实时热榜
OpenReview
- 01Simplifying Multi-Task Architectures Through Task-Specific Normalization期刊：Transactions on Machine Learning Research · 摘要：Multi-task learning (MTL) aims to leverage shared knowledge across tasks to improve generalization and parameter efficiency, yet balancing resources and mitigating interference remain open challenges. Architectural solutions often introduce elaborate task-specific modules or routing schemes, increasing complexity and overhead. In this work, we show that normalization layers alone are sufficient to address many of these challenges. Simply replacing shared normalization with task-specific variants already yields competitive performance, questioning the need for complex designs. Building on this… · 篇幅：Regular submission (no more than 12 pages of main content) · 代码：https://github.com/xapharius/TSsNorm · OpenReview ID：QNO893OXrSMihai Suteu, Ovidiu Serban
- 02A Player Selection Network for Scalable Game-Theoretic Prediction and Planning期刊：Transactions on Machine Learning Research · 摘要：While game-theoretic planning frameworks are effective at modeling multi-agent interactions, they require solving large optimization problems where the number of variables increases with the number of agents, resulting in long computation times that limit their use in large-scale, real-time systems. To address this issue, we propose i) PSN Game—a learning-based, game-theoretic prediction and planning framework that reduces game size by learning a Player Selection Network (PSN); and ii) a Goal Inference Network (GIN) that makes it possible to use the PSN in incomplete-information games where o… · 篇幅：Regular submission (no more than 12 pages of main content) · OpenReview ID：YvvB78ILSPTianyu Qiu, Eric Ouano, Fernando Palafox et al.
- 03Harmonizing Gradient Matching For Fairness期刊：Transactions on Machine Learning Research · 摘要：Ensuring fairness across demographic groups is critical for machine learning systems deployed in high-stakes applications. Most existing approaches enforce fairness by directly minimizing disparities in predefined fairness metrics between groups, focusing primarily on the final model outcome. However, differences in distributions across groups can lead to heterogeneous optimization signals during training, resulting in imbalanced parameter updates and unstable fairness–performance trade-offs. In this work, we propose Fair Gradient Matching (FairGM), a fairness-aware optimization framework tha… · 篇幅：Regular submission (no more than 12 pages of main content) · OpenReview ID：HMlK36YWWtZiwei Wu, Yikun Ban, Jingrui He
- 04Self-Supervised Learning via Flow-Guided Neural Operator on Time-Series Data期刊：Transactions on Machine Learning Research · 摘要：Self-supervised learning (SSL) is a powerful paradigm for learning from unlabeled time-series data. However, popular methods such as masked autoencoders (MAEs) rely on reconstructing inputs from a fixed, predetermined masking ratio. Instead of this static design, we propose treating the corruption level as a new degree of freedom for representation learning, enhancing flexibility and performance. To achieve this, we introduce the Flow-Guided Neural Operator (FGNO), a novel framework combining operator learning with flow matching for SSL training. FGNO learns the frequency functions represente… · 篇幅：Regular submission (no more than 12 pages of main content) · 代码：https://github.com/neuraloperator/FGNO · OpenReview ID：YAYW9Y173zDuy Nguyen, Jiachen Yao, Jiayun Wang et al.
- 05Tensor-Decomposed RNNs for Marked Temporal Point Processes期刊：Transactions on Machine Learning Research · 摘要：We study parameter-efficient neural Marked Temporal Point Processes (MTPPs) for high-dimensional mark and exogenous feature spaces. Building on tensor-train (TT) factorization of recurrent kernels, we propose mark-aware TT shaping that aligns TT cores with known multi-way domain structure (e.g., asset/venue/side in finance). We provide a conditional intensity function-consistent training recipe and evaluate both accuracy and calibration (mark reliability and time-rescaling diagnostics). Across finance and public MTPP benchmarks, TT-compressed RNNs reduce parameters by 40-70% while matching de… · 篇幅：Regular submission (no more than 12 pages of main content) · OpenReview ID：Y0up92GcjFTimothy Mulumba
- 06SinGLU: Sinusoidal Gated Linear Units Improve Classification Accuracy of Small Vision Transformers期刊：Transactions on Machine Learning Research · 摘要：Gated Linear Unit (GLU) variants such as SwiGLU are now widely used in modern Transformers. However, the GLU functions explored in the recent literature represent only a small fraction of the possible GLU design space. Starting from a systematic enumeration of a restricted family of zeroth-, first-, and second-order GLU-type formulas, we conduct a controlled study on ViT‑Tiny across CIFAR‑10, CIFAR‑100, SVHN and ImageNet‑64, instantiating each GLU formula with Sigmoid, Tanh and Sin activations. Under identical training recipes and matched parameter counts, our proposed first-order variant \te… · 篇幅：Long submission (more than 12 pages of main content) · 代码：https://github.com/Luke-Byrne-Eng/SinGLU-Experimental-Code · OpenReview ID：qq4yipldw2Luke Byrne, Paul Murray
- 07Foundation Models in Robotics: A Comprehensive Review of Methods, Models, Datasets, Challenges and Future Research Directions期刊：Transactions on Machine Learning Research · 摘要：Over the recent years, the field of robotics has been undergoing a transformative paradigm shift from fixed, single-task, domain-specific solutions towards adaptive, multi-function, general-purpose agents, capable of operating in complex, open-world, dynamic environments. This tremendous advancement is primarily driven by the emergence of Foundation Models (FMs), i.e., large-scale neural-network architectures trained on massive, internet-scale, heterogeneous datasets that provide unprecedented capabilities in multi-modal understanding/reasoning, long-horizon planning, and cross-embodiment gen… · 篇幅：Long submission (more than 12 pages of main content) · OpenReview ID：qF7vdPrpPkAggelos Psiris, Vasileios Argyriou, Evangelos K. Markakis et al.
- 08Discovering Generalizable Governing Equations for Graph Dynamical Systems with Interpretable Neural Networks期刊：Transactions on Machine Learning Research · 摘要：The discovery of symbolic governing equations is a central goal in science; yet, it remains challenging particularly for graph dynamical systems, where the network topology further shapes the system behavior. While artificial intelligence offers powerful tools for modeling these dynamics, the field lacks a rigorous comparative benchmark to assess the true scientific utility of the discovered laws. To address this challenge, this work proposes a novel evaluation pipeline designed to rigorously assess state-of-the-art symbolic regression models for graph equation discovery. Moving beyond simple… · 篇幅：Long submission (more than 12 pages of main content) · 代码：https://github.com/riccardocappi/Kan-for-Interpretable-Graph-Dynamics · OpenReview ID：a2mPNSSAYLRiccardo Cappi, Paolo Frazzetto, Nicolò Navarin et al.
- 09Realistic Evaluation of Model Merging for Compositional Generalization期刊：Transactions on Machine Learning Research · 摘要：Model merging has emerged as a practical and cost-effective approach for combining multiple pretrained models into a single model that inherits their capabilities and often achieves improved performance. Its growing popularity has led to the rapid development of numerous merging techniques. However, these methods are typically evaluated in disparate experimental settings and make differing assumptions about model architecture, data availability, and computational budget, making direct comparison difficult. In this work, we systematically characterize the relative strengths and limitations of… · 篇幅：Regular submission (no more than 12 pages of main content) · 代码：https://github.com/r-three/realistic_evaluation_of_model_merging_for_compositional_generalization · OpenReview ID：j7ye0nXvEmDerek Tam, Yash Kant, Brian Lester et al.
- 10LAW & ORDER: Adaptive Spatial Weighting for Medical Diffusion and Segmentation期刊：Transactions on Machine Learning Research · 摘要：Medical image analysis depends on accurate segmentation and controllable synthesis, but both tasks face severe spatial imbalance: lesions occupy small regions against large backgrounds. We study adaptive spatial weighting as a task-level design principle and instantiate it in two adapters. LAW learns per-pixel loss weights for mask-conditioned diffusion by modulating a ratio prior with a feature-dependent delta map, with normalization, clamping, and Dice regularization for stability. ORDER improves lightweight segmentation by adding selective bidirectional skip attention with stage-wise confi… · 篇幅：Regular submission (no more than 12 pages of main content) · OpenReview ID：sJXqzr3oLlAnugunj Naman, Ayushman Singh, Gaibo Zhang et al.
- 11SafeFix: Targeted Model Repair via Controlled Image Generation期刊：Transactions on Machine Learning Research · 摘要：Deep learning models for visual recognition often exhibit systematic errors due to underrepresented semantic subpopulations. While existing debugging frameworks can identify these failure slices, effectively repairing them remains difficult. Current solutions often rely on manually designed prompts to generate synthetic images—an approach that introduces distribution shift and semantic errors, often resulting in new bugs. To address these issues, we introduce SafeFix, a framework for distribution-consistent model repair via controlled generation that employs a diffusion model to generate sema… · 篇幅：Regular submission (no more than 12 pages of main content) · 代码：https://github.com/oxu2/SafeFix · OpenReview ID：TtpW6JiEiWOuyang Xu, Baoming Zhang, RUIYU MAO et al.
- 12Lipschitz Continuity in Deep Learning: A Systematic Review of Theoretical Foundations, Estimation Methods, Regularization Approaches, and Certifiable Robustness期刊：Transactions on Machine Learning Research · 摘要：Lipschitz continuity is a fundamental property of neural networks that characterizes their sensitivity to input perturbations. It plays a pivotal role in deep learning, governing robustness, generalization and optimization dynamics. Despite its importance, research on Lipschitz continuity is scattered across various domains, lacking a unified perspective. This paper addresses this gap by providing a systematic review of Lipschitz continuity in deep learning. We explore its theoretical foundations, estimation methods, regularization approaches, and certifiable robustness. By reviewing existing… · 篇幅：Long submission (more than 12 pages of main content) · 代码：https://github.com/roisincrtai/lipschitz_survey · OpenReview ID：pRZ0RKl11fRóisín Luo, James McDermott, Colm O'Riordan
- 13Achieving Adaptivity and Optimality for Multi-armed Bandits using Exponential-Kullback Leibler Maillard Sampling期刊：Transactions on Machine Learning Research · 摘要：We study the problem of $K$-armed bandits with reward distributions belonging to a one-parameter exponential distribution family. In the literature, several criteria have been proposed to evaluate the performance of such algorithms, including Asymptotic Optimality, Minimax Optimality, Sub-UCB, and variance-adaptive worst-case regret bound. Thompson Sampling-based and Upper Confidence Bound-based algorithms have been employed to achieve some of these criteria. However, none of these algorithms simultaneously satisfy all the aforementioned criteria. In this paper, we design an algorithm, Expone… · 篇幅：Regular submission (no more than 12 pages of main content) · OpenReview ID：IuVkRmecVpHao Qin, Kwang-Sung Jun, Chicheng Zhang
- 14Not All Structure Is Learned: Disentangling Inherited and Learned Representations in Recurrent Networks期刊：Transactions on Machine Learning Research · 摘要：Structure observed in trained recurrent networks may be inherited from input encodings rather than learned from data. We develop and apply a three-step decomposition to disentangle the two: (1) compare trained representations against untrained baselines to isolate input-driven structure, (2) compare against information-theoretic bounds to quantify what is achievable without learning, and (3) use causal interventions to test whether inherited and learned components are functionally used. Applied to GRUs trained via behavioral cloning on aliased navigation in a 127-node binary tree, the most pr… · 篇幅：Long submission (more than 12 pages of main content) · OpenReview ID：1RfgHzf5IAMark Alence
- 15CoDoL: Conditional Domain Prompt Learning for Out-of-Distribution Generalization期刊：Transactions on Machine Learning Research · 摘要：Recent advances in pre-training vision-language models (VLMs), e.g., contrastive language-image pre-training (CLIP) methods, have shown great potential in learning out-of-distribution (OOD) representations. Despite showing competitive performance, the prompt-based CLIP methods still suffer from: i) inaccurate text descriptions, which leads to degraded accuracy and robustness, and poses a challenge for zero-shot CLIP methods. ii) limited vision-language embedding alignment, which significantly affects the generalization performance. To tackle the above issues, this paper proposes a novel Condi… · 篇幅：Regular submission (no more than 12 pages of main content) · OpenReview ID：MDxhbeE21DMin Zhang, YUYIN WANG, Zhongxiang Dai et al.
- 16Do We Really Need to Approach the Entire Pareto Front in Many-Objective Bayesian Optimisation?期刊：Transactions on Machine Learning Research · 摘要：Many-objective optimisation, a subset of multi-objective optimisation, involves optimisation problems with more than three objectives. As the number of objectives increases, the number of solutions needed to adequately represent the entire Pareto front typically grows substantially. This makes it challenging, if not infeasible, to design a search algorithm capable of effectively exploring the entire Pareto front. This difficulty is particularly acute in the Bayesian optimisation paradigm, where sample efficiency is critical and only a limited number of solutions (often a few hundred) are eval… · 篇幅：Regular submission (no more than 12 pages of main content) · 代码：https://github.com/ChaoJiang52/SPMO · OpenReview ID：ZsI4bmqUD8Chao Jiang, Jingyu Huang, Miqing Li
- 17A Survey on Efficient Protein Language Models期刊：Transactions on Machine Learning Research · 摘要：Protein language models (pLMs) have become indispensable tools in computational biology, driving advances in variant effect prediction, functional annotation, structure prediction, and engineering. However, their rapid expansion from millions to tens of billions of parameters introduces significant computational, accessibility, and sustainability challenges that limit practical application in environments constrained by GPU memory, hardware availability, and energy budgets. This survey presents the first comprehensive review of efficient pLMs, synthesizing recent advancements across four key… · 篇幅：Long submission (more than 12 pages of main content) · 代码：https://github.com/SR-A-W/efficient-protein-language-model-survey · OpenReview ID：PTReuOwsXzShouren Wang, Debargha Ganguly, Vinooth Rao Kulkarni et al.
- 18MEMETRON: Memetic Response Optimizer for Reward-Guided Post-Decoding Optimization of Large Language Models期刊：Transactions on Machine Learning Research · 摘要：Modern large language models (LLMs) are commonly optimized using scalar reward signals defined over completed responses, applied both during training and at inference time. However, most such reward-guided post-decoding methods remain one-shot: they independently sample a set of responses, score each once, and select the best. Staying shallow and narrow leaves higher-reward responses unrealized, while scaling up to shallow and wide sampling exacerbates reward hacking, making downstream selection methods such as Best-of-$N$ and Self-consistency unreliable. We propose MEMETRON, an anytime memet… · 篇幅：Long submission (more than 12 pages of main content) · OpenReview ID：QRW8OGn3vbSon The Nguyen, Theja Tulabandhula
- 19Random Logit Scaling: Defending Deep Neural Networks Against Black-Box Score-Based Adversarial Example Attacks期刊：Transactions on Machine Learning Research · 摘要：Machine learning models are increasingly adapted in various domains. However, adversarial examples pose a significant threat to the reliable deployment of these models. In recent years, some powerful adversarial example attacks have been proposed for the fast and query-efficient generation of adversarial examples, even in black-box scenarios, highlighting the need for scalable, low-cost, and powerful defenses. In this work, we present two contributions to the domain of black-box adversarial example attacks and defenses. First, we propose Random Logit Scaling (RLS), a randomization-based defen… · 篇幅：Regular submission (no more than 12 pages of main content) · 代码：https://github.com/HamidDasht/RLS-adversarial-defense · OpenReview ID：CXafPv4aAGHamid Dashtbani, Mehdi Dousti Gandomani, AmirMahdi Sadeghzadeh
- 20White-Box Sensitivity Auditing with Steering Vectors期刊：Transactions on Machine Learning Research · 摘要：Algorithmic audits are essential tools for examining systems for properties required by regulators or desired by operators. Current audits of large language models (LLMs) primarily rely on black-box evaluations that assess model behavior only through input–output testing. These methods are limited to tests constructed in the input space, often generated by heuristics. In addition, many socially relevant model properties (e.g., gender bias) are abstract and difficult to measure through text-based inputs alone. To address these limitations, we propose a white-box sensitivity auditing framework… · 篇幅：Long submission (more than 12 pages of main content) · 代码：https://github.com/hannahxchen/llm-steering-audit · OpenReview ID：EfinGGyQRzHannah Cyberey, Yangfeng Ji, David Evans
- 21ReDiTT: Retrieval Augmented Conditional Diffusion Transformers for Asynchronous Time Series期刊：Transactions on Machine Learning Research · 摘要：We present a diffusion based model for asynchronous time series prediction, where the goal is to predict the next inter event time and event type. To address the inherent uncertainty of future events, we introduce ReDiTT, a retrieval augmented conditional diffusion transformer that operates in latent space. ReDiTT retrieves structurally similar latent sequences from a memory bank during both training and inference and incorporates them as reference conditions through cross attention. This retrieval based conditioning allows the model to attend to relevant temporal dynamics and provides global… · 篇幅：Regular submission (no more than 12 pages of main content) · OpenReview ID：sv35KiCipbSaiyue Lyu, Zhitian Zhang, Ruizhi Deng et al.
- 22Task-Aware Model Merging via Fisher-Weighted Median期刊：Transactions on Machine Learning Research · 摘要：Fine-tuning large language models provides strong in-domain performance, but it can limit generalization and requires the storage of many specialized models. Retraining a unified multitask model is often infeasible due to data unavailability or high computational cost. Most model merging approaches perform arithmetic operations directly on model parameters. Although research on model merging has expanded significantly in recent years, two distinct directions have become dominant: (1) techniques that mitigate interference from redundant parameters and sign conflicts, and (2) techniques that ac… · 篇幅：Long submission (more than 12 pages of main content) · 代码：https://github.com/babangain/drift-median · OpenReview ID：tB6bb0ZosXBaban Gain, Saswati Dana, Udit Sharma et al.
- 23GE-FM: Geometry-aware Energy-based Flow Matching for Non Euclidean Manifolds期刊：Transactions on Machine Learning Research · 摘要：Flow Matching has emerged as a powerful framework for generative transport and denoising, yet existing formulations are inherently Euclidean, neglecting the curved and time-evolving geometry of diffusion manifolds. Recent higher-order extensions seek to recover curved transport by explicitly modeling higher derivatives, but these approaches introduce instability and accumulate discretization error, particularly in few-step ODE sampling regimes. We propose a strictly first-order-in-time, energy-based flow matching framework that incorporates geometry through Christoffel-adjusted dynamics. Our… · 篇幅：Regular submission (no more than 12 pages of main content) · 代码：https://github.com/AyushRoy2001/GE-FM.git · OpenReview ID：Wb9wwUF7aVAyush Roy, Arjun Ramesh Kaushik, Vishnu Suresh Lokhande et al.
- 24Bridging Reasoning to Learning: Unmasking Illusions using Complexity Out-of-Distribution Generalization期刊：Transactions on Machine Learning Research · 摘要：Recent progress has pushed AI frontiers from pattern-recognition tasks toward problems that require step-by-step, System-2-style reasoning, especially with large language models. Yet, unlike learning, where generalization and out‑of‑distribution (OoD) evaluation concepts are well formalized, there is no clear, consistent definition or metric for “reasoning ability.” We propose Complexity Out‑of‑Distribution (Complexity OoD) generalization as a framework and problem setting to measure reasoning. A model exhibits Complexity OoD generalization when it maintains performance on test instances whos… · 篇幅：Long submission (more than 12 pages of main content) · OpenReview ID：07fh13gWs0Mahdi Samiei, Arash Marioriyad, Arman Tahmasebi-Zadeh et al.
- 25Continual Test-Time Adaptation in Computer Vision: Methods, Benchmarks, and Future Directions期刊：Transactions on Machine Learning Research · 摘要：Deep neural nets achieve remarkable performance when training and test data share the same distribution, but this assumption frequently breaks in real-world deployment, where data undergoes continual distributional shifts. Continual Test-Time Adaptation (CTTA) addresses this challenge by adapting pretrained models to non-stationary target distributions on-the-fly, without access to source data or labeled targets, while mitigating two critical failure modes: catastrophic forgetting of source knowledge and error accumulation from noisy pseudo-labels over extended time horizons. In this comprehe… · 篇幅：Long submission (more than 12 pages of main content) · OpenReview ID：mM3r03Xw1VSarthak Kumar Maharana, Shambhavi Mishra, Yunbei Zhang et al.
- 26Investigating the limits of free-form debate as a scalable oversight strategy期刊：Transactions on Machine Learning Research · 摘要：Debate is a scalable oversight method involving two copies of a strong model trained to defend alternative responses to a question, with a judge with less task-relevant information, time, or domain-specific capability evaluating which answer is better supported. We replicate and extend a result from prior work demonstrating that training Llama3-8B-Instruct-262k as a debater led to increased performance of a GPT-4-class judge model on QuALITY, a question-answering task that grants the debaters a capability advantage via information asymmetry. When replicating the original setup as closely as p… · 篇幅：Regular submission (no more than 12 pages of main content) · 代码：https://github.com/garetht/nyu-debate-modeling · OpenReview ID：vRCGzuAhOMGareth Tan, Leonid Tsyplenkov, Edy Nastase et al.
- 27MatchEx: Model-Level GNN Explanations with Multi-Granular Insights期刊：Transactions on Machine Learning Research · 摘要：Graph Neural Networks (GNNs) are increasingly deployed in high-stakes domains where interpretability is crucial. Existing model-level explanation methods largely rely on generative models, which often produce motifs that fail to resemble real instances, cannot account for the diversity of discriminative motifs recognized by the classifier for a target class and lack mechanisms for translating global explanations to instance-level insights. We present MatchEx, a framework that discovers discriminative motifs directly from real instances by optimizing a novel matching objective. Unlike isomorph… · 篇幅：Regular submission (no more than 12 pages of main content) · OpenReview ID：YMETLG2WvMSayan Saha, Sanghamitra Bandyopadhyay
- 28Predicting integers from continuous parameters期刊：Transactions on Machine Learning Research · 摘要：We study the problem of predicting numeric labels that are constrained to the integers or to a subrange of the integers. For example, the number of up-votes on social media posts, or the number of bicycles available at a public rental station. While it is possible to model these as continuous values, and to apply traditional regression, this approach changes the underlying distribution on the labels from discrete to continuous. Discrete distributions have certain benefits, which leads us to the question whether such integer labels can be modeled directly by a discrete distribution, whose para… · 篇幅：Long submission (more than 12 pages of main content) · 代码：https://github.com/R3VST/predicting-integers-from-continuous-parameters · OpenReview ID：d1WKFlKFEaBas Joris Maat, Peter Bloem
- 29Extracting Common Components from Partially Observed Views Using Diffusion Geometry期刊：Transactions on Machine Learning Research · 摘要：Data acquired from multiple sensors or modalities, commonly referred to as multiview data, is prevalent in real-world applications. A core problem in multiview data analysis is finding representations of common components across views while filtering out view-specific nuisance factors. A widely spread assumption in existing methods is that the views are fully aligned, where each sample has measurements from all views. However, in practice, data is often partially aligned, where some samples have missing measurements from one or more views, and only a subset of the samples are fully aligned. I… · 篇幅：Regular submission (no more than 12 pages of main content) · 代码：https://github.com/barweiss1/adm_plus · OpenReview ID：yTHGIV8ToFBar Weiss, Hau-Tieng Wu, Ronen Talmon
- 30Don't Go Breaking My LLM: The Impact of Pruning Attention Layers on Explanation Faithfulness and Confidence Calibration期刊：Transactions on Machine Learning Research · 摘要：Pruning Large Language Models (LLMs) reduces memory and inference costs by removing parts of the network, producing smaller models that retain most of their accuracy. As attention layers are the most resource-intensive parts of LLMs, pruning them is a promising compression strategy. Prior work shows that up to 33% of attention layers can be pruned with minimal accuracy loss. Nevertheless, the impact of attention pruning on model interpretability, specifically faithfulness and confidence calibration, remains unstudied. To address this gap, we study how pruning attention layers affects explanat… · 篇幅：Long submission (more than 12 pages of main content) · 代码：https://github.com/pietrotrope/Dont_Go_Breaking_My_LLM · OpenReview ID：VxZd6HfMOoPietro Tropeano, Maria Maistro, Tuukka Ruotsalo et al.
- 31Federated Learning with Projected Trajectory Regularization期刊：Transactions on Machine Learning Research · 摘要：Federated learning enables joint training of machine learning models from distributed clients without sharing their local data. One key challenge in federated learning is to handle non-identically distributed data across the clients, which leads to deteriorated model training performance. Prior works in this line of research mainly focus on utilizing last-step global model parameters/gradients or the linear combinations of the past model parameters/gradients, which do not fully exploit the potential of global information from the model training trajectory. In this paper, we propose a novel fe… · 篇幅：Regular submission (no more than 12 pages of main content) · OpenReview ID：vfCztZvcP3Tiejin Chen, Yuanpu Cao, Yujia Wang et al.
- 32End-to-End 4D Heart Mesh Recovery Across Full-Stack and Sparse Cardiac MRI期刊：Transactions on Machine Learning Research · 摘要：Reconstructing cardiac motion from CMR sequences is critical for diagnosis, prognosis, and intervention. Existing methods rely on complete CMR stacks to infer full heart motion, limiting their applicability during intervention when only sparse observations are available. We present TetHeart, the first end-to-end framework for unified 4D heart mesh recovery from both offline full-stack and intra-procedural sparse-slice observations. Our method leverages deformable tetrahedra to capture shape and motion in a coherent space shared across cardiac structures. Before a procedure, it initializes det… · 篇幅：Regular submission (no more than 12 pages of main content) · 代码：https://github.com/Scalsol/TetHeart · OpenReview ID：9k00kN5yk2Yihong Chen, Jiancheng Yang, Deniz Sayin Mercadier et al.
- 33POPS: Recovering Unlearned Multi-Modality Knowledge in MLLMs with Prompt-Optimized Parameter Shaking期刊：Transactions on Machine Learning Research · 摘要：Multimodal Large Language Models (MLLMs) have demonstrated impressive performance on cross-modal tasks by jointly training on large-scale textual and visual data, where privacy-sensitive examples could be unintentionally encoded, raising concerns about privacy or copyright violation. To this end, Multi-modality Machine Unlearning (MMU) was proposed as a mitigation that can effectively force MLLMs to forget private information. However, the robustness of such unlearning methods is not fully exploited when the model is published and accessible to malicious users. In this paper, we propose a nov… · 篇幅：Regular submission (no more than 12 pages of main content) · OpenReview ID：wMiEcH84l9Zhangheng LI, Jianing Zhu, Junyuan Hong et al.
- 34Neural Networks Performance Prediction using Weights and Gradients Analysis期刊：Transactions on Machine Learning Research · 摘要：Neural network performance predictors are widely used to accelerate neural architecture search, but existing methods face a persistent trade-off: learning-based predictors require costly per-dataset initialization, while lightweight proxies are fast yet struggle to exploit prior experience and often degrade under dataset shift. We introduce NAP2, a hybrid performance predictor that models early training dynamics. NAP2 tracks the temporal evolution of layer-wise weight and gradient statistics over a small number of mini-batches, producing accurate rankings from as little as 100 mini-batches pe… · 篇幅：Regular submission (no more than 12 pages of main content) · OpenReview ID：51TWh8tlSyMichael Bohadana, Alon Schneider, Gilad Katz
- 35Sliding Window Recurrences for Sequence Models期刊：Transactions on Machine Learning Research · 摘要：Multi-hybrid architectures are poised to take over language modeling due to better quality and performance. We introduce a hierarchical decomposition framework for linear recur- rences that yields Sliding Window Recurrences, a windowed training mode for recurrence layers in hybrid models. Unlike sliding-window attention, SWR is derived from the transfer structure of recurrences: it truncates the carrier system induced by the decomposition while preserving dense local recurrence dynamics. We focus specifically on hardware-aligned win- dows which are naturally jagged, limiting costly inter-warp… · 篇幅：Long submission (more than 12 pages of main content) · 代码：https://github.com/radicalnumerics/spear · OpenReview ID：V09uO70ouzDragos Secrieru, Garyk Brixi, Yoshua Bengio et al.
- 36RA-CoA: Training-free Fashion Image Captioning via Retrieval-Augmented Chain-of-Attributes期刊：Transactions on Machine Learning Research · 摘要：Fashion Image Captioning (FIC) plays a vital role in enhancing user experience and product search in e-commerce platforms. Unlike natural scene image captioning, FIC requires fine-grained visual reasoning and knowledge of domain-specific terminology to capture subtle attributes such as neckline and closure types, graphic patterns, and dress silhouettes. Moreover, as fashion inventories evolve rapidly with new trends, styles, and frequently emerging vocabulary, developing training-free captioning solution becomes essential for scalability and real-world adaptability. Instruction-tuned vision-l… · 篇幅：Long submission (more than 12 pages of main content) · 代码：https://github.com/vl2g/RACoA · OpenReview ID：PpkOrVUpJ6Abhirama Subramanyam Penamakuri, Shreya Shukla, Anand Mishra
- 37A Monte Carlo Framework for Calibrated Uncertainty Estimation in Sequence Prediction期刊：Transactions on Machine Learning Research · 摘要：Probabilistic prediction of sequences from images and other high-dimensional data remains a key challenge---particularly in safety-critical domains. In these settings, it is often desirable to quantify the uncertainty associated with a prediction in addition to determining the most likely sequence. In this paper, we consider a Monte Carlo framework to estimate probabilities and confidence intervals associated with sequences. The framework uses a Monte Carlo simulator, implemented as an autoregressively trained neural network, to sample sequences conditioned on an image input. We then use thes… · 篇幅：Regular submission (no more than 12 pages of main content) · 代码：https://github.com/qy707/Deep_Probability · OpenReview ID：sJE59flFC1Qidong Yang, Weicheng Zhu, Joseph Keslin et al.
- 38When Does Multimodality Lead to Better Time Series Forecasting?期刊：Transactions on Machine Learning Research · 摘要：Recently, there has been growing interest in incorporating textual information into foundation models for time series forecasting. However, it remains unclear whether and under what conditions such multimodal text integration consistently yields gains. We systematically investigate these questions across a diverse benchmark of 16 forecasting tasks spanning 7 domains, including health, environment, and economics. We evaluate two popular multimodal forecasting paradigms: aligning-based methods, which align time series and text representations; and prompting-based methods, which directly prompt… · 篇幅：Regular submission (no more than 12 pages of main content) · OpenReview ID：RggcWYWR3NXiyuan Zhang, Boran Han, Haoyang Fang et al.
- 39Low-Rank Filtering & Smoothing for Sequential Deep Learning期刊：Transactions on Machine Learning Research · 摘要：Learning multiple tasks sequentially requires neural networks to balance retaining knowledge, yet being flexible enough to adapt to new tasks. Regularizing network parameters is a common approach, but it rarely incorporates prior knowledge about task relationships, and limits information flow to future tasks only. We propose a Bayesian framework that treats the network's parameters as the state space of a nonlinear Gaussian model, unlocking two key capabilities: (1) A principled way to encode domain knowledge about task relationships, allowing, e.g., control over which layers should adapt bet… · 篇幅：Regular submission (no more than 12 pages of main content) · OpenReview ID：1TJXpLHLKGJoanna Sliwa, Frank Schneider, Nathanael Bosch et al.
- 40Hierarchy-Aware Multimodal Unlearning for Medical AI期刊：Transactions on Machine Learning Research · 摘要：Multimodal large language models (MLLMs) are increasingly used in sensitive domains such as medical AI, where privacy regulations, including HIPAA and GDPR, require the removal of specific individuals' or institutions' data. This motivates machine unlearning, which aims to remove the influence of target data from a trained model. However, existing unlearning benchmarks fail to reflect the hierarchical and multimodal structure of real-world medical data, limiting their ability to properly evaluate unlearning in practice. Therefore, we introduce MedForget, a hierarchy-aware multimodal unlearnin… · 篇幅：Regular submission (no more than 12 pages of main content) · OpenReview ID：TVSIhLqIkfFengli Wu, Vaidehi Patil, Jaehong Yoon et al.
- 41Data-Dependent Regret and Polyak Corrections for Constrained Online Convex Optimization期刊：Transactions on Machine Learning Research · 摘要：In constrained online convex optimization, the learner must minimize regret against adversarially chosen convex costs while satisfying a convex constraint at every round, a requirement that arises naturally in safety-critical domains such as power systems, autonomous control, and clinical decision-making. A natural and computationally efficient approach augments online gradient descent with a Polyak feasibility step: a closed-form half-space projection requiring only one constraint evaluation and one subgradient per round. This approach is known to achieve $O(\sqrt{T})$ regret with per-round… · 篇幅：Regular submission (no more than 12 pages of main content) · OpenReview ID：tBM4UrAPf8Wentao Zhang
- 42LENS: Learning to Navigate with Active Search for Partially Observable MAPF in Unknown Environments期刊：Transactions on Machine Learning Research · 摘要：Classical Multi-Agent Path Finding (MAPF) solvers guarantee collision-free coordination but rely on perfect global knowledge, limiting their applicability in strictly unknown environments. Consequently, modern learning-based approaches face a dichotomy: decentralized reactive heuristics scale under partial observability but fail at structured deadlocks due to limited horizons and weak interaction inductive biases, while neural foundation models (e.g., MAPF-GPT) provide topological awareness but require pre-computed global heuristics and prohibitive training data. We address Centralized Collab… · 篇幅：Regular submission (no more than 12 pages of main content) · OpenReview ID：DuNU6ZN5GGDi Yang, Yuheng Li, Yanhai Xiong
- 43Hallucination Detection and Mitigation with Diffusion in Multi-Variate Time-Series Foundation Models期刊：Transactions on Machine Learning Research · 摘要：Foundation models (FMs) for natural language processing have many coherent definitions of hallucination and methods for its detection and mitigation. However, analogous definitions and methods do not exist for multi-variate time-series (MVTS) FMs. We propose new definitions for MVTS hallucination, along with new detection and mitigation methods using a diffusion model to estimate hallucination levels. We derive relational datasets from popular time-series datasets to benchmark these relational hallucination levels. Using these definitions and models, we find that open-source pre-trained MVTS… · 篇幅：Regular submission (no more than 12 pages of main content) · 代码：https://github.com/vijja-w/mvts-rhallu · OpenReview ID：fHGQ7hZlb5Vijja Wichitwechkarn, Charles Fox, R. Choudhary
- 44Conditional Risk-Averse Constrained Reinforcement Learning期刊：Transactions on Machine Learning Research · 摘要：In Risk-averse Constrained Reinforcement Learning (RaCRL), the optimal tolerance for risk often depends on a preference over the trade-off between reward and safety. This trade-off is influenced by environmental uncertainty, which is generally difficult to quantify, in turn making its effect on an agent's performance difficult to predict at the outset of training. Conventional RaCRL approaches typically train agents under a fixed risk level, set at the beginning of training, leading to an agent with a fixed, often conservative, reward-safety trade-off at deployment time. In this paper, we int… · 篇幅：Regular submission (no more than 12 pages of main content) · OpenReview ID：JJFFx1HVHiJames McCarthy, Radu Marinescu, Elizabeth M. Daly et al.
- 45Exposing Long-Tail Safety Failures in Large Language Models through Efficient Diverse Response Sampling期刊：Transactions on Machine Learning Research · 摘要：Safety tuning through supervised fine-tuning and reinforcement learning from human feedback has substantially improved the robustness of large language models (LLMs). However, it typically suppresses rather than eliminates unsafe behaviors, leaving rare but critical failures hidden in the long tail of the output distribution. While most red-teaming work emphasizes adversarial prompt search (input-space search), we show that these hidden risks can be systematically exposed through diverse response generation (output-space search). Specifically, we show that, for a fixed safety-critical prompt,… · 篇幅：Long submission (more than 12 pages of main content) · 代码：https://github.com/PalGitts/PDPS · OpenReview ID：tHfAskovWISuvadeep Hajra, Palash Nandi, Tanmoy Chakraborty
- 46Aggregation-Free Heterogeneous Federated Learning with Data-Free Knowledge Exchange期刊：Transactions on Machine Learning Research · 摘要：Heterogeneous Federated Learning (HFL) is a decentralized machine learning paradigm that enables participants to leverage distributed knowledge from diversified environments while safeguarding individual privacy. Recent works that address both data and model heterogeneity still require aggregating model parameters, which restricts architectural flexibility. Knowledge Distillation (KD) has been adopted in HFL to circumvent direct model aggregation by aggregating knowledge, but it depends on a public dataset and may incur information loss when redistributing knowledge from the global model. We… · 篇幅：Regular submission (no more than 12 pages of main content) · 代码：https://github.com/HaowenGuan/DFFKE · OpenReview ID：BLQ9JulhQmHaowen Guan, Xuan Zhao, Chengjie Zheng et al.
- 47Regularity and Stability Properties of Selective SSMs with Discontinuous Gating期刊：Transactions on Machine Learning Research · 摘要：Selective State-Space Models (SSMs) such as Mamba have become central to long-sequence modeling. Still, their stability is poorly understood: their state-space coefficients are modulated online by a token-dependent gating signal, making the recurrence neither linear time-invariant nor classically nonlinear. We study continuous-time selective SSMs through passivity, dissipativity, and Input-to-State Stability (ISS), explicitly separating the selection signal $x(\cdot)$ from the driving input $u(\cdot)$. We obtain four results: exponential forgetting under strict dissipativity; a canonical $\ma… · 篇幅：Long submission (more than 12 pages of main content) · OpenReview ID：7Vav53cDeNNikola Zubic, Davide Scaramuzza
- 48Progressive Checkerboards for Autoregressive Multiscale Image Generation期刊：Transactions on Machine Learning Research · 摘要：A key challenge in autoregressive image generation is to efficiently sample independent locations in parallel, while still modeling mutual dependencies with serial conditioning. Some recent works have addressed this by conditioning between scales in a multiscale pyramid. Others have looked at parallelizing samples in a single image using regular partitions or randomized orders. In this work we examine a flexible, fixed ordering based on progressive checkerboards for multiscale autoregressive image generation. Our ordering draws samples in parallel from evenly spaced regions at each scale, mai… · 篇幅：Regular submission (no more than 12 pages of main content) · 代码：https://github.com/deigen/checkerboardgen · OpenReview ID：wkCTCXo1a9David Eigen
- 49Lens: A Knowledge-Guided Foundation Model for Network Traffic期刊：Transactions on Machine Learning Research · 摘要：Network traffic refers to the amount of data being sent and received over the Internet or any system that connects computers. Analyzing network traffic is vital for security and management, yet remains challenging due to the heterogeneity of plain-text packet headers and encrypted payloads. To capture the latent semantics of traffic, recent studies have adopted Transformer-based pretraining techniques to learn network representations from massive traffic data. However, these methods pre-train on data-driven tasks but overlook network knowledge, such as masking partial digits of the indivisibl… · 篇幅：Regular submission (no more than 12 pages of main content) · 代码：https://github.com/V-Enzo/Lens-foundation-model-for-network-traffic · OpenReview ID：cGDwTgnJIRXiaochang Li, Chen Qian, Qineng Wang et al.
- 50Interpretable factorization of clinical questionnaires to identify latent factors of psychopathology期刊：Transactions on Machine Learning Research · 摘要：Psychiatry research seeks to understand the manifestations of psychopathology in behavior, as measured in questionnaire data, by identifying a small number of latent factors that explain them. While factor analysis is the canonical tool for this purpose, the resulting factors may not be interpretable, and may also be subject to confounding variables. Moreover, missing data are common, and explicit imputation is often required. To overcome these limitations, we introduce Interpretability Constrained Questionnaire Factorization (ICQF), a non-negative matrix factorization method with regularizat… · 篇幅：Regular submission (no more than 12 pages of main content) · 代码：https://github.com/jefferykclam/ICQF · OpenReview ID：1Yq6INJwiOKa Chun Lam, Francisco Pereira, Armin Raznahan et al.
共 50 条热搜查看完整榜单