-03-29T20:43:12.431562-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=Cje2pdPusB", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "172", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13576, "uid": "edfbe1afcf9246bb0d40eb4d8027d90f", "name": "Transportability without Graphs: A Bayesian Approach to Identifying s-Admissible Backdoor Sets", "authors": [{"id": 22407, "fullname": "Konstantina Lelova", "url": "http://virtual.aistats.org/api/miniconf/users/22407?format=json", "institution": "University of Crete"}, {"id": 9747, "fullname": "Gregory Cooper", "url": "http://virtual.aistats.org/api/miniconf/users/9747?format=json", "institution": "University of PIttsburgh"}, {"id": 22408, "fullname": "Sofia Triantafillou", "url": "http://virtual.aistats.org/api/miniconf/users/22408?format=json", "institution": "University of Crete"}], "abstract": "Transporting causal information across populations is a critical challenge in clinical decision-making. Causal modeling provides criteria for identifiability and transportability, but these require knowledge of the causal graph, which rarely holds in practice. We propose a Bayesian method that combines observational data from the target domain with experimental data from a different domain to identify s-admissible backdoor sets, which enable unbiased estimation of causal effects across populations, without requiring the causal graph. We prove that if such a set exists, we can always find one within the Markov boundary of the outcome, narrowing the search space, and we establish asymptotic convergence guarantees for our method.  We develop a greedy algorithm that reframes transportability as a feature selection problem, selecting conditioning sets that maximize the marginal likelihood of experimental data given observational data. In simulated and semi-synthetic data, our method correctly identifies transportability bias, improves causal effect estimation, and performs favorably against alternatives.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13576", "url": null, "sourceid": 740, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=XUK8q1ISCv", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11269, "modified": "2026-03-29T20:43:04.234530-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=XUK8q1ISCv", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "172", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13442, "uid": "ba3866600c3540f67c1e9575e213be0a", "name": "Sparse Offline Reinforcement Learning with Corruption Robustness", "authors": [{"id": 22114, "fullname": "Nam Tran", "url": "http://virtual.aistats.org/api/miniconf/users/22114?format=json", "institution": "University of Warwick"}, {"id": 5595, "fullname": "Andi Nika", "url": "http://virtual.aistats.org/api/miniconf/users/5595?format=json", "institution": "MPI-SWS"}, {"id": 9167, "fullname": "Goran Radanovic", "url": "http://virtual.aistats.org/api/miniconf/users/9167?format=json", "institution": "Max Planck Institute for Software Systems"}, {"id": 22115, "fullname": "Long Tran-Thanh", "url": "http://virtual.aistats.org/api/miniconf/users/22115?format=json", "institution": "The university of Warwick"}, {"id": 10953, "fullname": "Debmalya Mandal", "url": "http://virtual.aistats.org/api/miniconf/users/10953?format=json", "institution": "University of Warwick"}], "abstract": "We investigate robustness to strong data corruption in offline sparse reinforcement learning (RL). In our setting, an adversary may arbitrarily perturb a fraction of the collected trajectories from a high-dimensional but sparse Markov decision process, and our goal is to estimate a near-optimal policy. The main challenge is that, in the high-dimensional regime where the number of samples $N$ is smaller than the feature dimension $d$, exploiting sparsity is essential for obtaining non-vacuous guarantees but has not been systematically studied in offline RL. We analyse the problem under uniform coverage and sparse single-concentrability assumptions. While Least Square Value Iteration (LSVI), a standard approach for robust offline RL, performs well under uniform coverage, we show that integrating sparsity into LSVI is unnatural, and its analysis may break down due to overly pessimistic bonuses.  To overcome this, we propose actor\u2013critic methods with sparse robust estimator oracles, which avoid the use of pointwise pessimistic bonuses and provide the first non-vacuous guarantees for sparse offline RL under single-policy concentrability coverage. Moreover, we extend our results to the contaminated setting and show that our algorithm remains robust under strong contamination. Our results provide the first non-vacuous guarantees in high-dimensional sparse MDPs with single-policy concentrability coverage and corruption, showing that learning near-optimal policy remains possible in regimes where traditional robust offline RL techniques may fail.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13442", "url": null, "sourceid": 732, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=lrQkwrvUFW", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11135, "modified": "2026-03-29T20:42:59.088792-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=lrQkwrvUFW", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "172", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13360, "uid": "4888241374e8c62ddd9b4c3cfd091f96", "name": "TexTSC: Class-Texture Preserving Data Condensation for Time Series Classification", "authors": [{"id": 19888, "fullname": "Pouya Hosseinzadeh", "url": "http://virtual.aistats.org/api/miniconf/users/19888?format=json", "institution": "Utah State University"}, {"id": 21932, "fullname": "Peiyu Li", "url": "http://virtual.aistats.org/api/miniconf/users/21932?format=json", "institution": "Utah State University"}, {"id": 21933, "fullname": "Omar Bahri", "url": "http://virtual.aistats.org/api/miniconf/users/21933?format=json", "institution": "Utah State University"}, {"id": 21934, "fullname": "Soukaina Filali Boubrahimi", "url": "http://virtual.aistats.org/api/miniconf/users/21934?format=json", "institution": "Utah State University"}, {"id": 21935, "fullname": "Shah Muhammad Hamdi", "url": "http://virtual.aistats.org/api/miniconf/users/21935?format=json", "institution": "Utah State University"}], "abstract": "Dataset condensation seeks to generate a small set of synthetic examples that can replace large real datasets for training, but existing methods for time series often rely on unstable training-trajectory matching or capture only limited signal structure. We present TexTSC, a condensation framework that preserves class structure using spectro-temporal second-order statistics instead of trajectory replay. TexTSC models each class\u2019s \u201ctexture\u201d as the co-activation pattern among intermediate teacher features, aligning Gram matrices of activations in time to capture temporal correlations and in frequency to capture spectral envelopes and harmonics. A short-lag autocorrelation term stabilizes local rhythm, while a lightweight gradient anchor at the final layer ensures discriminative power. TexTSC optimizes synthetic sequences directly, remains model-agnostic, and requires only closed-form statistics, making it simple and stable. Experiments on standard benchmarks show that TexTSC produces compact datasets that retain class-conditional structure and achieve higher classification accuracy than first-order or single-domain baselines.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13360", "url": null, "sourceid": 1843, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=wS87abrnaZ", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11053, "modified": "2026-03-29T20:42:55.756934-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=wS87abrnaZ", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "174", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13353, "uid": "5fa9e41bfec0725742cc9d15ef594120", "name": "The Cross-Context Threshold Test: Detecting Discrimination Under Environmental Shifts", "authors": [{"id": 21919, "fullname": "Jun Yuan", "url": "http://virtual.aistats.org/api/miniconf/users/21919?format=json", "institution": "University of Virginia, Charlottesville"}, {"id": 21920, "fullname": "Xinyue Ye", "url": "http://virtual.aistats.org/api/miniconf/users/21920?format=json", "institution": "The University of Alabama"}], "abstract": "We study how threshold tests for detecting discrimination under environmental shifts, focusing on the Veil-of-Darkness (VoD) setting where visibility changes between daylight and darkness. We show that standard threshold tests, when applied separately to daylight and darkness data, violate key assumptions: risk distributions drift across contexts and thresholds fluctuate arbitrarily. We propose a cross-context threshold test that enforces distributional invariance and monotonic threshold decay. Using New York City stop-and-frisk data and synthetic experiments, we demonstrate that this model yields more reliable thresholds, improves bias detection, and aligns with the counterfactual logic of the VoD test. Our framework generalizes to fairness auditing whenever environmental context influences decisions.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13353", "url": null, "sourceid": 1894, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=x1qKvXw6n0", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11046, "modified": "2026-03-29T20:42:55.461531-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=x1qKvXw6n0", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "175", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13861, "uid": "0e4e946668cf2afc4299b462b812caca", "name": "Unmixing Mean Embeddings for Domain Adaptation with Target Label Proportion", "authors": [{"id": 4053, "fullname": "Alain Rakotomamonjy", "url": "http://virtual.aistats.org/api/miniconf/users/4053?format=json", "institution": "University of Rouen / Criteo AI Lab"}, {"id": 22999, "fullname": "Maxime Berar", "url": "http://virtual.aistats.org/api/miniconf/users/22999?format=json", "institution": "Universit\u00e9 de Rouen Normandie"}, {"id": 11035, "fullname": "Mokhtar Alaya", "url": "http://virtual.aistats.org/api/miniconf/users/11035?format=json", "institution": "LMAC - Universit\u00e9 de Technologie de Compi\u00e8gne"}], "abstract": "We introduce a novel approach to domain adaptation within the context of Learning from Label Proportions (LLP). We address the challenging scenario where labeled samples are available in the source domain,  but only bags of unlabeled samples with their corresponding label proportions  are accessible in the target domain.   Our proposed method, bagMME (Bag Matching Mean Embeddings),  tackles the distributional shift between domains by focusing on matching class-conditional distributions. A key contribution of bagMME is a simple yet effective unmixing strategy that  leverages the target label proportions to estimate the target class-conditional  mean embeddings. These estimated target means are then aligned with their corresponding source class-conditional means, thereby reducing the domain discrepancy. We theoretically  demonstrate the soundness of our approach and its effectiveness in mitigating distributional shifts.  Extensive experiments on various computer vision datasets showcase the superior performance of bagMME compared to state-of-the-art baselines. Our results highlight the critical role of incorporating target label  proportions into the learning process for improved generalization on the target domain.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13861", "url": null, "sourceid": 1412, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=5RnLbMM7A8", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11554, "modified": "2026-03-29T20:43:16.292909-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=5RnLbMM7A8", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "177", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13803, "uid": "7d04bbbe5494ae9d2f5a76aa1c00fa2f", "name": "Stochastic Bandits on Mixture Distributions: Metrics & Regret Bounds", "authors": [{"id": 22897, "fullname": "Adit Jain", "url": "http://virtual.aistats.org/api/miniconf/users/22897?format=json", "institution": "Cornell University"}, {"id": 9491, "fullname": "Sujay Bhatt", "url": "http://virtual.aistats.org/api/miniconf/users/9491?format=json", "institution": "JP Morgan"}, {"id": 18164, "fullname": "Alec Koppel", "url": "http://virtual.aistats.org/api/miniconf/users/18164?format=json", "institution": "J.P. Morgan Chase"}], "abstract": "Multimodal reward distributions naturally arise in real-world applications such as targeted recommendations to heterogeneous sub-populations and selective unit-level interventions. These settings challenge standard mean or risk-based bandit approaches, requiring metrics that quantify the merit of mixture parameters without prior mode knowledge. We consider the bandit setting where the reward associated with an arm is sampled from a finite mixture of Gaussians, which is strictly more general than the unimodal setting. We consider ranking arms using functions of the mixture parameters and propose methods to minimize the cumulative regret with respect to the induced ranking.  We show that the achievable pseudo-regret has a lower bound of the order $\\Omega(\\mathsf{T}^{1/2})$ and propose an explore and exploit based on expectation maximization (ETE-EM) algorithm which achieves a regret of $\\widetilde{\\mathsf{O}}(\\mathsf{T}^{2/3})$. Further, we show that the modification of Thompson sampling (TS-EM) achieves a Bayes regret of $\\widetilde{\\mathsf{O}}(\\mathsf{T}^{1/2})$. Experiments validate our approach in practice, where we benchmark against both algorithms designed for sub-Gaussian bandits and naive clustering-based extensions of empirical CDF methods, showing our approach achieves consistently lower regret across choice of metrics.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13803", "url": null, "sourceid": 486, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=APPzkBvwX5", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11496, "modified": "2026-03-29T20:43:13.661481-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=APPzkBvwX5", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "177", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13594, "uid": "e449b9317dad920c0dd5ad0a2a2d5e49", "name": "The Information Geometry of Local Generalization Dynamics", "authors": [{"id": 19743, "fullname": "Emmanouil-Marios Athanasakos", "url": "http://virtual.aistats.org/api/miniconf/users/19743?format=json", "institution": "National and Kapodistrian University of Athens, Dpt. of Informatics and Telecommunications"}], "abstract": "Information-theoretic bounds on generalization are foundational to learning theory, yet their static form offers limited insight into the dynamic, iterative nature of modern optimization. This gap is addressed herein by developing a local theory of generalization based on Euclidean Information Theory, where each update is modeled as a perturbation vector. The resulting analysis shows that the change in the generalization gap is bounded by the expected squared norm of this vector, a quantity interpreted as the local generalization cost. The proposed local bound is proven to be the first-order approximation of classic global bounds, revealing their underlying differential structure. Within this framework, it is further revealed that the optimal update direction is mathematically equivalent to the natural gradient, offering an information-geometric justification for natural gradient descent. Finally, the overall theory is validated through experiments showing that the derived bound closely tracks training dynamics.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13594", "url": null, "sourceid": 1663, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=VZVeNli1wd", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11287, "modified": "2026-03-29T20:43:04.914126-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=VZVeNli1wd", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "177", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13653, "uid": "2c89109d42178de8a367c0228f169bf8", "name": "Variational Grey-Box Dynamics Matching", "authors": [{"id": 19457, "fullname": "Gurjeet Singh", "url": "http://virtual.aistats.org/api/miniconf/users/19457?format=json", "institution": "University of Geneva"}, {"id": 22565, "fullname": "Frantzeska Lavda", "url": "http://virtual.aistats.org/api/miniconf/users/22565?format=json", "institution": "HES-SO : UAS Western Switzerland"}, {"id": 17834, "fullname": "Giangiacomo Mercatali", "url": "http://virtual.aistats.org/api/miniconf/users/17834?format=json", "institution": "Geneva School of Business Administration, HES-SO University of Applied Sciences of Western Switzerland;"}, {"id": 17816, "fullname": "Alexandros Kalousis", "url": "http://virtual.aistats.org/api/miniconf/users/17816?format=json", "institution": "University of Applied Sciences Western Switzerland"}], "abstract": "Deep generative models such as flow matching and diffusion models have shown great potential in learning complex distributions and dynamical systems, but often act as black-boxes, neglecting underlying physics. In contrast, physics-based simulation models described by ODEs/PDEs remain interpretable, but may have missing or unknown terms, unable to fully describe real-world observations. We bridge this gap with a novel grey-box method that integrates incomplete physics models directly into generative models. Our approach learns dynamics from observational trajectories alone, without ground-truth physics parameters, in a simulation-free manner that avoids scalability and stability issues of Neural ODEs. The core of our method lies in modelling a structured variational distribution within the flow matching framework, by using two latent encodings: one to model the missing stochasticity and multi-modal velocity, and a second to encode physics parameters as a latent variable with a physics-informed prior. Furthermore, we present an adaptation of the framework to handle second-order dynamics. Our experiments on representative ODE/PDE problems and real-world weather forecasting demonstrate that our method performs on par with or superior to fully data-driven approaches and previous grey-box baselines, while preserving the interpretability of the physics model. Our code is available at [https://github.com/DMML-Geneva/VGB-DM](https://github.com/DMML-Geneva/VGB-DM).", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13653", "url": null, "sourceid": 1244, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=NMuUPLBc84", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11346, "modified": "2026-03-29T20:43:07.423383-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=NMuUPLBc84", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "178", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13819, "uid": "d198bd736a97e7cecfdf8f4f2027ef80", "name": "The Unseen Adversaries: Robust and Generalized Defense Against Adversarial Patches", "authors": [{"id": 19885, "fullname": "Vishesh Kumar", "url": "http://virtual.aistats.org/api/miniconf/users/19885?format=json", "institution": "IISER Bhopal"}, {"id": 22931, "fullname": "Akshay Agarwal", "url": "http://virtual.aistats.org/api/miniconf/users/22931?format=json", "institution": "IISER Bhopal, India"}], "abstract": "The vulnerabilities of deep neural networks against singularities have raised serious concerns regarding their deployment in the physical world. One of the most prominent and impactful physical-world adversarial perturbations is the attachment of patches to clean images, known as an adversarial patch attack. Similarly, natural noises such as Gaussian and Salt\\&Pepper are highly prevalent in the real world. The current research need arises from the above vulnerabilities and the lack of efforts to tackle these two singularities independently and, especially, in combination. In this research, we have, for the first time, combined these two prominent singularities and proposed a novel dataset. Using this dataset, we have conducted a benchmark study of singularity data-point detection using features from several convolutional neural networks. For classification, rather than the popular neural network-based parameter tuning, we have used traditional yet effective machine learning classifiers. The extensive experiments across various in- and out-of-distribution (OOD) singularities reveal several interesting findings about the effectiveness of classifiers and show that it is hard to defend against adversaries when they are treated independently, and inefficient classifiers are selected.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13819", "url": null, "sourceid": 2034, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=8xSNoxxqsO", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11512, "modified": "2026-03-29T20:43:14.348824-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=8xSNoxxqsO", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "179", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13639, "uid": "4d6b3e38b952600251ee92fe603170ff", "name": "Three-Step Nav: A Hierarchical Global\u2013Local Planner for Zero-Shot Vision-and-Language Navigation", "authors": [{"id": 22532, "fullname": "Wanrong Zheng", "url": "http://virtual.aistats.org/api/miniconf/users/22532?format=json", "institution": "University of Southern California"}, {"id": 22533, "fullname": "Yunhao Ge", "url": "http://virtual.aistats.org/api/miniconf/users/22533?format=json", "institution": "NVIDIA"}, {"id": 4300, "fullname": "Laurent Itti", "url": "http://virtual.aistats.org/api/miniconf/users/4300?format=json", "institution": "University of Southern California"}], "abstract": "Breakthrough progress in vision-based navigation through unknown environments has been achieved by using multimodal large language models (MLLMs). These models can plan a sequence of motions by evaluating the current view at each time step against the task and goal given to the agent. However, current zero-shot Vision-and-Language Navigation (VLN) agents powered by MLLMs still tend to drift off course, halt prematurely, and achieve low overall success rates. We propose Three-Step Nav to counteract these failures with a three-view protocol: First, \"look forward\" to extract global landmarks and sketch a coarse plan. Then, \"look now\" to align the current visual observation with the next sub-goal for fine-grained guidance. Finally, \"look backward\" audits the entire trajectory to correct accumulated drift before stopping. Requiring no gradient updates or task-specific fine-tuning, our planner drops into existing VLN pipelines with minimal overhead. Three-Step Nav achieves state-of-the-art zero-shot performance on the R2R-CE and RxR-CE dataset.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13639", "url": null, "sourceid": 2087, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=OvYmtu0hp6", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11332, "modified": "2026-03-29T20:43:06.780707-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=OvYmtu0hp6", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "180", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13586, "uid": "7fec306d1e665bc9c748b5d2b99a6e97", "name": "Where the Score Lives: A Wavelet View of Diffusion", "authors": [{"id": 22436, "fullname": "Emma Finn", "url": "http://virtual.aistats.org/api/miniconf/users/22436?format=json", "institution": "Harvard University"}, {"id": 22437, "fullname": "Binxu Wang", "url": "http://virtual.aistats.org/api/miniconf/users/22437?format=json", "institution": "Harvard University"}, {"id": 22438, "fullname": "T. Keller", "url": "http://virtual.aistats.org/api/miniconf/users/22438?format=json", "institution": "Harvard University"}, {"id": 22439, "fullname": "Demba Ba", "url": "http://virtual.aistats.org/api/miniconf/users/22439?format=json", "institution": "Harvard University"}], "abstract": "Score-based generative models have had remarkable success over the last decade in generating a diverse set of visually plausible images.  A variety of architectures including CNNs,  U-Nets, and Transformers have been used as the score-approximation network in such diffusion modeling; however, to date, relatively little is known about how these architectural choices impact generative behavior. In this work, to provide insight into this area, we propose an analytically solvable parameterization of the score function using an expansion in a 2D orthogonal wavelet basis. In particular, we derive interpretable optimal score functions in terms of the moments of the data distribution. We use this parametrization to provide an architecture-agnostic, moment-based analysis that reveals which attributes of the data distribution tend to matter most for denoising. Our score machine is flexible enough to partially mimic the relevant inductive biases of multiple architectures, including U-Nets, and CNNs, taking a step towards understanding why different score architectures can exhibit distinct generative behavior. Since our score is solvable in terms of the moments of the data, we can begin to understand how the data distribution interacts with the score network to produce the behavior we observe in diffusion models.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13586", "url": null, "sourceid": 1053, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=WZz9xOHfZr", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11279, "modified": "2026-03-29T20:43:04.566954-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=WZz9xOHfZr", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "181", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13780, "uid": "6c8349cc7260ae62e3b1396831a8398f", "name": "TAS-EGNN: Task-Aware Spectral Ego-Graphs for Efficient GNNs-Based Classification", "authors": [{"id": 22847, "fullname": "Mebarka Allaoui", "url": "http://virtual.aistats.org/api/miniconf/users/22847?format=json", "institution": "Bishop&#x27;s University"}, {"id": 22848, "fullname": "Rachid Hedjam", "url": "http://virtual.aistats.org/api/miniconf/users/22848?format=json", "institution": "Bishop&#x27;s University"}, {"id": 22849, "fullname": "Sonia Gupta", "url": "http://virtual.aistats.org/api/miniconf/users/22849?format=json", "institution": "Indian Institute of Technology, Delhi"}], "abstract": "Graph Neural Networks (GNNs) achieve strong accuracy but remain costly to train on large graphs and in resource-constrained settings. Coreset selection mitigates this by training on a compact, representative node subset, yet many existing methods rely on expensive spectral routines or bilevel and iterative optimizations. We propose a Task-Aware Spectral Ego-Graph Neural Network (TAS-EGNN) that scores nodes within lightweight ego-graphs by combining (i) local spectral complexity, (ii) predictive uncertainty, and (iii) supervised error signals, followed by a greedy coverage step to avoid redundancy. TAS-EGNN circumvents heavy optimization, using only local spectra (or moment proxies) and a single model forward pass to obtain task signals. We evaluate TAS-EGNN across three benchmark tasks: citation networks, social networks, and graph-based bank transaction fraud detection. The third task, in particular, underscores the algorithm\u2019s effectiveness in anomaly detection for highly imbalanced settings. TAS-EGNN matches or surpasses state-of-the-art reduction baselines, across \\emph{budgets} (i.e., the allowed size of the selected training subset, controlled via the coreset ratio), including condensation, coarsening, and ego-graph selection, while delivering substantial wall-clock and peak-memory savings. Time and memory profiling show that TAS-EGNN tracks the lower envelope among structure-aware methods and scales to large graphs, whereas several other works reach OOT/OOM. These results indicate that efficiently encoded task-aware structural priors enable robust, scalable coreset selection for both standard node classification and fraud detection. The source code will be available on GitHub.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13780", "url": null, "sourceid": 45, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=C6GP4NM1FP", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11473, "modified": "2026-03-29T20:43:12.756070-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=C6GP4NM1FP", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "182", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13403, "uid": "5227b6aaf294f5f027273aebf16015f2", "name": "Tractable Gaussian Phase Retrieval with Heavy Tails and Adversarial Corruption with Near-Linear Sample Complexity", "authors": [{"id": 19880, "fullname": "Santanu Das", "url": "http://virtual.aistats.org/api/miniconf/users/19880?format=json", "institution": "Tata Institute Of Fundamental Research, Mumbai"}, {"id": 22015, "fullname": "jatin batra", "url": "http://virtual.aistats.org/api/miniconf/users/22015?format=json", "institution": "Tata institute of fundamental research, Mumbai"}], "abstract": "Phase retrieval is the classical problem of recovering a signal $x^* \\in \\mathbb{R}^n$ from its noisy phaseless measurements $y_i = \\langle a_i, x^* \\rangle^2 + \\zeta_i$ (where $\\zeta_i$ denotes noise, and $a_i$ is the sensing vector) for $i \\in [m]$. The problem of phase retrieval has a rich history, with a variety of applications such as optics, crystallography, heteroscedastic regression, astrophysics, etc. A major consideration in algorithms for phase retrieval is \\emph{robustness} against measurement errors. In recent breakthroughs in algorithmic robust statistics, efficient algorithms have been developed for several parameter estimation tasks such as mean estimation, covariance estimation, robust principal component analysis (PCA), etc. in the presence of heavy-tailed noise and adversarial corruptions. In this paper, we study efficient algorithms for robust phase retrieval with heavy-tailed noise when a constant fraction of both the measurements $y_i$ and the sensing vectors $a_i$ may be arbitrarily adversarially corrupted. For this problem, Buna and Rebeschini (AISTATS 2025) very recently gave an \\emph{exponential} time algorithm with sample complexity $O(n \\log n)$. Their algorithm needs a \\emph{robust spectral initialization}, specifically, a robust estimate of the top eigenvector of a covariance matrix, which they deemed to be beyond known efficient algorithmic techniques (similar spectral initializations are a key ingredient of a large family of phase retrieval algorithms). In this work, we make a connection between robust spectral initialization and recent algorithmic advances in robust PCA, yielding the first polynomial-time algorithms for robust phase retrieval with both heavy-tailed noise and adversarial corruptions, in fact with near-linear (in $n$) sample complexity.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13403", "url": null, "sourceid": 2127, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=pqzhavU710", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11096, "modified": "2026-03-29T20:42:57.506933-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=pqzhavU710", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "182", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13903, "uid": "3937230de3c8041e4da6ac3246a888e8", "name": "Two mathematical models of knowledge distillation", "authors": [{"id": 23084, "fullname": "Audrey Xie", "url": "http://virtual.aistats.org/api/miniconf/users/23084?format=json", "institution": "Computer Science Department, Stanford University"}, {"id": 23085, "fullname": "Ludwig Schmidt", "url": "http://virtual.aistats.org/api/miniconf/users/23085?format=json", "institution": "Stanford University"}, {"id": 539, "fullname": "John Duchi", "url": "http://virtual.aistats.org/api/miniconf/users/539?format=json", "institution": "Stanford University"}], "abstract": "Many hypotheses compete to explain the successes of knowledge distillation. To help address this, we propose and analyze a mathematical model of distillation, which suggests that distillation's performance comes not from obtaining better models but from easier to optimize landscapes. For generalized linear models trained with stochastic gradient descent, we prove that distillation fits performant student models asymptotically more quickly than non-distilled models. In rank-1 matrix approximation, we characterize conditions on the target matrix under which gradient descent with distillation converges strictly faster than training on the supervised objective. The theory helps delineate the ways distillation provides benefits (i.e., in optimization speed, not in generalization), and experiments on real datasets corroborate the theoretical predictions.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13903", "url": null, "sourceid": 2117, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=1EI4hUZ9oi", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11596, "modified": "2026-03-29T20:43:17.946425-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=1EI4hUZ9oi", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "183", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13678, "uid": "66808e327dc79d135ba18e051673d906", "name": "The Minimax Lower Bound of Kernel Stein Discrepancy Estimation", "authors": [{"id": 19787, "fullname": "Jose Cribeiro-Ramallo", "url": "http://virtual.aistats.org/api/miniconf/users/19787?format=json", "institution": "Karlsruhe Institute of Technology"}, {"id": 22612, "fullname": "Agnideep Aich", "url": "http://virtual.aistats.org/api/miniconf/users/22612?format=json", "institution": "University of Louisiana at Lafayette"}, {"id": 14692, "fullname": "Florian Kalinke", "url": "http://virtual.aistats.org/api/miniconf/users/14692?format=json", "institution": "Karlsruhe Institute of Technology (KIT)"}, {"id": 22613, "fullname": "Ashit Baran Aich", "url": "http://virtual.aistats.org/api/miniconf/users/22613?format=json", "institution": "Presidency University"}, {"id": 4962, "fullname": "Zoltan Szabo", "url": "http://virtual.aistats.org/api/miniconf/users/4962?format=json", "institution": "LSE"}], "abstract": "Kernel Stein discrepancies (KSDs) have emerged as a powerful tool for quantifying goodness-of-fit over the last decade, featuring numerous successful applications. To the best of our knowledge, all existing KSD estimators with known rate achieve $\\sqrt n$-convergence. In this work, we present two complementary results (with different proof strategies), establishing that the minimax lower bound of KSD estimation is $n^{-1/2}$ and settling the optimality of these estimators. Our first result focuses on KSD estimation on $\\mathbb R^d$ with the Langevin-Stein operator; our explicit constant for the Gaussian base kernel indicates that the difficulty of KSD estimation may increase exponentially with the dimensionality $d$. Our second result settles the minimax lower bound for KSD estimation on general domains.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13678", "url": null, "sourceid": 414, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=KT7IF27o7K", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11371, "modified": "2026-03-29T20:43:08.373619-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=KT7IF27o7K", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "184", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13712, "uid": "6e5025ccc7d638ae4e724da8938450a6", "name": "Understanding the Benefits of SimCLR Pre-Training in Two-Layer Convolutional Neural Networks", "authors": [{"id": 22702, "fullname": "Han Zhang", "url": "http://virtual.aistats.org/api/miniconf/users/22702?format=json", "institution": "University of Hong Kong"}, {"id": 17962, "fullname": "Yuan Cao", "url": "http://virtual.aistats.org/api/miniconf/users/17962?format=json", "institution": "University of Hong Kong"}], "abstract": "SimCLR is a popular contrastive learning method for vision tasks, renowned for its ability to pre-train neural networks to learn efficient representations. Despite its empirical effectiveness, the theoretical understanding of SimCLR is still very limited, even in the simplest learning scenarios. In this paper, we introduce a theoretical case study of SimCLR. Specifically, we consider training a two-layer convolutional neural network (CNN) to learn a toy image data model that has been considered in a series of recent works. For this particular learning task, we precisely characterize the label complexity under which SimCLR pre-training followed by supervised fine-tuning achieves approximately zero training loss and almost optimal test loss. Notably, the label complexity for SimCLR pre-training is far less demanding compared to direct supervised training, especially when the signal-to-noise ratio in the data is low. Our analysis sheds light on the benefits of SimCLR in learning with fewer labels.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13712", "url": null, "sourceid": 2163, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=HN1ItuXq5M", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11405, "modified": "2026-03-29T20:43:09.827002-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=HN1ItuXq5M", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "185", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13412, "uid": "248e844336797ec98478f85e7626de4a", "name": "The Riemannian Geometry Associated to Gradient Flows of Linear Convolutional Networks", "authors": [{"id": 22031, "fullname": "El Mehdi Achour", "url": "http://virtual.aistats.org/api/miniconf/users/22031?format=json", "institution": "University Mohammed VI Polytechnic (College of Computing)"}, {"id": 18407, "fullname": "Kathl\u00e9n Kohn", "url": "http://virtual.aistats.org/api/miniconf/users/18407?format=json", "institution": "KTH Royal Institute of Technology"}, {"id": 22032, "fullname": "Holger Rauhut", "url": "http://virtual.aistats.org/api/miniconf/users/22032?format=json", "institution": "Ludwig-Maximilians-Universit\u00e4t M\u00fcnchen"}], "abstract": "We study geometric properties of the gradient flow for learning deep linear convolutional networks. For linear fully connected networks, it has been shown recently that the corresponding gradient flow on parameter space can be written as a Riemannian gradient flow on function space (i.e., on the product of weight matrices) if the initialization satisfies a so-called balancedness condition. We establish that the gradient flow on parameter space for learning linear convolutional networks can be written as a Riemannian gradient flow on function space regardless of the initialization. This result holds for $D$-dimensional convolutions with $D \\geq 2$, and for $D =1$ it holds if all so-called strides of the convolutions are greater than one. The corresponding Riemannian metric depends on the initialization.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13412", "url": null, "sourceid": 432, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=otSGTt4PMw", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11105, "modified": "2026-03-29T20:42:57.855261-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=otSGTt4PMw", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "185", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13908, "uid": "59bcda7c438bad7d2afffe9e2fed00be", "name": "VIPaint: Image Inpainting with Pre-Trained Diffusion Models via Variational Inference", "authors": [{"id": 19829, "fullname": "Sakshi Agarwal", "url": "http://virtual.aistats.org/api/miniconf/users/19829?format=json", "institution": "University of California, Irvine"}, {"id": 23092, "fullname": "Gabriel Hope", "url": "http://virtual.aistats.org/api/miniconf/users/23092?format=json", "institution": "University of California, Irvine"}, {"id": 23093, "fullname": "Jimin Heo", "url": "http://virtual.aistats.org/api/miniconf/users/23093?format=json", "institution": "University of California, Irvine"}, {"id": 19924, "fullname": "Erik B. Sudderth", "url": "http://virtual.aistats.org/api/miniconf/users/19924?format=json", "institution": "University of California, Irvine"}], "abstract": "Diffusion probabilistic models learn to remove noise added during training, generating novel data (e.g., images) from Gaussian noise through sequential denoising. However, conditioning the generative process on corrupted or masked images is challenging. While various methods have been proposed for inpainting masked images with diffusion priors, they often fail to produce samples from the true conditional distribution, especially for large masked regions. Additionally, many can't be applied to latent diffusion models which have been demonstrated to generate high-quality images at a significantly lower computational cost. We propose a hierarchical variational inference algorithm that optimizes a non-Gaussian Markov approximation of the true diffusion posterior. Our VIPaint method outperforms existing approaches to inpainting, producing diverse high-quality imputations, while also being effective for other inverse problems like deblurring and superresolution.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13908", "url": null, "sourceid": 1845, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=0ehuNXBslr", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11601, "modified": "2026-03-29T20:43:18.147417-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=0ehuNXBslr", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "188", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13466, "uid": "019d385eb67632a7e958e23f24bd07d7", "name": "Train Less, Infer Faster: Efficient Model Finetuning and Compression via Structured Sparsity", "authors": [{"id": 22165, "fullname": "Jonathan Svirsky", "url": "http://virtual.aistats.org/api/miniconf/users/22165?format=json", "institution": "Bar-Ilan University"}, {"id": 22166, "fullname": "Yehonathan Refael", "url": "http://virtual.aistats.org/api/miniconf/users/22166?format=json", "institution": "Google"}, {"id": 13458, "fullname": "Ofir Lindenbaum", "url": "http://virtual.aistats.org/api/miniconf/users/13458?format=json", "institution": "Bar Ilan University"}], "abstract": "Fully finetuning foundation language models (LMs) with billions of parameters is often impractical due to high computational costs, memory requirements, and the risk of overfitting. Although methods like low-rank adapters help address these challenges by adding small trainable modules to the frozen LM, they also increase memory usage and do not reduce inference latency. We uncover an intriguing phenomenon: sparsifying specific model rows and columns enables efficient task adaptation without requiring weight tuning. We propose a scheme for effective finetuning via sparsification using training stochastic gates, which requires minimal trainable parameters, reduces inference time, and removes 20--40\\% of model parameters without significant accuracy loss. Empirical results show it outperforms recent finetuning baselines in efficiency and performance. Additionally, we provide theoretical guarantees for the convergence of this stochastic gating process, and show that our method admits a simpler and better-conditioned optimization landscape compared to LoRA. Our results highlight sparsity as a compelling mechanism for task-specific adaptation in LMs.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13466", "url": null, "sourceid": 433, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=jU4ERfrjpH", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11159, "modified": "2026-03-29T20:43:00.040154-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=jU4ERfrjpH", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "188", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13805, "uid": "90599c8fdd2f6e7a03ad173e2f535751", "name": "WSBD: Freezing-Based Optimizer for Quantum Neural Networks", "authors": [{"id": 22901, "fullname": "Christopher Kverne", "url": "http://virtual.aistats.org/api/miniconf/users/22901?format=json", "institution": "Florida International University"}, {"id": 22902, "fullname": "Mayur Akewar", "url": "http://virtual.aistats.org/api/miniconf/users/22902?format=json", "institution": "Florida International University"}, {"id": 22903, "fullname": "Yuqian Huo", "url": "http://virtual.aistats.org/api/miniconf/users/22903?format=json", "institution": "Rice University"}, {"id": 22904, "fullname": "Tirthak Patel", "url": "http://virtual.aistats.org/api/miniconf/users/22904?format=json", "institution": "Rice University"}, {"id": 22905, "fullname": "Janki Bhimani", "url": "http://virtual.aistats.org/api/miniconf/users/22905?format=json", "institution": "Florida International University"}], "abstract": "The training of Quantum Neural Networks (QNNs) is hindered by the high computational cost of gradient estimation and the barren plateau problem, where optimization landscapes become intractably flat. To address these challenges, we introduce Weighted Stochastic Block Descent (WSBD), a novel optimizer with a dynamic, parameter-wise freezing strategy. WSBD intelligently focuses computational resources by identifying and temporarily freezing less influential parameters based on a gradient-derived importance score. This approach significantly reduces the number of forward passes required per training step and helps navigate the optimization landscape more effectively. Unlike pruning or layer-wise freezing, WSBD maintains full expressive capacity while adapting throughout training. Our extensive evaluation shows that WSBD converges on average 63.9\\% faster than Adam for the popular ground-state-energy problem, an advantage that grows with QNN size. We provide a formal convergence proof for WSBD and show that parameter-wise freezing outperforms traditional layer-wise approaches in QNNs. Project page: https://github.com/Damrl-lab/WSBD-Stochastic-Freezing-Optimizer.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13805", "url": null, "sourceid": 1796, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=AKoOSM6d3r", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11498, "modified": "2026-03-29T20:43:13.737719-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=AKoOSM6d3r", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "190", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13501, "uid": "67e103b0761e60683e83c559be18d40c", "name": "Uncertainty-Aware Meta-Learning with Analytically Tractable Posterior", "authors": [{"id": 22248, "fullname": "Young-Jin Park", "url": "http://virtual.aistats.org/api/miniconf/users/22248?format=json", "institution": "MIT"}, {"id": 22249, "fullname": "Cesar Almecija", "url": "http://virtual.aistats.org/api/miniconf/users/22249?format=json", "institution": "Mines Paris-PSL"}, {"id": 22250, "fullname": "Apoorva Sharma", "url": "http://virtual.aistats.org/api/miniconf/users/22250?format=json", "institution": "NVIDIA"}, {"id": 4501, "fullname": "Navid Azizan", "url": "http://virtual.aistats.org/api/miniconf/users/4501?format=json", "institution": " Massachusetts Institute of Technology"}], "abstract": "Meta-learning is a popular approach for learning new tasks with limited data by leveraging the commonalities among different tasks. However, meta-learned models can perform poorly when context data is too limited, or when data is drawn from an out-of-distribution (OoD) task. Especially in safety-critical settings, this necessitates an uncertainty-aware approach to meta-learning. In addition, the often multimodal nature of task distributions can pose unique challenges to meta-learning methods. In this work, we present UNLIMITED, a meta-learning method that (1) makes probabilistic predictions on in-distribution tasks efficiently, (2) is capable of detecting OoD context data, and (3) handles heterogeneous, multimodal task distributions effectively. The strength of our framework lies in its solid theoretical basis, enabling exact Bayesian inference for principled uncertainty estimation and robust generalization. We achieve this by adopting a probabilistic perspective and training a parametric, tunable task distribution via Bayesian inference on a linearized neural network, leveraging Gaussian process theory. Moreover, we make our approach computationally tractable by leveraging a low-rank prior covariance learning scheme based on the Fisher Information Matrix. Our numerical analysis demonstrates that UNLIMITED quickly adapts to new tasks and remains accurate even in low-data regimes, it effectively detects OoD tasks, and that both of these properties continue to hold for multimodal task distributions.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13501", "url": null, "sourceid": 641, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=ftAZdcNSmL", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11194, "modified": "2026-03-29T20:43:01.348368-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=ftAZdcNSmL", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "190", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13568, "uid": "d67d8ab4f4c10bf22aa353e27879133c", "name": "Unsupervised Ensemble Learning Through Deep Energy-based Models", "authors": [{"id": 22390, "fullname": "Ariel Maymon", "url": "http://virtual.aistats.org/api/miniconf/users/22390?format=json", "institution": "Bar-Ilan University"}, {"id": 22391, "fullname": "Yanir Buznah", "url": "http://virtual.aistats.org/api/miniconf/users/22391?format=json", "institution": "Bar-Ilan University"}, {"id": 19861, "fullname": "Uri Shaham", "url": "http://virtual.aistats.org/api/miniconf/users/19861?format=json", "institution": "Bar Ilan University"}], "abstract": "Unsupervised ensemble learning emerged to address the challenge of combining multiple learners' predictions without access to ground truth labels or additional data. This paradigm is crucial in scenarios where evaluating individual classifier performance or understanding their strengths is challenging due to limited information. We propose a novel deep energy-based method for constructing an accurate meta-learner using only the predictions of individual learners, potentially capable of capturing complex dependence structures between them. Our approach requires no labeled data, learner features, or problem-specific information, and has theoretical guarantees for when learners are conditionally independent. We demonstrate superior performance across diverse ensemble scenarios, including challenging mixture of experts settings. Our experiments span standard ensemble datasets and curated datasets designed to test how the model fuses expertise from multiple sources. These results highlight the potential of unsupervised ensemble learning to harness collective intelligence, especially in data-scarce or privacy-sensitive environments.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13568", "url": null, "sourceid": 39, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=YF1ObZwFnk", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11261, "modified": "2026-03-29T20:43:03.897712-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=YF1ObZwFnk", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "192", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13899, "uid": "6081594975a764c8e3a691fa2b3a321d", "name": "Welfare-Centric Clustering", "authors": [{"id": 23079, "fullname": "Claire Jie Zhang", "url": "http://virtual.aistats.org/api/miniconf/users/23079?format=json", "institution": "Department of Computer Science, University of Washington"}, {"id": 18418, "fullname": "Seyed Esmaeili", "url": "http://virtual.aistats.org/api/miniconf/users/18418?format=json", "institution": "University of Chicago"}, {"id": 12739, "fullname": "Jamie Morgenstern", "url": "http://virtual.aistats.org/api/miniconf/users/12739?format=json", "institution": "U Washington"}], "abstract": "Fair clustering has traditionally focused on ensuring equitable group representation or equalizing group-specific clustering costs. However, \\citet{dickerson2024fair} recently showed that these fairness notions may yield undesirable or unintuitive clustering outcomes and advocated for a welfare-centric clustering approach that models the utilities of the groups. In this work, we model group utilities based on both distances and proportional representation and formalize two optimization objectives based on welfare-centric clustering: the Rawlsian (Egalitarian) objective and the Utilitarian objective. We introduce novel algorithms for both objectives and prove theoretical guarantees for them. Empirical evaluations on multiple real-world datasets demonstrate that our methods significantly outperform existing fair clustering baselines.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13899", "url": null, "sourceid": 711, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=1YcI04VDUr", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11592, "modified": "2026-03-29T20:43:17.802560-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=1YcI04VDUr", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "196", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13441, "uid": "b73ce398c39f506af761d2277d853a92", "name": "When Can Federated Learning Match Centralized Learning? A PAC-Bayesian Generalization Gap Analysis", "authors": [{"id": 19797, "fullname": "Xuanyu Chen", "url": "http://virtual.aistats.org/api/miniconf/users/19797?format=json", "institution": "The University of Sydney"}, {"id": 22113, "fullname": "Shuai Wang", "url": "http://virtual.aistats.org/api/miniconf/users/22113?format=json", "institution": "Northwest Polytechnical University Xi&#x27;an"}, {"id": 21908, "fullname": "NAN YANG", "url": "http://virtual.aistats.org/api/miniconf/users/21908?format=json", "institution": "University of Sydney"}, {"id": 21911, "fullname": "Dong Yuan", "url": "http://virtual.aistats.org/api/miniconf/users/21911?format=json", "institution": "University of Sydney"}], "abstract": "The growing focus on distributed data and privacy has spurred the rise of Federated Learning (FL). Empirical studies show that, under equal resources, FL often underperforms centralized training, but the reasons behind this gap remain theoretically unclear. This lack of understanding leaves open whether FL is inherently inferior in generalization and how the gap might be closed. We address this by formulating FL as a server-based SGD optimization problem over distributed data and analyzing the generalization gap within the PAC-Bayesian framework. Our analysis derives non-vacuous bounds on this gap, showing that such a gap necessarily exists under equal resources and depends on training parameters. We further prove that the gap can be fully eliminated only by introducing new clients or adding new data to existing clients, with the latter being more efficient. In contrast, allowing FL to have advantages in other resources, such as larger models or more communication rounds, cannot close the gap. As a complementary analysis, we also confirm from a stability perspective that centralized FL holds a generalization advantage over decentralized FL, justifying our FL formulation choice. Extensive experiments across different model architectures and datasets validate our theory.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13441", "url": null, "sourceid": 160, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=m436nG28pv", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11134, "modified": "2026-03-29T20:42:59.053167-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=m436nG28pv", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "197", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13509, "uid": "812b4ba287f5ee0bc9d43bbf5bbe87fb", "name": "Where You Place the Norm Matters: From Prejudiced to Neutral Initializations", "authors": [{"id": 22273, "fullname": "Emanuele Francazi", "url": "http://virtual.aistats.org/api/miniconf/users/22273?format=json", "institution": "EPFL - EPF Lausanne"}, {"id": 22274, "fullname": "Francesco Pinto", "url": "http://virtual.aistats.org/api/miniconf/users/22274?format=json", "institution": "Delft University of Technology"}, {"id": 3603, "fullname": "Aurelien Lucchi", "url": "http://virtual.aistats.org/api/miniconf/users/3603?format=json", "institution": "ETH Zurich"}, {"id": 17793, "fullname": "Marco Baity-Jesi", "url": "http://virtual.aistats.org/api/miniconf/users/17793?format=json", "institution": "Eawag"}], "abstract": "Normalization layers, such as Batch Normalization and Layer Normalization, are central components in modern neural networks, widely adopted to improve training stability and generalization. While their practical effectiveness is well documented, a detailed theoretical understanding of how normalization affects model behavior\u2014starting from initialization\u2014remains an important open question. In this work, we investigate how both the presence and placement of normalization within hidden layers influence the statistical properties of network predictions before training begins. In particular, we study how these choices shape the distribution of class predictions at initialization, which can range from unbiased (Neutral) to highly concentrated (Prejudiced) toward a subset of classes. Our analysis shows that normalization placement induces systematic differences in the initial prediction behavior of neural networks, which in turn shape the dynamics of learning. By linking architectural choices to prediction statistics at initialization, our work provides a principled understanding of how normalization can influence early training behavior and offers guidance for more controlled and interpretable network design.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13509", "url": null, "sourceid": 95, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=egBZyQbUjt", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11202, "modified": "2026-03-29T20:43:01.654356-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=egBZyQbUjt", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "198", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13476, "uid": "3cf166c6b73f030b4f67eeaeba301103", "name": "On Relation-Aware Slicing in Cross-Domain Alignment", "authors": [{"id": 22187, "fullname": "Dhruv Sarkar", "url": "http://virtual.aistats.org/api/miniconf/users/22187?format=json", "institution": "Indian Institute of Technology Kharagpur"}, {"id": 22188, "fullname": "Aprameyo Chakrabartty", "url": "http://virtual.aistats.org/api/miniconf/users/22188?format=json", "institution": "Indian Institute of Technology Kharagpur"}, {"id": 22189, "fullname": "Anish Chakrabarty", "url": "http://virtual.aistats.org/api/miniconf/users/22189?format=json", "institution": "Indian Statistical Institute, Kolkata"}, {"id": 10097, "fullname": "Swagatam Das", "url": "http://virtual.aistats.org/api/miniconf/users/10097?format=json", "institution": "Indian Statistical Institute"}], "abstract": "The Sliced Gromov-Wasserstein (SGW) distance, aiming to relieve the computational cost of solving a non-convex quadratic program that is the Gromov-Wasserstein distance, utilizes projecting directions sampled uniformly from unit hyperspheres. This slicing mechanism incurs unnecessary computational costs due to uninformative directions, which also affects the representative power of the distance. However, finding a more appropriate distribution over the projecting directions (_slicing distribution_) is often an optimization problem in itself that comes with its own computational cost. In addition, with more intricate distributions, the sampling itself may be expensive. As a remedy, we propose an optimization-free slicing distribution that provides fast sampling for the Monte Carlo approximation. We do so by introducing the Relation-Aware Projecting Direction (RAPD), effectively capturing the pairwise association of each of two pairs of random vectors, each following their ambient law. This enables us to derive the Relation-Aware Slicing Distribution (RASD), a location-scale law corresponding to sampled RAPDs. Finally, we introduce the RASGW distance and its variants, e.g., IWRASGW (Importance Weighted RASGW), which overcome the shortcomings experienced by SGW. We theoretically analyze its properties and substantiate its empirical prowess using extensive experiments on various alignment tasks.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13476", "url": null, "sourceid": 499, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=hawDS7QpCr", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11169, "modified": "2026-03-29T20:43:00.390951-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=hawDS7QpCr", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "123", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13892, "uid": "310dcbbf4cce62f762a2aaa148d556bd", "name": "$k$-PCA for (non-squared) Euclidean Distances: Deterministic Polynomial Time Approximation", "authors": [{"id": 19350, "fullname": "Daniel Greenhut", "url": "http://virtual.aistats.org/api/miniconf/users/19350?format=json", "institution": "University of Haifa"}, {"id": 4076, "fullname": "Dan Feldman", "url": "http://virtual.aistats.org/api/miniconf/users/4076?format=json", "institution": "The University of Haifa"}], "abstract": "Given an integer $k\\geq1$ and a set $P$ of $n$ points in $\\mathbb{R}^d$, the classic $k$-PCA (Principal Component Analysis) approximates the affine \\emph{$k$-subspace mean} of $P$, which is the $k$-dimensional affine linear subspace that minimizes its sum of squared Euclidean distances ($\\ell_{2,2}$-norm) over the points of $P$, i.e., the mean of these distances. The \\emph{$k$-subspace median} is the subspace that minimizes its sum of (non-squared) Euclidean distances ($\\ell_{2,1}$-mixed norm), i.e., their median. The median subspace is usually more sparse and robust to noise/outliers than the mean, but also much harder to approximate since, unlike the $\\ell_{z,z}$ (non-mixed) norms, it is non-convex for $k<d-1$.  We provide the first polynomial-time deterministic algorithm whose both running time and approximation factor are not exponential in $k$. More precisely, the multiplicative approximation factor is $\\sqrt{d}$, and the running time is polynomial in the size of the input. We expect that our technique would be useful for many other related problems, such as $\\ell_{2,z}$ norm of distances for $z\\not \\in \\{1,2\\}$, e.g., $z=\\infty$, and handling outliers/sparsity.  Open code and experimental results on real-world datasets are also provided.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13892", "url": null, "sourceid": 333, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=2DjZFQ7oia", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11585, "modified": "2026-03-29T20:43:17.537435-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=2DjZFQ7oia", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "1", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13444, "uid": "f9028faec74be6ec9b852b0a542e2f39", "name": "A Bayesian Information-Theoretic Approach to Data Attribution", "authors": [{"id": 13256, "fullname": "Dharmesh Tailor", "url": "http://virtual.aistats.org/api/miniconf/users/13256?format=json", "institution": "University of Amsterdam"}, {"id": 22119, "fullname": "Nicol\u00f2 Felicioni", "url": "http://virtual.aistats.org/api/miniconf/users/22119?format=json", "institution": "Spotify"}, {"id": 22120, "fullname": "Kamil Ciosek", "url": "http://virtual.aistats.org/api/miniconf/users/22120?format=json", "institution": "Spotify"}], "abstract": "Training Data Attribution (TDA) seeks to trace model predictions back to influential training examples, enhancing interpretability and safety. We formulate TDA as a Bayesian information-theoretic problem: subsets are scored by the information loss they induce\u2014the entropy increase at a query when removed. This criterion credits examples for resolving predictive uncertainty rather than label noise. To scale to modern networks, we approximate information loss using a Gaussian Process surrogate built from tangent features. For even larger-scale retrieval, we relax the information-gain objective and add a variance correction for scalable attribution in vector databases. Our method aligns with classical influence scores for single-example attribution, while promoting diversity for subsets. Experiments show competitive performance on counterfactual sensitivity and ground-truth retrieval, showing that our method scales to modern architectures bridging principled measures with practice.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13444", "url": null, "sourceid": 838, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=lkBZ0WtmAQ", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11137, "modified": "2026-03-29T20:42:59.156783-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=lkBZ0WtmAQ", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "1", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13906, "uid": "da8ce53cf0240070ce6c69c48cd588ee", "name": "A Divergence-Based Method for Weighting and Averaging Model Predictions", "authors": [{"id": 19808, "fullname": "Olav Benjamin Vassend", "url": "http://virtual.aistats.org/api/miniconf/users/19808?format=json", "institution": "University of Inland Norway"}], "abstract": "This paper uses a minimum divergence framework to introduce a new way of calculating model weights that can be used to average probabilistic predictions from statistical and machine learning models. The method is general and can be applied regardless of whether the models under consideration are fit to data using frequentist, Bayesian, or some other fitting method. The proposed method is motivated in two different ways and is shown empirically to perform better than or on a par with standard model averaging methods, including model stacking and model averaging that relies on Akaike-style negative exponentiated model weighting, especially when the sample size is small. Our theoretical analysis explains why the method has a small-sample advantage.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13906", "url": null, "sourceid": 793, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=0r6ifRf7Xw", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11599, "modified": "2026-03-29T20:43:18.073899-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=0r6ifRf7Xw", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "2", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13483, "uid": "3871bd64012152bfb53fdf04b401193f", "name": "A Covering Framework for Offline POMDPs Learning using Belief Space Metric", "authors": [{"id": 22202, "fullname": "Youheng Zhu", "url": "http://virtual.aistats.org/api/miniconf/users/22202?format=json", "institution": "Northwestern University"}, {"id": 17960, "fullname": "Yiping Lu", "url": "http://virtual.aistats.org/api/miniconf/users/17960?format=json", "institution": "Northwestern University"}], "abstract": "In off\u2011policy evaluation (OPE) for partially observable Markov decision processes (POMDPs), an agent must infer hidden states from past observations, which exacerbates both the curse of horizon and the curse of memory in existing OPE methods. This paper introduces a novel covering analysis framework that exploits the intrinsic metric structure of the belief space (distributions over latent states) to relax traditional coverage assumptions. By focusing on the policies with stability property, we derive error bounds that mitigate exponential blow-ups in horizon and memory length. Our unified analysis technique applies to a broad class of OPE algorithms, yielding concrete error bounds and coverage requirements expressed in terms of belief space metrics rather than raw history coverage. We illustrate the improved sample efficiency of this framework via case studies: the double sampling Bellman error minimization algorithm, and the memory-based future-dependent value functions (FDVF). In both cases, our coverage definition based on the belief\u2010space metric yields tighter bounds.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13483", "url": null, "sourceid": 622, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=h0S9W11dM2", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11176, "modified": "2026-03-29T20:43:00.658707-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=h0S9W11dM2", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "3", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13374, "uid": "a8f8f60264024dca151f164729b76c0b", "name": "A Geometric Approach to Optimal Experimental Design", "authors": [{"id": 21964, "fullname": "Gavin Kerrigan", "url": "http://virtual.aistats.org/api/miniconf/users/21964?format=json", "institution": "University of Oxford"}, {"id": 18665, "fullname": "Christian Andersson Naesseth", "url": "http://virtual.aistats.org/api/miniconf/users/18665?format=json", "institution": "University of Amsterdam"}, {"id": 1406, "fullname": "Tom Rainforth", "url": "http://virtual.aistats.org/api/miniconf/users/1406?format=json", "institution": "University of Oxford"}], "abstract": "We introduce a novel geometric framework for optimal experimental design (OED). Traditional OED approaches, such as those based on mutual information, rely explicitly on probability densities, leading to restrictive invariance properties. To address these limitations, we propose the mutual transport dependence (MTD), a measure of statistical dependence grounded in optimal transport theory which provides a geometric objective for optimizing designs. Unlike conventional approaches, the MTD can be tailored to specific downstream estimation problems by choosing appropriate geometries on the underlying spaces. We demonstrate that our framework produces high-quality designs while offering a flexible alternative to standard information-theoretic techniques.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13374", "url": null, "sourceid": 1437, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=u0aepMHQ5p", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11067, "modified": "2026-03-29T20:42:56.238061-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=u0aepMHQ5p", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "3", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13783, "uid": "fddd7938a71db5f81fcc621673ab67b7", "name": "A Gaussian Process View on Observation Noise and Initialization in Wide Neural Networks", "authors": [{"id": 22858, "fullname": "Sergio Calvo Ordo\u00f1ez", "url": "http://virtual.aistats.org/api/miniconf/users/22858?format=json", "institution": "University of Oxford"}, {"id": 22859, "fullname": "Jonathan Plenk", "url": "http://virtual.aistats.org/api/miniconf/users/22859?format=json", "institution": "University of Oxford"}, {"id": 22860, "fullname": "Richard Bergna", "url": "http://virtual.aistats.org/api/miniconf/users/22860?format=json", "institution": "University of Cambridge"}, {"id": 22861, "fullname": "Alvaro Cartea", "url": "http://virtual.aistats.org/api/miniconf/users/22861?format=json", "institution": "University of Oxford"}, {"id": 509, "fullname": "Jose Miguel Hernandez-Lobato", "url": "http://virtual.aistats.org/api/miniconf/users/509?format=json", "institution": "University of Cambridge"}, {"id": 22862, "fullname": "Konstantina Palla", "url": "http://virtual.aistats.org/api/miniconf/users/22862?format=json", "institution": "Spotify Research"}, {"id": 22120, "fullname": "Kamil Ciosek", "url": "http://virtual.aistats.org/api/miniconf/users/22120?format=json", "institution": "Spotify"}], "abstract": "Performing gradient descent in a wide neural network is equivalent to computing the posterior mean of a Gaussian Process with the Neural Tangent Kernel (NTK-GP), for a specific choice of prior mean and with zero observation noise. However, existing formulations of this result have two limitations: i) the resultant NTK-GP assumes no noise in the observed target variables, which can result in suboptimal predictions with noisy data; ii) it is unclear how to extend the equivalence to an arbitrary prior mean, a crucial aspect of formulating a well-specified model. To address the first limitation, we introduce a regularizer into the neural network's training objective, formally showing its correspondence to incorporating observation noise into the NTK-GP model. To address the second, we introduce a \\textit{shifted network} that enables arbitrary prior mean functions. This approach further allows us to obtain the posterior mean with gradient descent on a single neural network, without expensive ensembling or kernel matrix inversion. Our theoretical insights are validated empirically, with experiments exploring different values of observation noise, datasets, and network architectures. These results remove key obstacles that have limited the practical use of NTK-GP equivalence in applied Gaussian process modeling.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13783", "url": null, "sourceid": 1683, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=BwRIYQ7fkP", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11476, "modified": "2026-03-29T20:43:12.856840-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=BwRIYQ7fkP", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "4", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13724, "uid": "67f7fb873eaf29526a11a9b7ac33bfac", "name": "A Finite Time Analysis of Thompson Sampling for Bayesian Optimization with Preferential Feedback", "authors": [{"id": 308, "fullname": "Sattar Vakili", "url": "http://virtual.aistats.org/api/miniconf/users/308?format=json", "institution": "MediaTek Research"}, {"id": 14445, "fullname": "Joseph Lazzaro", "url": "http://virtual.aistats.org/api/miniconf/users/14445?format=json", "institution": "Imperial College London"}, {"id": 22724, "fullname": "Davide Buffelli", "url": "http://virtual.aistats.org/api/miniconf/users/22724?format=json", "institution": "MediaTek Research"}, {"id": 22725, "fullname": "Da-shan Shiu", "url": "http://virtual.aistats.org/api/miniconf/users/22725?format=json", "institution": "MediaTek"}], "abstract": "Preference feedback---pairwise comparisons instead of scalar scores---has seen growing use in applications such as human-, lab-, expert-in-the-loop design and scientific discoveries. We propose a Thompson Sampling (TS) approach to Bayesian optimization with preferential feedback that models comparisons through a monotone link on latent utility differences and leverages the dueling kernel induced by a base kernel. We give a finite-time analysis showing that its performance matches that of standard TS for conventional Bayesian optimization with scalar feedback. The analysis exploits TS's anchor-invariance for challenger selection and introduces a double-TS pairing variant. We also demonstrate the performance on both synthetic and real examples.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13724", "url": null, "sourceid": 445, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=G2UzUsM64N", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11417, "modified": "2026-03-29T20:43:10.320724-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=G2UzUsM64N", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "4", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13356, "uid": "8b0d268963dd0cfb808aac48a549829f", "name": "A Goemans-Williamson type algorithm for identifying subcohorts in clinical trials", "authors": [{"id": 21926, "fullname": "Pratik Worah", "url": "http://virtual.aistats.org/api/miniconf/users/21926?format=json", "institution": "Google"}], "abstract": "We design an efficient algorithm that outputs tests for identifying predominantly homogeneous subcohorts of patients from large in-homogeneous datasets. Our theoretical contribution is a rounding technique, similar to that of Goemans and Wiliamson (1995), which approximates the optimal solution within a factor of $0.82$. As an application, we use our algorithm to trade-off sensitivity for specificity to systematically identify clinically interesting homogeneous subcohorts of patients in the RNA microarray data set for breast cancer from Curtis et al. (2012). One identified subcohort suggests a link between LXR over-expression and BRCA2 and MSH6 methylation levels for patients in that subcohort.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13356", "url": null, "sourceid": 1471, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=wx9pulxj9k", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11049, "modified": "2026-03-29T20:42:55.574362-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=wx9pulxj9k", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "4", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13688, "uid": "b197ffdef2ddc3308584dce7afa3661b", "name": "Accelerated Distributed Optimization with Compression and Error Feedback", "authors": [{"id": 22631, "fullname": "Yuan Gao", "url": "http://virtual.aistats.org/api/miniconf/users/22631?format=json", "institution": "CISPA, saarland university, saarland informatics campus"}, {"id": 22632, "fullname": "Anton Rodomanov", "url": "http://virtual.aistats.org/api/miniconf/users/22632?format=json", "institution": "CISPA"}, {"id": 22633, "fullname": "Jeremy Rack", "url": "http://virtual.aistats.org/api/miniconf/users/22633?format=json", "institution": "Universit\u00e4t des Saarlandes"}, {"id": 5265, "fullname": "Sebastian Stich", "url": "http://virtual.aistats.org/api/miniconf/users/5265?format=json", "institution": "CISPA Helmholtz Center for Information Security gGmbH"}], "abstract": "Modern machine learning tasks often involve massive datasets and models, necessitating distributed optimization algorithms with reduced communication overhead. Communication compression, where clients transmit compressed updates to a central server, has emerged as a key technique to mitigate communication bottlenecks. However, the theoretical understanding of stochastic distributed optimization with contractive compression remains limited, particularly in conjunction with Nesterov acceleration---a cornerstone for achieving faster convergence in optimization. In this paper, we propose a novel algorithm, ADEF (**A**ccelerated **D**istributed **E**rror **F**eedback), which integrates Nesterov acceleration, contractive compression, error feedback, and gradient difference compression. We prove that ADEF achieves the first accelerated convergence rate for stochastic distributed optimization with contractive compression in the general convex regime. Numerical experiments validate our theoretical findings and demonstrate the practical efficacy of ADEF in reducing communication costs while maintaining fast convergence.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13688", "url": null, "sourceid": 1441, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=JdX5bJOJLC", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11381, "modified": "2026-03-29T20:43:08.815780-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=JdX5bJOJLC", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "6", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13577, "uid": "db6ebd0566994d14a1767f14eb6fba81", "name": "Accelerating PDE Surrogates via RL-Guided Mesh Optimization", "authors": [{"id": 19886, "fullname": "yang meng", "url": "http://virtual.aistats.org/api/miniconf/users/19886?format=json", "institution": "university of chicago"}, {"id": 23266, "fullname": "Ruoxi Jiang", "url": "http://virtual.aistats.org/api/miniconf/users/23266?format=json", "institution": "Fudan University"}, {"id": 22410, "fullname": "Zhuokai Zhao", "url": "http://virtual.aistats.org/api/miniconf/users/22410?format=json", "institution": "Meta"}, {"id": 18477, "fullname": "Chong Liu", "url": "http://virtual.aistats.org/api/miniconf/users/18477?format=json", "institution": "University at Albany, State University of New York"}, {"id": 950, "fullname": "Rebecca Willett", "url": "http://virtual.aistats.org/api/miniconf/users/950?format=json", "institution": "U Chicago"}, {"id": 943, "fullname": "Yuxin Chen", "url": "http://virtual.aistats.org/api/miniconf/users/943?format=json", "institution": "UChicago"}], "abstract": "Deep learning\u2013based surrogate models for parametric partial differential equations (PDEs) can deliver high-fidelity approximations but remain prohibitively data-hungry: training often requires thousands of fine-grid simulations, each incurring substantial computational cost. To address this challenge, we introduce RLMesh, an end-to-end framework for efficient surrogate training under limited simulation budget. The key idea is to use reinforcement learning (RL) to adaptively allocate mesh grid points non-uniformly within each simulation domain, focusing numerical resolution in regions most critical for accurate PDE solutions. A lightweight proxy model further accelerates RL training by providing efficient reward estimates without full surrogate retraining. Experiments on standard PDE benchmarks, including 1D Burgers\u2019 equation and 2D Darcy flow, demonstrate that RLMesh achieves competitive accuracy to baselines but with substantially fewer simulation queries. These results show that solver-level spatial adaptivity can dramatically improve the efficiency of surrogate training pipelines, enabling practical deployment of learning-based PDE surrogates across a wide range of problems.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13577", "url": null, "sourceid": 1438, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=XPhYq5Hty8", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11270, "modified": "2026-03-29T20:43:04.268514-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=XPhYq5Hty8", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "7", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13349, "uid": "8d34201a5b85900908db6cae92723617", "name": "A pure hypothesis test for inhomogeneous random graph models based on a kernelised Stein discrepancy", "authors": [{"id": 21913, "fullname": "Anum Fatima", "url": "http://virtual.aistats.org/api/miniconf/users/21913?format=json", "institution": "Oxofrd, University of Oxford"}, {"id": 415, "fullname": "Gesine Reinert", "url": "http://virtual.aistats.org/api/miniconf/users/415?format=json", "institution": "University of Oxford"}], "abstract": "Complex data are often represented as a graph, which in turn can often be viewed as a realisation of a random graph, such as an inhomogeneous random graph model (IRG). For general fast goodness-of-fit tests in high dimensions, kernelised Stein discrepancy (KSD) tests are a powerful tool. Here, we develop a KSD-type test for IRG models that can be carried out with a single observation of the network. The test applies to networks of any size, but is particularly relevant for small networks for which asymptotic tests are not warranted. We also provide theoretical guarantees.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13349", "url": null, "sourceid": 548, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=xTDNeJVfXB", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11042, "modified": "2026-03-29T20:42:55.312458-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=xTDNeJVfXB", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "7", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13410, "uid": "1e056d2b0ebd5c878c550da6ac5d3724", "name": "Active Subspaces in Infinite Dimension", "authors": [{"id": 22029, "fullname": "Poorbita Kundu", "url": "http://virtual.aistats.org/api/miniconf/users/22029?format=json", "institution": "Fred Hutchinson Cancer Research Center"}, {"id": 10995, "fullname": "Nathan Wycoff", "url": "http://virtual.aistats.org/api/miniconf/users/10995?format=json", "institution": "Georgetown University"}], "abstract": "Active subspace analysis uses the leading eigenspace of the gradient's second moment to conduct supervised dimension reduction. In this article, we extend this methodology to real-valued functionals on Hilbert space. We define an operator which coincides with the active subspace matrix when applied to a Euclidean space. We show that many of the desirable properties of Active Subspace analysis extend directly to the infinite dimensional setting. We also propose a Monte Carlo procedure and discuss its convergence properties. Finally, we deploy this methodology to create visualizations as well as improve modeling and optimization on complex test problems.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13410", "url": null, "sourceid": 918, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=pA5WWLL742", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11103, "modified": "2026-03-29T20:42:57.779449-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=pA5WWLL742", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "8", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13774, "uid": "c902b497eb972281fb5b4e206db38ee6", "name": "Accelerated Learning on Large Scale Screens using Generative Library Models", "authors": [{"id": 22830, "fullname": "Eli Weinstein", "url": "http://virtual.aistats.org/api/miniconf/users/22830?format=json", "institution": "Technical University of Denmark"}, {"id": 22831, "fullname": "Andrei Slabodkin", "url": "http://virtual.aistats.org/api/miniconf/users/22831?format=json", "institution": "JURA Bio, inc"}, {"id": 22832, "fullname": "Mattia Gollub", "url": "http://virtual.aistats.org/api/miniconf/users/22832?format=json", "institution": "JURA Bio, Inc."}, {"id": 22833, "fullname": "Elizabeth Wood", "url": "http://virtual.aistats.org/api/miniconf/users/22833?format=json", "institution": "JURA Bio, Inc."}], "abstract": "Biological machine learning is often bottlenecked by a lack of scaled data. One promising route to relieving data bottlenecks is through high throughput screens, which can experimentally test the activity of $10^6-10^{12}$ protein sequences in parallel. In this article, we introduce algorithms to optimize high throughput screens for data creation and model training. We focus on the large scale regime, where dataset sizes are limited by the cost of measurement and sequencing. We show that when active sequences are rare, we maximize information gain if we \\textit{only} collect positive examples of active sequences, i.e. $x$ with $y>0$. We can correct for the missing negative examples using a generative model of the library, producing a consistent and efficient estimate of the true $p(y\\mid x)$. We demonstrate this approach in simulation and on a large scale screen of antibodies. Overall, co-design of experiments and inference lets us accelerate learning.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13774", "url": null, "sourceid": 2133, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=CVQioqHorp", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11467, "modified": "2026-03-29T20:43:12.511090-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=CVQioqHorp", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "9", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13878, "uid": "3de2334a314a7a72721f1f74a6cb4cee", "name": "Adaptive Coverage Policies in Conformal Prediction", "authors": [{"id": 23038, "fullname": "Etienne Gauthier", "url": "http://virtual.aistats.org/api/miniconf/users/23038?format=json", "institution": "INRIA"}, {"id": 886, "fullname": "Francis Bach", "url": "http://virtual.aistats.org/api/miniconf/users/886?format=json", "institution": "INRIA - Ecole Normale Sup\u00e9rieure"}, {"id": 366, "fullname": "Michael Jordan", "url": "http://virtual.aistats.org/api/miniconf/users/366?format=json", "institution": "UC Berkeley"}], "abstract": "Traditional conformal prediction methods construct prediction sets such that the true label falls within the set with a user-specified coverage level. However, poorly chosen coverage levels can result in uninformative predictions, either producing overly conservative sets when the coverage level is too high, or empty sets when it is too low. Moreover, the fixed coverage level cannot adapt to the specific characteristics of each individual example, limiting the flexibility and efficiency of these methods. In this work, we leverage recent advances in e-values and post-hoc conformal inference, which allow the use of data-dependent coverage levels while maintaining valid statistical guarantees. We propose to optimize an adaptive coverage policy by training a neural network using a leave-one-out procedure on the calibration set, allowing the coverage level and the resulting prediction set size to vary with the difficulty of each individual example. We support our approach with theoretical coverage guarantees and demonstrate its practical benefits through a series of experiments.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13878", "url": null, "sourceid": 1405, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=3ud62gG8ad", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11571, "modified": "2026-03-29T20:43:16.995318-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=3ud62gG8ad", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "10", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13753, "uid": "23d2e1578544b172cca332ff74bddf5f", "name": "ACE-KT: Cascaded Cognitive Modeling for Stage-wise Knowledge Tracing", "authors": [{"id": 22768, "fullname": "Teng Guo", "url": "http://virtual.aistats.org/api/miniconf/users/22768?format=json", "institution": "Jinan University"}, {"id": 22769, "fullname": "Yubin Xia", "url": "http://virtual.aistats.org/api/miniconf/users/22769?format=json", "institution": "Jinan University"}, {"id": 22770, "fullname": "Jinsen Ke", "url": "http://virtual.aistats.org/api/miniconf/users/22770?format=json", "institution": "Jinan University"}, {"id": 22771, "fullname": "Mingliang Hou", "url": "http://virtual.aistats.org/api/miniconf/users/22771?format=json", "institution": "Jinan University"}, {"id": 22772, "fullname": "Jiaqi Zheng", "url": "http://virtual.aistats.org/api/miniconf/users/22772?format=json", "institution": "Jinan University"}, {"id": 22773, "fullname": "Zitao Liu", "url": "http://virtual.aistats.org/api/miniconf/users/22773?format=json", "institution": "Jinan University"}], "abstract": "Knowledge Tracing (KT) aims to predict students\u2019 academic performance by modeling their knowledge mastery over time, based on their historical learning interactions.  However, current KT models often oversimplify student interactions by treating them as standard time series rather than as cognitive processes. Consequently, modeling student learning as a process of cognitive transformation rather than as a mere sequence of time-stamped events remains a fundamental challenge in KT research. To address this issue, we propose **ACE-KT** (c**A**scaded **C**ognitive mod**E**ling for **K**nowledge **T**racing), a novel framework inspired by cognitive process theory, which shifts the focus from purely sequential modeling to cognitive representation learning. Specifically, we design a cascaded cognitive framework inspired by human cognitive processes in three sequential stages: convolution-based rhythm perception module, Transformer encoder-based contextual structuring module, and cognitive integration module implemented via a selective structured state space model. Extensive experiments on five real-world datasets demonstrate that **ACE-KT** consistently outperforms 21 SOTA KT baselines, demonstrating its effectiveness.  The source code is publicly available at our GitHub repository (https://github.com/AWord992/ACEKT.git).", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13753", "url": null, "sourceid": 2037, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=EO68tiePSY", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11446, "modified": "2026-03-29T20:43:11.679231-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=EO68tiePSY", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "10", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13684, "uid": "2b8a61594b1f4c4db0902a8a395ced93", "name": "Accelerating Byzantine-Robust Distributed Learning with Compressed Communication via Double Momentum and Variance Reduction", "authors": [{"id": 22622, "fullname": "Yanghao Li", "url": "http://virtual.aistats.org/api/miniconf/users/22622?format=json", "institution": "Sichuan University"}, {"id": 22623, "fullname": "Changxin Liu", "url": "http://virtual.aistats.org/api/miniconf/users/22623?format=json", "institution": "East China University of Science and Technology"}, {"id": 22624, "fullname": "Yuhao Yi", "url": "http://virtual.aistats.org/api/miniconf/users/22624?format=json", "institution": "Sichuan University"}], "abstract": "In collaborative and distributed learning, Byzantine robustness reflects a major facet of optimization algorithms. Such distributed algorithms are often accompanied by transmitting a large number of parameters, so communication compression is essential for an effective solution. In this paper, we propose Byz-DM21, a novel Byzantine-robust and communication-efficient stochastic distributed learning algorithm. Our key innovation is a novel gradient estimator based on a double-momentum mechanism, integrating recent advancements in error feedback techniques. Using this estimator, we design both standard and accelerated algorithms that eliminate the need for large batch sizes while maintaining robustness against Byzantine workers. We prove that the Byz-DM21 algorithm has a smaller neighborhood size and converges to $\\varepsilon$-stationary points in $\\mathcal{O}(\\varepsilon^{-4})$ iterations. To further enhance efficiency, we introduce a distributed variant called Byz-VR-DM21, which incorporates local variance reduction at each node to progressively eliminate variance from random approximations. We show that Byz-VR-DM21 provably converges to $\\varepsilon$-stationary points in $\\mathcal{O}(\\varepsilon^{-3 })$ iterations. Additionally, we extend our results to the case where the functions satisfy the Polyak-\u0141ojasiewicz condition. Finally, numerical experiments demonstrate the effectiveness of the proposed method.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13684", "url": null, "sourceid": 515, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=JzTseloW84", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11377, "modified": "2026-03-29T20:43:08.671991-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=JzTseloW84", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "11", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13563, "uid": "4abe17a1c80cbdd2aa241b70840879de", "name": "Active Measurement of Two-Point Correlations", "authors": [{"id": 22380, "fullname": "Max Hamilton", "url": "http://virtual.aistats.org/api/miniconf/users/22380?format=json", "institution": "University of Massachusetts at Amherst"}, {"id": 1210, "fullname": "Daniel Sheldon", "url": "http://virtual.aistats.org/api/miniconf/users/1210?format=json", "institution": "University of Massachusetts, Amherst"}, {"id": 22381, "fullname": "Subhransu Maji", "url": "http://virtual.aistats.org/api/miniconf/users/22381?format=json", "institution": "University of Massachusetts at Amherst"}], "abstract": "Two-point correlation functions (2PCF) are often used to characterize how points cluster together. In this work, we are interested in measuring the 2PCF among a large number of points, but restricted to a subset that satisfies some property of interest. An example comes from astronomy, where scientists measure the 2PCF of star clusters, which make up only a tiny subset of possible sources within a galaxy. This task typically requires careful labeling of sources to construct catalogs, which is time-consuming. We present a human-in-the-loop framework for efficient estimation of 2PCF of target sources. By leveraging a pre-trained classifier to guide sampling, our approach adaptively selects the most informative points for human labeling. After each annotation, it produces unbiased estimates of pair counts across multiple distance bins simultaneously. Compared to simple Monte Carlo approaches, our method achieves substantially lower variance while significantly reducing annotation effort. We introduce a novel unbiased estimator, sampling strategy, and confidence-interval construction that together enable scalable and statistically grounded measurement of two-point correlations in astronomy datasets.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13563", "url": null, "sourceid": 1773, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=YautjjI11j", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11256, "modified": "2026-03-29T20:43:03.683193-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=YautjjI11j", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "11", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13492, "uid": "4e4e53aa080247bc31d0eb4e7aeb07a0", "name": "Adversarial Robustness in One-Stage Learning-to-Defer", "authors": [{"id": 22148, "fullname": "Yannis Montreuil", "url": "http://virtual.aistats.org/api/miniconf/users/22148?format=json", "institution": "National University of Singapore"}, {"id": 22230, "fullname": "Letian Yu", "url": "http://virtual.aistats.org/api/miniconf/users/22230?format=json", "institution": "national university of singaore, National University of Singapore"}, {"id": 22151, "fullname": "Axel Carlier", "url": "http://virtual.aistats.org/api/miniconf/users/22151?format=json", "institution": "Institut Sup\u00e9rieur de l&#x27;A\u00e9ronautique et de l&#x27;Espace"}, {"id": 22150, "fullname": "Lai Xing Ng", "url": "http://virtual.aistats.org/api/miniconf/users/22150?format=json", "institution": "Institute for Infocomm Research (I2R), A*STAR"}, {"id": 22152, "fullname": "Wei Ooi", "url": "http://virtual.aistats.org/api/miniconf/users/22152?format=json", "institution": "National University of Singapore"}], "abstract": "Learning-to-Defer (L2D) enables hybrid decision-making by routing inputs either to a predictor or to external experts. While promising, L2D is highly vulnerable to adversarial perturbations, which can not only flip predictions but also manipulate deferral decisions. Prior robustness analyses focus solely on two-stage settings, leaving open the end-to-end (one-stage) case where predictor and allocation are trained jointly. We introduce the first framework for adversarial robustness in one-stage L2D, covering both classification and regression. Our approach formalizes attacks, proposes cost-sensitive adversarial surrogate losses, and establishes theoretical guarantees including $\\mathcal{H}$, $(\\mathcal{R }, \\mathcal{F})$, and Bayes consistency. Experiments on benchmark datasets confirm that our methods improve robustness against untargeted and targeted attacks while preserving clean performance.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13492", "url": null, "sourceid": 1207, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=gF8tv1SaR2", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11185, "modified": "2026-03-29T20:43:00.978918-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=gF8tv1SaR2", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "11", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13531, "uid": "a4f23670e1833f3fdb077ca70bbd5d66", "name": "Active learning for stochastic contextual linear bandits", "authors": [{"id": 1327, "fullname": "Emma Brunskill", "url": "http://virtual.aistats.org/api/miniconf/users/1327?format=json", "institution": "Stanford University"}, {"id": 22300, "fullname": "Ishani Karmarkar", "url": "http://virtual.aistats.org/api/miniconf/users/22300?format=json", "institution": "Stanford University"}, {"id": 22301, "fullname": "Zhaoqi Li", "url": "http://virtual.aistats.org/api/miniconf/users/22301?format=json", "institution": "Stanford University"}], "abstract": "A key goal in stochastic contextual linear bandits is to efficiently learn a near-optimal policy. Prior algorithms for this problem learn a policy by strategically sampling actions and naively (passively) sampling contexts from the underlying context distribution. However, in many practical scenarios---including online content recommendation, survey research, and clinical trials---practitioners can actively sample or recruit contexts based on prior knowledge of the context distribution. Despite this potential for _active learning_, the role of strategic context sampling in stochastic contextual linear bandits is underexplored. We propose an algorithm that learns a near-optimal policy by strategically sampling rewards of context-action pairs. We prove _instance-dependent_ theoretical guarantees demonstrating that our active context sampling strategy can improve over the minimax rate by up to a factor of $\\smash{\\sqrt{d}}$, where $d$ is the linear dimension. We also show empirically that our algorithm reduces the number of samples needed to learn a near-optimal policy, in tasks such as warfarin dose prediction and joke recommendation.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13531", "url": null, "sourceid": 260, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=cAeeSxkc80", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11224, "modified": "2026-03-29T20:43:02.454085-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=cAeeSxkc80", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "12", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13342, "uid": "7c4ede33a62160a19586f6e26eaefacf", "name": "Active Measuring in Reinforcement Learning With Delayed Negative Effects", "authors": [{"id": 21898, "fullname": "Daiqi Gao", "url": "http://virtual.aistats.org/api/miniconf/users/21898?format=json", "institution": "Harvard University"}, {"id": 12758, "fullname": "Ziping Xu", "url": "http://virtual.aistats.org/api/miniconf/users/12758?format=json", "institution": "Harvard University"}, {"id": 21899, "fullname": "Aseel Rawashdeh", "url": "http://virtual.aistats.org/api/miniconf/users/21899?format=json", "institution": "Harvard University"}, {"id": 18336, "fullname": "Predrag Klasnja", "url": "http://virtual.aistats.org/api/miniconf/users/18336?format=json", "institution": "University of Michigan - Ann Arbor"}, {"id": 12759, "fullname": "Susan Murphy", "url": "http://virtual.aistats.org/api/miniconf/users/12759?format=json", "institution": "Harvard University"}], "abstract": "Measuring states in reinforcement learning (RL) can be costly in real-world settings and may negatively influence future outcomes. We introduce the Actively Observable Markov Decision Process (AOMDP), where an agent not only selects control actions but also decides whether to measure the latent state. The measurement action reveals the true latent state but may have a negative delayed effect on the environment. We show that this reduced uncertainty enables sample-efficient learning and may increase the value of the optimal policy despite these costs. We formulate an AOMDP as a periodic partially observable MDP and propose an online RL algorithm based on belief states. To approximate the belief states, we further propose a sequential Monte Carlo method to jointly approximate the posterior of unknown static environment parameters and unobserved latent states. We evaluate the proposed algorithm in a digital health application, where the agent decides when to deliver digital interventions and when to assess users' psychological status through surveys.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13342", "url": null, "sourceid": 1689, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=yH8cGcqK65", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11035, "modified": "2026-03-29T20:42:54.989877-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=yH8cGcqK65", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "12", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13366, "uid": "c850371fda6892fbfd1c5a5b457e5777", "name": "An Evaluation of Cost Functions for Algorithmic Recourse", "authors": [{"id": 19884, "fullname": "Eoin Kenny", "url": "http://virtual.aistats.org/api/miniconf/users/19884?format=json", "institution": "JPMC AI Research"}, {"id": 21947, "fullname": "Allan Anzagira", "url": "http://virtual.aistats.org/api/miniconf/users/21947?format=json", "institution": "North Carolina A&amp;T State University"}, {"id": 21948, "fullname": "Tom Bewley", "url": "http://virtual.aistats.org/api/miniconf/users/21948?format=json", "institution": "Mistral AI"}, {"id": 21949, "fullname": "Freddy Lecue", "url": "http://virtual.aistats.org/api/miniconf/users/21949?format=json", "institution": "INRIA"}, {"id": 21950, "fullname": "Manuela Veloso", "url": "http://virtual.aistats.org/api/miniconf/users/21950?format=json", "institution": "School of Computer Science, Carnegie Mellon University"}], "abstract": "Algorithmic recourse is a field concerned with offering actionable recommendations to individuals who have received adverse outcomes from automated systems. Most recourse algorithms assume access to a cost function, which quantifies the effort involved in following these suggestions. However, to date, there has been no serious benchmarking of these functions both from a computational and human perspective. In this paper, we propose four metrics to evaluate whether currently popular cost functions in recourse satisfy the minimal requirements for meaningful distance calculations. In addition, we also propose extensions to current approaches using large-language models (LLMs) as surrogate human labellers, which are prompted with a cost-based desiderata. Experiments revealed that methods focused on the Bradley-Terry model perform best, but only when scaled up with our proposed LLM extensions, which would be the recommended choice in practice. We expect our insights to help practitioners in training and designing appropriate cost functions in the future.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13366", "url": null, "sourceid": 1270, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=vAawi6vLTG", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11059, "modified": "2026-03-29T20:42:55.961466-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=vAawi6vLTG", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "13", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13765, "uid": "5c04925674920eb58467fb52ce4ef728", "name": "ADOPT: Additive Optimal Transport Regression", "authors": [{"id": 19640, "fullname": "Wookyeong Song", "url": "http://virtual.aistats.org/api/miniconf/users/19640?format=json", "institution": "University of California, Davis"}, {"id": 22808, "fullname": "Hans-Georg M\u00fcller", "url": "http://virtual.aistats.org/api/miniconf/users/22808?format=json", "institution": "University of California, Davis"}], "abstract": "Regression models for responses $Y$ taking values in general metric spaces $(\\mathcal{M}, d)$, with Euclidean predictors $X \\in \\mathbb{R}^p,$ has attracted growing interest in recent years. While additive regression is a powerful tool for enhancing interpretability and mitigating the curse of dimensionality in the presence of multivariate predictors, its direct extension is hindered by the absence of vector space operations in general metric spaces. We propose a novel framework for additive optimal transport regression, which incorporates additive structure through optimal geodesic transports. A key idea is to extend the notion of optimal transports in Wasserstein spaces to general geodesic metric spaces. This unified approach accommodates a wide range of responses, including probability distributions, symmetric positive definite (SPD) matrices with various metrics and spherical data. The practical utility of the method is illustrated with correlation matrices derived from resting state fMRI brain imaging data.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13765", "url": null, "sourceid": 669, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=DBVYeyyUJi", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11458, "modified": "2026-03-29T20:43:12.139702-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=DBVYeyyUJi", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "14", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13829, "uid": "872488f88d1b2db54d55bc8bba2fad1b", "name": "Asymptotic optimality theory of confidence intervals of the mean", "authors": [{"id": 19917, "fullname": "Vikas Deep", "url": "http://virtual.aistats.org/api/miniconf/users/19917?format=json", "institution": "National University of Singapore"}, {"id": 22213, "fullname": "Achal Bassamboo", "url": "http://virtual.aistats.org/api/miniconf/users/22213?format=json", "institution": "Northwestern University"}, {"id": 22214, "fullname": "Sandeep Juneja", "url": "http://virtual.aistats.org/api/miniconf/users/22214?format=json", "institution": "Ashoka University"}], "abstract": "We address the classical problem of constructing confidence intervals (CIs) for the mean of a distribution, given $N$ i.i.d. samples, such that the CI contains the true mean with probability at least $1 - \\delta$, where $\\delta \\in (0,1)$. We characterize three distinct learning regimes based on the minimum achievable limiting width of any CI as the sample size $N_\\delta \\to \\infty$ and $\\delta \\to 0$. In the first regime, where $N_\\delta$ grows slower than $\\log(1/\\delta)$, the limiting width of any CI equals the width of the distribution\u2019s support, precluding meaningful inference. In the second regime, where $N_\\delta$ scales as $\\log(1/\\delta)$, we precisely characterize the minimum limiting width, which depends on the scaling constant. In the third regime, where $N_\\delta$ grows faster than $\\log(1/\\delta)$, complete learning is achievable, and the limiting width of the CI collapses to zero and CI converges to the true mean. We demonstrate that CIs derived from concentration inequalities based on Kullback-Leibler (KL) divergences achieve asymptotically optimal performance, attaining the minimum limiting width in both the sufficient and the complete learning regimes for distributions in three families: single-parameter exponential, bounded support and known bound on $(1+\\epsilon)^{\\rm th}$ moment. Additionally, these results extend to one-sided CIs, with the width notion adjusted appropriately. Finally, we generalize our findings to settings with random per-sample costs, motivated by practical applications such as stochastic simulators and cloud service selection. Instead of a fixed sample size, we consider a cost budget $C_\\delta$, identifying analogous learning regimes and characterizing the optimal CI construction policy.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13829", "url": null, "sourceid": 754, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=84nYl7LTjG", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11522, "modified": "2026-03-29T20:43:14.780253-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=84nYl7LTjG", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "15", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13440, "uid": "39461a19e9eddfb385ea76b26521ea48", "name": "Almost Sure Convergence of Differential Temporal Difference Learning for Average Reward Markov Decision Processes", "authors": [{"id": 12722, "fullname": "Ethan Blaser", "url": "http://virtual.aistats.org/api/miniconf/users/12722?format=json", "institution": "University of Virginia"}, {"id": 22111, "fullname": "Jiuqi Wang", "url": "http://virtual.aistats.org/api/miniconf/users/22111?format=json", "institution": "University of Virginia, Charlottesville"}, {"id": 22112, "fullname": "Shangtong Zhang", "url": "http://virtual.aistats.org/api/miniconf/users/22112?format=json", "institution": "University of Virginia"}], "abstract": "The average reward is a fundamental performance metric in reinforcement learning (RL) focusing on the long-run performance of an agent. Differential temporal difference (TD) learning algorithms are a major advance for average reward RL as they provide an efficient online method to learn the value functions associated with the average reward in both on-policy and off-policy settings. However, existing convergence guarantees require a local clock in learning rates tied to state visit counts, which practitioners do not use and does not extend beyond tabular settings. We address this limitation by proving the almost sure convergence of on-policy $n$-step differential TD for any $n$ using standard diminishing learning rates without a local clock. We then derive three sufficient conditions under which off-policy $n$-step differential TD also converges without a local clock. These results strengthen the theoretical foundations of differential TD and bring its convergence analysis closer to practical implementations.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13440", "url": null, "sourceid": 386, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=mBiPa311mz", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11133, "modified": "2026-03-29T20:42:59.011218-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=mBiPa311mz", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "15", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13587, "uid": "060ad92489947d410d897474079c1477", "name": "Amortized Safe Active Learning for Real-Time Data Acquisition: Pretrained Neural Policies from Simulated Nonparametric Functions", "authors": [{"id": 22440, "fullname": "Cen-You Li", "url": "http://virtual.aistats.org/api/miniconf/users/22440?format=json", "institution": "Finnish Center for Artificial Intelligence"}, {"id": 22441, "fullname": "Marc Toussaint", "url": "http://virtual.aistats.org/api/miniconf/users/22441?format=json", "institution": "TU Berlin"}, {"id": 431, "fullname": "Barbara Rakitsch", "url": "http://virtual.aistats.org/api/miniconf/users/431?format=json", "institution": "Bosch Center for Artificial Intelligence"}, {"id": 22442, "fullname": "Christoph Zimmer", "url": "http://virtual.aistats.org/api/miniconf/users/22442?format=json", "institution": "Duale Hochschule Baden-Wuerttemberg Mannheim"}], "abstract": "Safe active learning (AL) is a sequential scheme for learning unknown systems while respecting safety constraints during data acquisition. Existing methods often rely on Gaussian processes (GPs) to model the task and safety constraints, requiring repeated GP updates and constrained acquisition optimization--incurring significant computations which are challenging for real-time decision-making. We propose amortized AL for regression and amortized safe AL, replacing expensive online computations with a pretrained neural policy. Inspired by recent advances in amortized Bayesian experimental design, we leverage GPs as pretraining simulators. We train our policy prior to the AL deployment on simulated nonparametric functions, using Fourier feature-based GP sampling and a differentiable acquisition objective that is safety-aware in the safe AL setting. At deployment, our policy selects informative and (if desired) safe queries via a single forward pass, eliminating GP inference and acquisition optimization. This leads to magnitudes of speed improvements while preserving learning quality. Our framework is modular and, without the safety component, yields fast unconstrained AL for time-sensitive tasks.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13587", "url": null, "sourceid": 221, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=WWrdo1tfdw", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11280, "modified": "2026-03-29T20:43:04.598998-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=WWrdo1tfdw", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "16", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13667, "uid": "f50a6c02a3fc5a3a5d4d9391f05f3efc", "name": "Adversarial Debiasing for Parameter Recovery", "authors": [{"id": 19747, "fullname": "Luke Sanford", "url": "http://virtual.aistats.org/api/miniconf/users/19747?format=json", "institution": "Yale University School of the Environment"}, {"id": 22587, "fullname": "Megan Ayers", "url": "http://virtual.aistats.org/api/miniconf/users/22587?format=json", "institution": "Yale University"}, {"id": 22588, "fullname": "Matthew Gordon", "url": "http://virtual.aistats.org/api/miniconf/users/22588?format=json", "institution": "Paris School of Economics"}, {"id": 22589, "fullname": "Eliana Stone", "url": "http://virtual.aistats.org/api/miniconf/users/22589?format=json", "institution": "Yale University"}], "abstract": "Advances in machine learning and the increasing availability of high-dimensional data have led to the proliferation of social science research that uses the predictions of machine learning models as proxies for outcomes of interest. However, prediction errors from machine learning models can lead to bias in downstream estimation tasks, including regression. In this paper, we show how this bias can arise, propose a test for detecting bias, and demonstrate the use of an adversarial machine learning algorithm in order to generate predictions suitable for unbiased downstream estimation. Here, we focus on a setting where machine-learned predictions are the dependent variable in a regression. We conduct simulations and empirical exercises using ground truth and satellite data on forest cover in Africa. Using the predictions from a naive machine learning model leads to biased parameter estimates, while the predictions from the adversarial model recover the true coefficients. Our approach consistently matches or exceeds the performance of existing methods.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13667", "url": null, "sourceid": 1681, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=LtLmVk2CAx", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11360, "modified": "2026-03-29T20:43:07.950883-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=LtLmVk2CAx", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "17", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13385, "uid": "fe7ee8fc1959cc7214fa21c4840dff0a", "name": "Bayesian Inverse Transition Learning: Learning Dynamics from Near-Optimal Trajectories", "authors": [{"id": 17943, "fullname": "Leo Benac", "url": "http://virtual.aistats.org/api/miniconf/users/17943?format=json", "institution": "Harvard University, Harvard University"}, {"id": 21980, "fullname": "Abhishek Sharma", "url": "http://virtual.aistats.org/api/miniconf/users/21980?format=json", "institution": "Harvard University"}, {"id": 17956, "fullname": "Sonali Parbhoo", "url": "http://virtual.aistats.org/api/miniconf/users/17956?format=json", "institution": "Imperial College London, Imperial College London"}, {"id": 4070, "fullname": "Finale Doshi-Velez", "url": "http://virtual.aistats.org/api/miniconf/users/4070?format=json", "institution": "Harvard"}], "abstract": "We consider the problem of estimating the transition dynamics from near-optimal expert trajectories in the context of offline model-based reinforcement learning. We develop a novel constraint-based method, Inverse Transition Learning, that treats the limited coverage of the expert trajectories as a feature: we use the fact that the expert is near-optimal to inform our estimate of. We integrate our constraints into a Bayesian approach. Across both synthetic environments and real healthcare scenarios like Intensive Care Unit (ICU) patient management in hypotension, we demonstrate not only significant improvements in decision-making, but that our posterior can inform when transfer will be successful.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13385", "url": null, "sourceid": 986, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=sPU8IlOfb5", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11078, "modified": "2026-03-29T20:42:56.765559-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=sPU8IlOfb5", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "17", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13655, "uid": "d86ea612dec96096c5e0fcc8dd42ab6d", "name": "An Information-Theoretic Approach to Understanding Transformers' In-Context Learning of Variable-Order Markov Chains", "authors": [{"id": 12997, "fullname": "Ruida Zhou", "url": "http://virtual.aistats.org/api/miniconf/users/12997?format=json", "institution": "Texas A&amp;M University"}, {"id": 4292, "fullname": "Chao Tian", "url": "http://virtual.aistats.org/api/miniconf/users/4292?format=json", "institution": "Texas A&amp;M University"}, {"id": 17701, "fullname": "Suhas Diggavi", "url": "http://virtual.aistats.org/api/miniconf/users/17701?format=json", "institution": "University of California, Los Angeles"}], "abstract": "We study transformers' in-context learning of variable-length Markov chains (VOMCs), focusing on the finite-sample accuracy as the number of in-context examples increases. Compared to fixed-order Markov chains (FOMCs), learning VOMCs is substantially more challenging due to the additional structural learning component. The problem is naturally suited to a Bayesian formulation, where the context-tree weighting (CTW) algorithm, originally developed in the information theory community for universal data compression, provides an optimal solution. Empirically, we find that single-layer transformers fail to learn VOMCs in context, whereas transformers with two or more layers can succeed, with additional layers yielding modest but noticeable improvements. In contrast to prior results on FOMCs, attention-only networks appear insufficient for VOMCs. To explain these findings, we provide explicit transformer constructions: one with $D+2$ layers that can exactly implement CTW for VOMCs of maximum order $D$, and a simplified two-layer construction that uses partial information for approximate blending, shedding light on why two-layer transformers can perform well.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13655", "url": null, "sourceid": 603, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=N6DOObrBqq", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11348, "modified": "2026-03-29T20:43:07.486598-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=N6DOObrBqq", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "18", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13527, "uid": "efdf562ce2fb0ad460fd8e9d33e57f57", "name": "Aggregation on Learnable Manifolds for Asynchronous Federated Optimisation", "authors": [{"id": 19436, "fullname": "Archie Licudi", "url": "http://virtual.aistats.org/api/miniconf/users/19436?format=json", "institution": "Imperial College London"}, {"id": 9421, "fullname": "Anshul Thakur", "url": "http://virtual.aistats.org/api/miniconf/users/9421?format=json", "institution": "University of Oxford"}, {"id": 9419, "fullname": "Soheila Molaei", "url": "http://virtual.aistats.org/api/miniconf/users/9419?format=json", "institution": "University of Oxford"}, {"id": 18694, "fullname": "Danielle Belgrave", "url": "http://virtual.aistats.org/api/miniconf/users/18694?format=json", "institution": "GSK plc"}, {"id": 17824, "fullname": "David Clifton", "url": "http://virtual.aistats.org/api/miniconf/users/17824?format=json", "institution": "University of Oxford"}], "abstract": "Asynchronous federated learning (FL) with heterogeneous clients faces two key issues: curvature-induced loss barriers encountered by standard linear parameter interpolation techniques (e.g. FedAvg) and interference from stale updates misaligned with the server\u2019s current optimisation state. To alleviate these issues, we introduce a geometric framework that casts aggregation as curve learning in a Riemannian model space and decouples choice of update direction from staleness conflict resolution. Within this, we propose $\\textbf{AsyncBezier}$, which replaces linear aggregation with low-degree polynomial (B\u00e9zier) trajectories to bypass loss barriers, and $\\textbf{OrthoDC}$, which orthogonally projects delayed updates to reduce interference. We establish framework-level convergence guarantees covering each variant given simple assumptions on their components. On three datasets spanning general-purpose and healthcare domains, including LEAF Shakespeare and FEMNIST, our approach consistently improves accuracy and client fairness over strong asynchronous baselines; finally, we show that these gains are preserved even when other methods are allocated a higher local compute budget.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13527", "url": null, "sourceid": 2227, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=cXxlrTg6pe", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11220, "modified": "2026-03-29T20:43:02.313775-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=cXxlrTg6pe", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "19", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13560, "uid": "d6baf65e0b240ce177cf70da146c8dc8", "name": "Bandit-based Maximum Inner Product Search with Data-Dependent Confidence Intervals", "authors": [{"id": 22369, "fullname": "Yoichi Sasaki", "url": "http://virtual.aistats.org/api/miniconf/users/22369?format=json", "institution": "NEC"}, {"id": 22370, "fullname": "Yuzuru Okajima", "url": "http://virtual.aistats.org/api/miniconf/users/22370?format=json", "institution": "NEC Corporation"}], "abstract": "Maximum inner product search is a fundamental problem in recommender systems, information retrieval, and machine learning. Recently proposed bandit-based approaches have achieved high scalability with respect to dimensionality and offer favorable precision-speedup trade-offs. However, the lengths of their confidence intervals are determined independently of the actual reward distributions, which can lead to search inefficiency in practice. In this paper, we propose a data-dependent bandit-based algorithm in which the lengths of the confidence intervals are adaptively adjusted based on observed samples. Theoretical analysis demonstrates that our algorithm guarantees $\\delta$-correctness for a broad class of distributions, including all log-concave continuous  distributions, and that the sample complexity can be reduced adaptively according to individual reward distributions. In experiments, our approach outperformed existing algorithms on both synthetic and real-world datasets.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13560", "url": null, "sourceid": 264, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=YpRQk8CLq2", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11253, "modified": "2026-03-29T20:43:03.574771-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=YpRQk8CLq2", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "20", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13879, "uid": "e3251075554389fe91d17a794861d47b", "name": "Bounds and Identification of Joint Probabilities of Potential Outcomes and Observed Variables under Monotonicity Assumptions", "authors": [{"id": 23039, "fullname": "Naoya Hashimoto", "url": "http://virtual.aistats.org/api/miniconf/users/23039?format=json", "institution": "Mohamed bin Zayed University of Artificial Intelligence"}, {"id": 23040, "fullname": "Yuta Kawakami", "url": "http://virtual.aistats.org/api/miniconf/users/23040?format=json", "institution": "Mohamed bin Zayed University of Artificial Intelligence"}, {"id": 18642, "fullname": "Jin Tian", "url": "http://virtual.aistats.org/api/miniconf/users/18642?format=json", "institution": "Mohamed bin Zayed University of Artificial Intelligence"}], "abstract": "Evaluating  joint probabilities of potential outcomes and observed variables, and their linear combinations, is a fundamental challenge in causal inference. This paper addresses the bounding and identification of these probabilities in settings with discrete treatment and discrete outcome. We propose new families of monotonicity assumptions  and formulate the bounding problem as a linear programming problem. We further introduce a new monotonicity assumption specifically to achieve identification.  Finally, we present numerical experiments to validate our methods and demonstrate their application using real-world datasets.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13879", "url": null, "sourceid": 1256, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=3mZllPs2qf", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11572, "modified": "2026-03-29T20:43:17.035301-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=3mZllPs2qf", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "22", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13898, "uid": "2f37d10131f2a483a8dd005b3d14b0d9", "name": "Beyond Black-Box Predictions: Identifying Marginal Feature Effects in Tabular Transformer Networks", "authors": [{"id": 23077, "fullname": "Anton Thielmann", "url": "http://virtual.aistats.org/api/miniconf/users/23077?format=json", "institution": "Amazon Music"}, {"id": 23078, "fullname": "Arik Reuter", "url": "http://virtual.aistats.org/api/miniconf/users/23078?format=json", "institution": "University of Cambridge"}, {"id": 12406, "fullname": "Benjamin S\u00e4fken", "url": "http://virtual.aistats.org/api/miniconf/users/12406?format=json", "institution": "TU Clausthal"}], "abstract": "In recent years, deep neural networks have showcased their predictive power across a variety of tasks. Beyond natural language processing, the transformer architecture has proven efficient in addressing tabular data problems and challenges the previously dominant gradient-based decision trees in these areas. However, this predictive power comes at the cost of intelligibility: Marginal feature effects are almost completely lost in the black-box nature of deep tabular transformer networks. Alternative architectures that use the additivity constraints of classical statistical regression models can maintain intelligible marginal feature effects, but often fall short in predictive power compared to their more complex counterparts.  To bridge the gap between intelligibility and performance, we propose an adaptation of tabular transformer networks designed to identify marginal feature effects. We provide theoretical justifications that marginal feature effects can be accurately identified, and our ablation study demonstrates that the proposed model efficiently detects these effects, even amidst complex feature interactions. To demonstrate the model's predictive capabilities, we compare it to several interpretable as well as black-box models and find that it can match black-box performances while maintaining intelligibility. The source code is available at https://github.com/OpenTabular/NAMpy.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13898", "url": null, "sourceid": 658, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=1ef7TN7vJH", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11591, "modified": "2026-03-29T20:43:17.765844-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=1ef7TN7vJH", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "23", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13572, "uid": "fe8c15fed5f808006ce95eddb7366e35", "name": "Busemann Functions in the Wasserstein Space: Existence, Closed-Forms, and Applications to Slicing", "authors": [{"id": 22399, "fullname": "Cl\u00e9ment Bonet", "url": "http://virtual.aistats.org/api/miniconf/users/22399?format=json", "institution": "Ecole Polytechnique"}, {"id": 22400, "fullname": "Elsa Cazelles", "url": "http://virtual.aistats.org/api/miniconf/users/22400?format=json", "institution": "CNRS, IRIT, Universit\u00e9 de Toulouse"}, {"id": 22401, "fullname": "Lucas Drumetz", "url": "http://virtual.aistats.org/api/miniconf/users/22401?format=json", "institution": "IMT Atlantique"}, {"id": 22402, "fullname": "Nicolas Courty", "url": "http://virtual.aistats.org/api/miniconf/users/22402?format=json", "institution": "IRISA"}], "abstract": "The Busemann function has recently found many interests in a variety of geometric machine   learning problems, as it naturally defines projections onto geodesic rays of Riemannian manifolds and generalizes the notion of hyperplanes. As several sources of data can be conveniently modeled as probability distributions, it is natural to study this function in the Wasserstein space, which carries a rich formal Riemannian structure induced by Optimal Transport metrics. In this work, we investigate the existence and computation of Busemann functions in Wasserstein space, which admits geodesic rays. We establish closed-form expressions in two important cases: one-dimensional distributions and Gaussian measures. These results enable explicit projection schemes for probability distributions on $\\mathbb{R}$, which in turn allow us to define novel Sliced-Wasserstein distances over Gaussian mixtures and labeled datasets. We demonstrate the efficiency of those original schemes on synthetic datasets as well as transfer learning problems.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13572", "url": null, "sourceid": 849, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=Xpt0HEC3fO", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11265, "modified": "2026-03-29T20:43:04.127603-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=Xpt0HEC3fO", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "23", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13880, "uid": "bffc98347ee35b3ead06728d6f073c68", "name": "Canopy Tree Height Estimation using Quantile Regression: Modeling and Evaluating Uncertainty in Remote Sensing", "authors": [{"id": 19785, "fullname": "Karsten Schr\u00f6dter", "url": "http://virtual.aistats.org/api/miniconf/users/19785?format=json", "institution": "University of M\u00fcnster"}, {"id": 23041, "fullname": "Jan Pauls", "url": "http://virtual.aistats.org/api/miniconf/users/23041?format=json", "institution": "University of M\u00fcnster"}, {"id": 23042, "fullname": "Fabian Gieseke", "url": "http://virtual.aistats.org/api/miniconf/users/23042?format=json", "institution": "University of M\u00fcnster, Department of Information Systems"}], "abstract": "Accurate tree height estimation is vital for ecological monitoring and biomass assessment. We apply quantile regression to existing tree height estimation models based on satellite data to incorporate uncertainty quantification. Most current approaches on tree height estimation rely on point predictions, which limits their applicability in risk-sensitive scenarios. In this work, we show that with minor modifications to the prediction head, existing models can be adapted to provide statistically calibrated uncertainty estimates via quantile regression. Furthermore, we demonstrate how our results correlate with known challenges in remote sensing (e.g., terrain complexity, vegetation heterogeneity), indicating that the model is less confident in more challenging conditions.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13880", "url": null, "sourceid": 1483, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=3foK47Zc9y", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11573, "modified": "2026-03-29T20:43:17.073778-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=3foK47Zc9y", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "24", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13340, "uid": "a51c896c9cb81ecb5a199d51ac9fc3c5", "name": "Atlas-based Manifold Representations for Interpretable Riemannian Machine Learning", "authors": [{"id": 19793, "fullname": "Ryan Robinett", "url": "http://virtual.aistats.org/api/miniconf/users/19793?format=json", "institution": "University of Chicago"}, {"id": 21893, "fullname": "Sophia Madejski", "url": "http://virtual.aistats.org/api/miniconf/users/21893?format=json", "institution": "University of Chicago"}, {"id": 21894, "fullname": "Kyle Ruark", "url": "http://virtual.aistats.org/api/miniconf/users/21894?format=json", "institution": "Harvard University"}, {"id": 21895, "fullname": "Samantha Riesenfeld", "url": "http://virtual.aistats.org/api/miniconf/users/21895?format=json", "institution": "University of Chicago"}, {"id": 21896, "fullname": "Lorenzo Orecchia", "url": "http://virtual.aistats.org/api/miniconf/users/21896?format=json", "institution": "University of Chicago"}], "abstract": "Despite the popularity of the manifold hypothesis, current manifold-learning methods do not support machine learning directly on the latent $d$-dimensional data manifold, as they primarily aim to perform dimensionality reduction into $\\mathbb{R}^D$, losing key manifold features when the embedding dimension $D$ approaches $d$. On the other hand, methods that directly learn the latent manifold as a differentiable atlas have been relatively underexplored. In this paper, we aim to give a proof of concept of the effectiveness and potential of atlas-based methods. To this end, we implement a generic data structure to maintain a differentiable atlas that enables Riemannian optimization over the manifold. We complement this with an unsupervised heuristic that learns a differentiable atlas from point cloud data. We experimentally demonstrate that this approach has advantages in terms of efficiency and accuracy in selected settings.  Moreover, in a supervised classification task over the Klein bottle and in RNA velocity analysis of hematopoietic data, we showcase the improved interpretability and robustness of our approach.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13340", "url": null, "sourceid": 1712, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=yTGh29HLlp", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11033, "modified": "2026-03-29T20:42:54.909335-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=yTGh29HLlp", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "25", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13914, "uid": "584b98aac2dddf59ee2cf19ca4ccb75e", "name": "Batch-Adaptive Causal Annotations", "authors": [{"id": 23098, "fullname": "Ezinne Nwankwo", "url": "http://virtual.aistats.org/api/miniconf/users/23098?format=json", "institution": "University of California, Berkeley"}, {"id": 23099, "fullname": "Lauri Goldkind", "url": "http://virtual.aistats.org/api/miniconf/users/23099?format=json", "institution": "Fordham University"}, {"id": 22404, "fullname": "Angela Zhou", "url": "http://virtual.aistats.org/api/miniconf/users/22404?format=json", "institution": "University of Southern California"}], "abstract": "Estimating the causal effects of interventions is crucial to policy and decision-making, yet outcome data are often missing or subject to non-standard measurement error. While ground-truth outcomes can sometimes be obtained through costly data annotation or follow-up, budget constraints typically allow only a fraction of the dataset to be labeled. We address this challenge by optimizing \\textit{which data points should be sampled for outcome information} in order to improve efficiency in average treatment effect estimation with missing outcomes. We derive a closed-form solution for the optimal sampling probability in batches. We optimize the asymptotic variance of a doubly-robust estimator for causal inference with missing outcomes, and show the resulting asymptotic convergence to the optimal variance. Motivated by a collaboration with a street outreach provider generating millions of case notes, we also extend this framework to costly annotations of unstructured data, such as text or images, common in healthcare and social services. Across simulated and real-world datasets, including one on outreach interventions in homelessness services, our approach achieves substantially lower mean-squared error and recovers the AIPW estimate with fewer labels than existing baselines. In practice, we show that our method can match confidence intervals obtained with 361 random samples using only 90 optimized samples\u2014saving 75% of the labeling budget.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13914", "url": null, "sourceid": 2052, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=01w2ebNFwW", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11607, "modified": "2026-03-29T20:43:18.449685-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=01w2ebNFwW", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "29", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13351, "uid": "5705e1164a8394aace6018e27d20d237", "name": "Conditional Flow Matching for Bayesian Posterior Sampling", "authors": [{"id": 18199, "fullname": "Percy Zhai", "url": "http://virtual.aistats.org/api/miniconf/users/18199?format=json", "institution": "University of Chicago"}, {"id": 21916, "fullname": "Sowon Jeong", "url": "http://virtual.aistats.org/api/miniconf/users/21916?format=json", "institution": "The University of Chicago"}, {"id": 14878, "fullname": "Veronika Rockova", "url": "http://virtual.aistats.org/api/miniconf/users/14878?format=json", "institution": "University of Chicago"}], "abstract": "We propose a generative multivariate posterior sampling method via flow matching. The method learns a dynamic, block-triangular velocity field in the joint space of data and parameters, which results in a deterministic transport map from a source distribution to the desired posterior. We introduce a time-dependent extension of block-triangular maps for posterior sampling, offering built-in invertibility and scalability. It offers a simple training objective, and does not require the access to likelihood evaluation. The map is automatically invertible without explicit assumptions, a desirable property of the map. The inverse map, named vector rank, is accessible by reversibly integrating the velocity over time. It is advantageous to leverage the dynamic design: proper constraints on the velocity yield a monotone map, which leads to a conditional Brenier map, enabling a fast and simultaneous generation of Bayesian credible sets that agree with Monge-Kantorovich data depth. Our approach is computationally lighter compared to GAN-based and diffusion-based counterparts, and is capable of capturing complex posterior structures. Finally, frequentist theoretical guarantee on the consistency of the recovered posterior distribution, and of the corresponding Bayesian credible sets, is provided.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13351", "url": null, "sourceid": 897, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=x6J48heM3C", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11044, "modified": "2026-03-29T20:42:55.392135-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=x6J48heM3C", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "29", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13428, "uid": "58e4d44e550d0f7ee0a23d6b02d9b0db", "name": "ConMeZO: Adaptive Descent-Direction Sampling for Gradient-Free Finetuning of Large Language Models", "authors": [{"id": 20594, "fullname": "Lejs Deen Behric", "url": "http://virtual.aistats.org/api/miniconf/users/20594?format=json", "institution": "ETH Z\u00fcrich"}, {"id": 5159, "fullname": "Liang Zhang", "url": "http://virtual.aistats.org/api/miniconf/users/5159?format=json", "institution": "ETH Zurich"}, {"id": 22073, "fullname": "Bingcong Li", "url": "http://virtual.aistats.org/api/miniconf/users/22073?format=json", "institution": "ETHZ - ETH Zurich"}, {"id": 22074, "fullname": "Kiran Thekumparampil", "url": "http://virtual.aistats.org/api/miniconf/users/22074?format=json", "institution": "Amazon"}], "abstract": "Zeroth\u2011order or derivative-free optimization (MeZO) is an attractive strategy for finetuning large language models (LLMs) because it eliminates the memory overhead of backpropagation. However, it converges slowly due to the inherent curse of dimensionality when searching for descent directions in the high-dimensional parameter space of billion-scale LLMs. We propose ConMeZO, a novel zeroth\u2011order optimizer that accelerates convergence by adaptive directional sampling. Instead of drawing the direction uniformly at random, ConMeZO restricts the sampling to a cone centered around a momentum estimate. This concentrates the search in directions where the true gradient is more likely to lie and thus reduces the effect of high dimensions. We prove that ConMeZO achieves the same worst-case convergence rate as MeZO. Empirically, when finetuning LLMs on natural language tasks, ConMeZO is up to 2$\\times$ faster than MeZO while retaining the low\u2011memory footprint of zeroth-order methods.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13428", "url": null, "sourceid": 948, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=nLXCavoxjE", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11121, "modified": "2026-03-29T20:42:58.412259-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=nLXCavoxjE", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "30", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13384, "uid": "1a5b1e4daae265b790965a275b53ae50", "name": "Causal-DRF: Conditional Kernel Treatment Effect Estimation using Distributional Random Forest", "authors": [{"id": 19895, "fullname": "Jeffrey N\u00c4F", "url": "http://virtual.aistats.org/api/miniconf/users/19895?format=json", "institution": "University of Geneva"}, {"id": 21978, "fullname": "Junhyung Park", "url": "http://virtual.aistats.org/api/miniconf/users/21978?format=json", "institution": "Department of Computer Science, ETHZ - ETH Zurich"}, {"id": 21979, "fullname": "Herbert Susmann", "url": "http://virtual.aistats.org/api/miniconf/users/21979?format=json", "institution": "NYU Langone Health"}], "abstract": "The conditional average treatment effect (CATE) is a commonly targeted statistical parameter for measuring the effect of a treatment conditional on covariates. However, the CATE will fail to capture effects of treatments beyond differences in conditional expectations. Inspired by causal forests for CATE estimation, we develop a forest-based method to estimate the conditional kernel treatment effect (CKTE), based on the recently introduced Distributional Random Forest (DRF) algorithm. Adapting the splitting criterion of DRF, we show how one forest fit can be used to obtain a consistent and asymptotically normal estimator of the CKTE, as well as an approximation of its sampling distribution. This allows to study the difference in distribution between control and treatment group and thus yields a more comprehensive understanding of the treatment effect.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13384", "url": null, "sourceid": 446, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=sSBgDqNSsf", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11077, "modified": "2026-03-29T20:42:56.726206-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=sSBgDqNSsf", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "30", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13725, "uid": "f4a331b7a22d1b237565d8813a34d8ac", "name": "Connectome-Guided Optimization for Deep Networks", "authors": [{"id": 19893, "fullname": "Peilin He", "url": "http://virtual.aistats.org/api/miniconf/users/19893?format=json", "institution": "Duke Kunshan University"}, {"id": 22726, "fullname": "Tananun Songdechakraiwut", "url": "http://virtual.aistats.org/api/miniconf/users/22726?format=json", "institution": "Duke University"}], "abstract": "The human brain is highly adaptive: its functional connectivity reconfigures on multiple timescales during cognition and learning, enabling flexible information processing. By contrast, artificial neural networks typically rely on manually-tuned learning-rate schedules or generic adaptive optimizers whose hyperparameters remain largely agnostic to a model's internal dynamics. In this paper, we propose Connectome-Guided Automatic Learning Rate (CG-ALR) that dynamically constructs a functional connectome of the neural network from neuron co-activations at each training iteration and adjusts learning rates online as this connectome reconfigures. This connectomics-inspired mechanism adapts step sizes to the network's dynamic functional organization, slowing learning during unstable reconfiguration and accelerating it when stable organization emerges. Our results demonstrate that principles inspired by brain connectomes can inform the design of adaptive learning rates in deep learning, with particularly consistent improvements over traditional SGD-based schedules and competitive performance against Adam-family scheduled baselines and recent adaptive methods.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13725", "url": null, "sourceid": 1440, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=Fwt9bAhXln", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11418, "modified": "2026-03-29T20:43:10.357312-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=Fwt9bAhXln", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "31", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13634, "uid": "df263d996281d984952c07998dc54358", "name": "Closed-Form Coordinate Ascent Variational Inference for Student-t Process Regression with Student-t Likelihood", "authors": [{"id": 20596, "fullname": "Keisuke Onoue", "url": "http://virtual.aistats.org/api/miniconf/users/20596?format=json", "institution": "Nara Institute of Science and Technology"}, {"id": 20593, "fullname": "Takatomi Kubo", "url": "http://virtual.aistats.org/api/miniconf/users/20593?format=json", "institution": "Nara Institute of Science and Technology"}, {"id": 22526, "fullname": "Kazushi Ikeda", "url": "http://virtual.aistats.org/api/miniconf/users/22526?format=json", "institution": "Nara Institute of Science and Technology (NAIST)"}], "abstract": "Combining a Student-t Process prior with a Student-t likelihood yields a doubly robust regression model whose intractable posterior has prevented its practical use.  We introduce the first tractable variational inference framework for this model.  Leveraging the Student-t distribution's scale-mixture representation, we design a structured variational family that affords an analytic evidence lower bound.  To overcome the non-conjugacy of this family, which precludes closed-form updates, we devise a novel projection-based optimization:  we find the optimum in a simpler, factorized family and analytically project it back onto our structured one.  The framework is extended to a scalable sparse, stochastic setting.  Empirical results demonstrate strong performance, particularly in the full-batch setting, establishing this robust model as a practical and powerful tool.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13634", "url": null, "sourceid": 739, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=PN83VEfbQI", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11327, "modified": "2026-03-29T20:43:06.604978-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=PN83VEfbQI", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "31", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13734, "uid": "74563ba21a90da13dacf2a73e3ddefa7", "name": "Beyond Spectral Clustering: Probabilistic Cuts for Differentiable Graph Partitioning", "authors": [{"id": 14531, "fullname": "Ayoub Ghriss", "url": "http://virtual.aistats.org/api/miniconf/users/14531?format=json", "institution": "University of Colorado, Boulder"}], "abstract": "Probabilistic relaxations of graph cuts offer a differentiable alternative to spectral clustering, enabling end-to-end and online learning without eigendecompositions, yet prior work centered on RatioCut and lacked general guarantees and principled gradients. We present a unified probabilistic framework that covers a wide class of cuts, including Normalized Cut. Our framework provides tight analytic upper bounds on expected discrete cuts via integral representations and Gauss hypergeometric functions with closed-form forward and backward. Together, these results deliver a rigorous, numerically stable foundation for scalable, differentiable graph partitioning covering a wide range of clustering and contrastive learning objectives.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13734", "url": null, "sourceid": 1726, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=FN6QAT5Tmc", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11427, "modified": "2026-03-29T20:43:10.696918-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=FN6QAT5Tmc", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "32", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13471, "uid": "44c4c17332cace2124a1a836d9fc4b6f", "name": "Conditional Vendi Score: Prompt-Aware Diversity Evaluation for Text-Guided Generative AI Models", "authors": [{"id": 19764, "fullname": "Mohammad Jalali", "url": "http://virtual.aistats.org/api/miniconf/users/19764?format=json", "institution": "The Chinese University of Hong Kong"}, {"id": 19915, "fullname": "Azim Ospanov", "url": "http://virtual.aistats.org/api/miniconf/users/19915?format=json", "institution": "The Chinese University of Hong Kong"}, {"id": 22175, "fullname": "Amin Gohari", "url": "http://virtual.aistats.org/api/miniconf/users/22175?format=json", "institution": "The Chinese University of Hong Kong"}, {"id": 9929, "fullname": "Farzan Farnia", "url": "http://virtual.aistats.org/api/miniconf/users/9929?format=json", "institution": "The Chinese University of Hong Kong"}], "abstract": "Generative models guided by text prompts are widely evaluated for fidelity and prompt alignment, yet their ability to produce diverse outputs remains underexplored. Existing diversity metrics such as Vendi and RKE, which are based on the von Neumann and R\u00e9nyi entropies of kernel matrices, were developed for unconditional models and cannot distinguish prompt-induced from model-induced variability. We address this gap by introducing *Conditional-Vendi* and *Conditional-RKE*, diversity measures derived from the conditional entropy of positive semidefinite matrices. These scores isolate model-induced diversity in prompt-guided generation, with Conditional-RKE enjoying an $O(1/\\sqrt{n})$ convergence rate. For Conditional-Vendi, we introduce a truncated-spectrum approximation that yields scalable and consistent estimates. Experiments on text-to-image, image-captioning, and language generation tasks demonstrate that the conditional scores recover ground-truth diversity orderings and can also guide diffusion models toward more diverse generations.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13471", "url": null, "sourceid": 606, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=iDrZToIsyd", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11164, "modified": "2026-03-29T20:43:00.198050-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=iDrZToIsyd", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "33", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13423, "uid": "b20bb95ab626d93fd976af958fbc61ba", "name": "Counterfactual Explanations via Latent Structure for Time Series Classification", "authors": [{"id": 22058, "fullname": "Akihiro Yamaguchi", "url": "http://virtual.aistats.org/api/miniconf/users/22058?format=json", "institution": "Kyushu University"}, {"id": 22059, "fullname": "Shizuo Kaji", "url": "http://virtual.aistats.org/api/miniconf/users/22059?format=json", "institution": "Kyoto University"}, {"id": 22060, "fullname": "Kaname Matsue", "url": "http://virtual.aistats.org/api/miniconf/users/22060?format=json", "institution": "Kyushu University"}, {"id": 11066, "fullname": "Ryusei Shingaki", "url": "http://virtual.aistats.org/api/miniconf/users/11066?format=json", "institution": "Yokohama National University"}], "abstract": "There is a growing need for explainability in time series classification. Counterfactual (CF) generation creates in-distribution synthetic instances that flip the prediction to a desired class. We propose CELT, a model-agnostic CF generation method for time-series classifiers, including non-differentiable and one-class models. In the development phase, CELT learns a structured latent space in which desired-class latent instances form clusters and other latent instances are pushed away. In addition, the design enables segment-wise, time-local edits. In the deployment phase, CELT efficiently generates CFs by editing a minimal number of time-local segments, guided by the learned structure. We formulate both phases as mathematically sound optimization problems that uniformly handle supervised and one-class classification, and we demonstrate effectiveness on UCR datasets.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13423", "url": null, "sourceid": 1191, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=oKUjFiuoVi", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11116, "modified": "2026-03-29T20:42:58.270772-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=oKUjFiuoVi", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "33", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13505, "uid": "d16509f6eaca1022bd8f28d6bc582cae", "name": "BOAT: Navigating The Sea of in Silico Predictors for Antibody Design via Multi-Objective Bayesian Optimization", "authors": [{"id": 22260, "fullname": "Jackie Rao", "url": "http://virtual.aistats.org/api/miniconf/users/22260?format=json", "institution": "MRC Biostatistics Unit, University of Cambridge"}, {"id": 22261, "fullname": "Ferran Gonzalez", "url": "http://virtual.aistats.org/api/miniconf/users/22261?format=json", "institution": "AstraZeneca"}, {"id": 22262, "fullname": "Leon Gerard", "url": "http://virtual.aistats.org/api/miniconf/users/22262?format=json", "institution": "AstraZeneca"}, {"id": 22263, "fullname": "Alexandra Gessner", "url": "http://virtual.aistats.org/api/miniconf/users/22263?format=json", "institution": "AstraZeneca"}], "abstract": "Antibody lead optimization is inherently a multi-objective challenge in drug discovery. Achieving a balance between different drug-like properties is crucial for the development of viable candidates, and this search becomes exponentially challenging as desired properties grow. The ever-growing zoo of sophisticated *in silico* tools for predicting antibody properties calls for an efficient joint optimization procedure to overcome resource-intensive sequential filtering pipelines. We present BOAT, a versatile Bayesian optimization framework for multi-property antibody engineering. Our 'plug-and-play' framework couples uncertainty-aware surrogate modeling with a genetic algorithm to jointly optimize various predicted antibody traits while enabling efficient exploration of sequence space. Through systematic benchmarking against genetic algorithms and newer generative learning approaches, we demonstrate competitive performance with state-of-the-art methods for multi-objective protein optimization. We identify clear regimes where surrogate-driven optimization outperforms expensive generative approaches and establish practical limits imposed by sequence dimensionality and oracle costs.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13505", "url": null, "sourceid": 2432, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=eqah46PWof", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11198, "modified": "2026-03-29T20:43:01.491088-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=eqah46PWof", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "34", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13391, "uid": "0ff39bbbf981ac0151d340c9aa40e63e", "name": "Conformal Robust Control of Linear Systems", "authors": [{"id": 12434, "fullname": "Yash Patel", "url": "http://virtual.aistats.org/api/miniconf/users/12434?format=json", "institution": "University of Michigan"}, {"id": 12435, "fullname": "Sahana Rayan", "url": "http://virtual.aistats.org/api/miniconf/users/12435?format=json", "institution": "University of Michigan"}, {"id": 324, "fullname": "Ambuj Tewari", "url": "http://virtual.aistats.org/api/miniconf/users/324?format=json", "institution": "University of Michigan"}], "abstract": "End-to-end engineering design pipelines, in which designs are evaluated using concurrently defined optimal controllers, are becoming increasingly common in practice. To discover designs that perform well even under the misspecification of system dynamics, such end-to-end pipelines have now begun evaluating designs with a robust control objective in place of the nominal optimal control setup. Current approaches of specifying such robust control subproblems, however, rely on hand specification of perturbations anticipated to be present upon deployment or margin methods that ignore problem structure, resulting in a lack of theoretical guarantees and overly conservative empirical performance. We, instead, propose a novel methodology for LQR systems that leverages conformal prediction to specify such uncertainty regions in a data-driven fashion. Such regions have distribution-free coverage guarantees on the true system dynamics, in turn allowing for a probabilistic characterization of the regret of the resulting robust controller. We then demonstrate that such a controller can be efficiently produced via a novel policy gradient method that has convergence guarantees. We finally demonstrate the superior empirical performance of our method over alternate robust control specifications, such as $H_\\infty$ and LQR with multiplicative noise, across a collection of engineering control systems.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13391", "url": null, "sourceid": 646, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=rdsOb7q6mw", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11084, "modified": "2026-03-29T20:42:57.012848-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=rdsOb7q6mw", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "35", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13570, "uid": "8fe0093bb30d6f8c31474bd0764e6ac0", "name": "Consistent PCA and Spectral Clustering", "authors": [{"id": 22396, "fullname": "Satoshi Hara", "url": "http://virtual.aistats.org/api/miniconf/users/22396?format=json", "institution": "The University of Electro-Communications"}, {"id": 22397, "fullname": "Yuichi Yoshida", "url": "http://virtual.aistats.org/api/miniconf/users/22397?format=json", "institution": "National Institute of Informatics"}], "abstract": "Principal component analysis (PCA) and spectral clustering are representative methods for extracting and interpreting the inherent structure of data. However, if the output results significantly change upon the addition of new data points, it can lead to several issues such as instability in the downstream task or a lack of trust in the findings. To address these problems, we consider online variants of PCA and spectral clustering, and show that a natural subspace-preserving regularizer provides provable approximation and consistency guarantees. Here, an algorithm is said to have a high consistency if the output change, with respect to an appropriate distance metric, is small when new data points are added. We empirically confirm the superiority of the proposed methods using real-world data.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13570", "url": null, "sourceid": 416, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=Y4BkiyOYA2", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11263, "modified": "2026-03-29T20:43:04.036432-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=Y4BkiyOYA2", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "36", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13567, "uid": "68a83eeb494a308fe5295da69428a507", "name": "Data Distribution Valuation Using Generalized Bayesian Inference", "authors": [{"id": 19772, "fullname": "Ngoc Cuong Nguyen", "url": "http://virtual.aistats.org/api/miniconf/users/19772?format=json", "institution": "Durham University"}, {"id": 22389, "fullname": "Cuong V. Nguyen", "url": "http://virtual.aistats.org/api/miniconf/users/22389?format=json", "institution": "Durham University"}], "abstract": "We investigate the data distribution valuation problem, which aims to quantify the values of data distributions from their samples. This is a recently proposed problem that is related to but different from classical data valuation and can be applied to various applications. For this problem, we develop a novel framework called *Generalized Bayes Valuation* that utilizes generalized Bayesian inference with a loss constructed from transferability measures. This framework allows us to solve, in a unified way, seemingly unrelated practical problems, such as annotator evaluation and data augmentation. Using the Bayesian principles, we further improve and enhance the applicability of our framework by extending it to the continuous data stream setting. Our experiment results confirm the effectiveness and efficiency of our framework in different real-world scenarios.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13567", "url": null, "sourceid": 1259, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=YVokfG01YJ", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11260, "modified": "2026-03-29T20:43:03.845791-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=YVokfG01YJ", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "36", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13799, "uid": "716e1b8c6cd17b771da77391355749f3", "name": "Calibrated Predictive Lower Bounds on Time-to-Unsafe-Sampling in LLMs", "authors": [{"id": 22892, "fullname": "Hen Davidov", "url": "http://virtual.aistats.org/api/miniconf/users/22892?format=json", "institution": "Oxofrd, University of Oxford"}, {"id": 22893, "fullname": "Shai Feldman", "url": "http://virtual.aistats.org/api/miniconf/users/22893?format=json", "institution": "Computer Science Departmen, Technion-Israel Institute of Technology"}, {"id": 22894, "fullname": "Gilad Freidkin", "url": "http://virtual.aistats.org/api/miniconf/users/22894?format=json", "institution": "Technion - Israel Institute of Technology, Technion - Israel Institute of Technology"}, {"id": 9369, "fullname": "Yaniv Romano", "url": "http://virtual.aistats.org/api/miniconf/users/9369?format=json", "institution": "Technion"}], "abstract": "We introduce time-to-unsafe-sampling, a novel safety measure for generative models, defined as the number of generations required by a large language model (LLM) to trigger an unsafe (e.g., toxic) response. While providing a new dimension for prompt-adaptive safety evaluation, quantifying time-to-unsafe-sampling is challenging: unsafe outputs are often rare in well-aligned models and thus may not be observed under any feasible sampling budget. To address this challenge, we frame this estimation problem as one of survival analysis. We build on recent developments in conformal prediction and propose a novel calibration technique to construct a lower predictive bound (LPB) on the time-to-unsafe-sampling of a given prompt with rigorous coverage guarantees. Our key technical innovation is an optimized sampling-budget allocation scheme that improves sample efficiency while maintaining distribution-free guarantees. Experiments on both synthetic and real data support our theoretical results and demonstrate the practical utility of our method for safety risk assessment in generative AI models.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13799", "url": null, "sourceid": 2283, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=Amw0R1RHgW", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11492, "modified": "2026-03-29T20:43:13.504176-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=Amw0R1RHgW", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "37", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13397, "uid": "bcc0d400288793e8bdcd7c19a8ac0c2b", "name": "Deformed Decomposition for Non-negative Tensors", "authors": [{"id": 22004, "fullname": "Kazu Ghalamkari", "url": "http://virtual.aistats.org/api/miniconf/users/22004?format=json", "institution": "Technical University of Denmark"}, {"id": 22005, "fullname": "Petr Taborsky", "url": "http://virtual.aistats.org/api/miniconf/users/22005?format=json", "institution": "Technical University of Denmark"}, {"id": 4932, "fullname": "Morten M\u00f8rup", "url": "http://virtual.aistats.org/api/miniconf/users/4932?format=json", "institution": "Technical University of Denmark, Section for Cognitive Systems"}], "abstract": "Non-negative tensor factorization finds widespread use in numerous applications, however, its global optimization has been a long-standing challenge. As such, the Frobenius norm minimization even in the rank-1 setting is an NP-hard problem. We presently reformulate tensor decompositions using deformed algebra, and show that the best rank-1 approximation thereby reduces to a convex optimization problem for the rich \u03c7-divergence family. Building on this foundation, we propose the deformed many-body approximation for non-negative tensors, which expands model capacity while maintaining global optimality by preserving the flatness of the model manifold. Introducing latent variables, for a subclass of \u03c7-divergences, we further develop an Expectation-Maximization-based framework for the deformed extension of traditional low-rank approximations as iterative convex subproblems. We empirically demonstrate in tensor-based discrete density estimation that the deformed decompositions induce regularization and robustness against noise and mislabelled data. Beyond ordinary tensor algebra, our findings provide a factorization framework that enables us to leverage various divergences with convex rank-1 and many-body approximations.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13397", "url": null, "sourceid": 1349, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=qORe5NKGOn", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11090, "modified": "2026-03-29T20:42:57.262612-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=qORe5NKGOn", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "37", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13770, "uid": "365d17770080c807a0e47ae9118d8641", "name": "Calibrated Principal Component Regression", "authors": [{"id": 19862, "fullname": "Yixuan Florence Wu", "url": "http://virtual.aistats.org/api/miniconf/users/19862?format=json", "institution": "Northwestern University"}, {"id": 22820, "fullname": "Yilun Zhu", "url": "http://virtual.aistats.org/api/miniconf/users/22820?format=json", "institution": "University of Michigan - Ann Arbor"}, {"id": 22821, "fullname": "Lei Cao", "url": "http://virtual.aistats.org/api/miniconf/users/22821?format=json", "institution": "Northwestern University"}, {"id": 22822, "fullname": "Naichen Shi", "url": "http://virtual.aistats.org/api/miniconf/users/22822?format=json", "institution": "Northwestern University"}], "abstract": "We propose a new method for statistical inference in generalized linear models. In the overparameterized regime, Principal Component Regression (PCR) reduces variance by projecting high-dimensional data to a low-dimensional principal subspace before fitting. However, PCR incurs truncation bias whenever the true regression vector has mass outside the retained principal components (PC). To mitigate the bias, we propose Calibrated Principal Component Regression (CPCR), which first learns a low-variance prior in the PC subspace and then calibrates the model in the original feature space via a centered Tikhonov step. CPCR leverages cross-fitting and controls the truncation bias by softening PCR's hard cutoff. Theoretically, we calculate the out-of-sample risk in the random matrix regime, which shows that CPCR outperforms standard PCR when the regression signal has non-negligible components in low-variance directions. Empirically, CPCR consistently improves prediction across multiple overparameterized problems. The results highlight CPCR's stability and flexibility in modern overparameterized settings.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13770", "url": null, "sourceid": 1784, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=CohzTlZ8GG", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11463, "modified": "2026-03-29T20:43:12.324264-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=CohzTlZ8GG", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "38", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13740, "uid": "d645920e395fedad7bbbed0eca3fe2e0", "name": "Convergence of projected stochastic natural gradient variational inference: Balancing speed, computational effort, and accuracy", "authors": [{"id": 11038, "fullname": "Thomas Guilmeau", "url": "http://virtual.aistats.org/api/miniconf/users/11038?format=json", "institution": "Universit\u00e9 Paris-Saclay, CentraleSup\u00e9lec, INRIA, CVN"}, {"id": 10365, "fullname": "Hadrien Hendrikx", "url": "http://virtual.aistats.org/api/miniconf/users/10365?format=json", "institution": "Inria"}, {"id": 22748, "fullname": "Florence Forbes", "url": "http://virtual.aistats.org/api/miniconf/users/22748?format=json", "institution": "INRIA"}], "abstract": "Stochastic natural gradient variational inference (NGVI) is a popular and efficient algorithm for Bayesian inference. Despite empirical success, the convergence of this method is still not fully understood. In this work, we define and study a projected stochastic NGVI when variational distributions form an exponential family. Stochasticity arises when either gradients are intractable expectations or large sums. We prove new non-asymptotic convergence results for combinations of constant or decreasing step sizes and constant or increasing sample/batch sizes. When all hyperparameters are fixed, NGVI is shown to converge geometrically to a neighborhood of the optimum, while we establish convergence to the optimum with rates of the form $\\mathcal{O}\\left(\\frac{1}{T^{\\rho}} \\right)$, possibly with $\\rho \\geq 1$, for all other combinations of step size and sample/batch size schedules. These rates apply when the target posterior distribution is close in some sense to the considered exponential family. Our theoretical results extend existing NGVI and stochastic optimization results and provide more flexibility to adjust, in a principled way, step sizes and sample/batch sizes in order to meet speed, resources, or accuracy constraints.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13740", "url": null, "sourceid": 40, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=ErbBuMVhRi", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11433, "modified": "2026-03-29T20:43:11.102995-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=ErbBuMVhRi", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "38", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13693, "uid": "37bc2f75bf1bcfe8450a1a41c200364c", "name": "Convexified Message-Passing Graph Neural Networks", "authors": [{"id": 22642, "fullname": "Saar Cohen", "url": "http://virtual.aistats.org/api/miniconf/users/22642?format=json", "institution": "Bar-Ilan University"}, {"id": 22643, "fullname": "Noa Agmon", "url": "http://virtual.aistats.org/api/miniconf/users/22643?format=json", "institution": "Bar-Ilan University"}, {"id": 19861, "fullname": "Uri Shaham", "url": "http://virtual.aistats.org/api/miniconf/users/19861?format=json", "institution": "Bar Ilan University"}], "abstract": "Graph Neural Networks (GNNs) are key tools for graph representation learning, demonstrating strong results across diverse prediction tasks. In this paper, we present **Convexified Message-Passing Graph Neural Networks** (CGNNs), a novel and general framework that combines the power of message-passing GNNs with the tractability of *convex* optimization. By mapping their nonlinear filters into a reproducing kernel Hilbert space, CGNNs transform training into a convex optimization problem, which projected gradient methods can solve both efficiently and optimally. Convexity further allows CGNNs' statistical properties to be analyzed accurately and rigorously. For two-layer CGNNs, we establish rigorous generalization guarantees, showing convergence to the performance of an optimal GNN. To scale to deeper architectures, we adopt a principled layer-wise training strategy. Experiments on benchmark datasets show that CGNNs significantly exceed the performance of leading GNN models, obtaining 10\u201340\\% higher accuracy in most cases, underscoring their promise as a powerful and principled method with strong theoretical foundations. In rare cases where improvements are not quantitatively substantial, the convex models either slightly exceed or match the baselines, stressing their robustness and wide applicability. Though over-parameterization is often used to enhance performance in non-convex models, we show that our CGNNs yield shallow convex models that can surpass non-convex ones in accuracy and model compactness.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13693", "url": null, "sourceid": 304, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=IrQumNXKy5", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11386, "modified": "2026-03-29T20:43:08.999023-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=IrQumNXKy5", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "39", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13845, "uid": "309fee4e541e51de2e41f21bebb342aa", "name": "Clustering-Based Edge Augmentation for Minimizing the Kirchhoff Index", "authors": [{"id": 22983, "fullname": "Prasanth Yalamanchili", "url": "http://virtual.aistats.org/api/miniconf/users/22983?format=json", "institution": "University of Utah"}, {"id": 1131, "fullname": "Aditya Bhaskara", "url": "http://virtual.aistats.org/api/miniconf/users/1131?format=json", "institution": "University of Utah"}], "abstract": "The Kirchhoff index ($\\mathcal{K}_{G}$), defined as the sum of effective resistances over all pairs of nodes in a connected undirected graph $G$, is a fundamental metric for real-world networks. It corresponds to average power consumption in electrical circuits, average commute time of random walks, and more relevantly to optimization, is equal to $\\text{Tr}(\\mathcal{L}^{\\dagger})$, where $\\mathcal{L}$ is the graph Laplacian. In this paper, we study the problem of augmenting a given graph by adding $k$ edges to minimize the Kirchhoff index. The problem was introduced in a work of Ghosh, Boyd, and Saberi (2008), and is known to be NP-hard; the state-of-the-art algorithms mostly employ greedy heuristics and have very weak guarantees. We design novel algorithms and show bi-criteria approximation guarantees, i.e., the algorithm adds $c \\cdot k$ edges and obtains an $\\alpha$ factor approximation to the optimum objective value with $k$ edges. Specifically, an algorithm based on $k$-median clustering with penalties achieves $c=2$ and $\\alpha = O(k)$. By using known submodularity ideas, we extend this to achieve $c=O(\\log k)$ and $\\alpha=(4+\\epsilon)$. The problem corresponds to an augmentation version of the classic A-optimal experimental design problem in statistics. We also prove strong integrality gaps for the natural convex relaxation and demonstrate the performance of our algorithm on real and synthetic graphs.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13845", "url": null, "sourceid": 1596, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=6lTwYO9dZk", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11538, "modified": "2026-03-29T20:43:15.465633-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=6lTwYO9dZk", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "40", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13600, "uid": "f2217062e9a397a1dca429e7d70bc6ca", "name": "Corruption Robust Thompson Sampling for Gaussian Bandits", "authors": [{"id": 22462, "fullname": "Yinglun Xu", "url": "http://virtual.aistats.org/api/miniconf/users/22462?format=json", "institution": "University of Illinois, Urbana Champaign"}, {"id": 22463, "fullname": "ZHIWEI WANG", "url": "http://virtual.aistats.org/api/miniconf/users/22463?format=json", "institution": "Tsinghua University"}, {"id": 22237, "fullname": "Gagandeep Singh", "url": "http://virtual.aistats.org/api/miniconf/users/22237?format=json", "institution": "University of Illinois, Urbana Champaign"}], "abstract": "Thompson sampling is one of the most popular learning algorithms for online sequential decision-making problems and has rich real-world applications. However, traditional Thompson sampling algorithms are limited by the assumption that the rewards received are uncorrupted, which may not hold in real-world applications where adversarial reward poisoning exists. To make Thompson sampling more reliable, our goal is to make it robust against adversarial reward poisoning. Particularly, we consider a strong attack threat model where an adversary applies corruption after observing the agent's actions. The main challenge is that one can no longer compute the actual posteriors for the true reward, as the agent can only observe the rewards after corruption. In this work, we solve this problem by computing pseudo-posteriors that are less likely to be manipulated by the attack. Particularly, we focus on two popular settings: stochastic bandits and contextual linear bandits with priors as Gaussian distributions. **We are the first** to propose robust algorithms based on Thompson sampling for the two bandit settings in both cases where the agent is aware or unaware of the attacker's budget. We theoretically show that our algorithms guarantee near-optimal regret under any attack strategy.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 1", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13600", "url": null, "sourceid": 149, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=UWTP7rymyF", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%201?format=json", "parent_id": 11473, "eventmedia": [{"id": 11293, "modified": "2026-03-29T20:43:05.219244-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=UWTP7rymyF", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "40", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13343, "uid": "03e0704b5690a2dee1861dc3ad3316c9", "name": "DIVERSED: Relaxed Speculative Decoding via Dynamic Ensemble Verification", "authors": [{"id": 21900, "fullname": "Ziyi Wang", "url": "http://virtual.aistats.org/api/miniconf/users/21900?format=json", "institution": "Purdue University"}, {"id": 18615, "fullname": "Siva Rajesh Kasa", "url": "http://virtual.aistats.org/api/miniconf/users/18615?format=json", "institution": "national university of singaore, National University of Singapore"}, {"id": 21901, "fullname": "Ankith M S", "url": "http://virtual.aistats.org/api/miniconf/users/21901?format=json", "institution": "Amazon"}, {"id": 18612, "fullname": "SANTHOSH KASA", "url": "http://virtual.aistats.org/api/miniconf/users/18612?format=json", "institution": "Carnegie Mellon University"}, {"id": 21902, "fullname": "Jiaru Zou", "url": "http://virtual.aistats.org/api/miniconf/users/21902?format=json", "institution": "Department of Computer Science, University of Illinois at Urbana-Champaign"}, {"id": 21904, "fullname": "Sumit Negi", "url": "http://virtual.aistats.org/api/miniconf/users/21904?format=json", "institution": "Amazon"}, {"id": 10220, "fullname": "Ruqi Zhang", "url": "http://virtual.aistats.org/api/miniconf/users/10220?format=json", "institution": "Purdue University"}, {"id": 21903, "fullname": "Nan Jiang", "url": "http://virtual.aistats.org/api/miniconf/users/21903?format=json", "institution": "University of Texas at El Paso"}, {"id": 438, "fullname": "Qifan Song", "url": "http://virtual.aistats.org/api/miniconf/users/438?format=json", "institution": "Purdue University "}], "abstract": "Speculative decoding is an effective technique for accelerating large language model inference by drafting multiple tokens in parallel. In practice, its speedup is often bottlenecked by a rigid verification step that strictly enforces the accepted token distribution to exactly match the target model. This constraint leads to the rejection of many plausible tokens, lowering the acceptance rate and limiting overall time speedup. To overcome this limitation, we propose DynamIc VErification RElaxed SpEculative Decoding (DIVERSED), a relaxed verification framework that improves time efficiency while preserving generation quality.  DIVERSED: learns an ensemble-based verifier that blends the draft and target model distributions with a task-dependent and context-dependent weight.  We provide theoretical justification for our approach and demonstrate empirically that DIVERSED achieves substantially higher inference efficiency compared to standard speculative decoding methods. Code is available at: \\url{https://github.com/comeusr/diversed}.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 2", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13343", "url": null, "sourceid": 1019, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=yCNt8ovyzl", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%202?format=json", "parent_id": 11472, "eventmedia": [{"id": 11036, "modified": "2026-03-29T20:42:55.032465-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=yCNt8ovyzl", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "40", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}, {"id": 13812, "uid": "b9f94c77652c9a76fc8a442748cd54bd", "name": "ConDiSim: Conditional Diffusion Models for Simulation-Based Inference", "authors": [{"id": 19833, "fullname": "Mayank Nautiyal", "url": "http://virtual.aistats.org/api/miniconf/users/19833?format=json", "institution": "Uppsala University"}, {"id": 22918, "fullname": "Andreas Hellander", "url": "http://virtual.aistats.org/api/miniconf/users/22918?format=json", "institution": "Uppsala University"}, {"id": 22919, "fullname": "Prashant Singh", "url": "http://virtual.aistats.org/api/miniconf/users/22919?format=json", "institution": "Uppsala University"}], "abstract": "We present ConDiSim, a conditional diffusion model for simulation-based inference in complex systems with intractable likelihoods. ConDiSim leverages denoising diffusion probabilistic models to approximate posterior distributions, consisting of a forward process that adds Gaussian noise to parameters, and a reverse process learning to denoise, conditioned on observed data. This approach effectively captures complex dependencies and multi-modalities within posteriors. ConDiSim is evaluated across ten benchmark problems and two real-world test problems, where it demonstrates effective posterior approximation accuracy while maintaining computational efficiency and stability in model training. ConDiSim provides a robust and extensible framework for simulation-based inference, well suited to parameter estimation tasks that demand fast methods for handling noisy, time series observations.", "topic": null, "keywords": [], "decision": "Accept (Poster)", "session": "Poster Session 3", "eventtype": "Poster", "event_type": "Poster", "room_name": null, "virtualsite_url": "/virtual/2026/poster/13812", "url": null, "sourceid": 1702, "sourceurl": "https://openreview.net/group?id=aistats.org/AISTATS/2026/Conference", "starttime": null, "endtime": null, "starttime2": null, "endtime2": null, "diversity_event": null, "paper_url": "https://openreview.net/forum?id=9u30uqg7hp", "paper_pdf_url": null, "children_url": null, "children": [], "children_ids": [], "parent": "http://virtual.aistats.org/api/miniconf/events/Conference%20Sessions%202026:%20Poster%20Session%203?format=json", "parent_id": 11471, "eventmedia": [{"id": 11505, "modified": "2026-03-29T20:43:14.039244-07:00", "display_section": 1, "type": "URL", "name": "OpenReview", "visible": true, "sortkey": 0, "is_live_content": false, "uri": "https://openreview.net/forum?id=9u30uqg7hp", "resourcetype": "UriEventmedia"}], "show_in_schedule_overview": false, "visible": true, "poster_position": "41", "schedule_html": "", "latitude": null, "longitude": null, "related_events": [], "related_events_ids": []}]}