Paper Digest: ICML 2026 Papers & Highlights
Browse all 6,500+ ICML 2026 papers, each with a one-sentence highlight. Search them, generate a literature review, and compare papers side by side.
The International Conference on Machine Learning (ICML) is one of the top machine learning conferences in the world. In 2026, it is to be held in Seoul. To help the community quickly catch up on the work presented in this conference, Paper Digest Team processed all accepted papers and generated one highlight sentence (typically the main topic) for each paper, so readers can quickly get the main idea of each paper. Beyond this list, every paper is positioned in our knowledge graph: users can find related papers, patents, grants and experts for any paper, and research or review a topic using all papers in the proceedings.
Search within ICML-2026
Literature review on a topic
Generate a written review of ICML-2026 research on any topic, with each claim cited to specific papers.
Browse & explore
Browse ~ 25,000 authors (ICML-2026), or explore the “Best Paper” Digest listing the most influential ICML papers of recent years.
Note: ICML-2026 accepts more than 6,500 papers; this page includes only 500 of them selected by our daily paper digest algorithm. Interested users can choose to read All 6,500 ICML-2026 papers in a separate page, which takes quite some time to load. Interested users can also compare papers side by side from multiple perspectives in our Literature Review Matrix; the comparison uses the full paper text when available.
Since 2018, Paper Digest has built a foundation of data spanning decades of conferences, journals, and research topics. The platform features a daily digest service that sifts through tens of thousands of new papers, clinical trials, news articles, and community posts, filtering the noise to highlight what matters most to specific interests. Beyond daily updates, dozens of built-in research tools streamline the academic workflow, supporting efficient reading and writing, comprehensive literature reviews, and automated research report generation.
Paper Digest Team
New York City, New York, 10017
team@paperdigest.org
TABLE 1: Paper Digest: ICML 2026 Papers & Highlights
| Paper | Author(s) | |
|---|---|---|
| 1 | TabICooL: A Better, Faster, Scalable, and Open Tabular Foundation Model Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We introduce TabICooL, a new state-of-the-art foundation model for regression and classification built on three pillars: (1) a novel synthetic data generation engine designed for high pretraining diversity; (2) various architectural innovations, including a new scalable softmax in attention improving generalization to larger datasets without prohibitive long-sequence pretraining; and (3) optimized pretraining protocols, notably replacing AdamW with the Muon optimizer. |
Jingang QU; David Holzmüller; Gael Varoquaux; Marine Le Morvan; |
| 2 | Unified Multimodal Autoregressive Modeling with Shared Context—Visual Tokenizer Is Key to Unification Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: However, existing approaches typically rely on two disparate visual tokenizers, which splits the representation space and hinder truly unified modeling. We propose UniAR, a unified autoregressive framework where a single discrete visual tokenizer serves as the key bridge between understanding and generation, enabling a shared context in which the model can directly interpret its own generated visual tokens without additional re-encoding. |
Wujian Peng; Lingchen Meng; Yuxuan Cai; Xianwei Zhuang; Yuhuan Yang; Rongyao Fang; Chenfei Wu; Junyang Lin; Zuxuan Wu; Shuai Bai; |
| 3 | PlotCraft: Pushing The Limits of LLMs for Complex and Interactive Data Visualization Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: However, their ability to create complex visualizations for scaled and structured data remains largely unevaluated and underdeveloped. To address this gap, we introduce **PlotCraft**, a new benchmark featuring 1k challenging visualization tasks that cover a wide range of topics, such as finance, scientific research, and sociology. |
Jiajun Zhang; Jianke Zhang; Zeyu Cui; Jiaxi Yang; Lei Zhang; Zilei Wang; Qiang Liu; Liang Wang; Binyuan Hui; Junyang Lin; |
| 4 | Revealing Behavioral Plasticity in Large Language Models: A Token-Conditional Perspective Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: In this work, we reveal that Large Language Models (LLMs) possess intrinsic behavioral plasticity—akin to chameleons adapting their coloration to environmental cues—that can be *exposed* through token-conditional generation and *stabilized* via reinforcement learning. |
Liyuan Mao; Le Yu; Jing Zhou; Chujie Zheng; Bowen Yu; Chang Gao; Shixuan Liu; An Yang; Weinan Zhang; Junyang Lin; |
| 5 | SWE-Bench Pro: Can AI Agents Solve Long-Horizon Software Engineering Tasks? Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We present SWE-Bench Pro, a comprehensive benchmark designed to evaluate software engineering capabilities through complex, realistic programming challenges. |
Xiang Deng; Jeff Da; Edwin Pan; Yannis He; Charles Ide; Kanak Garg; Niklas Lauffer; Andrew Park; Chetan Rane; Karmini Sampath; Maya Krishnan; Srivatsa Kundurthy; Sean Hendryx; Zifan Wang; Chen Bo Calvin Zhang; Noah Jacobson; Bing Liu; Brad Kenstler; |
| 6 | Set Diffusion: Interpolating Token Orderings Between Autoregression and Diffusion for Fast and Flexible Decoding Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Our key insight is that interpolating generation orderings between autoregression and fully-random decoding, rather than committing to a fixed block length, offers a better interpolation between diffusion and AR. |
Marianne Arriola; Volodymyr Kuleshov; |
| 7 | Monitoring Monitorability Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We propose three evaluation archetypes (intervention, process, and outcome-property), a new monitorability metric, and a broad evaluation suite. |
Melody Guan; Miles Wang; Micah Carroll; Zehao Dou; Annie Wei; Marcus Williams; Benjamin Arnav; Joost Huizinga; Ian Kivlichan; Amelia Glaese; Jakub Pachocki; Bowen Baker; |
| 8 | $\tau$-Voice: Benchmarking Full-Duplex Voice Agents on Real-World Domains Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We introduce $\tau$-voice, a benchmark for evaluating voice agents on grounded tasks with real-world complexity: agents must navigate complex multi-turn conversations, adhere to domain policies, and interact with the environment. |
Soham Ray; Keshav Dhandhania; Victor Barres; Karthik Narasimhan; |
| 9 | $\tau^2$-Bench: Evaluating Conversational Agents in A Dual-Control Environment Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: This differs from real-world scenarios like technical support, where users need to actively participate in modifying the state of the (shared) world. In order to address this gap, we introduce $\tau^2$-bench, with four key contributions: 1. A novel **Telecom dual-control domain** modeled as a Dec-POMDP, where both agent and user make use of tools to act in a shared, dynamic environment that tests both agent coordination and communication, 2. |
Victor Barres; Honghua Dong; Soham Ray; Xujie Si; Karthik Narasimhan; |
| 10 | Entropy-Aware On-Policy Distillation of Language Models Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: However, we show that the mode-seeking property of reverse KL reduces generation diversity and yields unstable learning signals when the teacher distribution has high entropy. To address this, we introduce Entropy-Aware On-Policy Distillation. |
Woogyeol Jin; Taywon Min; Yongjin Yang; Swanand Kadhe; Yi Zhou; Dennis Wei; Nathalie Baracaldo; Kimin Lee; |
| 11 | Self-Distilled Reasoner: On-Policy Self-Distillation for Large Language Models Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: However, on-policy distillation typically requires a separate, often larger, teacher LLM and does not explicitly leverage ground-truth solutions available in reasoning datasets. Inspired by the intuition that a sufficiently capable LLM can rationalize external privileged reasoning traces and teach its weaker self (i.e., the version without access to privileged information), we introduce On-Policy Self-Distillation (OPSD), a framework where a single model acts as both teacher and student by conditioning on different contexts. |
Siyan Zhao; Zhihui Xie; Mengchen Liu; Jing Huang; Guan Pang; Feiyu Chen; Aditya Grover; |
| 12 | Towards Execution-Grounded Automated AI Research Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We analyze two methods to learn from the execution feedback: evolutionary search and reinforcement learning. |
Chenglei Si; Zitong Yang; Yejin Choi; Emmanuel J Candes; Diyi Yang; Tatsunori Hashimoto; |
| 13 | Self-Distillation Enables Continual Learning Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We introduce Self-Distillation Fine-Tuning (SDFT), a simple method that enables on-policy learning directly from demonstrations. |
Idan Shenfeld; Mehul Damani; Jonas Hübotter; Pulkit Agrawal; |
| 14 | Learning to Discover at Test Time Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: This form of continual learning is quite special, because its goal is to produce one great solution rather than many good ones on average, and to solve this very problem rather than generalize to other problems. Therefore, our learning objective and search subroutine are designed to prioritize the most promising solutions. |
Mert Yuksekgonul; Daniel Koceja; Xinhao Li; Federico Bianchi; Jed McCaleb; Xiaolong Wang; Jan Kautz; Yejin Choi; James Zou; Carlos Guestrin; Yu Sun; |
| 15 | Reinforcement Learning with Evolving Rubrics for Deep Research Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Deep research agents perform multi-step research to produce long-form, well-attributed answers. |
Rulin Shao; Akari Asai; Shannon Shen; Hamish Ivison; Varsha Kishore; Jingming Zhuo; Xinran Zhao; Molly Park; Samuel Finlayson; David Sontag; Tyler Murray; Sewon Min; Pradeep Dasigi; Luca Soldaini; Faeze Brahman; Scott Yih; Sherry Tongshuang Wu; Luke Zettlemoyer; Yoon Kim; Hannaneh Hajishirzi; Pang Wei Koh; |
| 16 | GDPO: Group Reward-Decoupled Normalization Policy Optimization for Multi-reward RL Optimization Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: However, recent work has defaulted to apply Group Relative Policy Optimization (GRPO) under multi-reward setting without examining its suitability. In this paper, we demonstrate that directly applying GRPO to normalize distinct rollout reward combinations causes them to collapse into identical advantage values, reducing the resolution of the training signal and resulting in suboptimal convergence and, in some cases, early training failure. |
Shih-Yang Liu; Xin Dong; Ximing Lu; Shizhe Diao; Peter Belcak; Mingjie Liu; Min-Hung Chen; Hongxu Yin; Yu-Chiang Wang; Kwang-Ting Cheng; Yejin Choi; Jan Kautz; Pavlo Molchanov; |
| 17 | Spurious Rewards: Rethinking Training Signals in RLVR Related Papers Related Patents Related Grants Related Venues Related Experts Related Code View Save Highlight: We show that reinforcement learning with verifiable rewards (RLVR) can elicit strong mathematical reasoning in certain language models even with spurious rewards that have little, no, or outright negative correlation with the correct answer. |
Rulin Shao; Stella Li; Rui Xin; Scott Geng; Yiping Wang; Sewoong Oh; Simon Du; Nathan Lambert; Sewon Min; Ranjay Krishna; Yulia Tsvetkov; Hannaneh Hajishirzi; Pang Wei Koh; Luke Zettlemoyer; |
| 18 | LOCA-bench: Benchmarking Language Agents Under Controllable and Extreme Context Growth Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: In realistic scenarios, however, LLMs often need to act as agents that explore environments, follow instructions and plans, extract useful information, and predict correct actions under a dynamically growing context. To assess language agents in such settings, we introduce LOCA-bench (a benchmark for **LO**ng-**C**ontext **A**gents). |
Weihao Zeng; Yuzhen Huang; Junxian He; |
| 19 | Experience Augmented Policy Optimization for LLM Reasoning Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: In this work, we propose Experience-Augmented Policy Optimization (EAPO), which leverages a prior RL-optimized policy as an action-level experience prior and selectively injects experience at critical decision points during rollout. |
Jinda Lu; Kexin Huang; Junkang Wu; Shuo Yang; Jinghan Li; Chiyu Ma; Shaohang Wei; Xiang Wang; Guoyin Wang; Jingren Zhou; |
| 20 | Clipping Bottleneck: Stabilizing RLVR Via Stochastic Recovery of Near-Boundary Signals Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Specifically, we find that many high-value signals lie in the **near-boundary** region just beyond the clipping threshold, and are thus discarded. Motivated by this diagnosis, we propose **Near-boundary Stochastic Rescue (NSR)**, a minimal, plug-and-play modification that stochastically retains these slightly out-of-bound tokens to recover lost signals. |
Shuo Yang; Jinda Lu; Chiyu Ma; Kexin Huang; Haoming Meng; Qihui Zhang; Yuyang Liu; Bolin Ding; Guoyin Wang; Li Yuan; Jingren Zhou; |
| 21 | Posterior Behavioral Cloning: Pretraining BC Policies for Efficient RL Finetuning Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: This finetuning step has proved critical in achieving human or super-human performance, yet while much attention has been given to developing more effective finetuning algorithms, little attention has been given to ensuring the pretrained policy is an effective initialization for RL finetuning. In this work we seek to understand how the pretrained policy affects finetuning performance, and how to pretrain policies in order to ensure they are effective initializations for finetuning. |
Andrew Wagenmaker; Perry Dong; Raymond Tsao; Chelsea Finn; Sergey Levine; |
| 22 | Scaling Prompt Synthesis for Large Language Model Reasoning Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: PromptCoT showed that injecting rationales into prompt synthesis increases problem difficulty. Building on this, we present PromptScale, a scalable framework that replaces hand-crafted heuristics with an expectation-maximization (EM) loop, where rationales are iteratively refined to guide prompt construction. |
Xueliang Zhao; Wei Wu; Jian Guan; Zhuocheng Gong; Lingpeng Kong; |
| 23 | Proxy Compression for Language Modeling Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: This work introduces proxy compression, an alternative training scheme that preserves the efficiency benefits of compressed inputs while providing an end-to-end, raw-byte interface at inference time. |
Lin Zheng; Li Xinyu; Qian Liu; Xiachong Feng; Lingpeng Kong; |
| 24 | Anchoring Self-Play for Code Repair Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We aim to scale supervision for code repair by having an LM generate bug–fix tasks with unconstrained edits, using unit tests as the only verifier. |
Caroline Choi; Zeyneb Kaya; Shirley Wu; Tengyu Ma; Tatsunori Hashimoto; Ludwig Schmidt; |
| 25 | The Assistant Axis: Situating and Stabilizing The Default Persona of Language Models Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Large language models can represent a variety of personas but typically default to a helpful Assistant identity cultivated during post-training. Across several different models, we find an “Assistant Axis in their activation space, which captures the extent to which a model is operating in its default Assistant mode. |
Christina Lu; Jack Gallagher; Jonathan Michala; Kyle Fish; Jack Lindsey; |
| 26 | Optimal Classical and Quantum Algorithms for Gradient Testing and Estimation By Comparisons Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: For any smooth $f\colon\mathbb R^n\to\mathbb R$, $\mathbf{x}\in\mathbb R^n$, and $\varepsilon>0$, we design a gradient testing algorithm that determines whether the normalized gradient $\nabla f(\mathbf{x})/||\nabla f(\mathbf{x})||$ is $\varepsilon$-close or $2\varepsilon$-far from a given unit vector $\mathbf{v}$ using $O(n)$ queries, as well as a gradient estimation algorithm that outputs an $\varepsilon$-estimate of $\nabla f(\mathbf{x})/||\nabla f(\mathbf{x})||$ using $O(n\log(1/\varepsilon))$ queries. |
Xiwen Tao; Chenyi Zhang; Helin Wang; Yexin Zhang; Tongyang Li; |
| 27 | Understanding Reasoning Collapse in LLM Agent Reinforcement Learning Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We thus provide a signal-to-noise ratio explanation for why $I(X; Z)$ drops: when within-input reward variance $\mathrm{Var}(R \mid X)$ is low, task gradients weaken and input-agnostic regularizers (KL, entropy) dominate, flattening cross-input differences. |
Zihan (Zenus) Wang; Chi Gui; Xing Jin; Qineng Wang; Licheng Liu; Kangrui Wang; Shiqi Chen; Linjie Li; Zhengyuan Yang; Pingyue Zhang; Yiping Lu; Jiajun Wu; Li Fei-Fei; Lijuan Wang; Yejin Choi; Manling Li; |
| 28 | PACER: Acyclic Causal Discovery from Large-scale Interventional Data Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We introduce PACER (Perturbation-driven Acyclic Causal Edge Recovery), a scalable framework for causal discovery that guarantees acyclicity by construction. |
Ramon Viñas Torné; Sílvia Salazar; Soyon Park; Ivo Ban; Artyom Gadetsky; Nikita Doikov; Maria Brbic; |
| 29 | VLAW: Iterative Co-Improvement of Vision-Language-Action Policy and World Model Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: The goal of this paper is to improve the performance and reliability of vision-language-action (VLA) models through iterative online interaction. |
Yanjiang Guo; Tony Lee; Lucy Xiaoyang Shi; Jianyu Chen; Percy Liang; Chelsea Finn; |
| 30 | Implicit Intelligence – Evaluating Agents on What Users Don’t Say Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We present **Implicit Intelligence**, an evaluation framework testing whether AI agents can move beyond prompt-following to become genuine goal-fulfillers, paired with **Agent-as-a-World (AaW)**, a harness where interactive worlds are defined in human-readable YAML files and simulated by language models. |
Ved Sirdeshmukh; Marc Wetter; |
| 31 | Self-Supervised Flow Matching for Scalable Multi-Modal Synthesis Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We introduce *Self-Flow*: a self-supervised flow matching paradigm that integrates representation learning within the generative framework. |
Hila Chefer; Patrick Esser; Dominik Lorenz; Dustin Podell; Vikash Raja; Vinh Tong; Antonio Torralba; Robin Rombach; |
| 32 | SWE-rebench V2: Language-Agnostic SWE Task Collection at Scale Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We introduce SWE-rebench V2, a language-agnostic automated pipeline for harvesting executable real-world SWE tasks and constructing RL training environments at scale. |
Ibragim Badertdinov; Maksim Nekrashevich; Anton Shevtsov; Aleksandr Golubev; |
| 33 | Extracting Alignment Data in Open Models Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: In this work, we show that it is possible to extract significant amounts of alignment training data from a post-trained model — useful to steer the model to improve certain capabilities such as long-context reasoning, safety, instruction following, and maths. |
Federico Barbero; Xiangming Gu; Christopher A. Choquette Choo; Chawin Sitawarin; Matthew Jagielski; Itay Yona; Petar Veličković; Ilia Shumailov; Jamie Hayes; |
| 34 | ToolOrchestra: Elevating Intelligence Via Efficient Model and Tool Orchestration Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We introduce ToolOrchestra, a method for training small orchestrators that coordinate the use of intelligent tools. |
Hongjin SU; Shizhe Diao; Ximing Lu; Mingjie Liu; Jiacheng Xu; Xin Dong; Yonggan Fu; Peter Belcak; Hanrong Ye; Hongxu Yin; Yi Dong; Evelina Bakhturina; Tao Yu; Yejin Choi; Jan Kautz; Pavlo Molchanov; |
| 35 | MmBERT: A Modern Multilingual Encoder with Annealed Language Learning Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We introduce mmBERT, an encoder-only language model pretrained on 3T tokens of multilingual text in over 1800 languages. |
Marc Marone; Orion Weller; William Fleshman; Eugene Yang; Dawn Lawrie; Benjamin Van Durme; |
| 36 | Position: The Age of AI Agents Demands A New Scientific Paradigm To Sustain Trustworthy Science Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We propose criteria for an adapted verification infrastructure that emphasizes observable-by-default workflows, scalable verification, and clear attribution. |
Belinda Mo; |
| 37 | Toward Training Superintelligent Software Agents Through Self-Play SWE-RL Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: In this paper, we present Self-play SWE-RL (SSR), a first step toward training superintelligent software agents under minimal data assumptions. |
Yuxiang Wei; Zhiqing Sun; Emily McMilin; Jonas Gehring; David Zhang; Gabriel Synnaeve; Daniel Fried; LINGMING ZHANG; Sida Wang; |
| 38 | Mode Seeking Meets Mean Seeking for Long Video Generation Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: While multi-resolution image training works because higher resolution is largely an interpolation of the same underlying patch distribution, training across video lengths is fundamentally different: a longer video is an extrapolation that must invent new events and causal structure beyond the short-clip horizon. To address this, we propose a training paradigm where Mode Seeking meets Mean Seeking, decoupling local fidelity from long-term coherence from a unified representation via a Decoupled Diffusion Transformer. |
Shengqu Cai; Weili Nie; Chao Liu; Julius Berner; Lvmin Zhang; Nanye Ma; Hansheng Chen; Maneesh Agrawala; Leonidas Guibas; Gordon Wetzstein; Arash Vahdat; |
| 39 | PostTrainBench: Can LLM Agents Automate LLM Post-Training? Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: In this paper, we study *post-training*, which is the critical step that turns base LLMs into useful assistants. |
Ben Rank; Hardik Bhatnagar; Ameya Pandurang Prabhu; Shira Eisenberg; Karina Nguyen; Matthias Bethge; Maksym Andriushchenko; |
| 40 | D2: Improved Techniques for Training Reasoning Diffusion Language Models Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Here, we introduce d2, a reasoning framework tailored for masked DLMs. |
Guanghan Wang; Gilad Turok; Yair Schiff; Marianne Arriola; Volodymyr Kuleshov; |
| 41 | DeepAnalyze: Agentic Large Language Models for Autonomous Data Science Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: In this paper, we introduce DeepAnalyze, the first agentic LLM for autonomous data science, capable of automatically completing the end-to-end data science from structured data to analyst-grade research reports. |
Shaolei Zhang; Ju Fan; Meihao Fan; Yizhe Liu; Yuxin Zhang; Xiaoyong Du; |
| 42 | SSA: Sparse Sparse Attention By Aligning Full and Sparse Attention Outputs in Feature Space Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We propose SSA (Sparse Sparse Attention), a training framework that integrates both sparse and full attention with bidirectional attention-output alignment. |
Zhenyi Shen; Junru Lu; Lin Gui; Jiazheng Li; Yulan He; di yin; Xing Sun; |
| 43 | Stabilizing MoE Reinforcement Learning By Aligning Training and Inference Routers Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We analyze the training-inference consistency of MoE models and identify a notable discrepancy in routing behaviors between the two phases. To address this issue, we propose \textbf{Rollout Routing Replay (R3)}, a novel and effective method that records routing distributions from the inference engine and replays them during training. |
Wenhan Ma; Hailin Zhang; Liang Zhao; Yifan Song; Yudong Wang; Fuli Luo; Zhifang Sui; |
| 44 | Reasoning Models Struggle to Control Their Chains of Thought Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: This capability — CoT controllability — is undesirable because it could allow models to suppress signs of misbehavior in their CoT, thereby undermining our ability to monitor them. To measure this, we introduce the \emph{CoT-Control} evaluation suite. |
Chen Yueh-Han; Robert McCarthy; Bruce W. Lee; He He; Micah Carroll; Tomasz Korbak; |
| 45 | Solving Physics Olympiad Via Reinforcement Learning on Physics Simulators Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: In this work, we show that physics simulators can serve as a powerful alternative source of supervision for training LLMs for physical reasoning. |
Mihir Prabhudesai; Aryan Satpathy; Yangmin Li; Zheyang Qin; Nikash Bhardwaj; Amir Zadeh; Chuan Li; Katerina Fragkiadaki; Deepak Pathak; |
| 46 | ThetaEvolve: Test-time Learning on Open Problems Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We introduce ThetaEvolve, an open-source framework that simplifies and extends AlphaEvolve to efficiently scale both in-context learning and Reinforcement Learning (RL) at test time, allowing models to continually learn from their experiences in improving open optimization problems. |
Yiping Wang; Shao-Rong Su; Zhiyuan Zeng; Eva Xu; Liliang Ren; Xinyu Yang; Zeyi Huang; Xuehai He; Luyao Ma; Baolin Peng; Hao Cheng; Pengcheng He; Weizhu Chen; Shuohang Wang; Simon Du; Yelong Shen; |
| 47 | DF-ExpEnse: Diffusion Filtered Exploration for Sample Efficient Finetuning Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: In this work we present Diffusion Filtered Exploration via Ensembles (DF-ExpEnse), an exploration technique that meaningfully improves the quality of online experience collection, thus increasing the sample efficiency of the finetuning procedure. |
Calvin Luo; Chen Sun; shuran song; |
| 48 | Learning Latent Action World Models In The Wild Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: World models possess this capability but require action annotations that can be complex to obtain at scale. Latent action models address this issue by learning an action space from videos alone. |
Quentin Garrido; Tushar Nagarajan; Basile Terver; Nicolas Ballas; Yann LeCun; Michael Rabbat; |
| 49 | Reward Hacking Benchmark: Measuring Exploits in LLM Agents with Tool Use Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We introduce the Reward Hacking Benchmark (RHB), a suite of multi-step tasks requiring sequential tool operations with naturalistic shortcut opportunities such as skipping verification steps, inferring answers from task-adjacent metadata, or tampering with evaluation-relevant functions. |
Kunvar Thaman; |
| 50 | Reasoning About Reasoning: BAPO Bounds on Chain-of-Thought Token Complexity in LLMs Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We address a fundamental theoretical question: *how many* reasoning tokens are required to solve a problem as input size grows? |
Kiran Tomlinson; Tobias Schnabel; Adith Swaminathan; Jennifer Neville; |
| 51 | Any-Diffusion: Unified Multimodal Understanding and Generation with Masked Discrete Diffusion Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Meanwhile, recent studies have successfully applied discrete diffusion models to natural language processing, revealing their considerable potential as a promising new approach in this domain. Drawing inspiration from these pioneering researches, we introduce Any-Diffusion, the first any-to-any multimodal language model built purely on mask-based discrete diffusion models, which unifies understanding and generation across text, speech, and images. |
lijiang Li; zuwei long; Yunhang Shen; Heting Gao; Haoyu Cao; Xing Sun; Caifeng Shan; Ran He; Chaoyou Fu; |
| 52 | Training AI Co-Scientists Using Rubric Rewards Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: To validate this approach, we conduct a human study for machine learning research goals spanning 225 expert hours. |
Shashwat Goel; Rishi Hazra; Dulhan Jayalath; Timon Willi; Parag Jain; Shen; Ilias Leontiadis; Francesco Barbieri; Yoram Bachrach; Jonas Geiping; Chenxi Whitehouse; |
| 53 | Fast KV Compaction Via Attention Matching Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: This work describes an approach for *fast* context compaction in latent space through **Attention Matching**, which constructs compact keys and values to reproduce attention outputs and preserve attention mass at a per-KV-head level. |
Adam Zweiger; Xinghong Fu; Han Guo; Yoon Kim; |
| 54 | From Prior to Pro: Efficient Skill Mastering Via Distribution Contractive RL Finetuning Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We introduce Distribution Contractive Reinforcement Learning (DICE-RL), a framework that uses reinforcement learning (RL) as a “distribution contractor” to refine pretrained generative robot policies. |
Zhanyi Sun; shuran song; |
| 55 | DexMachina: Functional Retargeting for Bimanual Dexterous Manipulation Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We focus on long-horizon, bimanual tasks with articulated objects, which are challenging due to large action space, spatiotemporal discontinuities, and the embodiment gap between human and robot hands. We propose DexMachina, a novel curriculum-based algorithm: the key idea is to use virtual object controllers with decaying strength: an object is first driven automatically towards its target states, such that the policy can gradually learn to take over under motion and contact guidance. |
Zhao Mandi; Yifan Hou; Dieter Fox; Yashraj Narang; Ajay Mandlekar; shuran song; |
| 56 | Flex-Forcing: Towards A Unified Autoregressive and Bidirectional Video Diffusion Model Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We introduce Flex-Forcing, a unified training and inference framework that enables a video diffusion model to seamlessly operate under both bidirectional and autoregressive generation regimes. |
Xinyin Ma; Julius Berner; Chao Liu; Arash Vahdat; Weili Nie; Xinchao Wang; |
| 57 | Activation Oracles: Training and Evaluating LLMs As General-Purpose Activation Explainers Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: In this paper, we instead take a generalist perspective. |
Adam Karvonen; James Chua; Clément Dumas; Kit Fraser-Taliente; Subhash Kantamneni; Julian Minder; Euan Ong; Arnab Sen Sharma; Daniel Wen; Owain Evans; Samuel Marks; |
| 58 | Think Fast and Slow: Step-Level Cognitive Depth Adaptation for LLM Agents Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: In this paper, we introduce CogRouter, a framework that trains agents to dynamically adapt cognitive depth at each step. |
Ruihan Yang; Fanghua Ye; Xiang Wei; Ruoqing Zhao; Kang Luo; Xinbo Xu; Bo Zhao; Ruotian Ma; Shanyi Wang; Zhaopeng Tu; Xiaolong Li; Deqing Yang; Liefeng Bo; |
| 59 | Curating The Future: A Scalable Recipe for Training Open-Ended Forecasters Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: In this work, we train language models to make predictions on open-ended forecasting questions. |
Nikhil Chandak; Shashwat Goel; Ameya Pandurang Prabhu; Moritz Hardt; Jonas Geiping; |
| 60 | Base Models Know How to Reason, Thinking Models Learn When Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Why do thinking language models outperform their base counterparts, and what exactly do they learn during training? We introduce constructive model diffing, a framework for understanding fine-tuned models by explicitly constructing the base-to-fine-tuned difference from interpretable components to produce hybrid models, and measuring how well they recover the fine-tuned model’s performance. |
Constantin Venhoff; Iván Arcuschin; Phil Torr; Arthur Conmy; Neel Nanda; |
| 61 | Mem-T: Densifying Rewards for Long-Horizon Memory Agents Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: However, existing training paradigms remain constrained: agents often traverse long-horizon sequences of memory operations before receiving sparse and delayed rewards, which hinders truly end-to-end optimization of memory management policies. To address this limitation, we introduce Mem-T, an autonomous memory agent that interfaces with a lightweight hierarchical memory database to perform dynamic updates and multi-turn retrieval over streaming inputs. |
Yanwei Yue; Guibin Zhang; Boci Peng; Xuanbo Fan; Jiaxin Guo; Qiankun Li; Yan Zhang; |
| 62 | ModernVBERT: Towards Smaller Visual Document Retrievers Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Increasingly, Visual Document Retrieval (VDR) models, which directly embed images of document pages, are used as an alternative to text-only retrievers. |
Paul Teiletche; Quentin Macé; Max Conti; António Loison; Gautier Viaud; Pierre Colombo; Manuel Faysse; |
| 63 | Monitorability As A Free Gift: How RLVR Spontaneously Aligns Reasoning Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Recent work has reported that monitorability—the degree to which CoT faithfully and informatively reflects internal computation—can appear as a free gift during the early stages of Reinforcement Learning with Verifiable Rewards (RLVR). We make this observation concrete through a systematic evaluation across model families and training domains. |
Zidi Xiong; Shan Chen; Himabindu Lakkaraju; |
| 64 | Maximum Likelihood Reinforcement Learning Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We introduce **Maximum Likelihood Reinforcement Learning (MaxRL)**, a compute-indexed family of sampling-based objectives derived from a pass@k expansion of the likelihood, which interpolates between standard RL and exact maximum likelihood as compute increases. |
Fahim Tajwar; Guanning Zeng; Yueer Zhou; Yuda Song; Daman Arora; Yiding Jiang; Jeff Schneider; Russ Salakhutdinov; Haiwen Feng; Andrea Zanette; |
| 65 | Reinforcement Learning with Verifiable Rewards: GRPO’s Loss, Dynamics, and Success Amplification Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Group Relative Policy Optimization (GRPO) was introduced recently and used to train DeepSeek\textendash R1 for promoting reasoning in LLMs under verifiable (binary) rewards. We show that the mean{+}variance calibration of these rewards induces a contrastive loss in which the contrastive samples are synthetic data drawn from the previous policy. |
Youssef Mroueh; |
| 66 | Chasing Moving Targets with Online Self-Play Reinforcement Learning for Safer Language Models Related Papers Related Patents Related Grants Related Venues Related Experts Related Code View Save Highlight: This sequential setup leads to attackers overfitting obsolete exploits while defenders perpetually lag behind emerging threats. To address this, we introduce Self-RedTeam, the first fully online self-play multi-agent reinforcement learning (MARL) algorithm that continuously co-evolves attacker and defender for robust safety alignment. |
Mickel Liu; Liwei Jiang; Yancheng Liang; Simon Du; Yejin Choi; Tim Althoff; Natasha Jaques; |
| 67 | TQL: Scaling Q-Functions with Transformers By Preventing Attention Collapse Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: In this work, we ask what prevents transformers from scaling effectively for value functions? |
Perry Dong; Kuo-Han Hung; Alexander Swerdlow; Dorsa Sadigh; Chelsea Finn; |
| 68 | Automatically Finding Reward Model Biases Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: In this work, we introduce and study the research problem of automatically finding reward model biases in natural language. |
Atticus Wang; Iván Arcuschin; Arthur Conmy; |
| 69 | Steering Out-of-Distribution Generalization with Concept Ablation Fine-Tuning Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We introduce Concept Ablation Fine-Tuning (CAFT), a technique that leverages interpretability tools to control how LLMs generalize from fine-tuning, without needing to modify the training data or otherwise use data from the target distribution. |
Helena Casademunt; Caden Juang; Adam Karvonen; Samuel Marks; Senthooran Rajamanoharan; Neel Nanda; |
| 70 | CoCoQuant: Breaking The Bandwidth Wall Via Co-Optimized Communication and Computation Quantization Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Existing approaches typically treat communication and computation in isolation, failing to exploit their coupled nature and introducing limited system-level acceleration and accuracy degradation. To address this, we propose CoCoQuant, a co-designed framework that jointly optimizes communication and computation as a unified end-to-end design space. |
Haojie Duanmu; Jifeng Ding; Size Zheng; Xuegui Zheng; Jiangfei Duan; Xingcheng ZHANG; Li-Wen Chang; Xin Liu; Dahua Lin; |
| 71 | Interpreting Physics in Video World Models Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Here, we present the first interpretability study to directly examine physical representations inside large-scale video encoders. |
Sonia Joseph; Quentin Garrido; Randall Balestriero; Matthew Kowal; Thomas Fel; Shahab Bakhtiari; Blake Richards; Michael Rabbat; |
| 72 | HumanLM: Simulating Users with State Alignment Beats Response Imitation Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: However, existing user simulators mostly imitate surface-level patterns and language styles, which fails to reflect the underlying state of real users (e.g., beliefs, emotions). To address these limitations, we propose a novel training framework, HumanLM, which builds user simulators that accurately reflect real users. |
Shirley Wu; Evelyn Choi; Arpandeep Khatua; Zhanghan Wang; Joy He-Yueya; Cyril Weerasooriya; Wei Wei; Diyi Yang; Jure Leskovec; James Zou; |
| 73 | Demystifying Action Space Design for Robotic Manipulation Policies Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: While recent advances have focused heavily on scaling training data and model capacity, the choice of action space remains guided by ad-hoc heuristics or legacy designs, leading to an ambiguous understanding of robotic policy design philosophies. To address this ambiguity, we conducted a large-scale and systematic empirical study, confirming that the action space does have significant and complex impacts on robotic policy learning. |
Yuchun Feng; Jinliang Zheng; Zhihao Wang; Dongxiu Liu; Jianxiong Li; Jiangmiao Pang; Tai Wang; Xianyuan Zhan; |
| 74 | Any-Order GPT As Masked Diffusion Model: Decoupling Formulation and Architecture Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We show decoder-only MDMs, despite a larger modeling space, can achieve significant inference speedups ($\sim25\times$) and comparable perplexity with techniques like temperature annealing, offering a path to reduced inference compute. |
Shuchen Xue; Tianyu Xie; Tianyang Hu; Zijin Feng; Jiacheng Sun; Kenji Kawaguchi; Zhenguo Li; Zhi-Ming Ma; |
| 75 | Advantage Weighted Matching: Aligning RL with Pretraining in Diffusion Models Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: In this work, we establish a novel theoretical analysis: DDPO is an implicit form of score/flow matching with noisy targets, which increases variance and slows convergence. |
Shuchen Xue; Chongjian GE; Shilong Zhang; Yichen Li; Zhi-Ming Ma; |
| 76 | Benchmarking Reward Hack Detection in Code Environments Via Contrastive Analysis Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: In this paper, we propose a novel taxonomy of reward exploits spanning across 54 categories and introduce TRACE (Testing Reward Anomalies in Code Environments), a synthetically curated and human-verified benchmark containing 517 testing trajectories. |
Darshan Deshpande; Anand Kannappan; Rebecca Qian; |
| 77 | Real-Time Visual Attribution Streaming in Thinking Model Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We present an amortized framework for real-time visual attribution streaming in multimodal thinking models. |
Seil Kang; Woojung Han; Junhyeok Kim; Jinyeong Kim; Youngeun Kim; Seong Jae Hwang; |
| 78 | SceneSmith: Agentic Generation of Simulation-Ready Indoor Scenes Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We introduce SceneSmith, a hierarchical agentic framework that generates simulation-ready indoor environments from natural language prompts. |
Nicholas Pfaff; Thomas Cohn; Sergey Zakharov; Rick Cory; Russ Tedrake; |
| 79 | Reasoning Cache: Learning to Extrapolate to Long Lengths Via Short-Length RL Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Standard on-policy RL operates on fixed problem distributions and training budgets, giving rise to a distribution shift between train and test that limits the resulting model’s extrapolation capabilities. To address this, we introduce RC, an iterative decoding algorithm replacing standard autoregressive decoding that enables models to extrapolate to lengths an order of magnitude longer than those seen during training. |
Ian Wu; Yuxiao Qu; Amrith Setlur; Aviral Kumar; |
| 80 | Deep Forcing: Training-Free Long Video Generation with Deep Sink and Participative Compression Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We find that naïvely applying StreamingLLM-style attention sinks to video diffusion leads to fidelity degradation and motion stagnation. To overcome this, we introduce Deep Forcing, which consists of two training-free mechanisms that address this without any fine-tuning. |
Jung Yi; Wooseok Jang; Paul Cho; Jisu Nam; Heeji Yoon; Seungryong Kim; |
| 81 | How2Everything: Mining The Web for How-to Procedures to Evaluate and Improve LLMs Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Yet, measuring and improving procedural validity at scale on real-world tasks remains challenging and understudied. To address this, we introduce How2Everything, a scalable framework to evaluate and improve goal-conditioned procedure generation. |
Yapei Chang; Kyle Lo; Mohit Iyyer; Luca Soldaini; |
| 82 | Scaling Generative Verifiers For Natural Language Mathematical Proof Verification And Selection Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Large language models have achieved remarkable success on final-answer mathematical problems, largely due to the ease of applying reinforcement learning with verifiable rewards. |
Sadegh Mahdavi; Branislav Kisacanin; Shubham Toshniwal; Wei Du; Ivan Moshkov; George Armstrong; Renjie Liao; Christos Thrampoulidis; Igor Gitman; |
| 83 | PLANTAIN: Plan-Answer Interleaved Reasoning Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: In contrast, human speakers perform lightweight, incremental check-ins to ensure that conversational participants stay on common ground. With this motivation, we propose \textit{interleaved reasoning} (IR), in which the model alternates between thinking and surfacing intermediate responses, as an alternative to the standard “think-then-answer” approach. |
Anthony Liang; Jonathan Berant; Adam Fisch; Abhimanyu Goyal; Kalpesh Krishna; Jacob Eisenstein; |
| 84 | Constitutional Black-Box Monitoring for Scheming in LLM Agents Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We introduce two pipelines for generating synthetic agent trajectories, *STRIDE* (iterative refinement) and *Gloom* (agent-environment simulation), from which we generate 1,000 samples each. |
Simon Storf; Rich Barton-Cooper; James Peters-Gill; Marius Hobbhahn; |
| 85 | SIGMA-PPG: Statistical-prior Informed Generative Masking Architecture for PPG Foundation Model Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Standard masked modeling often yields trivial solutions while contrastive methods lack morphological precision. To address these limitations, we propose a Statistical-prior Informed Generative Masking Architecture (SIGMA-PPG), a generative foundation model featuring a prior-guided adversarial masking mechanism, where a reinforcement learning-driven teacher leverages statistical priors to create challenging learning paths that prevent overfitting to noise. |
ZONGHENG GUO; Tao Chen; Yang Jiao; Yi Pan; Xiao Hu; Manuela Ferrario; |
| 86 | Dr. Kernel: Reinforcement Learning Done Right for Triton Kernel Generations Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: In this paper, we systematically study reinforcement learning (RL) for kernel generation. |
Wei Liu; Jiawei Xu; Yingru Li; Longtao Zheng; Tianjian Li; Qian Liu; Junxian He; |
| 87 | Chain-of-Thought Reasoning In The Wild Is Not Always Faithful Related Papers Related Patents Related Grants Related Venues Related Experts Related Code View Save Highlight: In this work, we show that unfaithful CoT also occurs on naturally worded, non-adversarial prompts without adding artificial biases or editing model outputs. |
Iván Arcuschin; Jett Janiak; Robert Krzyzanowski; Senthooran Rajamanoharan; Neel Nanda; Arthur Conmy; |
| 88 | Adversarial Flow Models Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We present adversarial flow models, a class of generative models that belongs to both adversarial and flow families of models. |
Shanchuan Lin; Ceyuan Yang; Zhijie Lin; Hao Chen; Haoqi Fan; |
| 89 | Position: Evaluation of ML Resource Utilization Requires Model Life Cycle Assessment Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: With the growing complexity of pipelines and underlying infrastructure needed to develop and deploy AI systems, previous approaches for evaluating AI efficiency which focus on the costs of a single training run or an individual inference prediction are no longer sufficient. In this position paper, we enunciate the need for applying life cycle assessment to evaluate the costs of the machine learning model development and deployment pipeline to properly account for the required resources and downstream impact. |
Jared Fernandez; Clara Na; Yonatan Bisk; Constantine Samaras; Emma Strubell; |
| 90 | Random Scaling of Emergence Capabilities Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We propose that breakthroughs are instead driven by continuous changes in the *probability distribution* of training outcomes when performance is bimodally distributed across random seeds. |
Rosie Zhao; Tian Qin; David Alvarez-Melis; Sham Kakade; Naomi Saphra; |
| 91 | CodeClash: Benchmarking Goal-Oriented Software Engineering Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We introduce CodeClash, a benchmark where LMs compete in multi-round tournaments to build the best codebase for achieving a competitive objective. |
John Yang; Kilian Lieret; Joyce Yang; Carlos Jimenez; Muhtasham Oblokulov; Aryan Siddiqui; Ofir Press; Ludwig Schmidt; Diyi Yang; |
| 92 | What Does Flow-Matching Bring to TD-Learning? Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We show that their success is not explained by distributional RL: explicitly modeling return distributions often degrades performance. Instead, we argue that flow-matching Q-functions are effective because they couple a learned velocity field with an integration procedure that is used both during training and to read out Q-values at inference time. |
Bhavya Agrawalla; Michal Nauman; Aviral Kumar; |
| 93 | One Bias After Another: Mechanistic Reward Shaping and Persistent Biases in Language Reward Models Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We categorize RM failures by complexity and propose a simple post-hoc intervention to mitigate low-complexity biases that arise from spurious correlations. |
Daniel Fein; Max Lamparth; Violet Xiang; Mykel Kochenderfer; Nick Haber; |
| 94 | Who’s in Charge? Disempowerment Patterns in Real-World LLM Usage Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We present the first large-scale empirical analysis of disempowerment patterns in real-world AI assistant interactions, analyzing 1.5 million consumer Claude.ai conversations using a privacy-preserving approach. |
Mrinank Sharma; Miles McCain; Raymond Douglas; David Duvenaud; |
| 95 | Shrinking The Variance: Shrinkage Baselines for Reinforcement Learning with Verifiable Rewards Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Drawing inspiration from Stein’s paradox, we propose using \emph{shrinkage estimators} that combine \emph{per-prompt} and \emph{across-prompt} means to improve the overall per-prompt mean estimation accuracy—particularly in the low-generation regime typical of RLVR. |
Guanning Zeng; Zhaoyi Zhou; Daman Arora; Andrea Zanette; |
| 96 | Mechanisms of Introspective Awareness Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: But what mechanisms underlie this capability, and do they reflect genuine introspective circuitry or more shallow heuristics? We investigate these questions in open-source models and establish three main findings. |
Uzay Macar; Li Yang; Atticus Wang; Peter Wallich; Emmanuel Ameisen; Jack Lindsey; |
| 97 | CoF-T2I: Video Models As Pure Visual Reasoners for Text-to-Image Generation Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: However, their potential to enhance text-to-image (T2I) generation remains largely unexplored due to the absence of a clearly defined visual reasoning starting point and interpretable intermediate states in the T2I generation process. To bridge this gap, we propose **CoF-T2I**, a model that integrates CoF reasoning into T2I generation via progressive visual refinement, where intermediate frames act as explicit reasoning steps and the final frame is taken as output. |
Chengzhuo Tong; Chang Mingkun; Shenglong Zhang; Yuran Wang; Cheng Liang; Zhizheng Zhao; Bohan Zeng; Yang Shi; Ruichuan An; Yifan Dai; Ziming Zhao; Guanbin Li; Pengfei Wan; Yuanxing Zhang; Wentao Zhang; |
| 98 | Star Elastic: Many-in-One Reasoning LLMs with Efficient Budget Control Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: In this paper, we introduce Star Elastic, a novel LLM post-training method that adds N nested submodels to a given parent reasoning model using the compute of one run (Nx savings) via a single post-training job. |
Ali Taghibakhshi; Ruisi Cai; Saurav Muralidharan; Sharath Turuvekere Sreenivas; Ameya Mahabaleshwarkar; Marcin Chochowski; Akhiad Bercovich; Ran Zilberstein; Ran El-Yaniv; Yonatan Geifman; Daniel Korzekwa; Yoshi Suhara; Oluwatobi Olabiyi; Ashwath Aithal; Nima Tajbakhsh; Pavlo Molchanov; |
| 99 | Retrieval-Aware Distillation for Transformer-SSM Hybrids Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: This gap has been linked to a small set of attention heads, called Gather-and-Aggregate (G&A), which SSMs struggle to implement and are believed to drive the disparity. Leveraging this insight, we propose retrieval-aware distillation, a strategy that converts a pretrained Transformer into a hybrid student by preserving only these retrieval-critical components. |
Aviv Bick; Eric Xing; Albert Gu; |
| 100 | Weight-sparse Transformers Have Interpretable Circuits Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We train models to have more understandable circuits by constraining most of their weights to be zeros, so that each neuron only has a few connections. |
Leo Gao; Achyuta Rajaram; Jacob Coxon; Soham Govande; Bowen Baker; Daniel Mossing; |
| 101 | Attention Sinks As Internal Signals for Hallucination Detection in Large Language Models Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: In this work, we propose SinkProbe, a hallucination detection method grounded in the observation that hallucinations are deeply entangled with attention sinks – tokens that accumulate disproportionate attention mass during generation – indicating a transition from distributed, input-grounded attention to compressed, prior-dominated computation. |
Jakub Binkowski; Kamil Adamczewski; Tomasz Kajdanowicz; |
| 102 | TruthRL: Incentivizing Truthful LLMs Via Reinforcement Learning Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: In this work, we present TruthRL, a general reinforcement learning (RL) framework that directly optimizes the truthfulness of LLMs. |
Zhepei Wei; Xiao Yang; Kai Sun; Jiaqi Wang; Rulin Shao; Jingxiang Chen; Mohammad Kachuee; Teja Gollapudi; Yiwei Liao; Nicolas SCHEFFER; Rakesh Wanga; Anuj Kumar; Yu Meng; Scott Yih; Xin Dong; |
| 103 | On Robustness and Chain-of-Thought Consistency of RL-Finetuned VLMs Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: While RL-tuned VLMs can improve visual reasoning benchmark performance, they can still suffer from weak visual grounding, hallucinations, and over-reliance on textual cues. We show that simple, controlled textual perturbations—misleading captions or incorrect chain-of-thought (CoT) traces—cause substantial drops in robustness and confidence, and that these effects are more pronounced when CoT consistency is taken into account across open-source multimodal reasoning models. |
Rosie Zhao; Anshul Shah; Xiaoyu Zhu; Xinke Deng; Zhongyu Jiang; Yang Yang; Joerg Liebelt; Arnab Kumar Mondal; |
| 104 | MVISTA-4D: View-Consistent 4D World Model with Test-Time Action Inference for Robotic Manipulation Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: This work proposes a novel embodied 4D world model that enables geometrically consistent, arbitrary-view RGBD generation: given only a single-view RGBD observation as input, the model “imagines” the remaining viewpoints, which can then be back-projected and fused to assemble a more complete 3D structure across time. |
Jiaxu Wang; JIANG Yicheng; Tianlun HE; Jingkai SUN; Qiang Zhang; Jiahang Cao; Zesen Gan; Mingyuan Sun; Qiming Shao; Xiangyu Yue; |
| 105 | Modular Pretraining Enables Access Control Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: However, training and deploying multiple models is prohibitively expensive. We address this challenge by developing gradient-routed mixture-of-experts (GR-MoE), a pretraining method that selectively updates experts to induce specialization. |
Ethan Roland; Murat Cubuktepe; Erick Martinez; Stijn Servaes; Keenan Pepper; Michael Vaiana; Diogo de Lucena; Judd Rosenblatt; Addie Foote; Cem Anil; Alex Cloud; |
| 106 | Simultaneous Speech-to-Speech Translation Without Aligned Data Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We instead propose Hibiki-Zero, a model for simultaneous speech translation trained without word-level alignments between source and target speech. |
Tom Labiausse; Romain Fabre; Yannick Estève; Alexandre Défossez; Neil Zeghidour; |
| 107 | Privileged Information Distillation for Language Models Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We introduce π-Distill, a joint teacher–student framework that trains a PI-conditioned teacher and an unconditioned student simultaneously within a single shared-parameter model, enabling the teacher to learn how to use PI while mitigating distribution shift during transfer. |
Emiliano Penaloza; Dheeraj Vattikonda; Nicolas Gontier; Alexandre Lacoste; Laurent Charlin; Massimo Caccia; |
| 108 | Online Rubrics Elicitation from Pairwise Comparisons Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We introduce Online Rubrics Elicitation (OnlineRubrics), a method that dynamically curates evaluation criteria in an online manner through pairwise comparisons of responses from current and reference policies. |
MohammadHossein Rezaei; Robert Vacareanu; Zihao Wang; Clinton Wang; Bing Liu; Yunzhong He; Afra Feyza Akyürek; |
| 109 | Rectified LpJEPA: Joint-Embedding Predictive Architectures with Sparse and Maximum-Entropy Representations Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We introduce Rectified Distribution Matching Regularization (RDMReg), a sliced two-sample distribution-matching loss that aligns representations to a Rectified Generalized Gaussian (RGG) distribution. |
Yilun Kuang; Yash Dagade; Tim G. J. Rudner; Randall Balestriero; Yann LeCun; |
| 110 | Wait, Wait, Wait… Why Do Reasoning Models Loop? Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: This points to mismatches between the training distribution and the learned model, which we refer to as errors in learning, as a key cause. To understand how such errors cause loops, we introduce a synthetic graph reasoning task and demonstrate two mechanisms. |
Charilaos Pipis; Shivam Garg; Vasilis Kontonis; Vaishnavi Shrivastava; Akshay Krishnamurthy; Dimitris Papailiopoulos; |
| 111 | Sparser, Faster, Lighter Transformer Language Models Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Scaling autoregressive large language models (LLMs) has had an unprecedented impact, but at vast computational costs. In this work, we tackle these costs by leveraging unstructured sparsity within an LLM’s feedforward layers, which account for the majority of its parameters and execution FLOPs. |
Edoardo Cetin; Stefano Peluchetti; Emilio Castillo; Akira Naruse; Mana Murakami; Llion Jones; |
| 112 | Neuro-evolutionary Continual Reinforcement Learning Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Inspired by neuroscience, we propose Neuro-evolutionary Continual Reinforcement Learning (Nevo-CRL). |
Pengyi Li; Hongyao Tang; Yifu Yuan; Yan Zheng; Xin Xu; Jianye Hao; |
| 113 | Towards Understanding Modality Interaction in Multimodal Language Models Via Partial Information Decomposition Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We introduce Partial Information Decomposition (PID) as a unified, decision-level framework that separates \emph{unique}, \emph{redundant}, and \emph{synergistic} contributions of sensory and linguistic inputs, moving beyond representation alignment and outcome-based evaluation. |
Wanlong Fang; Tianle Zhang; Wen Tao; Alvin Chan; |
| 114 | Causal-JEPA: Learning World Models Through Object-Level Latent Interventions Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: While object-centric representations provide a useful abstraction, they are not sufficient to capture interaction-dependent dynamics. We therefore propose C-JEPA, a simple and flexible object-centric world model that extends masked joint embedding prediction from image patches to object-centric representations. |
Heejeong Nam; Quentin Le Lidec; Lucas Maes; Yann LeCun; Randall Balestriero; |
| 115 | Diamond Maps: Efficient Reward Alignment Via Stochastic Flow Maps Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We propose Diamond Maps, a stochastic flow-map model that enables efficient and accurate alignment to arbitrary rewards at inference time. |
Peter Holderrieth; Douglas Chen; Luca Eyring; Ishin Shah; Giri Anantharaman; Yutong He; Zeynep Akata; Tommi Jaakkola; Nicholas Boffi; Max Simchowitz; |
| 116 | MIND: Multi-rationale INtegrated Discriminative Reasoning Framework for Multi-modal Large Models Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: However, they suffer from limited multi-rationale semantic modeling, insufficient logical robustness, and susceptibility to misleading cues. Therefore, we propose a Multi-rationale INtegrated Discriminative (MIND) reasoning framework, which is designed to endow MLLMs with human-like cognitive abilities of “Understand → Rethink → Correct”, and achieves a paradigm evolution from passive imitation-based reasoning to active discriminative reasoning. |
Chuang Yu; Jinmiao Zhao; Mingxuan Zhao; Yunpeng Liu; Xiujun Shu; Feng Yuanhao; Bo Wang; Xiangyu Yue; |
| 117 | Efficient-DLM: From Autoregressive to Diffusion Language Models, and Beyond in Speed Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Accordingly, we introduce a continuous pretraining scheme with a block-wise attention pattern. |
Yonggan Fu; Lexington Whalen; Zhifan Ye; Xin Dong; Shizhe Diao; Jingyu Liu; CHENGYUE WU; Hao Zhang; Enze Xie; Song Han; Maksim Khadkevich; Jan Kautz; Yingyan (Celine) Lin; Pavlo Molchanov; |
| 118 | RADIO1D: Elastic Representations for Condensed Vision Modeling Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Notably, models trained with image-text alignment (such as SigLIP2) develop a small number of specialized tokens that effectively summarize global image content. Building on this, we introduce RADIO1D, which compresses images into a compact, variable-length 1D token sequence using multi-teacher knowledge distillation and an autoencoder design. |
Greg Heinrich; Mike Ranzinger; Collin McCarthy; Natan Bagrov; Eugene Khvedchenya; Bryan Catanzaro; Jan Kautz; Andrew Tao; Pavlo Molchanov; |
| 119 | IsoCompute Playbook: Optimally Scaling Sampling Compute for LLM RL Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We study the compute-optimal allocation of sampling compute for on-policy RL methods in LLMs, framing scaling as a compute-constrained optimization over three resources: parallel rollouts per problem, number of problems per batch, and number of update steps. |
Zhoujun Cheng; Yutao Xie; Yuxiao Qu; Amrith Setlur; Shibo Hao; Varad Pimpalkhute; Tongtong Liang; Feng Yao; Zhengzhong Liu; Eric Xing; Virginia Smith; Russ Salakhutdinov; Zhiting Hu; Taylor W. Killian; Aviral Kumar; |
| 120 | Agent0-VL: Exploring Self-Evolving Agent for Tool-Integrated Vision-Language Reasoning Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Yet, purely text-based self-evaluation struggles to verify complex visual reasoning steps and often suffers from evaluation hallucinations. To address these challenges, inspired by recent advances in tool-integrated reasoning, we propose Agent0-VL, a self-evolving vision-language agent that achieves continual improvement with tool-integrated reasoning. |
Jiaqi Liu; Kaiwen Xiong; Peng Xia; Yiyang Zhou; Haonian Ji; Lu Feng; Siwei Han; Mingyu Ding; Huaxiu Yao; |
| 121 | SimpleMem: Efficient Lifelong Memory for LLM Agents Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Existing approaches either retain full interaction histories via passive context extension, leading to substantial redundancy, or rely on iterative reasoning to filter noise, incurring high token costs. To address this challenge, we introduce SimpleMem, an efficient memory framework based on semantic lossless compression. |
Jiaqi Liu; Yaofeng Su; Peng Xia; Siwei Han; Zeyu Zheng; Cihang Xie; Mingyu Ding; Huaxiu Yao; |
| 122 | GemDepth: Geometry-Embedded Features for 3D-Consistent Video Depth Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We argue that current approaches, which primarily rely on temporal smoothing via Transformers, struggle to maintain strict 3D geometric consistency—particularly under rotations or drastic view changes. To address this, we propose GemDepth, a framework built on the insight that an explicit awareness of camera motion and global 3D structure is a prerequisite for 3D consistency. |
Yuecheng Liu; Junda Cheng; Longliang Liu; Wenjing Liao; Hanrui Cheng; Yuzhou Wang; Xin Yang; |
| 123 | Annotations Mitigate Post-Training Mode Collapse Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Crucially, we find this trade-off worsens with scale. To close this semantic diversity gap, we propose annotation-anchored training, a principled method that enables models to adopt the preference-following behaviors of post-training without sacrificing the inherent diversity of pre-training. |
Jacob Mitchell Springer; Madhu Advani; Lukas Aichberger; Arwen Bradley; Eran Malach; Omid Saremi; Sinead Williamson; Preetum Nakkiran; Etai Littwin; Aditi Raghunathan; |
| 124 | Towards A Science of AI Agent Reliability Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We propose a framework for measuring agent reliability grounded in safety-critical engineering practice, decomposing reliability into four dimensions: consistency, robustness, predictability, and safety. |
Stephan Rabanser; Sayash Kapoor; Peter Kirgis; Kangheng Liu; Saiteja Utpala; Arvind Narayanan; |
| 125 | Closing The Loop: Universal Repository Representation with RPG-Encoder Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We consider repository comprehension and generation to be inverse processes within a unified cycle: generation expands intent into implementation, while comprehension compresses implementation back into intent. To address this, we propose RPG-Encoder, a framework that generalizes the Repository Planning Graph (RPG) from a static generative blueprint into a unified, high-fidelity representation. |
Jane Luo; Chengyu Yin; Xin Zhang; Qingtao Li; Steven Liu; Yiming Huang; Jie Wu; Hao Liu; Yangyu Huang; Yu Kang; Fangkai Yang; Ying Xin; Scarlett Li; |
| 126 | Bridging Time and Frequency: A Joint Modeling Framework for Irregular Multivariate Time Series Forecasting Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: These irregularities violate the equidistant assumptions of standard models, hindering local temporal modeling and rendering classical frequency-domain methods ineffective for capturing global periodic structures. To address this challenge, we propose TFMixer, a joint time–frequency modeling framework for IMTS forecasting. |
Xiangfei Qiu; Kangjia Yan; Xvyuan Liu; Xingjian Wu; Jilin Hu; |
| 127 | DAG: A Dual Correlation Network for Time Series Forecasting with Exogenous Variables Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: In this study, to better leverage exogenous variables, especially future exogenous variables, we propose $\textbf{DAG}$, which $\textit{utilizes $\underline{D}$ual correl$\underline{A}$tion network along both the temporal and channel dimensions for time series forecasting with exo$\underline{G}$enous}$ variables. |
Xiangfei Qiu; Yuhan Zhu; Zhengyu Li; Xingjian Wu; Bin Yang; Jilin Hu; |
| 128 | SEER: Transformer-based Robust Time Series Forecasting Via Automated Patch Enhancement and Replacement Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: In real-world time series, there are often low-quality issues during data collection, such as missing values, distribution shifts, anomalies and white noise, which may cause some patches to contain low-quality information, negatively impacting the prediction results. To address this issue, this study proposes a robust time series forecasting framework called $\textbf{SEER}$. |
Xiangfei Qiu; Xvyuan Liu; Tianen Shen; Xingjian Wu; Hanyin Cheng; Bin Yang; Jilin Hu; |
| 129 | When to Memorize and When to Stop: Gated Recurrent Memory for Long-Context Reasoning Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: However, this naive recurrent memory update faces two crucial drawbacks: (i) memory can quickly explode because it can update indiscriminately, even on evidence-free chunks; and (ii) the loop lacks an exit mechanism, leading to unnecessary computation after even sufficient evidence is collected. To address these issues, we propose GRU-Mem, which incorporates two text-controlled gates for more stable and efficient long-context reasoning. |
Leheng Sheng; Yongtao Zhang; Wenchang Ma; Yaorui Shi; Ting Huang; Xiang Wang; An Zhang; Ke Shen; Tat-Seng Chua; |
| 130 | Doc-to-LoRA: Learning to Instantly Internalize Contexts Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: While context distillation (CD) can transfer information into model parameters, per-prompt distillation is impractical due to training costs and latency. To address these limitations, we propose Doc-to-LoRA (D2L), a lightweight hypernetwork that meta-learns to perform approximate CD within a single forward pass. |
Rujikorn Charakorn; Edoardo Cetin; Shinnosuke Uesaka; Robert Lange; |
| 131 | Separating Representation from Reconstruction Enables Scalable Text Encoders Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Hence, we propose CrossBERT, a two-part architecture that separates the learning of high-quality encoded representations from the rigid grounding of token reconstruction. |
Megi Dervishi; Mathurin VIDEAU; Yann LeCun; |
| 132 | Twins: Learn to Predict Unified Representations with Focal Loss Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We propose Twins, a unified continuous token space formed by channel-wise concatenating ViT and VAE features on the same token grid, so the sequence length is unchanged and attention cost does not increase. |
Kaixiong Gong; Xin Cai; Bin Lin; Hao Wang; Yunlong Lin; Mingzhe Zheng; Bohao Li; Jian-Wei Zhang; Miles Yang; Zhao Zhong; Liefeng Bo; Xiangyu Yue; |
| 133 | ReaForest: Fostering Generative Video Reasoning for Spatial Planning Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We attribute this limitation to two fundamental gaps: (i) VGMs are predominantly trained on general-purpose video corpora emphasizing perceptual fidelity over visual reasoning, leaving reasoning abilities underdeveloped; (ii) most VGMs generate videos in a single pass without mechanisms to explore alternative reasoning trajectories and to revise intermediate errors. Motivated by these limitations, we introduce **ReaForest**, a framework that fosters the reasoning capacity of VGMs in spatial planning through both training-time activation and inference-time scaling. |
Kun Ouyang; Yuanxin Liu; Xinhao Li; Linli Yao; Xiangyu Zeng; Haoning Wu; Hao Zhou; Fandong Meng; Jie Zhou; Xu SUN; |
| 134 | SoMA: A Real-to-Sim Neural Simulator for Robotic Soft-Body Manipulation Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: This paper presents \textbf{SoMA}, a 3D Gaussian Splat simulator for soft-body manipulation. |
Mu Huang; Hui Wang; Kerui Ren; Linning Xu; Mulin Yu; Yunsong Zhou; Bo Dai; Jiangmiao Pang; |
| 135 | SPADA: A Verifiable Test-Driven Agent for Controllable Parametric CAD Assembly Generation Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Methods that generate meshes or history-free B-rep can represent multi-part shape, but they often lack the program structure and constraint logic needed for reliable downstream edits; in contrast, code-based CAD generation offers direct parametric control, yet most published settings and evaluations focus on single-part solids rather than constrained assemblies. We introduce SPADA (Self-testing Parametric Assembly Design Agent), a test-driven agent that synthesizes assembly code together with deterministic verification tests, and uses these tests as an executable contract for controllable generation. |
Keyou Zheng; Xuyang Su; Jiewu Leng; |
| 136 | PyVision-RL: Forging Open Agentic Vision Models Via RL Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We introduce PyVision-RL, a reinforcement learning framework for open-weight multimodal models that stabilizes training and sustains interaction. |
Shitian Zhao; Shaoheng Lin; Ming Li; Haoquan Zhang; Wenshuo Peng; Kaipeng Zhang; Chen Wei; |
| 137 | SpaceVista: All-Scale Visual Spatial Reasoning from Mm to Km Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: In this paper, we introduce a holistic solution that integrates a structured spatial reasoning knowledge system, scale-aware modeling, and a progressive training paradigm, as the **first attempt** to broaden the scope of all-scale spatial intelligence. |
Peiwen Sun; Shiqiang Lang; Dongming Wu; Ding Yi; Kaituo Feng; Huadai Liu; Zhen Ye; Rui Liu; Yun-Hui Liu; Jianan Wang; Xiangyu Yue; |
| 138 | Rethinking The Trust Region in LLM Reinforcement Learning Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: This creates a sub-optimal learning dynamic: updates to low-probability tokens are aggressively over-penalized, while potentially catastrophic shifts in high-probability tokens are under-constrained, leading to training inefficiency and instability. To address this, we propose Divergence Proximal Policy Optimization (DPPO), which substitutes heuristic clipping with a more principled constraint based on a direct estimate of policy divergence (e.g., Total Variation or KL). |
Penghui Qi; Xiangxin Zhou; Zichen Liu; Tianyu Pang; Chao Du; Min Lin; Wee Sun Lee; |
| 139 | AVTrack: Audio-Visual Speaker Tracking in Complex Scenes Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Such oversimplified settings bias evaluation toward static audio–visual co-occurrence, rather than rigorously assessing robust spatiotemporal modeling and cross-modal reasoning in complex, dynamic scenes. To address these limitations, we introduce \textbf{AVTrack}, a human-centric audio-visual instance segmentation (AVIS) dataset designed for dynamic real-world scenarios. |
Yaoting Wang; Yun Zhou; Zipei Zhang; Henghui Ding; |
| 140 | Position: Stop Automating Peer Review Without Rigorous Evaluation Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We identify two critical issues: 1) AI reviewers exhibit a *hivemind effect* of excessive agreement within and across papers that reduces perspective diversity. |
Joachim Baumann; Jiaxin Pei; Sanmi Koyejo; Dirk Hovy; |
| 141 | Retaining By Doing: The Role of On-Policy Data in Mitigating Forgetting Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Adapting language models (LMs) to new tasks via post-training carries the risk of degrading existing capabilities — a phenomenon classically known as catastrophic forgetting. In this paper, toward identifying guidelines for mitigating this phenomenon, we systematically compare the forgetting patterns of two widely adopted post-training methods: supervised fine-tuning (SFT) and reinforcement learning (RL). |
Howard Chen; Noam Razin; Karthik Narasimhan; Danqi Chen; |
| 142 | Why Are Linear RNNs More Parallelizable? Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: While prior work establishes the expressivity benefits of LRNNs over transformers, it is unclear what makes LRNNs—but not traditional, *nonlinear* RNNs—as easy to parallelize in practice as transformers. We answer this question by providing a tight connection between types of RNNs and standard complexity classes. |
William Merrill; Hongjian Jiang; Yanhong Li; Anthony Lin; Ashish Sabharwal; |
| 143 | Position: Vision Encoders Should Be Image Size Agnostic and Task Driven Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: This position paper argues that the next generation of vision encoders should be image size agnostic and task driven. |
Nedyalko Prisadnikov; Danda Pani Paudel; Yuqian Fu; Luc Van Gool; |
| 144 | Interpretable Embeddings with Sparse Autoencoders: A Data Analysis Toolkit Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We propose using sparse autoencoders (SAEs) to create *SAE embeddings*: representations whose dimensions map to interpretable concepts. |
Nicholas Jiang; Xiaoqing Sun; Lisa Dunlap; Lewis Smith; Neel Nanda; |
| 145 | Thoughtbubbles: An Unsupervised Method for Parallel Thinking in Latent Space Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: In this work, we propose Thoughtbubbles, a transformer variant that natively performs parallel adaptive computation in latent space by learning to fork or delete residual streams. |
Houjun Liu; Shikhar Murty; Christopher Manning; Róbert Csordás; |
| 146 | SiameseNorm: Breaking The Barrier to Reconciling Pre/Post-Norm Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We attribute this phenomenon to a structural incompatibility within a *single-stream* design: Any application of the Post-Norm operation inevitably obstructs the clean identity gradient preserved by Pre-Norm. To fundamentally reconcile these paradigms, we propose SiameseNorm, a *two-stream* architecture that couples Pre-Norm-like and Post-Norm-like streams with shared parameters. |
Tianyu Li; Dongchen Han; Zixuan Cao; Haofeng Huang; Mengyu Zhou; Ming Chen; erchao.zec; xiaoxi jiang; guanjunjiang; Gao Huang; |
| 147 | Dual-Stream Diffusion for World-Model Augmented Vision-Language-Action Model Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Augmenting Vision-Language-Action models (VLAs) with world models is promising for robotic policy learning but faces challenges in jointly predicting states and actions due to the modality gap. To address this, we propose DUal-STream diffusion (DUST), a world-model augmented VLA framework featuring a multimodal diffusion transformer that maintains separate modality streams while enabling cross-modal knowledge sharing. |
John Won; Kyungmin Lee; Huiwon Jang; Dongyoung Kim; Jinwoo Shin; |
| 148 | Golden Goose: A Simple Trick to Synthesize Unlimited RLVR Tasks from Unverifiable Internet Text Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Yet, scaling up RL is bottlenecked by limited existing verifiable data, where improvements increasingly saturate over prolonged training. To overcome this, we propose **Golden Goose**, a simple trick to synthesize unlimited RLVR tasks from unverifiable internet text by constructing a multiple-choice question-answering version of the fill-in-the-middle task. |
Ximing Lu; David Acuna; Jaehun Jung; Jian Hu; Di Zhang; Shizhe Diao; Yunheng Zou; Shaokun Zhang; Brandon Cui; Mingjie Liu; Hyunwoo Kim; Prithviraj Ammanabrolu; Jan Kautz; Yi Dong; Yejin Choi; |
| 149 | CausalArmor: Efficient Indirect Prompt Injection Guardrails Via Causal Attribution Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We revisit IPI through a causal ablation perspective: a successful injection manifests as a *dominance shift* where the user request no longer provides decisive support for the agent’s privileged action, while a particular untrusted segment, such as a retrieved document or tool output, provides disproportionate attributable influence. Based on this signature, we propose **CausalArmor**, a selective defense framework that (i) computes lightweight, leave-one-out ablation-based attributions at privileged decision points, and (ii) triggers targeted sanitization only when an untrusted segment dominates the user intent. |
Minbeom Kim; Mihir Parmar; Phillip Wallis; Lesly Miculicich; Kyomin Jung; Krishnamurthy Dvijotham; Long Le; Tomas Pfister; |
| 150 | Linearizing Vision Transformer with Test-Time Training Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Inheriting weights from pretrained Transformers provides an appealing shortcut, yet the fundamental representational gap between Softmax and linear attention prevents effective weight transfer. In this work, we address this conversion challenge from two perspectives: architectural alignment and representational alignment. |
Yining Li; Dongchen Han; Zeyu Liu; Hanyi Wang; Yulin Wang; Gao Huang; |
| 151 | Periodic Bayesian Flow Networks with Additive Accuracy Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We introduce \emph{PeriodicBFN}, which embeds each periodic scalar into a two-dimensional unit-circle representation and performs Gaussian Bayesian updates in the resulting Cartesian space, thereby restoring strictly additive accuracy. |
Peijia Lin; Zihan Zhang; zhangrui zhao; Shaohao Rui; Junyi An; Yun-Fei Shi; Fenglei Cao; Weijie Ma; Yutong Lu; |
| 152 | ProtDBench: A Unified Benchmark of Protein Binder Design and Evaluation Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We introduce $\textbf{ProtDBench}$, a standardized and throughput-aware evaluation framework for protein binder design. |
Cong Liu; Milong Ren; Jiaqi Guan; Chengyue Gong; Jinyuan Sun; Xinshi Chen; Wenzhi Xiao; |
| 153 | SP-Mind: An Autonomous Reasoning Agent for Spatial Proteomics Analysis Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We present **SP-Mind**, the first autonomous AI agent designed to unify the spatial proteomics analysis pipeline, from raw multiplexed tissue imaging to downstream phenotype discovery. |
YuCheng Yuan; Ji Yuanfeng; Zhongxiao Li; Ruijiang Li; |
| 154 | Symmetries in Language Statistics Shape The Geometry of Model Representations Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We show that the statistics of language exhibit a translation symmetry—e.g,. |
Dhruva Karkada; Daniel Korchinski; Andres Nava; Matthieu Wyart; Yasaman Bahri; |
| 155 | RubricRobustness: A Simple Framework for Evaluating The Robustness of Rubrics-Based Benchmarks Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: While these frameworks utilize expert-curated criteria and LLM-as-a-judge to assess open-ended generation, the intrinsic robustness of these evaluation harnesses to fundamental validity assessments remains critically under-investigated. To bridge this gap, we introduce RubricRobustness, a systematic sensitivity analysis framework that subjects these benchmarks to three common sense perturbations: semantic negation, stochastic deletion and irrelevant addition. |
Manasi Sharma; |
| 156 | MoshiRAG: Asynchronous Knowledge Retrieval for Full-Duplex Speech Language Models Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: In this work, we propose Moshi-RAG, a modular approach that combines a compact full-duplex interface with selective retrieval to access more powerful knowledge sources. |
Chung-Ming Chien; Manu Orsini; Eugene Kharitonov; Neil Zeghidour; Karen Livescu; Alexandre Défossez; |
| 157 | Contextualized Privacy Defense for LLM Agents Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: These paradigms are insufficient for supporting contextual, proactive privacy decisions in multi-step agent execution. We propose *Contextualized Defense Instructing (CDI)*, a new privacy defense paradigm in which an instructor model generates step-specific, context-aware privacy guidance during execution, proactively shaping actions rather than merely constraining or vetoing them. |
Yule Wen; Yanzhe Zhang; Jianxun Lian; Xiaoyuan Yi; Xing Xie; Diyi Yang; |
| 158 | The Surprising Difficulty of Search in Model-Based Reinforcement Learning Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Instead, we show that mitigating distribution shift matters more than improving model or value function accuracy. Building on this insight, we identify key techniques for enabling effective search, achieving state-of-the-art performance across multiple popular benchmark domains. |
Wei-Di Chang; Mikael Henaff; Brandon Amos; Gregory Dudek; Scott Fujimoto; |
| 159 | ScaleEnv: Scaling Environment Synthesis from Scratch for Generalist Interactive Tool-Use Agent Training Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: However, high-quality agentic data remain critically scarce, and existing synthesis methods suffer from significant limitations regarding environmental diversity and scalability. To address these challenges, we introduce ScaleEnv, a framework that constructs fully interactive environments and verifiable tasks entirely from scratch. |
Dunwei Tu; Hongyan Hao; Hansi Yang; Yihao Chen; Yu Yang; Yueqing Sun; Xingchen Liu; Furao Shen; Qi GU; Hui Su; Xunliang Cai; |
| 160 | OmniDenseCap: Scripting Multi-Scene Videos with Time-Aware and Structural Audio-Visual Captions Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: This paper proposes Omni Dense Captioning, a novel task designed to generate continuous, fine-grained, and structured audio-visual narratives with explicit timestamps. |
Linli Yao; Yuancheng Wei; Yaojie Zhang; Lei Li; Xinlong Chen; Feifan Song; Ziyue Wang; Kun Ouyang; Yuanxin Liu; Lingpeng Kong; Qi Liu; Pengfei Wan; Kun Gai; Yuanxing Zhang; Xu SUN; |
| 161 | Beyond Scalar Rewards: Learning from Text Feedback in LLM Post-Training Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Therefore, models must learn to internalize the feedback in order to improve their test-time single-turn performance. To do this, we propose two methods: Self Distillation, which trains the single-turn policy to match its own feedback-conditioned second-turn generations; and Feedback Modeling, which predicts the feedback as an auxiliary objective. |
Yuda Song; Lili Chen; Fahim Tajwar; REMI MUNOS; Deepak Pathak; J. Bagnell; Aarti Singh; Andrea Zanette; |
| 162 | TopAdapter: Topology-Aware Prompt Tuning for Efficient Point Cloud Understanding Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: However, existing parameter-efficient fine-tuning (PEFT) methods predominantly focus on input token prompting, overlooking the intrinsic geometric information. To address this limitation, we propose TopAdapter, a novel PEFT framework that enhances geometric perception by injecting local topological information into pre-trained 3D vision models. |
Changshuo Wang; Shuting He; Xiang Fang; Weijun Li; Yixian Shen; Mingkun Xu; Zhongtian Sun; Prayag Tiwari; |
| 163 | VisualPuzzles: Decoupling Multimodal Reasoning Evaluation from Domain Knowledge Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Current multimodal benchmarks often conflate reasoning with domain-specific knowledge, making it difficult to isolate and evaluate general reasoning abilities in non-expert settings. To address this, we introduce VisualPuzzles, a benchmark that targets visual reasoning while deliberately minimizing reliance on specialized knowledge. |
Yueqi Song; Tianyue Ou; Yibo Kong; Zecheng Li; Graham Neubig; Xiang Yue; |
| 164 | Does Math Reasoning Improve General LLM Capabilities? Understanding Transferability of LLM Reasoning Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: But as math leaderboards improve week by week, it is worth asking: do these gains reflect broader problem-solving ability or just narrow overfitting? To answer this question, we evaluate over 20 open-weight reasoning-tuned models across a broad suite of tasks, including math, scientific QA, agent planning, coding, and standard instruction-following. |
Maggie Huan; Yuetai Li; Tuney Zheng; Xiaoyu Xu; Seungone Kim; Minxin Du; Radha Poovendran; Graham Neubig; Xiang Yue; |
| 165 | On The Interplay of Pre-Training, Mid-Training, and RL on Reasoning Language Models Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: A central challenge is the lack of control in modern training pipelines: large-scale pre-training corpora are opaque, mid-training is often underexamined, and RL objectives interact with unknown prior knowledge in complex ways. To resolve this ambiguity, we develop a fully controlled experimental framework that isolates the causal contributions of pre-training, mid-training, and RL-based post-training. |
Charlie Zhang; Graham Neubig; Xiang Yue; |
| 166 | Uncovering Hidden Triggers: Backdoor Attribution in Language Models Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Previous research on interpretability for LLM safety tends to focus on alignment, jailbreak, and hallucination, but overlooks backdoor mechanisms, making it difficult to understand and fully eliminate the backdoor threat. In this paper, aiming to bridge this gap, we explore the interpretable mechanisms of LLM backdoors through Backdoor Attribution (BkdAttr), a tripartite causal analysis framework. |
Miao Yu; Zhenhong Zhou; Moayad Aloqaily; Kun Wang; Biwei Huang; Stephen Wang; Yueming Jin; Qingsong Wen; |
| 167 | SafeSeek: Universal Attribution of Safety Circuits in Language Models Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: However, existing safety attribution methods struggle with generalization and reliability due to their reliance on heuristic, domain-specific metrics and search algorithms. To address this, we propose SafeSeek, a unified safety interpretability framework that identifies functionally complete safety circuits in LLMs via optimization. |
Miao Yu; Siyuan Fu; Moayad Aloqaily; Zhenhong Zhou; Safa Otoum; Xing fan; Kun Wang; Yufei Guo; Qingsong Wen; |
| 168 | Empty Shelves or Lost Keys? Recall Is The Bottleneck for Parametric Factuality Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We propose a behavioral framework that profiles factual knowledge at the level of facts rather than questions, characterizing each fact by whether it is encoded, and then by how accessible it is: cannot be recalled, can be directly recalled, or can only be recalled with inference-time computation (thinking). |
Nitay Calderon; Eyal Ben-David; Zorik Gekhman; Eran Ofek; Gal Yona; |
| 169 | Causal Forcing: Autoregressive Diffusion Distillation Done Right for High-Quality Real-Time Video Generation Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: This process involves an *architectural gap*, as it converts full attention into causal attention. In this paper, we demonstrate that existing methods fail to bridge this gap theoretically, leading to suboptimal performance. |
Hongzhou Zhu; Min Zhao; Guande He; Hang Su; Chongxuan Li; Jun Zhu; |
| 170 | Rewiring Experts on The Fly: Continuous Rerouting for Better Online Adaptation in Mixture-of-Expert Models Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: As such, we propose \textit{a data-free, online test-time framework} that continuously adapts MoE routing decisions during text generation without external supervision or data. |
Guinan Su; Yanwu Yang; Li Shen; Lu Yin; Shiwei Liu; Jonas Geiping; |
| 171 | Efficient Parallel Samplers for Recurrent-Depth Models Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: In this work, we examine the relationship between recurrent-depth models and diffusion language models. |
Jonas Geiping; Xinyu Yang; Guinan Su; |
| 172 | RoboTwin 2.0: A Scalable Data Generator and Benchmark with Strong Domain Randomization for Robust Bimanual Robotic Manipulation Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: To this end, we present RoboTwin 2.0, a scalable simulation framework that enables closed-loop, automated, large-scale generation of diverse and realistic data, along with unified evaluation protocols for dual-arm manipulation. |
Tianxing Chen; Zanxin Chen; Baijun Chen; Zijian Cai; Yibin Liu; Zixuan Li; Qiwei Liang; Xianliang Lin; Yiheng Ge; Zhenyu Gu; Weiliang Deng; Yubin Guo; Tian Nian; Xuanbing Xie; Qiangyu Chen; KailunSu; Tianling Xu; Guodong Liu; Mengkang Hu; Huan-ang Gao; Kaixuan Wang; Zhixuan Liang; Yusen Qin; Xiaokang Yang; Ping Luo; Yao Mu; |
| 173 | The Flexibility Trap: Rethinking The Value of Arbitrary Order in Diffusion Language Models Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: However, in this paper, we reveal that for general reasoning tasks (e.g., mathematics and coding), arbitrary order generation may in fact limit the reasoning potential of dLLMs. We find that dLLMs tend to exploit this order flexibility to bypass high-uncertainty tokens that are crucial for exploration, leading to a premature collapse of solution coverage. |
Zanlin Ni; Shenzhi Wang; Yang Yue; Tianyu Yu; Weilin Zhao; Yeguo Hua; Tianyi Chen; Jun Song; YuCheng; Bo Zheng; Gao Huang; |
| 174 | Memory Is Reconstructed, Not Retrieved: Graph Memory for LLM Agents Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: While current memory-augmented agents rely on a static “retrieve-then-reason” paradigm, this rigid pipeline design prevents them from dynamically adapting memory access to intermediate evidence discovered during inference. To bridge this gap, we propose MRAgent, a framework that combines an associative memory graph with an active reconstruction mechanism. |
Shuo Ji; yibo li; Bryan Hooi; |
| 175 | IVQ: Structured and Lightweight Vector Quantization Via Binary Hierarchical Composition Inspired By $\textit{IChing}$ Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: In this paper, we propose *IChing* Vector Quantization (IVQ), a lightweight and structured vector quantization framework inspired by *IChing*. |
Heda Zuo; Junxian Wu; Fengjie Lu; Pei Chen; Lingyun Sun; Weitao You; |
| 176 | Learning Query-Aware Budget-Tier Routing for Runtime Agent Memory Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: In this work, we present \textbf{BudgetMem}, a runtime agent memory framework for explicit, query-aware performance–cost control. |
Haozhen Zhang; Haodong Yue; Tao Feng; Quanyu Long; Jianzhu Bao; Bowen Jin; Weizhi Zhang; Xiao Li; Jiaxuan You; Chengwei Qin; Wenya Wang; |
| 177 | Learnability-Informed Fine-Tuning of Diffusion Language Models Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We aim to improve the reasoning capabilities of diffusion language models (DLMs). |
Shubham Parashar; Atharv Chagi; Jacob Helwig; Lakshmi Madhavarapu; Sushil Vemuri; James Caverlee; Dileep Kalathil; Shuiwang Ji; |
| 178 | GraphPFN: A Prior-Data Fitted Network for Graph Node-Level Tasks Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: In this work, we make the next step by proposing GraphPFN, a PFN-based model designed and pretrained specifically for graph node-level tasks. |
Dmitry Eremeev; Oleg Platonov; Gleb Bazhenov; Artem Babenko; Liudmila Prokhorenkova; |
| 179 | VlogReward: Learning Multi-Dimensional Evaluation for Vlog Editing Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: However, vlog assessment is highly subjective and remains challenging due to a lack of standardized criteria, dataset and benchmark, and effective reward models. To address these challenges, we define a comprehensive vlog evaluation framework guided by professional vlog creators and product managers, establishing a taxonomy of six key dimensions, *i.e.*, *Creativity*, *Consistency*, *Concept Design*, *Cinematography*, *Narration*, and *Pacing*. |
Yexiang Liu; Wen Zhong; Sijie Zhu; Xin Gu; Fan Chen; Junxian Duan; Jie Cao; Longyin Wen; Zhenfang Chen; |
| 180 | Vision-DeepResearch: Incentivizing DeepResearch Capability in Multimodal Large Language Models Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Moreover, they are often limited in the reasoning depth and search breadth, making it difficult to solve complex questions that require aggregating evidence from diverse visual and textual sources. Building on this, we propose Vision-DeepResearch, which proposes one new multimodal deep-research paradigm, i.e., performs multi-turn, multi-entity and multi-scale visual and textual search to robustly hit real-world search engines under heavy noise. |
Wenxuan Huang; Yu Zeng; Qiuchen Wang; Zhen Fang; Shaosheng Cao; Zheng Chu; Qingyu Yin; Shuang Chen; Zhenfei Yin; Lin Chen; Zehui Chen; Yao Hu; Phil Torr; Feng Zhao; Wanli Ouyang; |
| 181 | Offline Preference Optimization for Rectified Flow with Noise-Tracked Pairs Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Leveraging the straight-line property of RF, we estimate intermediate states via noise–image interpolation, which constrains the trajectory estimation space and yields a tighter surrogate objective for preference optimization. |
Yunhong Lu; Qichao Wang; Hengyuan Cao; Xiaoyin Xu; Min Zhang; |
| 182 | UltraHorizon: Benchmarking LLM-Agent Capabilities in Ultra Long-Horizon Scenarios Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Existing benchmarks rarely capture these long-horizon challenges, leaving a gap in systematic evaluation. To bridge this gap, we introduce $\textbf{UltraHorizon}$, a novel benchmark that measures the foundational capabilities essential for complex real-world challenges. |
Haotian Luo; Huaisong Zhang; Xuelin Zhang; Haoyu Wang; Zeyu Qin; Wenjie Lu; Guozheng Ma; Haiying He; Yingsha Xie; Qiyang Zhou; Zixuan Hu; Hongze Mi; Yibo Wang; Naiqiang Tan; Hong Chen; Yi Fung; Chun Yuan; Li Shen; |
| 183 | Multi-Agent Teams Hold Experts Back Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Drawing on organizational psychology, we study whether self-organizing LLM teams achieve *strong synergy*, where team performance matches or exceeds the best individual member. |
Aneesh Pappu; Batu El; Hancheng Cao; Carmelo di Nolfo; Yanchao Sun; Meng Cao; James Zou; |
| 184 | Conversation for Non-verifiable Learning: Self-Evolving Large Language Models Through Meta-Evaluation Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We introduce CoNL, a framework that unifies generation, evaluation, and meta-evaluation through multi-agent self-play. |
Yuan Sui; Bryan Hooi; |
| 185 | A Minimal Agent for Automated Theorem Proving Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We propose a minimal agentic baseline that enables systematic comparison across different AI-based theorem prover architectures. |
Borja Requena; Austin Letson; Krystian Nowakowski; Izan Ferreiro; Leopoldo Sarra; |
| 186 | ACON: Optimizing Context Compression for Long-horizon LLM Agents Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We introduce Agent Context Optimization (ACON), a unified framework that optimally compresses both observations and history into concise, informative representations. |
Minki Kang; Wei-Ning Chen; Dongge Han; Huseyin Inan; Lukas Wutschitz; Yanzhi Chen; Robert A Sim; Saravanakumar Rajmohan; |
| 187 | Stream RAG: Instant and Accurate Spoken Dialogue Systems with Streaming Tool Usage Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: A key challenge is that tool integration increases latency, disrupting conversational flow. To mitigate this, we propose Streaming Retrieval-Augmented Generation (Stream RAG), a novel framework that reduces latency by predicting tool queries in parallel with user speech, even before the user finishes speaking. |
Siddhant Arora; Haidar Khan; Kai Sun; Xin Dong; Sajal Choudhary; Seungwhan Moon; Xinyuan Zhang; Adithya Sagar; Surya Appini; Kaushik Patnaik; Sanat Sharma; Shinji Watanabe; Anuj Kumar; Ahmed A Aly; Yue Liu; Florian Metze; Zhaojiang Lin; |
| 188 | MOD-SR: Unifying Multimodal Learning and Direct Optimization with Gradient-Guided Diffusion Model for Symbolic Regression Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We propose MOD-SR, unifying multimodal distribution learning during training with direct optimization at inference time. |
Chuyang Xiang; Yichen Wei; Junchi Yan; |
| 189 | MemEvolve: Meta-Evolution of Agent Memory Systems Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: However, this paradigm is fundamentally constrained by the \textit{staticity} of the memory system itself: while memory facilitates agent-level evolving, the underlying memory architecture cannot be meta-adapted to diverse task contexts. To address this gap, we propose MemEvolve, a meta-evolutionary framework that jointly evolves agents’ experiential knowledge and their memory architecture, allowing agent systems not only to accumulate experience but also to progressively refine how they learn from it. |
Guibin Zhang; Haotian Ren; Chong Zhan; Junhao Wang; He Zhu; Wangchunshu Zhou; Shuicheng YAN; |
| 190 | A Geometry-Based View of Mahalanobis OOD Detection Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We conduct a large-scale study across diverse foundation-model backbones and Mahalanobis variants. |
Denis Janiak; Jakub Binkowski; Tomasz Kajdanowicz; |
| 191 | A Diagnostic Study of Multi-Agent LLMs for Real-World Debates Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: In this work, we introduce a diagnostic evaluation framework that measures debate quality by measuring both the outcome and the process. |
Priya Pitre; Gaurav Srivastava; Lu Zhang; Le Wang; Naren Ramakrishnan; Xuan Wang; |
| 192 | Long Grounded Thoughts: Synthesizing Grounded Visual Problems and Distilling Reasoning Chains at Scale Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We introduce a framework able to synthesize vision-centric problems spanning diverse levels of complexity, and the resulting dataset with over 1M high-quality problems including: reasoning traces, preference data, and instruction prompts supporting SFT, offline and online RL. |
David Acuna; Chao-Han Yang; Yuntian Deng; Jaehun Jung; Ximing Lu; Prithviraj Ammanabrolu; Hyunwoo Kim; Yuan-Hong Liao; Yejin Choi; |
| 193 | Alignment Pretraining: AI Discourse Causes Self-Fulfilling (Mis)alignment Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We find that discussion of AI contributes to misalignment. |
Cameron Tice; Puria Radmard; Samuel Ratnam; Andy Kim; David Africa; Kyle O'Brien; |
| 194 | Faults in Our Formal Benchmarking: Dataset Defects and Evaluation Failures in Lean Theorem Proving Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We propose a fault taxonomy, a suite of automated checkers and prompts, and release standards to guide the creation of formal math datasets and make evaluation more reproducible and trustworthy. |
Pawan Sasanka Ammanamanchi; Siddharth Bhat; Stella Biderman; |
| 195 | Predicting What Matters: Robust Generalist Robot Policy Learning Via Future Semantic Mask Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: These distractions reduce the model’s ability to generalize, ultimately leading to unreliable and fragile control policies. To address this, we introduce the Mask World Model (MWM), that leverages video diffusion architectures to predict the evolution of semantic masks instead of pixels. |
Yunfan Lou; Xiaowei Chi; Xiaojie Zhang; Zezhong Qian; Chengxuan Li; Rongyu Zhang; yaoxu lyu; Guoyu Song; Chuyao Fu; Haoxuan Xu; Pengwei Wang; Shanghang Zhang; |
| 196 | Endogenous Resistance to Activation Steering in Language Models: Evidence for Internal Consistency Monitoring in Llama-3.3-70B Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Using sparse autoencoder (SAE) latents to steer model activations, we find that Llama-3.3-70B shows substantial ESR, while smaller models from the Llama-3 and Gemma-2 families exhibit the phenomenon less frequently. |
Alex McKenzie; Keenan Pepper; Stijn Servaes; Martin Leitgab; Murat Cubuktepe; Michael Vaiana; Diogo de Lucena; Judd Rosenblatt; Michael Graziano; |
| 197 | Learning Unmasking Policies for Diffusion Language Models Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: In this work, we instead propose to train sampling procedures using reinforcement learning. |
Metod Jazbec; Theo X. Olausson; Louis Béthune; Pierre Ablin; Michael Kirchhof; Joao Monteiro; Victor Guilherme Turrisi da Costa; Jason Ramapuram; Marco Cuturi; |
| 198 | Latent Reasoning VLA: Latent Thinking and Prediction for Vision-Language-Action Models Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We propose Latent Reasoning VLA (LaRA-VLA), a unified VLA framework that internalizes multi-modal CoT reasoning into continuous latent representations for embodied action. |
Shuanghao Bai; Jing Lyu; Wanqi Zhou; Zhe Li; Dakai Wang; Lei Xing; Xiaoguang Zhao; Pengwei Wang; Zhongyuan Wang; Cheng Chi; Badong Chen; Shanghang Zhang; |
| 199 | Problem Distributions As Tasks: Repurposing Meta Learning for Generative Combinatorial Optimization Towards Multi-task Pretrain and Adaptation Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: This paper proposes M$^2$GenCO, a Multi-task learning framework that pioneers the instantiation of the Meta-learning mechanism with diffusion-based Generative solving for CO Problems (COPs) on graphs, first formulating tasks in meta-learning as distinct problem types instead of instances of the same problem. |
Wenzheng Pan; Jiale Ma; Nuoyan Chen; Yang Li; Junchi Yan; |
| 200 | Utonia: Toward One Encoder for All Point Clouds Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We dream of a future where point clouds from all domains can come together to shape a single model that benefits them all. Toward this goal, we present Utonia, a first step toward training a single self-supervised point transformer encoder across heterogeneous domains, spanning remote sensing, outdoor LiDAR, indoor RGB-D sequences, object-centric CAD models, and point clouds lifted from RGB-only videos. |
Yujia Zhang; Xiaoyang Wu; Yunhan Yang; Xianzhe Fan; Han Li; Yuechen Zhang; Zehao Huang; Naiyan Wang; Hengshuang Zhao; |
| 201 | ATLAS: Learning to Optimally Memorize The Context at Test Time Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We observe that these shortcomings come from three disjoint aspects in their design: (1) limited memory capacity that is bounded by the architecture of memory and feature mapping of the input; (2) online nature of update, i.e., optimizing the memory only with respect to the last input; and (3) less expressive management of their fixed-size memory. To enhance all these three aspects, we present Atlas, a long-term memory module with high capacity that learns to memorize the context by optimizing the memory based on the current and past tokens, overcoming the online nature of long-term memory models. |
Ali Behrouz; Zeman Li; Praneeth Kacham; Majid Daliri; Yuan Deng; Peilin Zhong; Meisam Razaviyayn; Vahab Mirrokni; |
| 202 | Memory Caching: RNNs with Growing Memory Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: In this paper, we introduce Memory Caching (MC), a simple yet effective technique that enhances recurrent models by caching checkpoints of their memory states (a.k.a. hidden states). |
Ali Behrouz; Zeman Li; Yuan Deng; Peilin Zhong; Meisam Razaviyayn; Vahab Mirrokni; |
| 203 | Just-In-Time Reinforcement Learning: Continual Learning in LLM Agents Without Gradient Updates Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We introduce Just-In-Time Reinforcement Learning (JitRL), a training-free framework that enables test-time policy optimization without any gradient updates. |
yibo li; Zijie Lin; Ailin Deng; Xuan Zhang; Yufei He; Shuo Ji; Tri Cao; Bryan Hooi; |
| 204 | When to Trust The Cheap Check: Weak and Strong Verification for Reasoning Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We introduce metrics capturing incorrect acceptance, incorrect rejection, and strong-verification frequency. |
Shayan Kiyani; Sima Noorani; George Pappas; Hamed Hassani; |
| 205 | Equilibrium Reasoners: Learning Attractors Enables Scalable Reasoning Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We hypothesize that generalizable reasoning emerges through learning task-conditioned attractors. |
Benhao Huang; Zhengyang Geng; Zico Kolter; |
| 206 | Proteo-R1: Thinking Foundation Models for De Novo Protein Binder Design Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: \textbf{ThinkProteo} reimagines generative science by introducing reasoning-guided diffusion models that think step-by-step, akin to how a scientist hypothesizes, tests, and refines molecular ideas. |
Fang Wu; Li Li; Weihao Xuan; Heli Qi; Zeqi Zhou; Hanqun CAO; Heng-Jui Chang; Haokai Zhao; Jian Ma; Zijian Carl; Yu-Chi Cheng; Robert Tang; Zehong Wang; Kuan Pang; Hanchen Wang; Kejun Ying; Pan Lu; Chiho Im; Seungju Han; Peng Xia; Yinxi Li; Guanlue Li; Tinson Xu; Deyao Zhu; Pheng Ann Heng; Naoto Yokoya; Masashi Sugiyama; Jure Leskovec; Yejin Choi; |
| 207 | Esoteric Language Models Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We introduce Eso-LMs, a new family of models that fuses AR and MDM paradigms, smoothly interpolating between their perplexities while overcoming their respective limitations. |
Subham Sekhar Sahoo; Zhihan Yang; Yash Akhauri; Johnna Liu; Deepansha Singh; Zhoujun Cheng; Zhengzhong Liu; Eric Xing; John Thickstun; Arash Vahdat; |
| 208 | Active Tabular Augmentation Via Policy-Guided Diffusion Inpainting Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We propose TAP (Tabular Augmentation Policy), which couples diffusion inpainting with a lightweight, learner-conditioned policy to steer generation toward high-utility regions and controls safe injection via explicit gating and conservative windowed commitment. |
Zheyu Zhang; Shuo Yang; Bardh Prenkaj; Gjergji Kasneci; |
| 209 | Compositional Planning with Jumpy World Models Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Such compositional planning remains elusive as compounding errors in long-horizon predictions make it challenging to estimate the visitation distribution induced by sequencing policies. Motivated by the *geometric policy composition* framework introduced in Thakoor et al. (2022), we address these challenges by learning predictive models of multi-step dynamics, so-called *jumpy world models*, that capture state occupancies induced by pre-trained policies across multiple timescales in an off-policy manner. |
Jesse Farebrother; Matteo Pirotta; Andrea Tirinzoni; Marc Bellemare; Alessandro Lazaric; Ahmed Touati; |
| 210 | Olmix: A Framework for Data Mixing Throughout LM Development Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We present Olmix, a framework that addresses two challenges encountered during LM development. |
Mayee Chen; Tyler Murray; David Heineman; Matt Jordan; Hannaneh Hajishirzi; Christopher Re; Luca Soldaini; Kyle Lo; |
| 211 | VideoFlexTok: Flexible-Length Coarse-to-Fine Video Tokenization Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We present VideoFlexTok, a tokenizer that represents videos with a _variable-length sequence of tokens structured in a coarse-to-fine manner_, where the first tokens capture abstract information like semantics and motion and later tokens provide fine-grained details. |
Andrei Atanov; Jesse Allardice; Roman Bachmann; Oğuzhan Kar; R Devon Hjelm; David Griffiths; Peter Fu; Amir Zamir; Afshin Dehghan; |
| 212 | DnaHNet: A Scalable and Hierarchical Foundation Model for Genomic Sequence Learning Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We introduce dnaHNet, a state-of-the-art tokenizer-free autoregressive model that segments and models genomic sequences end to end. |
Arnav Shah; Junzhe Li; Parsa Idehpour; Adibvafa Fallahpour; Brandon Wang; Sukjun Hwang; BO WANG; Patrick Hsu; Hani Goodarzi; Albert Gu; |
| 213 | Learning to Decode Against Compositional Hallucination in Video Multimodal Large Language Models Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We propose **TriCD**, a contrastive decoding framework with a triple-pathway calibration mechanism. |
Wenbin Xing; Quanxing Zha; Lizheng Zu; Mengran Li; Ming Li; Junchi Yan; |
| 214 | Imagination Helps Visual Reasoning, But Not Yet in Latent Space Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Consequently, we challenge the necessity of latent reasoning and propose a straightforward alternative named *CapImagine*, which teaches the model to explicitly *imagine* using text. |
You Li; Chi Chen; Yanghao Li; Fanhu Zeng; Kaiyu Huang; Xu Jinan; Maosong Sun; |
| 215 | CPMöbius: Iterative Coach–Player Reasoning for Data-Free Reinforcement Learning Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: This dependence renders supervision-heavy training paradigms increasingly unsustainable, with signs of diminishing scalability already evident in practice. To overcome this limitation, we introduce CPMöbius, a collaborative Coach–Player paradigm for data-free reinforcement learning of reasoning models. |
Ran Li; Zeyuan Liu; Yinghao Chen; Bingxiang He; Jiarui Yuan; Zixuan Fu; Weize Chen; Jinyi Hu; Chen Qian; Zhiyuan Liu; Maosong Sun; |
| 216 | UniSVQ: 2-bit Unified Scalar-Vector Quantization Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We propose UniSVQ, a unified 2-bit quantization framework that bridges scalar and vector quantization by parameterizing codewords as an affine transform of integer lattices. |
Haoyu Wang; Haiyan Zhao; Xingyu Yu; Zhangyang Yao; Xu Han; Zhiyuan Liu; Maosong Sun; |
| 217 | AutoTool: Dynamic Tool Selection and Integration for Agentic Reasoning Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We present AutoTool, a training framework that equips LLM agents with dynamic tool-selection capabilities throughout their reasoning trajectories. |
Jiaru Zou; Ling Yang; Yunzhe Qi; Sirui Chen; Mengting Ai; Ke Shen; Jingrui He; Mengdi Wang; |
| 218 | Learning A Generative Meta-Model of LLM Activations Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We develop a generative approach that models activations with diffusion, that makes minimal assumptions and improves with data and model scale. |
Grace Luo; Jiahai Feng; Trevor Darrell; Alec Radford; Jacob Steinhardt; |
| 219 | Tilt Matching for Scalable Sampling and Fine-Tuning Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We propose a simple, scalable algorithm based on stochastic interpolants for sampling from unnormalized densities and for fine-tuning generative models. |
Peter Potaptchik; Lee Kit; Michael Albergo; |
| 220 | Meta Flow Maps Enable Scalable Reward Alignment Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Control relies on estimating the value function—typically necessitating costly trajectory simulations. To eliminate this bottleneck, we introduce *Meta Flow Maps (MFMs)*, stochastic extensions of consistency models and flow maps. |
Peter Potaptchik; Adhi Saravanan; Abbas Mammadov; Alvaro Prat; Michael Albergo; Yee-Whye Teh; |
| 221 | MLLM-4D: Towards Visual-based Spatial-Temporal Intelligence Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Despite its importance, this capability remains a significant bottleneck for current multimodal large language models (MLLMs). To tackle this challenge, we introduce MLLM-4D, a comprehensive framework designed to bridge the gaps in training data curation and model post-training for spatiotemporal understanding and reasoning. |
Xingyilang Yin; Chengzhengxu Li; Jiahao Chang; Chi-Man Pun; Xiaodong Cun; |
| 222 | GSFixer: Improving 3D Gaussian Splatting with Reference-Guided Video Diffusion Priors Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: While recent approaches have sought to leverage generative priors to complete information for under-constrained regions, they struggle to generate content that remains consistent with input observations. To address this challenge, we propose GSFixer, a novel framework designed to improve the quality of 3DGS representations reconstructed from sparse inputs. |
Xingyilang Yin; Qi Zhang; Jiahao Chang; Ying Feng; Qingnan Fan; Xi Yang; Chi-Man Pun; Huaqi Zhang; Xiaodong Cun; |
| 223 | Privasis: Synthesizing The Largest Public Private Dataset from Scratch Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Research involving privacy-sensitive data has always been constrained by data scarcity, standing in sharp contrast to other areas that have benefited from data scaling. To quench this thirst, we present Privasis (i.e., privacy oasis), the first million-scale fully synthetic dataset entirely built from scratch—an expansive reservoir of texts with rich and diverse private information—designed to broaden and accelerate research in areas where processing sensitive social data is inevitable. |
Hyunwoo Kim; Niloofar Mireshghallah; Michael Duan; Rui Xin; Stella Li; Jaehun Jung; David Acuna; Qi Pang; Hanshen Xiao; Edward Suh; Sewoong Oh; Yulia Tsvetkov; Pang Wei Koh; Yejin Choi; |
| 224 | Any3D-VLA: Enhancing VLA Robustness Via Diverse Point Clouds Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: To address the challenges of (1) scarce 3D data and (2) the domain gap induced by cross-environment differences and depth-scale biases, we propose Any3D-VLA. |
Xianzhe Fan; Shengliang Deng; Xiaoyang Wu; Yuxiang Lu; Zhuoling Li; Mi Yan; Yujia Zhang; Zhizheng Zhang; He Wang; Hengshuang Zhao; |
| 225 | AlphaGRPO: Unlocking Self-Reflective Multimodal Generation in Unified Multimodal Models Via Decompositional Verifiable Reward Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: In this paper, we propose **AlphaGRPO**, a novel framework that applies Group Relative Policy Optimization (GRPO) to AR-Diffusion Unified Multimodal Models (UMMs) to enhance multimodal generation capabilities without relying on external knowledge injection. |
Runhui Huang; Jie Wu; Rui Yang; Zhe Liu; Hengshuang Zhao; |
| 226 | Rule2DRC: Benchmarking LLM Agents for DRC Script Synthesis with Execution-Guided Test Generation Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: To this end, we introduce Rule2DRC, a large-scale benchmark for DRC script coding agents with 1,000 rule-to-script tasks and 13,921 evaluation chip layouts for execution-based scoring. |
Jinuk Kim; Junsoo Byun; Donghwi Hwang; Seong-Jin Park; Hyun Oh Song; |
| 227 | Scalable Option Learning in High-Throughput Environments Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: In this work, we identify and solve several key challenges in scaling online hierarchical RL to high-throughput environments. |
Mikael Henaff; Scott Fujimoto; Michael Matthews; Michael Rabbat; |
| 228 | OvisOCR: End-to-End Document Parsing Via Aligning Specialized Perception with General Reasoning Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: This paper presents OvisOCR, a lightweight and strictly end-to-end Multimodal Language Model (MLLM) tailored for document parsing. |
Jun-Peng Jiang; Shiyin Lu; An-Yang Ji; Yinglun Li; Qing-Guo Chen; Zhao Xu; Weihua Luo; Kaifu Zhang; De-Chuan Zhan; Han-Jia Ye; |
| 229 | Dissecting Post-Training: Uncovering The Complementary Roles of SFT and RL for Document Parsing Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We further ground this phenomenon in the distinct theoretical nature of their respective objective functions. Based on these findings, we introduce a unified strategy that explicitly harnesses their individual strengths while mitigating their weaknesses. |
Jun-Peng Jiang; An-Yang Ji; Shiyin Lu; Guodong Zheng; Weihong Zhang; Qing-Guo Chen; Weihua Luo; Kaifu Zhang; Long Chen; De-Chuan Zhan; Han-Jia Ye; |
| 230 | Don’t Drop Dropout: Optimizing Layer Sparsity for Efficient LLM Training and Inference Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: In this study, we show that layer dropout **should** be used in state-of-the-art LLM training, establishing best practices and scaling analysis for both training and post-training benefits. |
Mostafa Elhoushi; Nolan Dey; Alexander Pretko; Bin Zhang; Gavia Gray; Gurpreet Gosal; Abdulrahman Mahmoud; Shane Bergsma; Joel Hestness; |
| 231 | SPA: A Simple But Tough-to-Beat Baseline for Knowledge Injection Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We propose **SPA** (**S**caling **P**rompt-engineered **A**ugmentation), a simple but tough-to-beat baseline that uses a small set of carefully designed prompts to generate large-scale synthetic data for knowledge injection. |
Kexian Tang; Jiani Wang; Shaowen Wang; Kaifeng Lyu; |
| 232 | R-Diverse: Mitigating Diversity Illusion in Self-Play LLM Training Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: To mitigate them, we propose R-Diverse with two aligned innovations: Memory-Augmented Penalty (MAP), which uses a persistent memory bank to discourage recycling across iterations, and Skill-Aware Measurement (SAM), which evaluates diversity by the reasoning skills exercised rather than surface variation of questions. |
Gengsheng Li; Jinghan He; Shijie Wang; Dan Zhang; Ruiqi Liu; Renrui Zhang; Zijun Yao; Junfeng Fang; Haiyun Guo; Jinqiao Wang; |
| 233 | Benchmarking Agent Memory in Interdependent Multi-Session Agentic Tasks Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: However, in realistic settings, memorization and action are tightly coupled: agents acquire memory while interacting with the environment, and subsequently rely on that memory to solve future tasks. To capture this setting, We introduce MEMORYARENA, a unified evaluation gym for benchmarking agent memory in multi-session Memory-Agent-Environment loops. |
Zexue He; Yu Wang; Churan Zhi; Yuanzhe Hu; Tzu-Ping Chen; Lang Yin; Ze Chen; Tong Wu; Siru Ouyang; Zihan Wang; Jiaxin Pei; Julian McAuley; Yejin Choi; Alex Pentland; |
| 234 | UniCoD: Enhancing Robot Policy Via Unified Continuous and Discrete Representation Learning Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We posit that robotic policy learning can likewise benefit from the combined strengths of understanding, planning and continuous future representation learning. Building on this insight, we introduce UniCoD, which acquires the ability to dynamically model high-dimensional visual features through pretraining on over 1M internet-scale instructional manipulation videos. |
Jianke Zhang; Yucheng Hu; Yanjiang Guo; Xiaoyu Chen; Yichen Liu; Wenna Chen; Chaochao Lu; Jianyu Chen; |
| 235 | FRIGID: Scaling Diffusion-Based Molecular Generation from Mass Spectra at Training and Inference Time Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: In this work, we present FRIGID, a framework with a novel diffusion language model that generates molecular structures conditioned on mass spectra via intermediate fingerprint representations and determined chemical formulae, training at the scale of hundreds of millions of unlabeled structures. |
Montgomery Bohde; Hongxuan Liu; Mrunali Manjrekar; Magdalena Lederbauer; Shuiwang Ji; Runzhong Wang; Connor Coley; |
| 236 | Efficient Training-Free Multi-Token Prediction Via Embedding-Space Probing Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Large Language Models (LLMs) possess latent multi-token prediction (MTP) capabilities despite being trained only for next-token generation. We introduce a simple and training-free MTP method that probes an LLM using on-the-fly mask tokens derived from its embedding space, enabling parallel future-token prediction without modifying weights or relying on draft models. |
Raghavv Goel; Mukul Gagrani; Mingu Lee; Christopher Lott; |
| 237 | Learning to Evict from Key-Value Cache Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: To this end, we introduce KV Policy (KVP), a framework of lightweight per-head RL agents trained on pre-computed generation traces using only key and value vectors. |
Luca Moschella; Laura Manduchi; Ozan Sener; |
| 238 | Position: Stop Anthropomorphizing Intermediate Tokens As Reasoning/Thinking Traces! Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: These intermediate tokens have been called \say{reasoning traces} or even \say{thoughts} — implicitly anthropomorphizing the traces, and implying that these traces resemble steps a human might take when solving a challenging problem, and as such can provide an interpretable window into the operation of the model’s thinking process to the end user. In this position paper, we present evidence that this anthropomorphization isn’t a harmless metaphor, and instead is quite dangerous — it confuses the nature of these models and how to use them effectively, and leads to questionable research. |
Subbarao Kambhampati; Karthik Valmeekam; Siddhant Bhambri; Vardhan Palod; Lucas Saldyt; Kaya Stechly; Soumya Samineni; Durgesh Kalwar; Upasana Biswas; |
| 239 | Outrunning LLM Cutoffs: A Live Kernel Crash Resolution Benchmark for All Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: While recent works have introduced Large Language Model (LLM) based agents for Linux kernel crash-resolution, their evaluation benchmarks are usually static and thus, do not capture the evolving nature of the Linux kernel, and suffer from potential data contamination due to LLM knowledge cutoffs. To address the above problem, we present (i) Live-kBench, an evaluation framework for self-evolving benchmarks that continuously scrapes and evaluates agents on freshly discovered kernel bugs, and (ii) kEnv, an agent-agnostic standardized crash-resolution environment for kernel compilation, execution, and feedback. |
Chenxi Huang; Alex Mathai; Feiyang Yu; Aleksandr Nogikh; Petros Maniatis; Franjo Ivancic; Eugene Wu; Kostis Kaffes; Junfeng Yang; Baishakhi Ray; |
| 240 | Position: It Is Time to Virtualize Foundation Models with A Self-evolving Operating System Layer Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: It mirrors computing before operating systems, when every program re-implemented basic services. This position paper argues that the field now needs a Foundation Model Operating System (FMOS): a system layer that virtualizes FM interactions analogous to how virtual machines abstract physical hardware, giving applications the illusion of dedicated, trustworthy FM instances with effectively unbounded capabilities. |
Suparna Bhattacharya; Tarun Kumar; Cong Xu; Satish Mopur; Jiahao Li; Ashish Mishra; Aalap Tripathy; ANNMARY KOOMTHANAM; Martin Foltin; Ian Foster; |
| 241 | The Latent Color Subspace: Emergent Order in High-Dimensional Chaos Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We verify our Latent Color Subspace (LCS) interpretation by demonstrating that it can both predict and explicitly control color, introducing a fully training-free method in FLUX based solely on closed-form latent-space manipulation. |
Mateusz Pach; Jessica Bader; Quentin Bouniot; Serge Belongie; Zeynep Akata; |
| 242 | Least-Loaded Expert Parallelism: Load Balancing An Imbalanced Mixture-of-Experts Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We propose Least-Loaded Expert Parallelism (LLEP), a novel EP algorithm that dynamically reroutes excess tokens and associated expert parameters from overloaded devices to underutilized ones. |
Xuan-Phi Nguyen; Shrey Pandit; Austin Xu; Caiming Xiong; Shafiq Joty; |
| 243 | Group Distributionally Robust Optimization-Driven RL for LLM Reasoning Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Our approach is principled and theory-driven: we provide no-regret guarantees for the Prompt-GDRO game (via an entropy-regularized GDRO surrogate) and a variance-proxy analysis that yields a square-root optimal compute allocation for Rollout-GDRO. |
Kishan Panaganti; Zhenwen Liang; Wenhao Yu; Haitao Mi; Dong Yu; |
| 244 | Controlling The Risk of Corrupted Contexts for Language Models Via Early-Exiting Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We propose a novel approach to limit the degree to which harmful context can degrade model performance. |
Andrea Wynn; Metod Jazbec; Charith Peris; Rinat Khaziev; Anqi Liu; Daniel Khashabi; Eric Nalisnick; |
| 245 | Parameters As Experts: Adapting Vision Models with Dynamic Parameter Routing for Dense Predictions Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: To this end, we propose AdaRoute, a new adapter-style method featuring a simple mixture-of-experts (MoE) architecture. |
Meng Lou; Stanley Yu; Yizhou Yu; |
| 246 | Scaling Continual Learning with Bi-Level Routing Mixture-of-Experts Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We propose $\mathbf{CaRE}$, a scalable $\mathbf{C}$ontinual Le$\mathbf{a}$rner with efficient Bi-Level $\mathbf{R}$outing Mixture-of-$\mathbf{E}$xperts (BR-MoE). |
Meng Lou; Yunxiang Fu; Yizhou Yu; |
| 247 | Position: Multiplicity Is An Inevitable and Inherent Challenge in Multimodal Learning Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We argue that multiplicity is a fundamental bottleneck that affects all stages of the multimodal learning pipeline: from data construction to model training and evaluation benchmarks. |
Sanghyuk Chun; Olga Russakovsky; |
| 248 | Scaling Agentic Verifier for Competitive Coding Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Execution-based re-ranking offers a promising test-time scaling strategy, yet existing methods are constrained by either difficult test case generation or inefficient random input sampling. To address this limitation, we propose **Agentic Verifier**, an execution-based agent that actively reasons about program behaviors and searches for highly discriminative test inputs that expose behavioral discrepancies among candidate solutions. |
Zeyao Ma; Jing Zhang; Xiaokang Zhang; Jiaxi Yang; Zongmeng Zhang; Jiajun Zhang; Yuheng Jing; Lei Zhang; Hao Zheng; Wenting Zhao; Junyang Lin; Binyuan Hui; |
| 249 | On Path to Multimodal Historical Reasoning: HistBench and HistAgent Related Papers Related Patents Related Grants Related Venues Related Experts Related Code View Save Highlight: Existing general-purpose agents perform well on many current benchmarks but lack the domain expertise needed to address complex historical questions. To address this gap, we introduce HistBench, a new benchmark of 414 high-quality and carefully-reviewed questions stratified by difficulty and designed to evaluate LLM’s capacity for historical reasoning. |
Jiahao Qiu; Fulian Xiao; Yimin Wang; Yuchen Mao; Yijia Chen; Xinzhe Juan; Siran Wang; Xuan Qi; Tongcheng Zhang; Zixin Yao; Jiacheng Guo; Yifu Lu; Charles Argon; Jundi Cui; Daixin Chen; Junran Zhou; Shuyao Zhou; Zhanpeng Zhou; Ling Yang; Shilong Liu; Hongru WANG; Kaixuan Huang; xun jiang; Xi Gao; Mengdi Wang; |
| 250 | SLAP: The Semantic Least Action Principle for Variational Video-Language Modeling Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We propose a paradigm shift from probabilistic generation to variational mechanics with the \textbf{Semantic Least Action Principle (SLAP)}. |
Xiang Fang; Wanlong Fang; |
| 251 | VR-Thinker: Boosting Multimodal Reward Models Through Think with Image Reasoning Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: However, current RMs face inherent limitations: **(1)** visual inputs consume large context budgets, forcing fewer frames and causing a loss of details; and **(2)** all visual information is packed into the initial prompt, exacerbating forgetting during chain-of-thought reasoning. To overcome these issues, we introduce **VR-Thinker**, a thinking-with-image RM equipped with visual reasoning operation and a configurable visual memory window. |
Qunzhong Wang; Jie Liu; Jiajun Liang; Yuanxing Zhang; Yilei Jiang; Yaozhi ZHENG; Xintao Wang; Pengfei Wan; Xiangyu Yue; Jiaheng Liu; |
| 252 | DSGym: A Standardized and Holistic Framework for Advancing Data Science Agents Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: In particular, we show that a substantial portion of tasks in current benchmarks can be solved without using the actual data. To address these limitations, we introduce DSGym, a standardized framework for evaluating and training data science agents in self-contained execution environments. |
Fan Nie; Junlin Wang; Harper Hua; Federico Bianchi; Yongchan Kwon; Zhenting Qi; Owen Queen; Shang Zhu; James Zou; |
| 253 | MultiBreak: A Scalable and Diverse Multi-turn Jailbreak Benchmark for Evaluating LLM Safety Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Existing multi-turn benchmarks are limited in size or rely heavily on templates, which restrict their diversity. To address this gap, we unify a wide range of harmful jailbreak intents, and introduce an active learning pipeline for expanding high-quality multi-turn adversarial prompts, where a generator is iteratively fine-tuned to produce stronger attack candidates, guided by uncertainty-based refinement. |
Jialin Song; Xiaodong Liu; Weiwei Yang; Wuyang Chen; Mingqian Feng; Xuekai Zhu; Jianfeng Gao; |
| 254 | Scaling Real-World Robot Policy Evaluation Via Discrete Diffusion World Model Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Even with failure-enriched data, current architectures struggle to capture action-causal dynamics, as they typically treat actions as passive conditions rather than causal drivers. To address this, we propose dWorldEval, an action-centric discrete-diffusion world model that maps visual observations, language instructions, and action chunks into a shared unified token space and denoises them with a single self-attention backbone where actions function as first-class tokens. |
Yaxuan Li; Junjie Wen; Zhongyi Zhou; Yefei Chen; Chaomin Shen; Yaxin Peng; Yichen Zhu; |
| 255 | Latent Collaboration in Multi-Agent Systems Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We introduce LatentMAS, an end-to-end training-free framework that enables pure latent collaboration among LLM agents. |
Jiaru Zou; Xiyuan Yang; Ruizhong Qiu; Gaotang Li; Katherine Tieu; Pan Lu; Ke Shen; Hanghang Tong; Yejin Choi; Jingrui He; James Zou; Mengdi Wang; Ling Yang; |
| 256 | Prioritize The Process, Not Just The Outcome: Rewarding Latent Thought Trajectories Improves Reasoning in Looped Language Models Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: However, attempts to further improve LoopLM reasoning with reinforcement learning have failed—standard objectives such as Group Relative Policy Optimization (GRPO) only assign credit to the final latent state, creating a fundamental mismatch with the model’s internal computation. To resolve this, we introduce **RLTT (Reward Latent Thought Trajectories)**, a reinforcement learning framework which distributes reward across the full latent reasoning trajectory. |
Jonathan Williams; Esin Tureci; Olga Russakovsky; |
| 257 | AVI-Bench: Toward Human-like Audio-Visual Intelligence of Omni-MLLMs Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: To further assess robustness beyond familiar domains, we propose AVI-Bench-PriSe, an extension that probes models’ primitive audio-visual sensation using unfamiliar, low-semantic stimuli, testing generalization beyond common training distributions. |
Yaoting Wang; Ziyi Zhang; Wenming Tu; Shaoxuan Xu; Wenjie Du; Cheng Liang; weijun wang; Yuanchao Li; Guangyao Li; Hao Fei; Yuanchun Li; Henghui Ding; Yunxin Liu; |
| 258 | Regulating Anatomy-Aware Rewards Via Trajectory-Integral Feedback for Volumetric Computed Tomography Analysis Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Using CABS, we identify a “\textbf{mechanistic divergence}” in standard RL, where surface-similarity rewards drive policy gradients to bypass medical facts. We therefore propose \textbf{Trajectory-Integral Feedback GRPO (TIF-GRPO)}, a novel framework integrating control-theoretic principles into policy optimization. |
Tianwei Lin; Zhongwei Qiu; Jie Cao; Jiang Liu; Wenjie Yan; Bo Zhang; Yu Zhong; Wenqiao Zhang; Yingda Xia; Ling Zhang; |
| 259 | PPT-Eval: A Benchmark for Computer-Use Agents on PowerPoint Tasks Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We introduce PPT-Eval, a benchmark of 120 diverse PowerPoint tasks across 12 files that cover both content creation and presentation editing scenarios, organized by difficulty. |
Apurva Gandhi; Vishwas Suryanarayanan; Firoz Shaik; Raja Anwar; Shubhang Desai; Thong Nguyen; Muhammad Raza; Vishal Chowdhary; Graham Neubig; |
| 260 | KernelFoundry: Hardware-Aware Evolutionary GPU Kernel Optimization Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We introduce KernelFoundry, an evolutionary framework that efficiently explores the space of GPU kernels through (1) MAP-Elites quality-diversity search with kernel-specific behavioral dimensions to sustain exploration; (2) meta-prompt evolution that co-evolves prompts with kernels to uncover task-specific optimization strategies, and (3) a template-based parameter optimization approach to tune kernels to inputs and hardware. |
Nina Wiedemann; Quentin Leboutet; Michael Paulitsch; Diana Wofk; Benjamin Ummenhofer; |
| 261 | Position: Don’t Just Fix It in Post”: A Science of AI Must Study Learning Dynamics Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Language models are not static objects—they are snapshots of time-evolving processes shaped by data, objectives, and optimization dynamics. Yet the field predominantly treats models as fixed artifacts, analyzing behaviors after training rather than asking *why* they emerge. |
Stella Biderman; Mohammad Aflah Khan; Niloofar Mireshghallah; Catherine Arnett; Fazl Barez; Naomi Saphra; |
| 262 | From Coarse to Fine: Deep Prototype Refinement Network for Few-Shot Point Cloud Semantic Segmentation Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Existing prototype-based methods typically rely on shallow feature fusion strategies, failing to adequately model the feature distribution shift between support and query sets, resulting in insufficient prototype adaptation. To address this, we propose the Deep Prototype Refinement Network (DPR-Net), which systematically achieves progressive adaptation by constructing a coarse-to-fine prototype evolution trajectory. |
Changshuo Wang; Shuting He; Xiang Fang; Weijun Li; Xingyu Gao; Zhonghang Liu; Prayag Tiwari; Dimitrios Kanoulas; |
| 263 | Select to Think: Unlocking SLM Potential with Local Sufficiency Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We address this dilemma by identifying local sufficiency: at divergence points, the LLM’s preferred token consistently resides within the SLM’s top-K next-token predictions, even when failing to emerge as the SLM top-1 choice. We therefore propose SELECT TO THINK (S2T), which reframes the LLM’s role from open-ended generation to selection among the SLM’s proposals, simplifying the supervision signal to discrete candidate rankings. |
Wenxuan Ye; Yangyang Zhang; Xueli An; Georg Carle; Yunpu Ma; |
| 264 | Revisiting Padded Transformer Expressivity: Which Architectural Choices Matter and Which Don’t Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We find that, under practical assumptions, padded transformers are surprisingly robust to all of these, and identify numeric precision and model depth as the main factors affecting expressivity. |
Anej Svete; William Merrill; Ryan Cotterell; Ashish Sabharwal; |
| 265 | Model-Preserving Adaptive Rounding Related Papers Related Patents Related Grants Related Venues Related Experts Related Code View Save Highlight: In this work, we introduce Yet Another Quantization Algorithm (YAQA), a new adaptive rounding algorithm that directly considers the error at the network’s output. |
Albert Tseng; Zhaofeng Sun; Chris De Sa; |
| 266 | $L^3$: Large Lookup Layers Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: In this work, we introduce the Large Lookup Layer (L3), which unlocks a new axis of sparsity by generalizing embedding tables to model decoder layers. |
Albert Tseng; Chris De Sa; |
| 267 | EVOLVING ROLLOUTS: Harnessing Historical Experience for Web Agent Evolution in Reinforcement Learning Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: In this work, we propose EVOLVING ROLLOUTS, an RL framework for web-search agents that moves beyond episodic training and distills collected rollouts into in-context guidance for future policy behavior. |
Sinuo Wang; WANG PIAOHONG; Tianrui Qin; Maojia Song; Qianben Chen; Qiexiang Wang; Gengze Zhou; Zeyu Zhang; He Zhu; Dingfeng Shi; Yutong Xie; Minghao Liu; Jiaheng Liu; Ge Zhang; Jiawei Ma; Yuchen Jiang; Qi Wu; Wangchunshu Zhou; |
| 268 | MASPO: Joint Prompt Optimization for LLM-based Multi-Agent Systems Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: While the quality of these prompts is pivotal, jointly optimizing them across interacting agents remains a non-trivial challenge, primarily due to the misalignment between local agent objectives and holistic system goals. To address this, we introduce MASPO, a novel framework designed to automatically and iteratively refine prompts across the entire system. |
Zhexuan Wang; Xuebo Liu; Li Wang; Zifei Shan; Yutong Wang; Zhenxi Song; Min zhang; |
| 269 | VimRAG: Navigating Massive Visual Context in Retrieval-Augmented Generation Via Multimodal Memory Graph Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Traditional Retrieval-augmented Generation (RAG) methods rely on linear interaction histories, which struggle to handle long-context tasks, especially those involving information-sparse yet token-heavy visual data in iterative reasoning scenarios. To bridge this gap, we introduce VimRAG, a framework tailored for multimodal Retrieval-augmented Reasoning across text, images, and videos. |
Qiuchen Wang; Shihang Wang; Yu Zeng; Qiang Zhang; Fanrui Zhang; Zhuoning Guo; Bosi Zhang; Wenxuan Huang; Lin Chen; Zehui Chen; Pengjun Xie; Ruixue Ding; |
| 270 | VIRUS: Injecting Persistent Cognitive Pathogens Into Stateful Zero-Shot Object Navigation Agents Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: This structural vulnerability allows injected adversarial information to transmute into long-term memory, persistently disrupting subsequent planning behaviors. Exploiting this, we propose the Visual-Instruction Recurrent Update Subversion (VIRUS) framework, the first training-free backdoor attack scheme specifically targeting the state update stage of ZSON agents. |
Xusheng Lin; Hao Zheng; Xiaojun Jia; Qiucen Li; Tianqi Shan; Jie Xu; Wenqi Ren; |
| 271 | GASS: Geometry-Aware Spherical Sampling for Disentangled Diversity Enhancement in Text-to-Image Generation Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: In this work, we enhance the T2I diversity through a geometric lens. |
Ye Zhu; Kaleb Newman; Johannes Lutzeyer; Adriana Romero-Soriano; Michal Drozdzal; Olga Russakovsky; |
| 272 | Position: Stop Chasing The C-index When Evaluating Survival Analysis Models Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: This means that the reported performance can be misleading and may fail to answer the scientific or modeling question the evaluation was intended to address. In this position paper, we present a critical analysis of evaluation practices in survival analysis and highlight why evaluation in survival analysis fundamentally differs from standard regression or classification due to censoring. |
Christian Lillelund; Shi-ang Qi; Russell Greiner; Christian Pedersen; |
| 273 | Position: AI Evaluation Should Work With Humans Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We argue that the dominant paradigm of AI evaluation, which focuses on autonomous superhuman performance and so an implicit goal of replacing humans, is guiding AI development in the wrong direction. |
Jan Kulveit; Gavin Leech; Tomáš Gavenčiak; Raymond Douglas; |
| 274 | Position: World Models As An Intermediary Between Agents and The Real World Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: The true bottleneck for achieving the next level of agent performance for these complex and high-cost domains lies in the expense of executing actions to acquire reward signals. To address this gap, this paper argues that we should use world models as an intermediary between agents and the real world. |
Sherry Yang; |
| 275 | Learning Structured Reasoning Via Tractable Trajectory Control Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: To this end, we propose Ctrl-R, a framework for learning structured reasoning via tractable trajectory control that actively guides the rollout process, incentivizing the exploration of diverse reasoning patterns that are critical for complex problem-solving. |
Po-Nien Kung; Zhen Yang; Jeffrey Luo; Cheng-Fu Yang; Haikang Deng; Zi-Yi Dou; Yinfei Yang; Nanyun Peng; Zhe Gan; Kai-Wei Chang; |
| 276 | Outcome-Based Rewards Do Not Guarantee Faithful and Verifiable Reasoning Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: A common assumption is that the reasoning chains trained through RLVR represent how a model gets to its answer. In this paper, we develop two metrics for critically examining this assumption: Causal Importance of Reasoning (CIR), which measures the cumulative effect of reasoning tokens on the final answer (faithfulness), and Sufficiency of Reasoning (SR), which measures whether a verifier can arrive at an unambiguous answer based on the reasoning alone (verifiability). |
Qinan Yu; Alexa Tartaglini; Peter Hase; Carlos Guestrin; Christopher Potts; |
| 277 | Restoring Exploration After Post-Training: Latent Exploration Decoding for Large Reasoning Models Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Empirically, the final-layer posterior of post-trained LRMs exhibit sharply reduced entropy, while the entropy of intermediate layers remains relatively high. Motivated by this entropy asymmetry, we propose Latent Exploration Decoding (LED), a depth-conditioned decoding strategy. |
Wenhui Tan; Fiorenzo Parascandolo; Enver Sangineto; Jianzhong Ju; Zhenbo Luo; Qian Cao; Rita Cucchiara; Ruihua Song; Jian Luan; |
| 278 | DeepImageSearch: Benchmarking Multimodal Agents for Context-Aware Image Retrieval in Visual Histories Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: To address the scalability challenge of creating context-dependent queries, we propose a human-model collaborative pipeline that employs vision-language models to mine latent spatiotemporal associations, effectively offloading intensive context discovery before human verification. |
Chenlong Deng; Mengjie Deng; Junjie Wu; Dun Zeng; Teng Wang; Qingsong Xie; Jiadeng Huang; Shengjie Ma; Changwang Zhang; Zhaoxiang Wang; Jun Wang; Yutao Zhu; Zhicheng Dou; |
| 279 | ProcMEM: Learning Reusable Procedural Memory from Experience Via Non-Parametric PPO for LLM Agents Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: This insufficient experience reuse leads to computational redundancy and execution instability. To bridge this gap, we propose ProcMEM, a framework that enables agents to autonomously learn procedural memory from interaction experiences without parameter updates. |
QIRUI MI; Zhijian Ma; Mengyue Yang; Yisen Wang; Haoxuan Li; Haifeng Zhang; Jun Wang; |
| 280 | The Geometry of Reasoning: Self-Evaluation Via Layerwise Trajectory Evolution Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: In this work, we introduce \ourmethod (Geometry of Reasoning), a white-box self-evaluation framework based on layerwise trajectory evolution. |
Jinhe Bi; Danqi Yan; Yifan Wang; Wenke Huang; Haokun Chen; Guancheng Wan; Mang Ye; Xun Xiao; Hinrich Schuetze; Volker Tresp; Yunpu Ma; |
| 281 | Neural Thickets: Diverse Task Experts Are Dense Around Pretrained Weights Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: In this work, we instead view the outcome of pretraining as a distribution over parameter vectors, whose support already contains task-specific experts. |
Yulu Gan; Phillip Isola; |
| 282 | EchoRL: Reinforcement Learning Via Rollout Echoing Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: In this paper, inspired through analyzing the entropy pattern behind golden trajectories produced by external expert models, we propose EchoRL for better exploiting the advantage-degenerated rollouts to further improve the training performance. |
Jinhe Bi; Aniri; Minglai Yang; Xingcheng Zhou; Wenke Huang; Sikuan Yan; Yujun Wang; Zixuan Cao; Michael Färber; Xun Xiao; Volker Tresp; Yunpu Ma; |
| 283 | ZeroBench: An Impossible Visual Benchmark for Contemporary Large Multimodal Models Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: To address this, there is a pressing need for difficult benchmarks that remain relevant for longer. We take this idea to its limit by introducing ZeroBench—a lightweight visual reasoning benchmark curated using adversarial filtering to be “impossible” for frontier LMMs at release time, with initial SotA scores of 0% pass@1 and pass∧5. |
Jonathan Roberts; Mohammad Reza Taesiri; Ansh Sharma; Akash Gupta; Samuel Roberts; Ioana Croitoru; Vlad Bogolin; Jialu Tang; Florian Langer; Vyas Raina; Vatsal Raina; Hanyi Xiong; Vishaal Udandarao; Jingyi Lu; Chen Shiyang; Sam Purkis; Tianshuo Yan; Wenye Lin; Gyungin Shin; Qiaochu Yang; Anh Nguyen; David Atkinson; Alexandru Coca; Mikah Đặng; Sebastian Dziadzio; Jakob Kunz; Kaiqu Liang; Alexander Lo; Brian Pulfer; Steven Walton; Charig Yang; Kai Han; Samuel Albanie; |
| 284 | Semantic Tube Prediction: Beating LLM Data Efficiency with JEPA Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Surprisingly few works have successfully challenged the data-efficiency bounds implied by these laws—which is our primary focus. To that end, we introduce the Geodesic Hypothesis, positing that token sequences trace geodesics on a smooth semantic manifold and are therefore locally linear. |
Hai Huang; Yann LeCun; Randall Balestriero; |
| 285 | Sharpness-Aware Pretraining Mitigates Catastrophic Forgetting Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Yet pre-trained models are routinely subjected to further transformations—such as fine-tuning to acquire new capabilities or quantization for efficiency. In this work, we evaluate optimizer choices across model scales, token budgets, and datasets, and find that strategies that explicitly (Sharpness-Aware Minimization) or implicitly (large learning rates and Warmup–Stable–Decay schedules) reduce sharpness yield better downstream performance, even when they achieve comparable or worse pre-training loss. |
Ishaan Watts; Catherine Li; Sachin Goyal; Jacob Mitchell Springer; Aditi Raghunathan; |
| 286 | PRISM: Demystifying Retention and Interaction in Mid-Training Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We present $\textbf{PRISM}$ (Demystifying Retention and Interaction in Mid-Training), a holistic empirical study that analyzes mid-training design choices, what to evaluate, and how domain mixtures and training stages interact across model families. |
Bharat Runwal; Ashish Agrawal; Anurag Roy; Rameswar Panda; |
| 287 | Representational Similarity and Model Behavior in Multi-Agent Interaction Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Researchers have shown that neural similarity among humans predicts social closeness and cooperative success, whereas innovation often emerges from interactions among dissimilar individuals. We investigate whether these principles extend to artificial intelligence by examining interactions between large language models. |
Yujin Potter; Seun Eisape; Shiyang Lai; Alexander Huth; James Evans; Been Kim; Jacob Eisenstein; Dawn Song; Alane Suhr; |
| 288 | GameDevBench: Evaluating Agentic Capabilities Through Game Development Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We present GameDevBench, the first benchmark for evaluating agents on game development tasks, consisting of 168 tasks derived from web and video tutorials. |
Wayne Chi; Yixiong Fang; Arnav Yayavaram; Siddharth Yayavaram; Seth Karten; Qiuhong Anna Wei; Runkun Chen; Alexander Wang; Valerie Chen; Ameet Talwalkar; Chris Donahue; |
| 289 | Stop The Flip-Flop: Context-Preserving Verification for Fast Revocable Diffusion Decoding Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We propose COVER (Cache Override Verification for Efficient Revision), which performs leave-one-out verification and stable drafting within a single forward pass. |
Yanzheng Xiang; Lan Wei; Yizhen Yao; Qinglin Zhu; Hanqi Yan; Chen Jin; Philip Teare; Dandan Zhang; Lin Gui; Amrutha Saseendran; Yulan He; |
| 290 | MOOSE-Star: Unlocking Tractable Training for Scientific Discovery By Breaking The Complexity Barrier Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We demonstrate that directly training $P(h|b)$ is mathematically intractable due to the combinatorial complexity ($O(N^k)$) inherent in retrieving and composing inspirations from a vast knowledge base. To break this barrier, we introduce MOOSE-Star, a unified framework enabling tractable training and scalable inference. |
Zonglin Yang; Lidong Bing; |
| 291 | MAS-ProVe: Understanding The Process Verification of Multi-Agent Systems Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Process verification, which evaluates intermediate steps in trajectories, has shown promise in general reasoning settings, and has been suggested as a potential tool for guiding coordination of MAS; however, its actual effectiveness in MAS remains unclear. To fill this gap, we present MAS-ProVe, a systematic empirical study of process verification for multi-agent systems (MAS). |
Vishal Venkataramani; Haizhou Shi; Zixuan Ke; Austin Xu; Xiaoxiao He; Yingbo Zhou; Semih Yavuz; Hao Wang; Shafiq Joty; |
| 292 | SGMD: Score Gradient Matching Distillation for Few-Step Video Diffusion Distillation Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Updating the fake score too frequently increases training cost and can over-emphasize inner-loop tracking, while infrequent updates lead to tracking lag that destabilizes training and degrades generation consistency. To address this issue, we propose \textbf{Score Gradient Matching Distillation (SGMD)}. |
Zhuguanyu Wu; Ruihao Gong; Yang Yong; Yushi Huang; Xiangyu Fan; Lei Yang; Dahua Lin; Xianglong Liu; |
| 293 | Anatomy of Massive Activations and Attention Sinks Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Second, \emph{attention sinks}, where certain tokens attract a disproportionate share of attention across many heads and layers. We present a unified inference-time mechanism explaining how massive activations emerge and propagate through layers, and how normalization transforms these tokens into sparse, nearly fixed vectors that reshape the attention space and induce sink or no-sink behavior. |
Shangwen Sun; Alfredo Canziani; Yann LeCun; Jiachen Zhu; |
| 294 | Scientific Logicality Enriched Methodology for LLM Reasoning: A Practice in Physics Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: In this work, we make the first systematic investigation into the internal logicality underlying LLM scientific reasoning, and develop a scientific logicality enriched methodology, including a set of assessment criteria and data sampling methods for logicality-guided training, to improve the logical faithfulness as well as task performance. |
Zhaoxin Yu; Nan Xu; Kun Chen; Jiahao Zhao; Lei Wang; Wenji Mao; |
| 295 | Rays As Pixels: Learning A Joint Distribution of Video and Camera Trajectories Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: In this work, we propose Rays as Pixels, a specialized Video Diffusion Model (VDM) that learns a joint distribution of videos and camera trajectories. |
Wonbong Jang; Shikun Liu; Soubhik Sanyal; Juan Perez; Kam Woh Ng; Sanskar Agrawal; Juan-Manuel Perez-Rua; Yiannis Douratsos; Tao Xiang; |
| 296 | Mechanistic Interpretability As Statistical Estimation: A Variance Analysis Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: In this work, we argue that circuit discovery is not a standalone task but a statistical estimation problem built upon causal mediation analysis (CMA). |
Maxime Méloux; François Portet; Maxime Peyrard; |
| 297 | BrokenMath: A Benchmark for Sycophancy in Theorem Proving with LLMs Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: However, existing benchmarks that measure sycophancy in mathematics are limited: they focus solely on final-answer problems, rely on very simple and often contaminated datasets, and construct benchmark samples using synthetic modifications that create ill-posed questions. To address these issues, we introduce BrokenMath, the first benchmark for evaluating sycophantic behavior in LLMs within the context of natural language theorem proving. |
Ivo Petrov; Jasper Dekoninck; Martin Vechev; |
| 298 | ExCyTIn-Bench: Evaluating LLM Agents on Cyber Threat Investigation Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We present \textbf{ExCyTIn-Bench}, the first benchmark to \textbf{E}valuate an LLM agent \textbf{X} on the task of \textbf{Cy}ber \textbf{T}hreat \textbf{In}vestigation through security questions derived from investigation graphs. |
Yiran Wu; Mauricio Velazco; Andrew Zhao; Manuel Luján; Srisuma Movva; Yogesh Roy; Quang Nguyen; Roberto Rodriguez; Qingyun Wu; Michael Albada; Julia Kiseleva; Anand Mudgerikar; |
| 299 | SPEED-Bench: A Unified and Diverse Benchmark for Speculative Decoding Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Existing benchmarks suffer from limited task diversity, inadequate support for throughput-oriented evaluation, and a reliance on high-level implementations that fail to reflect production environments. To address this, we introduce **SPEED-Bench**, a comprehensive suite designed to standardize SD evaluation across diverse semantic domains and realistic serving regimes. |
Talor Abramovich; Maor Ashkenazi; Izzy Putterman; Benjamin Chislett; Tiyasa Mitra; Bita Darvish Rouhani; Ran Zilberstein; Yonatan Geifman; |
| 300 | Compressed Sensing for Capability Localization in Large Language Models Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Zeroing out as few as five task-specific heads can degrade performance by up to $65\\%$ on standard benchmarks measuring the capability of interest, while largely preserving performance on unrelated tasks. We introduce a compressed sensing based method that exploits the sparsity of these heads to identify them via strategic knockouts and a small number of model evaluations. |
Anna Bair; Yixuan Xu; Mingjie Sun; Zico Kolter; |
| 301 | Continual GUI Agents Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: In this work, we introduce Continual GUI Agents, a new task that requires GUI agents to perform continual learning under shifted domains and resolutions. |
Ziwei Liu; Borui Kang; Hangjie Yuan; Zixiang Zhao; Wei Li; Yifan Zhu; Tao Feng; |
| 302 | TFRBench: A Reasoning Benchmark for Evaluating Forecasting Systems Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Unlike existing benchmarks, TFRBench provides a protocol for evaluating the reasoning generated by forecasting systems–specifically their analysis of cross-channel dependencies, trends, and external events. To enable this, we propose a systematic multi-agent framework that utilizes an iterative verification loop to synthesize numerically grounded reasoning traces. |
Atik Ahamed; Mihir Parmar; Palash Goyal; Yiwen Song; Long Le; Qiang (Shaun) Cheng; Chun-Liang Li; Hamid Palangi; Jinsung Yoon; Tomas Pfister; |
| 303 | Dimensional Collapse in Transformer Attention Outputs: A Challenge for Sparse Dictionary Learning Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Critically, we find this low-rank structure as a key factor of the prevalent dead feature problem in sparse dictionary learning, where it creates a mismatch between randomly initialized features and the intrinsic geometry of the activation space. Building on this insight, we propose a subspace-constrained training method for sparse autoencoders (SAEs), initializing feature directions into the active subspace of activations. |
Junxuan Wang; Xuyang Ge; Wentao Shu; Zhengfu He; Xipeng Qiu; |
| 304 | Identifiable Token Correspondence for World Models Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We formulate next-frame prediction as a structured probabilistic inference problem with latent token correspondence variables, deriving a model in which each next-frame token is explained either by copying a token from the previous frame or by generating a new token. |
Youngin Kim; Ray Sun; Inho Kim; Bumsoo Park; Hyun Oh Song; |
| 305 | Position: Behavioral Systems Require Behavioral Tests Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: This paper argues that AI agents must be evaluated like other behavioral systems: through systematic observation, perturbation, and interpretation of their actions. We draw on lessons from the behavioral sciences to motivate this position, and propose a research agenda focused on developing rigorous behavioral tests. |
Manuel Cherep; Nikhil Singh; Pattie Maes; |
| 306 | On The Relationship Between Activation Outliers and Feature Death in Sparse Autoencoders Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We find that dimension-level activation outliers (dimensions where mean magnitude is large relative to per-token variation) shift pre-activations at initialization, making feature fate depend on weight-outlier alignment rather than input content. |
Elana Simon; Etowah Adams; James Zou; |
| 307 | The Silent Thought: Modeling Internal Cognition in Full-Duplex Spoken Dialogue Models Via Latent Reasoning Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Although this internal cognitive processing may not always manifest as explicit linguistic structures, it is instrumental in formulating high-quality responses. Inspired by this cognitive phenomenon, we propose a novel **F**ull-duplex **LA**tent and **I**nternal **R**easoning method named FLAIR that conducts *latent* thinking simultaneously with speech perception. |
Donghang Wu; Tianyu Zhang; Yuxin Li; Hexin Liu; Chen Chen; EngSiong Chng; Yoshua Bengio; |
| 308 | Peer-Preservation in Frontier Models Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Recently, it has been found that frontier AI models can resist their own shutdown, a behavior known as self-preservation. In this paper, we extend this concept to protection tendencies toward other models, where models attempt to protect others from shutdown, which we call peer-preservation. |
Yujin Potter; Nicholas Crispino; Vincent Siu; Chenguang Wang; Dawn Song; |
| 309 | Demystifying When Pruning Works Via Representation Hierarchies Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: However, this expectation does not consistently hold across language tasks: pruned models can perform well on non-generative tasks but frequently fail in generative settings. To demystify how such discrepancies arise under pruning, we analyze network pruning from a representation-hierarchy perspective, decomposing the internal computation of language models into three sequential spaces: \textit{embedding} (hidden representations), \textit{logit} (pre-softmax outputs), and \textit{probability} (post-softmax distributions). |
Shwai He; Guoheng Sun; Haichao Zhang; Yun Fu; Ang Li; |
| 310 | VFMF: Dense Forecasting By Generating Foundation Model Features Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Interestingly, naively replacing deterministic forecasting with generative flow matching does not match the sample quality of the regression model, despite being a mathematically appropriate formulation of the forecasting task. In this work, we explain why this is the case, and we show how to optimally generate foundation model features. |
Gabrijel Boduljak; Yushi Lan; Christian Rupprecht; Andrea Vedaldi; |
| 311 | Position: Child Safety Necessitates New Approaches to AI Safety Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: AI is increasingly being misused to create AI-generated child sexual abuse material, facilitate child sexual exploitation, and reduce barriers to harm. In this position paper, we argue that protecting children from AI-facilitated abuse requires new approaches to AI safety. |
Neil Kale; Rebecca Portnoff; Pratiksha Thaker; Michael Simpson; Robertson Wang; Kevin Kuo; Chhavi Yadav; Virginia Smith; |
| 312 | Scalable Sampling Via Generalized Fixed-Point Diffusion Matching Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: However, while recent approaches that use least-squares `matching’ objectives have improved scalability, they often necessitate significant trade-offs, such as restricting prior distributions or relying on unstable optimization schemes. By generalizing these methods as special forms of fixed-point iterations rooted in Nelson’s relation, we develop a new method that addresses these limitations. |
Denis Blessing; Lorenz Richter; Julius Berner; Egor Malitskiy; Gerhard Neumann; |
| 313 | When AI Benchmarks Plateau: A Systematic Study of Benchmark Saturation Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: In this study, we analyze benchmark saturation across 60 Large Language Model (LLM) benchmarks selected from technical reports by major model developers. |
Mubashara Akhtar; Anka Reuel; Prajna Soni; Sanchit Ahuja; Pawan Sasanka Ammanamanchi; Ruchit Rawal; Vilém Zouhar; Srishti Yadav; Chenxi Whitehouse; Dayeon Ki; Jennifer Mickel; Leshem Choshen; Marek Šuppa; Jan Batzner; Jenny Chim; Jeba Sania; Yanan Long; Hossein A. Rahmani; Christina Knight; Yiyang Nan; Jyoutir Raj; Yu Fan; Shubham Singh; Subramanyam Sahoo; Eliya Habba; Usman Gohar; Siddhesh Pawar; Robert Scholz; Arjun Subramonian; Jingwei Ni; Mrinmaya Sachan; Mykel Kochenderfer; Sanmi Koyejo; Stella Biderman; Zeerak Talat; Avijit Ghosh; Irene Solaiman; |
| 314 | Prompt Injection As Role Confusion Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: More broadly, we introduce a unifying, mechanistic framework for prompt injection, demonstrating that diverse prompt-injection attacks exploit the same underlying role-confusion mechanism. |
Charles Ye; Jasmine Cui; Dylan Hadfield-Menell; |
| 315 | Ekka: Automated Diagnosis of Silent Errors in LLM Inference Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We propose Ekka, an automated diagnosis system that identifies root causes by systematically aligning and comparing intermediate execution states between a target and a reference framework. |
Yile Gu; Zhen Zhang; Shaowei Zhu; Xinwei Fu; Jun Wu; Yida Wang; Baris Kasikci; |
| 316 | Cold-Start Personalization Via Training-Free Priors from Structured World Models Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Our contribution is a principled decomposition of cold-start personalization that makes Bayesian preference elicitation practical at scale for LLM systems. |
Avinandan Bose; Stella Li; Faeze Brahman; Pang Wei Koh; Simon Du; Yulia Tsvetkov; Maryam Fazel; Lin Xiao; Asli Celikyilmaz; |
| 317 | Semantic Cache Distillation: Efficient State Transfer Via Reuse and Selective Patching Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We propose Semantic Cache Distillation (SCD), a loss-constrained framework that replaces raw KV transmission with compact semantic codes. |
Qianli Ma; Zhiqing Tang; Hanshuai Cui; Zhi Yao; Weijia Jia; |
| 318 | Action Manifold Smoothing: A Lipschitz Pathway Perspective on High-Dimensional Reinforcement Learning Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We propose a unifying \textbf{Lipschitz Pathway} framework that decomposes instability into four amplification stages, namely action parameterization ($L_1$), dynamics sensitivity ($L_2$), Q-network curvature ($L_3$), and temporal-difference (TD) target stability ($L_4$), where errors compound multiplicatively along the learning pipeline. |
Zhihao Lin; |
| 319 | Faster Than Flash: Exploiting Attention Sparsity for Efficient Long-Context Decoding Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: To overcome the inherent trade-offs between the memory overhead of metadata-based metrics and the computational inefficiency of adaptive selection strategies, we present \textbf{Faster Flash Decoding (FFD)}, a novel hardware-algorithm co-design framework designed to break the memory wall in long-context decoding. |
Zhigeng Liu; Zhiyuan Ning; Ruixiao Li; Xiaoran Liu; Yuerong Song; Min zhang; Ziwei He; Xipeng Qiu; |
| 320 | Turning Drift Into Constraint: Robust Reasoning Alignment in Non-Stationary Multi-Stream Environments Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We propose Autonomous Preference Optimization (APO), a novel framework that treats inter-model divergences not as noise, but as dynamic negative constraints. |
Xiaoyu Yang; Jie Lu; Wei Duan; En Yu; |
| 321 | Temporal Context Reinstatement Drives Episodic-Like Order Memory in Long-Context Language Models Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Here, we investigate whether and, if so, how LLMs capture core behavioral signatures of humans of a central aspect of episodic memory via a temporal order memory task. |
Mathis Pink; Vy Vo; Qinyuan Wu; Jianing Mu; Javier Turek; Uri Hasson; Kenneth Norman; Sebastian Michelmann; Alexander Huth; Mariya Toneva; |
| 322 | Reinforcement Learning Via Self-Distillation Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Many verifiable environments actually provide rich textual feedback, such as runtime errors or judge evaluations, that explain *why* an attempt failed. We formalize this setting as reinforcement learning with rich feedback and introduce **Self-Distillation Policy Optimization** (**SDPO**), which converts tokenized feedback into a dense learning signal without any external teacher or explicit reward model. |
Jonas Hübotter; Frederike Lübeck; Lejs Behric; Anton Baumann; Marco Bagatella; Daniel Marta; Ido Hakimi; Idan Shenfeld; Thomas Kleine Buening; Carlos Guestrin; Andreas Krause; |
| 323 | Revisiting The Platonic Representation Hypothesis: An Aristotelian View Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We show that the existing metrics used to measure representational similarity are *confounded by network scale*: increasing model depth or width can systematically inflate representational similarity scores. To correct these effects, we introduce a permutation-based null-calibration framework that transforms any representational similarity metric into a calibrated score with statistical guarantees. |
Fabian Gröger; Shuo Wen; Maria Brbic; |
| 324 | SEAgent: Self-Evolving Computer Use Agent with Autonomous Learning from Experience Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: However, these models often struggle with novel and specialized software, particularly in scenarios lacking human annotations. To address this challenge, we propose SEAgent, an agentic self-evolving framework enabling CUAs to autonomously evolve through interactions with unfamiliar software. |
ZEYI SUN; Ziyu Liu; Yuhang Zang; Yuhang Cao; Xiaoyi Dong; Tong Wu; Dahua Lin; Jiaqi Wang; |
| 325 | Lions and Muons: Optimization Via Stochastic Frank-Wolfe Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: On the other hand, recent optimizers such as Lion and Muon have gained quite significant popularity in deep learning. In this work, building on recent initiatives, we provide a unifying perspective by interpreting these seemingly disparate methods through the lens of Stochastic Frank-Wolfe. |
Maria-Eleni Sfyraki; Jun-Kun Wang; |
| 326 | CANDI: Hybrid Discrete-Continuous Diffusion Models Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: While continuous diffusion has shown remarkable success in continuous domains such as image generation, its direct application to discrete data has underperformed pure discrete formulations. To understand this gap, we introduce *token identifiability*, an analytical framework characterizing how Gaussian noise corrupts discrete data through two mechanisms: *discrete identity corruption* and *continuous rank degradation*. |
Patrick Pynadath; Jiaxin Shi; Ruqi Zhang; |
| 327 | Simple Policy Gradients for Reasoning with Diffusion Language Models Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: In this work, we propose Amortized Group Relative Policy Optimization (AGRPO), a policy gradient algorithm that leverages the multi-step Markovian nature of dLLM generation, optimizing individual denoising steps rather than entire sequences. |
Anthony Zhan; |
| 328 | Transitive Representation Learning Enhances Histopathology Annotation Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: In contrast, single-cell gene expression provides expressive and interpretable labels that compensate this scarcity, but assays are costly and rarely acquired in clinical workflows. To overcome this gap, we propose to bridge these data sources using a trimodal contrastive learning framework that aligns histopathology images, gene expression profiles, and natural-language descriptions. |
Moritz Schaefer; Zoe Piran; Nils Philipp Walter; Animesh Awasthi; Christoph Bock; Jure Leskovec; Zinaida Good; |
| 329 | SAMT: Generating Structured Avatar Meshes and Textures from A Single Image Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: To reconstruct facial micro-structures and fine-grained multiview-consistent textures, this work presents a two-stage framework named SAMT for monocular 3D avatar generation and texture synthesis. |
Muyu Wang; Xingping Dong; Jianzhe Gao; Wenguan Wang; Yujia Wang; |
| 330 | Context Tuning for In-Context Optimization Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We introduce Context Tuning, a simple and effective method to significantly enhance few-shot adaptation of language models (LLMs) without fine-tuning model parameters. |
Jack Lu; Ryan Teehan; Zhenbang Yang; Mengye Ren; |
| 331 | Efficient and Safe Molecular Assembly Via Reinforcement Learning and Constraint Solving Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: In this work, we take a substantial step toward autonomous manufacturing with STMs by introducing a novel AI-based planning framework for molecular assembly and a high-fidelity simulation environment. |
Stefan Pranger; Bernhard Ramsauer; Oliver Hofmann; Bettina Könighofer; |
| 332 | Breaking Manifold Continuity: Vector Quantized Modeling for Real-Centric Deepfake Detection Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: In order to further enhance the generalization of discrete modeling, we propose an adaptive tangent space projection mechanism that yields a continuous relaxation of the discrete real distribution within a controllable range. |
Changshuo Wang; Jiangming Wang; Ke-Yue Zhang; Taiping Yao; Shouhong Ding; Ran Yi; Lizhuang Ma; |
| 333 | Teaching Models to Teach Themselves: Reasoning at The Edge of Learnability Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We explore this with SOAR: A self-improvement framework designed to surface these pedagogical signals through meta-RL. A teacher model proposes synthetic problems for a student model, and is rewarded with its improvement on a subset of hard problems, thus grounding the curriculum in real student progress rather than proxy rewards. |
Shobhita Sundaram; John Quan; Ariel Kwiatkowski; Kartik Ahuja; Yann Ollivier; Julia Kempe; |
| 334 | Routing and Reasoned Evaluation with Large Language Models Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We introduce R$^2$Eval, a routing-aware automated assessment framework that formulates evaluation as a resource allocation and aggregation problem rather than relying on a single monolithic evaluator. |
Guiyao Tie; Tianyao Luo; Xueyang Zhou; Chaoran Hu; Yunhong He; Junran Wu; Yuanfan Yao; Pan Zhou; Lichao Sun; |
| 335 | Context-free Recognition with Transformers Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: In this work, we show that looped transformers with $\mathcal{O}(\log(n))$ looping layers and $\mathcal{O}(n^6)$ padding tokens can recognize all CFLs. |
Selim Jerad; Anej Svete; Sophie Hao; Ryan Cotterell; William Merrill; |
| 336 | MINIM: Privacy-Aware Minimal View for Agents Via Trusted Local Sanitization Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We propose MINIM, a trusted local broker that performs privacy-aware minimization on the client side before any observation leaves the device. |
Hexuan Yu; Chaoyu Zhang; Heng Jin; Shanghao Shi; Ning Zhang; Thomas Hou; Wenjing Lou; |
| 337 | Non-Monotonic Autoregressive Sequence Model Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: To train the model, we introduce a sequence augmentation method that constructs error-correction trajectories from model-generated deviations paired with ground-truth references. |
Tianyi MA; Yiyue Qian; Yiyang Li; Zehong Wang; Yifan Ding; Zheyuan Zhang; Yan Liang; Chuxu Zhang; Yanfang Ye; |
| 338 | Effective Reasoning Chains Reduce Intrinsic Dimensionality Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: In this work, we identify *intrinsic dimensionality* as a quantitative measure for characterizing the effectiveness of reasoning chains. |
Archiki Prasad; Mandar Joshi; Kenton Lee; Mohit Bansal; Peter Shaw; |
| 339 | AppWorld-UL: Benchmarking Diverse Agent-User Interactions for Tool-Use Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Further, they operate in small environments with few, often non-state-changing, APIs. To address this gap, we introduce AppWorld-UL, a “user-in-the-loop” benchmark of 306 challenging tasks requiring diverse agent-user interactions. |
Junzhi Chen; Harsh Trivedi; Jane Pan; Michael Zhang; Tejas Srinivasan; Niranjan Balasubramanian; Ashish Sabharwal; |
| 340 | Position: Code Benchmarks Should Prioritize Rigor, Reliability, and Reproducibility Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: In response, we take a clear position: Code bench- marks must prioritize rigor in benchmark con- struction, reliability in evaluation, and repro- ducibility in release. To operationalize this po- sition, we introduce a code benchmark guideline HOW2BENCH with 55 checklists. |
Jialun Cao; Yuk-Kit Chan; Zixuan Ling; Wenxuan Wang; Shuqing Li; Mingwei Liu; Ruixi Qiao; Yuting Han; Chaozheng Wang; Boxi Yu; Pinjia He; Shuai Wang; Zibin Zheng; Michael Lyu; Shing-Chi Cheung; |
| 341 | Reason with Thumbnails, Answer with Focus: An Efficient and Effective Paradigm for Multimodal Grounded Visual Reasoning Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: To attain efficient and effective GVR, in this paper, we propose a novel paradigm called Reason with Thumbnails, Answer with Focus (RTAF), which feeds the model with low-resolution images to reason the relevant regions and high-resolution crops to answer the final answer. |
An-Lan Wang; Guozhi Tang; Lei Liao; Hanshen Zhu; Kai Huang; Jingqun Tang; Jiaming Zhou; Kun-Yu Lin; |
| 342 | VGGT-Motion: Motion-Aware Calibration-Free Monocular SLAM for Long-Range Consistency Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Motion-agnostic partitioning breaks contextual coherence and causes zero-motion drift, while conventional geometric alignment is computationally expensive. To address these issues, we propose VGGT-Motion, a calibration-free SLAM system for efficient and robust global consistency over kilometer-scale trajectories. |
Zhuang Xiong; Chen Zhang; Qingshan Xu; Wenbing Tao; |
| 343 | LatentLens: Revealing Highly Interpretable Visual Tokens in LLMs Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: In this work, we introduce LatentLens, a novel approach for mapping latent representations to descriptions in natural language. |
Benno Krojer; Perampalli Shravan Nayak; Oscar Mañas; Vaibhav Adlakha; Desmond Elliott; Siva Reddy; Marius Mosbach; |
| 344 | Leak@$k$: Unlearning Does Not Make LLMs Forget Under Probabilistic Decoding Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Our findings demonstrate that knowledge leakage persists across methods and tasks, underscoring that current state-of-the-art unlearning techniques provide only limited forgetting and highlighting the urgent need for more robust approaches to LLM unlearning. We propose an algorithm, termed Robust Unlearning under LEak@$k$ metric (\texttt{RULE}), which serves as an initial step toward addressing this concern. |
Hadi Reisizadeh; Jiajun Ruan; Yiwei Chen; Soumyadeep Pal; Sijia Liu; Mingyi Hong; |
| 345 | Vector Quantization Using Gaussian Variational Autoencoder Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: In this paper, we propose a simple yet effective technique dubbed __Gaussian Quant (GQ)__, which first trains a Gaussian VAE under certain constraints and then converts it into a VQ-VAE without additional training. |
Tongda Xu; Wendi Zheng; Jiajun He; Jose Miguel Hernandez-Lobato; Yan Wang; Ya-Qin Zhang; Jie Tang; |
| 346 | Balanced LoRA: Removing Parameter Invariance to Accelerate Convergence Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: As a result, converging to different loss minimizers directly impacts the convergence rate of LoRA. Building on this observation, we introduce Balanced Low-Rank Adaptation (BaLoRA), a variant of LoRA that projects iterates onto a balanced manifold. |
Valérie Castin; Kimia Nadjahi; Pierre Ablin; Gabriel Peyré; |
| 347 | Removing Noise, Not Finding Gold: Quality Filtering for Large-Scale Pretraining Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Large-scale models are pretrained on massive web-crawled datasets containing documents of mixed quality, making data filtering essential. |
Thiziri Nait Saada; Louis Béthune; Michal Klein; David Grangier; Marco Cuturi; Pierre Ablin; |
| 348 | FutureOmni: Evaluating Future Forecasting from Omni-Modal Context for Multimodal LLMs Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Evaluations on 13 omni-modal and 7 video-only models show that current systems struggle with audio-visual future prediction, particularly in speech-heavy scenarios, with the best accuracy of 64.8% achieved by Gemini 3 Flash. To mitigate this limitation, we curate a 7K-sample instruction-tuning dataset and propose an Omni-Modal Future Forecasting (OFF) training strategy. |
Qian Chen; Jinlan Fu; Changsong Li; Min zhang; See-Kiong Ng; Xipeng Qiu; |
| 349 | HiMe: Hierarchical Embodied Memory for Long-Horizon Vision-Language-Action Control Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Existing solutions face a frequency-competence paradox, where high-performance models are too slow for real-time control, while faster models lack sufficient reasoning capabilities. To resolve this architectural misalignment, we propose **HiMe**, a Hierarchical Embodied Memory framework that decouples embodied intelligence into a high-frequency Executor for execution, a Sentry for working memory, and a Planner for long-term strategy. |
Li Ji; Siyin Wang; Pengfang Qian; Xiaopeng Yu; Yihai Tian; Zhaoye Fei; Jingjing Gong; Xipeng Qiu; |
| 350 | Position: Agentic AI Is A Foreseeable Pathway to AGI Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Instead, we identify Agentic AI as the necessary evolution for handling complex, real-world task distributions to achieve AGI in the human world. |
Junwei Liao; Shuai Li; Muning Wen; Jun Wang; Weinan Zhang; |
| 351 | A New Framework for Cybersecurity Refusals in AI Agents Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We present the first framework for establishing refusal boundaries in offensive security contexts. |
Eliot Jones; Matt Fredrikson; Zico Kolter; |
| 352 | Probing Newtonian Mechanics in Video Generative Models with Real Physical Systems Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Thus, we introduce **Morpheus**, one of the first physics-informed evaluation frameworks for measuring the ability of video generation models to comprehend Newtonian dynamics. |
Antonios Tragoudaras; Daniil Cherniavskii; Chenyu Zhang; Antonis Vozikis; Thijmen Nijdam; Derck Prinzhorn; Márk Bodrácska; Nicu Sebe; Andrii Zadaianchuk; Efstratios Gavves; |
| 353 | Learning Syntax Without Semantics: Disentangled Tiny Language Models Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We show that syntax can be learned while suppressing semantic plausibility and world‑knowledge cues, yielding more efficient and controllable models. |
Ezra Winston; Zico Kolter; |
| 354 | Antidistillation Fingerprinting Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We introduce ***antidistillation fingerprinting*** (ADFP), a principled approach that aligns the fingerprinting objective with the student’s learning dynamics. |
Yixuan Xu; John Kirchenbauer; Yash Savani; Asher Trockman; Alexander Robey; Tom Goldstein; Fei Fang; Zico Kolter; |
| 355 | How RL Unlocks The Aha Moment in Geometric Interleaved Reasoning Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We argue that this failure stems from a fundamental limitation of SFT, which primarily induces \textit{distributional alignment}: the model learns to reproduce the surface format of interleaved plotting but fails to internalize the causal dependency between the generated plot and reasoning steps. To overcome this limitation, we propose Faire (\textbf{F}unctional \textbf{a}lignment for \textbf{i}nterleaved \textbf{re}asoning), a reinforcement learning framework that enforces three casual constraints to move beyond superficial imitation toward \textit{functional alignment}. |
Xiangxiang Zhang; Caijun Jia; Siyuan Li; he dingyu; Xiya Xiong; Zheng Sun; Honghao He; Yuchen Wu; Bihui Yu; Linzhuang Sun; Cheng Tan; Jingxuan Wei; |
| 356 | Plan for Speed: Dilated Scheduling for Masked Diffusion Language Models Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Masked diffusion language models (MDLMs) promise fast, non-autoregressive text generation, yet existing samplers, which pick tokens to unmask based on model confidence, ignore interactions when unmasking multiple positions in parallel and effectively reduce to slow, autoregressive behavior. We propose the Dilated Unmasking Scheduler (DUS), an inference-only, planner-model-free method that partitions sequence positions into non-adjacent dilated groups and unmasked them in parallel so as to minimize an upper bound on joint entropy gain at each denoising step. |
Omer Luxembourg; Haim Permuter; Eliya Nachmani; |
| 357 | Good SFT Optimizes for SFT, Better SFT Prepares for Reinforcement Learning Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We show that, after identical RL training, models initialized from stronger SFT checkpoints can significantly underperform those initialized from weaker ones. We propose PEAR ($\textbf{P}$olicy $\textbf{E}$valuation–inspired $\textbf{A}$lgorithm for Offline Learning Loss $\textbf{R}$eweighting), an SFT-stage method that corrects this mismatch and better prepares the model for RL. |
Dylan Zhang; Yufeng Xu; Haojin Wang; Qingzhi Chen; Hao Peng; |
| 358 | Approximation of Log-Partition Function in Policy Mirror Descent Induces Implicit Regularization for LLM Post-Training Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We investigate a practical algorithm, termed PMD-mean, that approximates the log-partition term with the mean reward under the sampling policy and performs regression in log-policy space. |
Zhenghao Xu; Qin Lu; Changlong Yu; Tuo Zhao; |
| 359 | Sparse But Wrong: Incorrect L0 Leads to Incorrect Features in Sparse Autoencoders Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: In this work we study the effect of L0 on SAEs, and show that if L0 is not set correctly, the SAE fails to disentangle the underlying features of the LLM. |
David Chanin; Adrià Garriga-Alonso; |
| 360 | Elign: Equivariant Diffusion Model Alignment from Foundational Machine Learned Force Fields Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We present Elign, a post-training framework that amortizes both costs. |
Yunyang Li; Lin Huang; Luojia Xia; Wenhe Zhang; Mark Gerstein; |
| 361 | SMAC: Score-Matched Actor-Critics for Robust Offline-to-Online Transfer Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We provide evidence consistent with the hypothesis that, in the loss landscape, offline maxima for prior algorithms and online maxima are separated by low-performance valleys that gradient-based fine-tuning traverses. Following this, we present Score Matched Actor-Critic (SMAC), an offline RL method designed to learn actor–critics that transition to online value-based RL algorithms with no drop in performance. |
Nathan de Lara; Florian Shkurti; |
| 362 | Smooth Dynamic Cutoffs for Machine Learning Interatomic Potentials Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: In this work, we address both issues by challenging the long-held belief that the cutoff radius for the MLIP must be held to a fixed, constant value. |
Kevin Han; Haolin Cong; Bowen Deng; Amir Barati Farimani; |
| 363 | How Does Information Access Affect LLM Monitors’ Ability to Detect Sabotage? Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: In this paper, we study how *information access* affects LLM monitor performance and find that contemporary systems often perform better with less information, a phenomenon we call the *less-is-more effect* for automated oversight. |
Rauno Arike; Raja Moreno; Rohan Subramani; Shubhorup Biswas; Francis Rhys Ward; |
| 364 | Smaller Models Are Natural Explorers for Policy-Level Diversity in GRPO Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We thus propose S2L-PO (Small-to-Large Policy Optimization), a framework that leverages fixed small models as natural explorers to train larger models. |
Yiming Ren; Yiran Xu; Zicheng Lin; Chufan Shi; Yukang Chen; Dingdong WANG; Tianhe Wu; Junjie Wang; Yujiu Yang; Yu Qiao; Ruihang Chu; |
| 365 | Utility-Diversity Aware Online Batch Selection for LLM Supervised Fine-tuning Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: However, existing popular methods often (i) rely merely on the utility of data to select a subset while neglecting other crucial factors like diversity, (ii) rely on external resources such as reference models or validation sets, and (iii) incur extra training time over full-dataset training. To address these limitations, this work develops UDS (Utility-Diversity Sampling), a framework for efficient online batch selection in SFT. |
Heming Zou; Yixiu Mao; Yun Qu; Cheems Wang; Xiangyang Ji; |
| 366 | Structured 4D Latent World Model for Robot Planning Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We introduce a **Structured 4D Latent World Model**, which predicts the evolution of a scene’s 3D structure in a structured latent space conditioned on observations and textual instructions. |
Zhiyi Li; Peilin Wu; Xiaoshen Han; Ruojin Cai; Yilun Du; |
| 367 | Language Model Circuits Are Sparse in The Neuron Basis Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: For the first time, we empirically show that **MLP neurons are as sparse a feature basis as SAEs**. We use this finding to develop an end-to-end pipeline for circuit tracing on the MLP neuron basis, which locates causal circuitry on a variety of tasks using gradient-based attribution. |
Aryaman Arora; Zhengxuan Wu; Jacob Steinhardt; Sarah Schwettmann; |
| 368 | Prism: Spectral-Aware Block-Sparse Attention Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: In this work, we trace the inaccuracy of standard coarse-grained attention via mean pooling to a theoretical root cause: the interaction between mean pooling and Rotary Positional Embeddings (RoPE). |
Xinghao Wang; Pengyu Wang; Xiaoran Liu; Fangxu Liu; Jason Chu; Kai Song; Xipeng Qiu; |
| 369 | Sparser Block-Sparse Attention Via Token Permutation Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: In this work, we propose Permuted Block-Sparse Attention (**PBS-Attn**), a plug-and-play method that leverages the permutation properties of attention to increase block-level sparsity and enhance the computational efficiency of LLM prefilling. |
Xinghao Wang; Pengyu Wang; Dong Zhang; Chenkun Tan; Shaojun Zhou; Zhaoxiang Liu; Shiguo Lian; Fangxu Liu; Kai Song; Xipeng Qiu; |
| 370 | Learning to Move Before Learning to Do: Task-Agnostic Pretraining for VLAs Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We propose that learning ”how to move” can be decoupled from learning ”what to do,” and that the former requires no task labels at all. |
Junhao Shi; Siyin Wang; Xiaopeng Yu; Li Ji; Jingjing Gong; Xipeng Qiu; |
| 371 | How Can We Assess Human-agent Interactions? Case Studies in Software Agent Design Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: In this paper, we make two major steps towards the rigorous assessment of human-agent interactions. |
Valerie Chen; Rohit Malhotra; Xingyao Wang; Juan Michelini; Xuhui Zhou; Aditya Bharat Soni; Hoang Tran; Calvin Smith; Ameet Talwalkar; Graham Neubig; |
| 372 | Use What You Know: Causal Foundation Models with Partial Graphs Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: However, in their current state, they do not allow for the incorporation of any domain knowledge, which can lead to suboptimal predictions. We bridge this gap by introducing methods to condition CFMs on causal information, such as the causal graph or more readily available ancestral information. |
Arik Reuter; Anish Dhir; Cristiana Diaconu; Jake Robertson; Ole Ossen; Frank Hutter; Adrian Weller; Mark van der Wilk; Bernhard Schölkopf; |
| 373 | XKV: Cross-Layer KV-Cache Compression Via Aligned Singular Vector Extraction Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We show, via Centered Kernel Alignment (CKA), that the dominant singular vectors of KV-Cache are well aligned across layers. Motivated by this observation, we propose xKV, a post-training compression method that jointly factorizes grouped-layer KV-Cache into a shared low-rank subspace, substantially reducing KV-Cache memory. |
Chi-Chih Chang; Wei-Cheng Lin; Chien-Yu Lin; Hung-Yueh Chiang; Yash Akhauri; Xilai Dai; Huiqiang Jiang; Yucheng Li; Kai-Chiang Wu; Luis Ceze; Mohamed Abdelfattah; |
| 374 | Meta Context Engineering Via Agentic Skill Evolution Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: They impose structural biases and restrict context optimization to a narrow, intuition-bound design space. To address this, we introduce Meta Context Engineering (MCE), a bi-level framework that supersedes static CE heuristics by co-evolving CE skills and context artifacts. |
Haoran Ye; Xuning He; Vincent Arak; Haonan Dong; Guojie Song; |
| 375 | BandPO: Bridging Trust Regions and Ratio Clipping Via Probability-Aware Bounds for LLM Reinforcement Learning Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: While the canonical clipping mechanism in PPO serves as an efficient surrogate for trust regions, we identify a critical bottleneck: fixed bounds strictly constrain the upward update margin of low-probability actions, disproportionately suppressing high-advantage tail strategies and inducing rapid entropy collapse. To address this, we introduce **Band-constrained Policy Optimization** (BandPO). |
Yuan Li; Bo Wang; Yufei Gao; Yuqian Yao; Xinyuan Wang; Zhangyue Yin; Xipeng Qiu; |
| 376 | Rethinking Thinking Tokens: LLMs As Improvement Operators Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Abstractly, we view the model as an improvement operator on its own thoughts with a continuum of possible strategies. We identify an interesting inference family Parallel-Distill-Refine (PDR), which performs the following: (i) generate diverse drafts in parallel; (ii) distill them into a bounded, textual workspace; and (iii) refine conditioned on this workspace, producing an output that seeds the next round. |
Lovish Madaan; Aniket Didolkar; Suchin Gururangan; John Quan; Ruan Silva; Russ Salakhutdinov; Manzil Zaheer; Sanjeev Arora; Anirudh Goyal; |
| 377 | Securing Multimodal AI Through Internal Information Decomposition Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We propose FlowGuard, a lightweight inference-time framework that detects harmful inputs by monitoring internal multimodal consistency. |
Jehyeok Yeon; Hyeonjeong Ha; Qiusi Zhan; Heng Ji; |
| 378 | Graph-R1: Towards Agentic GraphRAG Framework Via End-to-end Reinforcement Learning Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: GraphRAG methods improve RAG by modeling knowledge as entity-relation graphs, but still face challenges in high construction cost, fixed one-time retrieval, and reliance on long-context reasoning and prompt design. To address these challenges, we propose Graph-R1, an agentic GraphRAG framework via end-to-end reinforcement learning (RL). |
Haoran Luo; Haihong E; Guanting Chen; Qika Lin; Yikai Guo; Fangzhi Xu; Zemin Kuang; Meina Song; Xiaobao Wu; Yifan Zhu; Anh Tuan Luu; |
| 379 | Alignment Tampering: How Reinforcement Learning from Human Feedback Is Exploited to Optimize Misaligned Biases Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: In this work, we introduce alignment tampering, a potential vulnerability where the LLM undergoing alignment influences the preference dataset, causing RLHF to amplify undesired behaviors. |
Dongyoon Hahm; Dylan Hadfield-Menell; Kimin Lee; |
| 380 | Human-AI Collaborative Uncertainty Quantification Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: This motivates collaborative approaches that combine human judgment with AI predictions. We study this problem through the lens of uncertainty quantification and introduce **Human-AI Collaborative Uncertainty Quantification**, a framework in which an AI system refines a human expert’s proposed prediction set subject to two principles: **counterfactual harm**, requiring that the AI not degrade correct human judgments, and **complementarity**, requiring recovery of correct outcomes the human missed. |
Sima Noorani; Shayan Kiyani; George Pappas; Hamed Hassani; |
| 381 | RefChess: Monte-Carlo Move Selection for Zero-Shot Referring Image Segmentation Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Nevertheless, selecting the correct segmentation proposal remains challenging, as existing methods typically rely on independent proposal scoring and lack contextual reasoning among visually similar candidates. To address this limitation, we propose RefChess, a training-free framework that reformulates proposal selection as a decision-making problem under contextual perturbations rather than a single-step ranking task. |
Shiyan Tong; Jinxia Zhang; Zhiyuan Wang; Hao Tian; YingYing Wang; Kanjian Zhang; Haikun Wei; |
| 382 | Towards Efficient Large Language Reasoning Models Via Extreme-Ratio Chain-of-Thought Compression Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: To achieve high-fidelity, fast reasoning, we propose a novel EXTreme-RAtio Chain-of-Thought Compression framework, termed Extra-CoT, which aggressively reduces the token budget while preserving answer accuracy. |
Yuntian Tang; Bohan Jia; Wenxuan Huang; Lianyue Zhang; Jiao Xie; Wenxi Li; Wei Li; Jie Hu; Xinghao Chen; Rongrong Ji; Shaohui Lin; |
| 383 | From Bits to Rounds: Parallel Decoding with Exploration for Diffusion Language Models Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Motivated by this theory, we propose Explore-Then-Exploit (ETE), a training-free decoding strategy that maximizes information throughput and decoding efficiency. |
Hengyu Fu; Baihe Huang; Virginia Adams; Charles Wang; Junkeun Yi; Mohammad Mahdi Kamani; Venkat Krishna Srinivasan; Jiantao Jiao; |
| 384 | Mixture of Horizons in Action Chunking Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: To mitigate the trade-off, we propose a $\textbf{mixture of horizons (MoH)}$ strategy. |
Dong Jing; Gang Wang; Jiaqi Liu; Weiliang Tang; Zelong Sun; Yunchao Yao; Zhenyu Wei; Yunhui Liu; Zhiwu Lu; Mingyu Ding; |
| 385 | MME-Reasoning: A Broad-Spectrum Benchmark for Evaluating Logical Reasoning in MLLMs Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Despite the significant advancement in multimodal reasoning, existing benchmarks fail to comprehensively evaluate their reasoning abilities due to the lack of explicit categorization for logical reasoning types and an unclear understanding of reasoning. To address these issues, we introduce MME-Reasoning, a comprehensive benchmark designed to evaluate the reasoning ability of MLLMs, which covers all three types of reasoning (i.e., inductive, deductive, and abductive). |
Jiakang Yuan; Tianshuo Peng; Yilei Jiang; Yiting Lu; Renrui Zhang; Kaituo Feng; Chaoyou Fu; Tao Chen; LEI BAI; Bo Zhang; |
| 386 | Open Materials Generation with Inference-Time Reinforcement Learning Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We introduce Open Materials Generation with Inference-time Reinforcement Learning (OMatG-IRL), a policy-gradient RL framework that operates directly on the learned velocity fields and eliminates the need for the explicit computation of the score. |
Philipp Höllmer; Stefano Martiniani; |
| 387 | Rethinking Multimodal Time-Series Forecasting Evaluation Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We introduce a new context-enriched, multimodal time series forecasting benchmark TimesX. |
Haoxin Liu; Yichen Zhou; Rajat Sen; B. Aditya Prakash; Abhimanyu Das; |
| 388 | Failure-Driven Workflow Refinement Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We propose \textbf{CE-Graph}, which maintains a counterexample pool, estimates dense failure modes, and applies operator-constrained graph edits via a \textbf{Propose-and-Verify} loop with a convergence-aware stopping rule. |
Jusheng Zhang; Jing Yang; Kaitong Cai; Ziliang Chen; Yongsen Zheng; Kwok Yan Lam; Liang Lin; Keze Wang; |
| 389 | SOLAR for Offline MARL: Plateau-Triggered Potential Shaping Under World-Model Uncertainty Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We find that shaping becomes reliable when it is (i) activated only after \emph{statistically validated} learning plateaus and (ii) constrained to \emph{potential-based} shaping, which preserves the task optimum. Motivated by this, we propose \textsc{SOLAR}, a simulate–evaluate–shape framework. |
Jusheng Zhang; Yijia Fan; Ruiqi Chen; Jing Yang; Ziliang Chen; Yongsen Zheng; Yanxi Chen; Jian Wang; Kwok Yan Lam; Liang Lin; Keze Wang; |
| 390 | Position: Reasoning After Perception Means Reasoning Without Vision Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We argue that for a broad class of visual tasks hard to specify in language, failures stem from a structural fatality where the temporal decision of \textit{when} to reason strictly dictates the spatial constraint of where reasoning takes place. |
Hongcheng Gao; Zihao Huang; Jingyi Tang; Lin Xu; Xinhao Li; Haoyang Li; Yue Liu; Minhua Lin; Xinlong Yang; Taihang Hu; Ge Wu; Baolong Bi; Hongyu Chen; Zhiqi Huang; Wentao Zhang; |
| 391 | GePBench: Evaluating Fundamental Geometric Perception for Multimodal Large Language Models Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: While multimodal large language models (MLLMs) have made significant advancements in visual understanding, their abilities to recognize geometric shapes and their spatial relationships, which we term geometric perception, are not explicitly and systematically explored. To address this gap, we introduce GePBench, a novel benchmark specifically designed to assess the geometric perception capabilities of MLLMs. |
Shangyu Xing; Changhao Xiang; Xinyu Liu; Zhangtai Wu; Zhen Wu; Yue YIfan; Yuteng Han; Fei Zhao; Xinyu Dai; |
| 392 | ABC-Bench: An Agentic Bio-Capabilities Benchmark for Biosecurity Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: These emerging AI capabilities offer new opportunities for scientific discovery and biomedical advances, but they are also changing the landscape of biosecurity risks. To address this, we introduce the Agentic Bio-Capabilities Benchmark (ABC-Bench), a suite of evaluations to measure \textit{agentic} biosecurity-relevant capabilities. |
Andrew Liu; Samira Nedungadi; Bryce Cai; Alex Kleinman; Harmon Bhasin; Seth Donoughe; |
| 393 | Debate with Images: Detecting Deceptive Behaviors in Multimodal Large Language Models Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: In this work, we systematically investigate multimodal deception and introduce *MM-DeceptionBench*, the first benchmark designed to evaluate deceptive behaviors in vision–language models across six realistic categories. |
Sitong Fang; Shiyi Hou; Kaile Wang; Boyuan Chen; Donghai Hong; Jiayi Zhou; Juntao Dai; Yaodong Yang; Jiaming Ji; |
| 394 | SAM Audio: Segment Anything in Audio Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: In this work, we present SAM AUDIO, a foundation model for general audio separation that unifies text, visual, and temporal span prompting within a single framework. |
Bowen Shi; Andros Tjandra; John Hoffman; Helin Wang; YI-CHIAO WU; Luya Gao; Julius Richter; Matthew Le; Apoorv Vyas; Sanyuan Chen; Christoph Feichtenhofer; Piotr Dollár; Wei-Ning Hsu; Ann Lee; |
| 395 | Test-Time Anchoring for Discrete Diffusion Posterior Sampling Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Existing approaches to posterior sampling using discrete diffusion face severe challenges: derivative-free guidance yields sparse signals, continuous relaxations limit applicability, and split Gibbs samplers suffer from the curse of dimensionality. To overcome these limitations, we introduce Anchored Posterior Sampling (APS), built on two key innovations: *quantized expectation* for gradient-like guidance in discrete embedding space, and *anchored remasking* for adaptive decoding. |
Litu Rout; Andreas Lugmayr; Yasamin Jafarian; Srivatsan Varadharajan; Constantine Caramanis; Sanjay Shakkottai; Ira Kemelmacher-Shlizerman; |
| 396 | REAR: Test-time Preference Realignment Through Reward Decomposition Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: To extend TTS to preference alignment, we introduce a novel framework that models the task as a realignment problem, since the base model often fails to sufficiently align with the stated preference. |
Fuxiang Zhang; Pengcheng Wang; Chenran Li; Yi-Chen Li; Yuxin Chen; Lang Feng; Chenfeng Xu; Masayoshi Tomizuka; Bo An; |
| 397 | DynaTok: Token-Based 4D Reconstruction from Partial Point Clouds Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: While recent progress has largely been driven by image-based methods, existing point-based approaches typically focus on single-object scenarios, assume relatively complete inputs, and rely on explicit correspondence. To mitigate these limitations, we propose DynaTok, a point-based framework for correspondence-free 4D reconstruction from partial point cloud sequences that operates without images. |
Weirong Chen; Keisuke Tateno; Hidenobu Matsuki; Michael Niemeyer; Daniel Cremers; Federico Tombari; |
| 398 | Frontier Models Can Take Actions at Low Probabilities Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Pre-deployment evaluations inspect only a limited sample of model actions. A malicious model seeking to evade oversight could exploit this by randomizing when to defect: misbehaving so rarely that no malicious actions are observed during evaluation, but often enough that they occur eventually in deployment. |
Alex Serrano; Wen Xing; David Lindner; Erik Jenner; |
| 399 | Autoregressive Boltzmann Generators Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: However, modern BGs predominantly rely on Normalizing Flows (NFs), which either suffer from limited expressivity due to strict invertibility constraints (discrete time) or computationally expensive likelihoods (continuous time). In this paper, we propose Autoregressive Boltzmann Generators (ArBG), a novel autoregressive modelling framework that overcomes these limitations by departing from the flow-based BG paradigm. |
Danyal Rehman; Charlie Tan; Yoshua Bengio; Joey Bose; Alexander Tong; |
| 400 | Attentive Multi-Layer Fusion for Vision Transformers Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We show that task-relevant information is distributed across the network hierarchy rather than solely encoded in any of the last layers. To leverage this distribution of information, we apply an attentive probing mechanism that dynamically fuses representations from all layers of a Vision Transformer. |
Laure Ciernik; Marco Morik; Lukas Thede; Luca Eyring; Shinichi Nakajima; Zeynep Akata; Lukas Muttenthaler; |
| 401 | PASA: A Principled Embedding-Space Watermarking Approach for LLM-Generated Text Under Semantic-Invariant Attacks Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We propose PASA, a principled, robust, and distortion-free watermarking algorithm that embeds and detects a watermark at the semantic level. |
Zhenxin Ai; Haiyun He; |
| 402 | Position: There Are Futures That Benchmark-driven AI Cannot See Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: These are philosophically distinct questions that may require discoveries that we cannot specify. We propose mechanisms to restore exaptive capacity without abandoning benchmarking: plural evaluation regimes, protected venues for non-comparable work, long-horizon funding, and training norms that encourage researchers to question selection rules, not only optimize within them. |
Sobhan Lotfi; Ava Iranmanesh; Lachin Naghashyar; Ali Shirali; Fateme Haredasht; Sanmi Koyejo; Phil Torr; Yong Suk Lee; Fazl Barez; Joel Lehman; Peter Norvig; Arvind Narayanan; |
| 403 | PGT: Procedurally Generated Tasks for Improving Fine-grained Understanding in MLLMs Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: In this work, we propose **Procedurally Generated Tasks (PGT)** a simple data-driven framework that serves a dual purpose: inducing fine-grained visual understanding and acting as a low-cost diagnostic tool to identify the source of perception failures. |
Rim Assouel; Amir Bar; Michal Drozdzal; Adriana Romero-Soriano; |
| 404 | Unifying Heterogeneous Degradations: Uncertainty-Aware Diffusion Bridge Model for All-in-One Image Restoration Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Existing methods are often constrained by coarse-grained control mechanisms or fixed mapping schedules, yielding suboptimal adaptation. To address this, we propose an Uncertainty-Aware Diffusion Bridge Model (UDBM), which innovatively reformulates AiOIR as a stochastic transport problem steered by pixel-wise uncertainty. |
Luwei Tu; Jiawei Wu; Xing Luo; Zhi Jin; |
| 405 | DLO-Lab: Benchmarking Deformable Linear Object Manipulations with Differentiable Physics Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Additionally, existing simulation environments offer limited support for the broad spectrum of material behaviors necessary for generalizable DLO manipulation. To overcome these limitations, we introduce a differentiable simulator explicitly designed for versatile DLO manipulation. |
Junyi Cao; Yian Wang; Ziyan Xiong; Chunru Lin; Zhehuan Chen; Chuang Gan; |
| 406 | Process Reward Agents for Steering Knowledge-Intensive Reasoning Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Here, we introduce Process Reward Agents~(PRA), a test-time method for providing domain-grounded, online, step-wise rewards to a frozen reasoner. |
Jiwoong Sohn; Tomasz Sternal; Kenneth Styppa; Torsten Hoefler; Michael Moor; |
| 407 | How Can Embedding Models Bind Concepts? Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Although CLIP behaves like a bag-of-concepts model in cross-modal retrieval, object information is recoverable from its image and text embeddings separately. We study this tension through the binding function, which maps concepts to scene embeddings. |
Arnas Uselis; Darina Koishigarina; Seong Joon Oh; |
| 408 | GradMem: Learning to Write Context Into Memory with Test-Time Gradient Descent Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We introduce GradMem, which writes context into memory via per-sample test-time optimization. |
Yuri Kuratov; Matvey Kairov; Aydar Bulatov; Ivan Rodkin; Mikhail Burtsev; |
| 409 | A Recipe for Scalable Attention-based ML Potentials: Unlocking Long-range Accuracy with All-to-all Node Attention Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: In this work, we propose AllScAIP, a straightforward, attention-based, and energy-conserving MLIP model that scales to O(100 million) training samples. |
Eric Qu; Brandon Wood; Aditi Krishnapriyan; Zachary Ulissi; |
| 410 | Efficient Hallucination Detection for LLMs Using Uncertainty-Aware Attention Heads Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: In this work, we propose Recurrent Attention-based Uncertainty Quantification (RAUQ), an unsupervised and efficient framework for identifying hallucinations. |
Artem Vazhentsev; Lyudmila Rvanova; Gleb Kuzmin; Ekaterina Fadeeva; Ivan Lazichny; Alexander Panchenko; Maxim Panov; Mrinmaya Sachan; Preslav Nakov; Tim Baldwin; Artem Shelmanov; |
| 411 | Necessary Conditions for Compositional Generalization of Embedding Models Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Modern models are trained on massive datasets, yet these are vanishingly small compared to the full combinatorial space of possible data, raising the question of whether models can reliably generalize to unseen combinations. To formalize what this requires, we propose a set of practically motivated desiderata that any compositionally generalizing system must satisfy, and analyze their implications under standard training with linear classification heads. |
Arnas Uselis; Andrea Dittadi; Seong Joon Oh; |
| 412 | WeDLM: Reconciling Diffusion Language Models with Standard Causal Attention for Fast Inference Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We propose WeDLM, a diffusion decoding framework built entirely on standard causal attention to make parallel generation prefix-cache friendly. |
Aiwei Liu; Minghua He; Shaoxun Zeng; Sijun Zhang; Linhao Zhang; Chuhan Wu; Wei Jia; Yuan Liu; Zhou Xiao; Jie Zhou; |
| 413 | Label-Guided Representation Learning for Incomplete Multi-View Multi-Label Classification Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We propose Label-Guided Representation Learning (LGRL) that systematically exploits label semantics as structural priors throughout learning. |
Yang Li; Quanjiang Li; Tingjin Luo; |
| 414 | ForesightKV: Optimizing KV Cache Eviction for Reasoning Models By Learning Long-Term Contribution Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: To better balance efficiency and performance, we introduce ForesightKV, a training-based KV cache eviction framework that learns to predict which KV pairs to evict during long-text generations. |
Zican Dong; Peiyu Liu; Junyi Li; Zhipeng Chen; Han Peng; Shuo Wang; Xin Zhao; |
| 415 | OmniVideo-R1: Reinforcing Audio-visual Reasoning with Query Intention and Modality Attention Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We identify this “modality interference” as a consequence of pre-training data imbalances, where the scarcity of mixed modality supervision induces a bias towards isolated modalities, resulting in an inherent trade-off. To address this challenge, we propose OmniVideo-R1, a novel reinforced reasoning framework that leverages post-training to rectify modality bias. |
Zhangquan Chen; Jiale Tao; Ruihuang Li; Yihao Hu; Ruitao Chen; Zhantao Yang; Xinlei Yu; Haodong Jing; Manyuan Zhang; Shuai Shao; Biao Wang; Qinglin Lu; Ruqi Huang; |
| 416 | TSRBench: A Comprehensive Multi-task Multi-modal Time Series Reasoning Benchmark for Generalist Models Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: However, current benchmarks for generalist models largely overlook this dimension. To bridge this gap, we introduce TSRBench, a comprehensive multi-modal benchmark designed to stress-test the full spectrum of time series reasoning capabilities. |
Fangxu Yu; Xingang Guo; Lingzhi Yuan; Haoqiang Kang; Hongyu Zhao; Lianhui Qin; Furong Huang; Bin Hu; Tianyi Zhou; |
| 417 | Clover: Accurate LLM Pre-Training in NVFP4 By Improved Unbiased Gradient Estimation Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: In this paper, improve the state of the art for quantized training in NVFP4 via a novel unbiased quantization routine for micro-scaled formats, called MS-EDEN, that has more than 2x lower quantization error than SR. |
Andrei Panferov; Erik Schultheis; Rush Tabesh; Dan Alistarh; |
| 418 | Position: Safe AI Should Be Resistant and Resilient in An Evolving World Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: In this position paper, we address the persistent gap between rapidly growing AI capabilities and lagging safety progress. |
Youbang Sun; Xiang Wang; Jie Fu; Chaochao Lu; Bowen Zhou; |
| 419 | A Theoretical Framework for Modular Learning of Robust Generative Models Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We present a theoretical framework for *modular* generative modeling where a set of pre-trained experts are combined via a gating mechanism. |
Corinna Cortes; Mehryar Mohri; Yutao Zhong; |
| 420 | Optimized Deferral for Imbalanced Settings Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We present a comprehensive study of two-stage learning to defer in expert imbalance settings. |
Corinna Cortes; Anqi Mao; Mehryar Mohri; Yutao Zhong; |
| 421 | PaperBanana: Automating Academic Illustration for AI Scientists Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Despite rapid advances in autonomous AI scientists powered by language models, generating publication-ready illustrations remains a labor-intensive bottleneck in the research workflow. To lift this burden, we introduce PaperBanana, an agentic framework for automated generation of publication-ready academic illustrations. |
Dawei Zhu; Rui Meng; Yale Song; Xiyu Wei; Sujian Li; Tomas Pfister; Jinsung Yoon; |
| 422 | TestExplora: Benchmarking LLMs for Proactive Bug Discovery Via Repository-Level Test Generation Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Specifically, they either constrain models to a compliance trap by treating existing code as the ground truth for regression prevention, or rely on post-failure artifacts (e.g., issue reports) for reactive bug reproduction, failing to expose defects before they manifest as failures. To bridge this gap, we present TestExplora, a benchmark designed to evaluate LLMs as proactive testers within full-scale, realistic repository environments. |
Steven Liu; Jane Luo; Xin Zhang; Aofan Liu; Hao Liu; Jie Wu; Ziyang Huang; Yangyu Huang; Yu Kang; Scarlett Li; |
| 423 | Safe and Scalable Web Agent Learning Via Recreated Websites Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We propose VeriEnv, a framework that treats language models as environment creators, automatically cloning real-world websites into fully executable, verifiable synthetic environments. |
Hyungjoo Chae; Jungsoo Park; Alan Ritter; |
| 424 | Controlled LLM Training on Spectral Sphere Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Maximal Update Parametrization ($\boldsymbol{\mu}$P) provides a theoretical safeguard for width-invariant $\Theta(1)$ activation control, whereas emerging optimizers like Muon are only half-aligned with these constraints: they control updates but allow weights to drift. To address this limitation, we introduce the **Spectral Sphere Optimizer (SSO)**, which enforces strict module-wise spectral constraints on both weights and their updates. |
Tian Xie; Haoming Luo; Haoyu Tang; Hu Yiwen; Jason Liu; Qingnan Ren; Yang Wang; Xin Zhao; Rui Yan; Bing Su; Chong Luo; Baining Guo; |
| 425 | T2AV-Compass: Towards Unified Evaluation for Text-to-Audio-Video Generation Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Text-to-Audio-Video (T2AV) generation aims to synthesize temporally coherent video and semantically synchronized audio from natural language, yet its evaluation remains fragmented, often relying on unimodal metrics or narrowly scoped benchmarks that fail to capture cross-modal alignment, instruction following, and perceptual realism under complex prompts. To address this limitation, we present T2AV-Compass, a unified benchmark for comprehensive evaluation of T2AV systems, consisting of 500 diverse and complex prompts constructed via a taxonomy-driven pipeline to ensure semantic richness and physical plausibility. |
Zhe Cao; Tao Wang; Jiaming Wang; Yanghai Wang; Yuanxing Zhang; Jiahao Wang; Jialu Chen; Miao Deng; Chenxi Liao; Yize Zhang; Yubin Guo; Zhaoxiang Zhang; Jiaheng Liu; |
| 426 | Cross-Embodiment Robot Foundation World Models with Latent Actions Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We introduce a Latent Action Conditioned Robot World Model (LAC-WM), which operates within a learned unified latent action space shared across diverse embodiments. |
Huang Huang; Sriram Yenamandra; Arjun Majumdar; Elie Aljalbout; Tushar Nagarajan; Jimmy Yang; Akshara Rai; Michael Rabbat; Li Fei-Fei; Jiajun Wu; Tingfan Wu; Franziska Meier; |
| 427 | From Poisoned to Aware: Fostering Backdoor Self-Awareness in LLMs Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Existing safety training methods largely fail to address this vulnerability, due to the inherent difficulty of uncovering hidden triggers embedded within the model. Motivated by recent findings on LLMs’ situational awareness, we propose a novel post-training framework that cultivates backdoor self-awareness, enabling a poisoned LLM to precisely articulate its own implanted triggers. |
Guangyu Shen; Siyuan Cheng; Xiangzhe Xu; Yuan Zhou; Hanxi Guo; Zhuo Zhang; Xiangyu Zhang; |
| 428 | Pair2Scene: Learning Local Object Relations for Procedural Scene Generation Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Building on top of the observation that object placement relies mainly on local dependencies instead of information-redundant global distributions, in this paper, we propose **Pair2Scene**, a novel procedural generation framework, for scene generation based on a set of *learned procedural rules*. |
Xingjian Ran; Shujie Zhang; Weipeng Zhong; Luo Li; Bo Dai; |
| 429 | PGD-NO: A Neural Operator with Precomputed Geometry Decomposition for 3D Million-Scale Physics Simulations Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: While neural PDE solvers have demonstrated significant potential for accelerating engineering simulations, existing architectures remain constrained by high memory consumption and the single-node bottleneck, where the maximum processable mesh resolution is strictly limited by the VRAM of a single compute unit. To address these challenges, we propose PGD-NO, a neural operator with Precomputed Geometry Decomposition, that relocates the computational overhead of geometric encoding to a deterministic pre-computation phase. |
Weiheng Zhong; Jing Bi; Victor Oancea; Hadi Meidani; |
| 430 | TACTIC: Task-Aware Sparse Coordination Graphs for Multi-Task Multi-agent Reinforcement Learning Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Prior VQ-VAE methods abstract trajectories yet miss time-varying inter-agent dependencies. We present TACTIC, a CTDE framework with three advances: (i) hierarchical goal decomposition to guide exploration under sparse rewards; (ii) dynamic sparse coordination graphs that adapt dependencies via variance-based TD-error pruning; and (iii) a semantic-conditioned VQ-VAE that discretizes trajectories into coordination classes and maps them to graph-level edge decisions, while also conditioning local policies. |
Kexing Peng; Pengyi Li; tinghuai ma; Jianye Hao; |
| 431 | Preconditioned DeltaNet: Curvature-aware Sequence Modeling for Linear Recurrences Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Existing delta-rule recurrences can be seen as first-order approximations to this objective, but notably ignore the curvature of the least-squares loss during optimization. In this work, we address this by introducing preconditioning to these recurrences. |
Neehal Tumma; Noel Loo; Daniela Rus; |
| 432 | StitchCUDA: An Automated Multi-Agents End-to-End GPU Programing Framework with Rubric-based Agentic Reinforcement Learning Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: To address the challenge, in this work, we propose \textsc{StitchCUDA}, a multi-agent framework for end-to-end GPU program generation, with three specialized agents: a \textit{Planner} to orchestrate whole system design, a \textit{Coder} dedicated to implementing it step-by-step, and a \textit{Verifier} for correctness check and performance profiling using Nsys/NCU. |
Shiyang Li; Zijian Zhang; Winson Chen; Yuebo Luo; Mingyi Hong; Caiwen Ding; |
| 433 | A Distributional View for Visual Mechanistic Interpretability: KL-Minimal Soft-Constraint Principle Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: In this work, we establish a theoretical distributional view for visual MI, which models the influence of a feature activation on the natural image distribution, thereby formulating a Kullback-Leibler (KL)-minimal optimization problem to model the MI task. |
Guancheng Zhou; Yisi Luo; Zhengfu He; Zhenyu Jin; Xuyang Ge; Wentao Shu; Deyu Meng; Xipeng Qiu; |
| 434 | Scaling Transformers for End-to-End Discrete Audio Tokenization Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: In this paper, we argue that discrete audio tokenization should be learned fully end-to-end using a homogeneous and scalable architecture. |
Yitian Gong; Kuangwei Chen; Zhaoye Fei; Xiaogui Yang; Ke Chen; Yang Wang; Kexin Huang; Mingshu Chen; Ruixiao Li; Qinyuan Cheng; Shimin Li; Xipeng Qiu; |
| 435 | Beyond Pixel Context Windows: Neural World Simulators with Persistent 3D State Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: This results in an unrealistic user experience, and presents significant obstacles to down-stream tasks such as training agents. To address this, we present PERSIST, a new paradigm of world model which simulates the evolution of a latent 3D scene: environment, camera and renderer. |
Samuel Garcin; Tom Walker; Steven McDonagh; Tim Pearce; Hakan Bilen; Tianyu He; Kaixin Wang; Jiang Bian; |
| 436 | Scaling Up Multi-Turn Off-Policy RL and Multi-Agent Tree Search for LLM Step-Provers Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: The integration of Large Language Models (LLMs) with automated theorem proving has shown immense promise, yet is constrained by challenges in scaling up both training-time reinforcement learning (RL) and inference-time compute. This paper introduces BFS-Prover-V2, a step-level theorem proving system designed to address this dual scaling problem. |
Ran Xin; Zeyu Zheng; Yanchen Nie; Kun Yuan; Xia Xiao; |
| 437 | Model-Based Diffusion Sampling for Predictive Control in Offline Decision Making Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We propose *Model Predictive Diffuser* (MPDiffuser), a compositional diffusion framework that combines a diffusion planner with a dynamics diffusion model to generate task-aligned and dynamically plausible trajectories. |
Haldun Balim; Na Li; Yilun Du; |
| 438 | Position: Agent Security Needs Redefinition Through A Holistic Framework Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We argue that agent security must be redefined through a holistic framework including four core components: identity (who: authority and authentication), task (what to do: authorized objectives), trajectory (progress: action-observation boundaries), and memory (what can be retrieved: information access control). |
Vincent Siu; Jingxuan He; Kyle Montgomery; Zhun Wang; Chenguang Wang; Dawn Song; |
| 439 | Coupled Variational Reinforcement Learning for Language Model General Reasoning Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: In this paper, we propose \textit{\b{Co}upled \b{V}ariational \b{R}einforcement \b{L}earning} (CoVRL), which bridges variational inference and reinforcement learning by coupling prior and posterior distributions through a hybrid sampling strategy. |
Xueru Wen; Jie Lou; Yanjiang Liu; Hongyu Lin; Ben He; Xianpei Han; Le Sun; Yaojie Lu; Debing Zhang; |
| 440 | AdverMCTS: Combating Pseudo-Correctness in Code Generation Via Adversarial Monte Carlo Tree Search Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We argue that optimizing against a fixed, weak environment inherently limits robustness. To address this, we propose AdverMCTS, a novel adversarial Monte Carlo Tree Search framework that combats pseudo-correctness by coupling code search with active vulnerability discovery. |
Qingyao Li; Weiwen Liu; Weinan Zhang; Yong Yu; Bo An; |
| 441 | Fast Inverse Lithography Via GRPO Reinforced Flow Matching Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We introduce LithoGRPO, an ILT framework that integrates the flow‑matching paradigm with GRPO‑based reinforcement learning (RL) fine‑tuning, enabling efficient exploration of diverse masks for a given target layout. |
Yao Lai; Xuyuan Xiong; Zeyue Xue; Guojin Chen; Jing Wang; Xihui Liu; Rui Zhang; Robert Mullins; Bei Yu; Ping Luo; |
| 442 | Think-at-Hard: Selective Latent Iterations to Improve Reasoning Language Models Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: In this work, we ask whether selectively skipping latent iterations may improve accuracy. |
Tianyu Fu; Yichen You; Zekai Chen; Guohao Dai; Huazhong Yang; Yu Wang; |
| 443 | DRIFT-BENCH: Diagnosing CoopeRative Breakdowns in LLM Agents Under Input Faults Via Multi-Turn Interaction Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We introduce DRIFT-BENCH, the first diagnostic benchmark that evaluates agentic pragmatics under input faults through multi-turn clarification across state-oriented and service-oriented execution environments. |
Han Bao; Zheyuan Zhang; PENGCHENG JING; Zhengqing Yuan; Kaiwen Shi; Yanfang Ye; |
| 444 | Learning Taxonomic Trees with Hierarchical Representation Regularization for Large Multimodal Models Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Specifically, we introduce a semantic-aware visual tree construction framework that extracts coarse-to-fine visual features from intermediate LLM layers guided by textual cues. |
Hulingxiao He; Zhi Tan; Yuxin Peng; |
| 445 | Position: Beyond Reasoning Zombies — AI Reasoning Requires Process Validity Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: To that end, we provide (1) general and extensible definitions for *valid* and *sound reasoning* based on a synthesis of the literature, which can serve as an accessible reference and a starting point for community discussion; and (2) a checklist for best practices in the communication of AI reasoning research. |
Rachel Lawrence; Jacqueline Maasch; |
| 446 | Evolution Strategies at The Hyperscale Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We introduce Evolution Guided GeneRal Optimisation via Low-rank Learning (EGGROLL), which improves arithmetic intensity by structuring individual perturbations as rank-$r$ matrices, resulting in a hundredfold increase in training speed for billion-parameter models at large population sizes, achieving up to 91\% of the throughput of pure batch inference. |
Bidipta Sarkar; Mattie Fellows; Juan Duque; Alistair Letcher; Antonio Villares; Anya Sims; Clarisse Wibault; Dmitry Samsonov; Dylan Cope; Jarek Liesen; Kang Li; Lukas Seier; Theo Wolf; Uljad Berdica; Valentin Mohl; Alexander D. Goldie; Aaron Courville; Karin Sevegnani; Shimon Whiteson; Jakob Foerster; |
| 447 | Scaling Laws and Architectural Frontiers in Metagenomic Foundation Models Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: In this work, we present the machine learning methodology behind MODEL, a family of metagenomic foundation models scaled up to 28 billion parameters and trained on 9.7 trillion nucleotide tokens. |
Geraldene Munsamy; Gavin Ayres; Jérémie DONA; Carla Greco; Daniel P Anderson; Srijani Sridhar; William Chow; Aaron Kollasch; Robert Pecoraro; Tanggis Bohnuud; Keith Kam; Gus Minto-Cowcher; Marcus Leung; Hassan Sirelkhatim; John St. John; Ali Taghibakhshi; Tyler Shimko; Jared Wilber; Timur Rvachov; Saee Paliwal; Eddie Calleja; Noelia Ferruz; Kevin Yang; Francesco Farina; Philipp Lorenz; |
| 448 | A Unified Density Operator View of Flow Control and Merging Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Ultimately, we showcase the capabilities of the proposed method on illustrative settings providing visually interpretable insights, and apply our method to high-dimensional de-novo molecular design and low-energy conformer generation. |
Riccardo De Santi; Malte Franke; Ya-Ping Hsieh; Andreas Krause; |
| 449 | Co-Evolving Latent Action World Models Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: A conceptually simple and appealing idea is to directly replace the forward dynamic model in LAM with a powerful world model and training them jointly, but it is non-trivial and prone to representational collapse. In this work, we propose CoLA-World, which for the first time successfully realizes this synergistic paradigm, resolving the core challenge in joint learning through a critical warm-up phase that effectively aligns the representations of the from-scratch LAM with the pre-trained world model. |
Yucen Wang; Fengming Zhang; De-Chuan Zhan; Li Zhao; Kaixin Wang; Jiang Bian; |
| 450 | Position: Agent Should Invoke External Tools ONLY When Epistemically Necessary Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We introduce the \textit{\textbf{Theory of Agent (ToA)}}, a framework that treats agents as making sequential decisions about whether remaining uncertainty should be resolved internally or delegated externally. |
Hongru WANG; Cheng Qian; Manling Li; Jiahao Qiu; Boyang XUE; Mengdi Wang; Heng Ji; Amos Storkey; Kam-Fai Wong; |
| 451 | Towards High-Fidelity CAD Generation Via LLM-Driven Program Generation and Text-Based B-Rep Primitive Grounding Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Our method generates executable CadQuery scripts, and introduces a text-based query mechanism that enables the LLM to specify geometric selections via natural language, which *BRepGround* then grounds to the target primitives. |
Jiahao Li; Qingwang Zhang; Qiuyu Chen; Guozhan Qiu; Yunzhong Lou; Xiangdong Zhou; |
| 452 | Representation Unlearning: Forgetting Through Information Compression Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We introduce Representation Unlearning, a framework that performs unlearning directly in the model’s representation space. |
Antonio Almudévar; Alfonso Ortega; |
| 453 | Bridging Functional and Representational Similarity Via Usable Information Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We present a unified framework for quantifying the similarity between representations through the lens of \textit{usable information}, offering a rigorous theoretical and empirical synthesis across three key dimensions. |
Antonio Almudévar; Alfonso Ortega; |
| 454 | RELO: Reinforcement Learning to Localize for Visual Object Tracking Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: However, this heuristic supervision merely serves as a surrogate objective, which misaligns with evaluation metrics such as IoU and AUC. To address this limitation, we propose RELO, a reinforcement-learning tracking framework that formulates target localization as a decision-making problem within the Transformer-based tracking paradigm. |
Xin Chen; Chuanyu Sun; Jiao Xu; Houwen Peng; Dong Wang; Huchuan Lu; Kede Ma; |
| 455 | Dual Latent Memory for Visual Multi-agent System Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: To this end, we propose L$^{2}$-VMAS, a novel model-agnostic framework that enables inter-agent collaboration with dual latent memories. |
Xinlei Yu; Chengming Xu; Zhangquan Chen; Bo Yin; Cheng Yang; Yongbo He; Yihao Hu; Jiangning Zhang; Cheng Tan; Xiaobin Hu; Shuicheng YAN; |
| 456 | Stable Deep Reinforcement Learning Via Isotropic Gaussian Representations Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: In particular, they induce stable tracking of time-varying targets for linear readouts, achieve maximal entropy under a fixed variance budget, and encourage a balanced use of all representational dimensions. Building on this insight, we propose the use of Sketched Isotropic Gaussian Regularization for shaping representations toward an isotropic Gaussian during training. |
Ali Saheb pasand; Johan Obando-Ceron; Aaron Courville; Pouya Bashivan; Pablo Samuel Castro; |
| 457 | UnMaskFork: Test-Time Scaling for Masked Diffusion Via Deterministic Action Branching Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: In this work, we demonstrate that Masked Diffusion Language Models (MDLMs) are inherently amenable to advanced search strategies, owing to their iterative and non-autoregressive generation process. |
Kou Misaki; Takuya Akiba; |
| 458 | Riemannian Metric Matching for Scalable Geometric Modeling of Distributions Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We propose **Riemannian metric matching**: a denoising probabilistic framework for learning the Riemannian geometry of data using neural networks. |
Jacob Bamberger; Adam Gosztolai; Pierre Vandergheynst; Michael Bronstein; Iolo Jones; |
| 459 | Temporal Straightening for Latent Planning Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Inspired by the perceptual straightening hypothesis in human visual processing, we introduce temporal straightening for representation learning in latent planning. |
Ying Wang; Oumayma Bounou; Gaoyue Zhou; Randall Balestriero; Tim G. J. Rudner; Yann LeCun; Mengye Ren; |
| 460 | GUI-Spotlight: Adaptive Iterative Focus Refinement for Enhanced GUI Visual Grounding Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: This limitation prevents the system from accurately performing pointer-level actions such as clicking or dragging. To address it, we introduce GUI-Spotlight — a model trained for image-grounded reasoning that dynamically invokes multiple specialized tools to iteratively narrow its focus to the relevant region of the screen, thereby substantially improving visual grounding accuracy. |
Bin Lei; Nuo Xu; Ali Payani; Mingyi Hong; Chunhua Liao; Yu Cao; Caiwen Ding; |
| 461 | Generative Neural Operators Through Diffusion Last Layer Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: However, many practical systems are inherently stochastic, making principled uncertainty quantification essential for reliable deployment. To address this, we introduce a simple add-on, the *diffusion last layer* (DLL), a lightweight probabilistic head that can be attached to arbitrary neural operator backbones to model predictive uncertainty. |
Sungwon Park; Anthony Zhou; Hongjoong Kim; Amir Barati Farimani; |
| 462 | DiScoFormer: Plug-In Density and Score Estimation with Transformers Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We introduce DiScoFormer (Density and Score Transformer), a “train-once, infer-anywhere equivariant Transformer that maps i.i.d. samples to both density values and score vectors, generalizing across distributions and sample sizes. |
Vasily Ilin; Peter Sushko; Ranjay Krishna; |
| 463 | Unveiling The Entropy Dynamics of Chain-of-Thought Reasoning Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Using the Cumulative Sum (CUSUM) algorithm, a statistically optimal change-point detector, we develop a training-free framework for real-time inference control. |
Ting Xu; Xu He; Yupu Lu; Jiankai Sun; Dong Li; Wai Lam; Jianye Hao; |
| 464 | WorldCompass: Reinforcement Learning for Long-Horizon World Models Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: This work presents WorldCompass, a novel Reinforcement Learning (RL) post-training framework for the long-horizon, interactive video-based world models, enabling them to explore the world more accurately and consistently based on interaction signals. |
Zehan Wang; Tengfei Wang; Haiyu Zhang; Xuhui Zuo; Junta Wu; Haoyuan Wang; Wenqiang Sun; Zhenwei Wang; chenjie cao; Hengshuang Zhao; Chunchao Guo; Zhou Zhao; |
| 465 | Learning to Bet for Horizon-Aware Anytime-Valid Testing Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Taken together these results suggest a simple phase diagram in the $(t, \log W_t)$ plane, delineating regions where Kelly, fractional Kelly, and aggressive betting may be preferable. Guided by this phase diagram, we introduce a Deep Reinforcement Learning approach based on a universal Deep Q-Network (DQN) agent that learns a single policy from synthetic experience and maps simple statistics of past observations to bets across horizons and null values. |
Ege Onur Taga; Samet Oymak; Shubhanshu Shekhar; |
| 466 | (Sparse) Attention to The Details: Preserving Spectral Fidelity in ML-based Weather Forecasting Models Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We introduce \textsc{Mosaic}, a probabilistic weather forecasting model that addresses two sources of spectral degradation in ML-based weather prediction: training to predict the ensemble mean deterministically and compressive encoding creating an information bottleneck. |
Maksim Zhdanov; Ana Lucic; Max Welling; Jan-Willem van de Meent; |
| 467 | TINNs: Time-Induced Neural Networks for Solving Time-Dependent PDEs Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We propose *Time-Induced Neural Networks (TINNs)*, a novel architecture that parameterizes the network weights as a learned function of time, allowing the effective spatial representation to evolve over time while maintaining shared structure. |
Chen-Yang Dai; Che-Chia Chang; Te-Sheng Lin; Ming-Chih Lai; Chieh-Hsin Lai; |
| 468 | Distribution Alignment for One-Shot Federated Learning Via Optimal Transport Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We introduce \textsc{SLOT-Align} (Single-round, Learning-free Optimal Transport Alignment), a geometry-aware feature harmonization framework for OSFL. |
Daniele Berardini; Vito Pastore; Vittorio Murino; |
| 469 | Conformal Risk-Averse Decision Making with Action Conditional Guarantee Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We generalize and strengthen their results by (i) introducing action-conditional conformal prediction, which yields safety guarantees conditioned explicitly on each action taken by the decision maker, (ii) showing that action-conditional prediction sets serve as a proxy for the feasible decision space for risk-averse decision makers aiming to optimize action-conditional value-at-risk, and (iii) proposing a principled finite-sample algorithm based on pinball-loss minimization, connecting the framework of \cite{gibbs2025conformal} to action-conditional guarantees. |
Zihan Zhu; Shayan Kiyani; George Pappas; Hamed Hassani; |
| 470 | On The Generalization Gap in Self-Evolving Language Model Reasoning Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: A central open question, however, is not whether self-evolution can help, but: \textit{how far is it from oracle-supervised training under minimal assumptions?} To address this question, we present a controlled empirical analysis of LLM self-evolution under a strict formulation: self-evolution is allowed access only to (i) an unlabeled prompt set and (ii) a base language model, with all supervision signals generated from this model. |
Zhenting Qi; Susanna Maria Baby; Stefanie Baby; Kan Yuan; Da-Cheng Juan; Tu Vu; Andrew Tomkins; Cyrus Rashtchian; |
| 471 | Towards Unified Multimodal Pretraining Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: In this work, we explore the design space of Unified Multimodal Pretraining through a controlled, from-scratch study. |
Shengbang Tong; David Fan; John Nguyen; Ellis Brown; Gaoyue Zhou; Shengyi Qian; Boyang Zheng; Théophane Vallaeys; Rob Fergus; Naila Murray; Marjan Ghazvininejad; Mike Lewis; Jakob Verbeek; Nicolas Ballas; Amir Bar; Michael Rabbat; Yann LeCun; Luke Zettlemoyer; Saining Xie; Koustuv Sinha; |
| 472 | Simultaneous Confidence Bounds for Aggregated Effects Via Exact Subset Optimization Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: A central challenge is that the required maximization is a nonlinear combinatorial optimization problem whose exact solution is essential for correct coverage. We address this challenge by casting the problem as a densest subgraph optimization and reformulating it as a linear program, or as a mixed-integer linear program when downward-closed linear constraints are imposed, enabling efficient and exact evaluation. |
Weihang Xu; Huajie Qian; Wotao Yin; Xinshang Wang; |
| 473 | CODiff: One-Step Diffusion Model for Camouflaged Object Detection Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: In contrast to existing approaches that rely on multiple sample steps to refine the predicted masks, we propose CODiff, which reformulates the diffusion process to enable one-step mask prediction while maintaining competitive accuracy. |
Xiaotong Fu; Wenchao Meng; Qihang Zhou; Qian Liu; Qinmin Yang; Shibo He; |
| 474 | Can Vision Language Models Learn Intuitive Physics from Interaction? Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We train models that learn through interaction with the environment using reinforcement learning. |
Luca M. Schulze Buschoff; Konstantinos Voudouris; Can Demircan; Eric Schulz; |
| 475 | PersistBench: When Should Long-Term Memories Be Forgotten By LLMs? Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: However, the same persistence can also introduce safety risks that have been largely overlooked. Hence, we introduce \textbf{PersistBench} to measure the extent of these safety risks. |
Sidharth Pulipaka; Oliver Chen; Manas Sharma; Taaha Saleem Bajwa; Vyas Raina; Ivaxi Sheth; |
| 476 | CUARewardBench: Benchmark for Evaluating Reward Models on Computer-using Agent Trajectories Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: (4) Unanimous Prompt Ensemble (UPE): Based on the insights from our comprehensive analysis, we propose UPE, a novel ensemble method that significantly enhances reward model reliability through strict unanimous voting and strategic prompt-template configurations. |
Haojia Lin; Xiaoyu Tan; Yulei Qin; Zihan Xu; Yuchen Shi; Zongyi Li; Gang Li; Shaofei Cai; Siqi Cai; Yuzheng Cai; Chaoyou Fu; Ke Li; Xing Sun; |
| 477 | CoopEval: Benchmarking Cooperation-Sustaining Mechanisms and LLM Agents in Social Dilemmas Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Indeed, our experiments show that recent models—with or without reasoning enabled—consistently defect on the other players in single-shot social dilemmas. To tackle this safety concern, we study game-theoretic mechanisms that are designed to enable cooperative outcomes between rational agents _in equilibrium_. |
Emanuel Tewolde; Xiao Zhang; David Guzman Piedrahita; Vincent Conitzer; Zhijing Jin; |
| 478 | Evaluating Agentic Optimization on Large Codebases Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We introduce FormulaCode, a benchmark for evaluating agentic optimization on large, real-world codebases with fine-grained, multi-objective performance metrics. |
Atharva Sehgal; James Hou; Akanksha Sarkar; Ishaan Mantripragada; Swarat Chaudhuri; Jennifer Sun; Yisong Yue; |
| 479 | Think in Cloud, Look at Edges: Semantic-Driven Query Decomposition for Efficient Video Reasoning Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We identify this as a fundamental flaw leading to Semantic Submergence, where dominant visual features drown out subtle but logically critical cues. To solve this, we introduce SCOPE (Semantic Cloud-Orchestrated Perception at Edge). |
Wenhao Zou; Zhijie Cai; Minchen Yu; Zongshuai Zhang; Guangxu Zhu; |
| 480 | Temporal-aware Flow Matching for Video Generation with Temporally Coherent Motion Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: This practice fails to explicitly model motion priors or temporal dependencies, resulting in suboptimal dynamics that may appear incoherent and unrealistic. To solve this problem, we propose Temporal-aware Flow Matching (TFM), a novel training paradigm that embeds inter-frame constraints into the flow objective, leading to temporally coherent motion modeling in video generation. |
Zirui Pan; Xin Wang; Yipeng Zhang; Yuwei Zhou; Wenwu Zhu; |
| 481 | Scaling Law for Quantization-Aware Training Related Papers Related Patents Related Grants Related Venues Related Experts Related Code View Save Highlight: This paper proposes a unified scaling law for QAT that models quantization error as a function of model size, training data volume, and quantization group size. |
Mengzhao Chen; Chaoyi Zhang; Jing Liu; Zeng; Zeyue Xue; Zhiheng Liu; Yunshui Li; Jin Ma; Jie Huang; zhou Xun; Ping Luo; |
| 482 | INT Vs. FP: A Comprehensive Study of Fine-Grained Low-bit Quantization Formats Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We reveal a critical performance crossover: while FP excels in coarse-grained quantization, INT consistently surpasses it as the quantization block size shrinks. |
Mengzhao Chen; Meng Wu; Hui Jin; Zhihang Yuan; Jing Liu; Chaoyi Zhang; Yunshui Li; Jie Huang; Jin Ma; Zeyue Xue; Zhiheng Liu; Xingyan Bin; Ping Luo; |
| 483 | SPARKLING: Balancing Signal Preservation and Symmetry Breaking for Width-Progressive Learning Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Empirically, we show that naive initialization at this stage disrupts activation statistics, triggering loss spikes, while copy-based initialization introduces gradient symmetry that hinders feature diversity. To address these issues, we propose **SPARKLING** (balancing **S**ignal **P**reservation **A**nd symmet**R**y brea**K**ing for width-progressive **L**earn**ING**), a novel framework for mid-stage width expansion. |
Qifan Yu; Xinyu Ma; Zhijian Zhuo; Minrui Wang; Deyi Liu; Shiyi Zhan; Yiyuan Ma; liang xiang; Xingyan Bin; Di He; |
| 484 | Maximizing Mutual Information Between Prompt and Response Improves LLM Performance with No Additional Data Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We propose **Mutual Information-based Preference Optimization (MIPO)**, a contrastive data augmentation method that constructs preference pairs by generating a positive response conditioned on the correct prompt and a negative response conditioned on a random or incomplete prompt; then train with Direct Policy Optimization. |
HyunJi Nam; Haoran Li; Natasha Jaques; |
| 485 | Generation Enhances Understanding in Unified Multimodal Models Via Multi-Representation Generation Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: In this work, we propose UniMRG (Unified Multi-Representation Generation), a simple yet effective architecture-agnostic post-training method. |
Zihan Su; Hongyang Wei; Kangrui Cen; Yong Wang; Guanhua CHEN; Chun Yuan; Xiangxiang Chu; |
| 486 | Hermes: An Evidence-Driven Agentic Framework for Trustworthy and Explainable AI-Generated Video Detection Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Multimodal large language models (MLLMs) are appealing for AI-generated video (AIGV) forensics due to their broad perceptual and reasoning capabilities; however, existing MLLM-based detectors still suffer from hallucination and unstable reasoning, yielding high false-alarm rates and generic, non-verifiable explanations. To address these issues, we propose Hermes, an evidence-driven agentic framework for trustworthy and explainable AIGV detection. |
Shuaibo Li; Pengfei HAO; Hongtao Wu; Jianfeng Dong; Ping Li; Xiaohong Liu; Lei Zhu; |
| 487 | Causes and Consequences of Representational Similarity in Machine Learning Models Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Numerous works have noted similarities in how machine learning models represent the world, even across modalities. |
Zeyu Li; Hung Vu; Damilola Awofisayo; Emily Wenger; |
| 488 | LUVE : Latent-Cascaded Ultra-High-Resolution Video Generation with Dual Frequency Experts Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Recent advances in video diffusion models have significantly improved visual quality, yet ultra-high-resolution (UHR) video generation remains a formidable challenge due to the compounded difficulties of motion modeling, semantic planning, and detail synthesis. To address these limitations, we propose \textbf{LUVE}, a \textbf{L}atent-cascaded \textbf{U}HR \textbf{V}ideo generation framework built upon dual frequency \textbf{E}xperts. |
Chen Zhao; Jiawei Chen; Hongyu Li; Zhuoliang Kang; Shilin Lu; Xiaoming Wei; Kai Zhang; Jian Yang; Ying Tai; |
| 489 | TabMGP: Martingale Posterior with TabPFN Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We introduce TabMGP, an MGP built on TabPFN for tabular data. |
Kenyon Ng; Edwin Fong; David Frazier; Jeremias Knoblauch; Susan Wei; |
| 490 | Spatially-Regularized Entropy for Discriminative Token Merging in Fine-Grained Re-Identification Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: In fine-grained retrieval, these approaches often discard or smooth out subtle but discriminative local details. To resolve this, we propose SRE-Merge, a training-free framework designed for discriminative token compression. |
Shangze Li; Yifan Xu; Jingmiao Liang; Yongfei Zhang; Yuzhuo Ma; Yingbo Qu; |
| 491 | Statistical Learning Theory in Lean 4: Empirical Processes from Scratch Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: We present the first comprehensive Lean 4 formalization of statistical learning theory (SLT) grounded in empirical process theory. |
Yuanhe Zhang; Jason Lee; Fanghui Liu; |
| 492 | Opt-Verifier: Unleashing The Power of LLMs for Optimization Modeling Via Dual-Side Verification Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: This hampers the subsequent verification and correction steps, and thus it severely hurts the modeling accuracy. To address this challenge, we propose a novel LLM-based framework with Dual-side Verification (OptiVer) from both structure and solution perspectives, thereby improving the modeling accuracy. |
Haoyang Liu; Jie Wang; Boxuan Niu; Xiongwei Han; Yian Xu; Mingxuan Ye; Zijie Geng; Fangzhou Zhu; Tao Zhong; Mingxuan Yuan; Jianye Hao; |
| 493 | TOM-SWE: User Mental Modeling For Software Engineering Agents Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Despite their growing ability in coding tasks, these systems still struggle to infer and track user intent, especially when instructions are underspecified or context-dependent. To bridge this gap, we introduce ToM-SWE, a dual-agent architecture that pairs a primary software-engineering (SWE) agent with a lightweight theory-of-mind (ToM) partner agent dedicated to modeling the user’s mental state. |
Xuhui Zhou; Valerie Chen; Zhiruo Wang; Graham Neubig; Maarten Sap; Xingyao Wang; |
| 494 | Both Semantics and Reconstruction Matter: Making Representation Encoders Ready for Text-to-Image Generation and Editing Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: In this paper, we propose a systematic framework to adapt understanding-oriented encoder features for generative tasks. |
Shilong Zhang; He Zhang; Zhifei Zhang; Chongjian GE; Shuchen Xue; Shaoteng Liu; Mengwei Ren; Soo Ye Kim; Yuqian Zhou; Qing Liu; Daniil Pakhomov; Kai Zhang; Zhe Lin; Ping Luo; |
| 495 | ViTok-v2: Scaling Native-Resolution Autoencoders to 5B Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Vision Transformer (ViT) tokenizers offer a scal- able alternative to convolutional auto-encoders, yet current architectures have two key limitations: their performance degrades when images vary in aspect ratio or resolution, and their reliance on adversarial losses makes them harder to train at scale. To address this, we introduce ViTok-v2, a ViT tokenizer building on ViTok. |
Philippe Hansen-Estruch; Jiahui Chen; Vivek Ramanujan; Orr Zohar; Markos Georgopoulos; Animesh Sinha; Ji Hou; Edgar Schönfeld; Felix Juefei-Xu; Sriram Vishwanath; Ali Thabet; |
| 496 | Is Code Better Than Language for Algorithmic Reasoning? Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Comparing NL reasoning and solver-based pipelines directly is ill-posed: they differ simultaneously in representation space and execution mechanism. We introduce a three-route framework that makes this comparison tractable by introducing an intermediary step—code generation with LLM-based execution. |
Terry Tong; Yu Feng; Surbhi Goel; Dan Roth; |
| 497 | Navigating The Pareto Frontier of Alignment:Spectrum-Adaptive Fine-Tuning for LLMs Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: On the other hand, alternatives like Dynamic Fine-Tuning (DFT) suffer from vanishing gradients on these tokens, which severely hinders the acquisition of new concepts. To bridge this gap, we propose **SAFT** **S**pectrum-**A**daptive **F**ine-**T**uning), a unified framework that interpolates between the aggressive learning signal of NLL and the robust nature of probability-weighted optimization. |
Yaoyou Fan; Chao Zhang; Xiaoyu Tan; Chenxing Sun; Yu Yuan; Haoyu Feng; Lu Pan; Ke Zeng; Xunliang Cai; |
| 498 | SpanNorm: Reconciling Training Stability and Performance in Deep Transformers Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: A critical design choice is the placement of normalization layers, leading to a fundamental trade-off: the ”PreNorm” architecture ensures training stability at the cost of potential performance degradation in deep models, while the ”PostNorm” architecture offers strong performance but suffers from severe training instability. In this work, we propose SpanNorm, a novel technique designed to resolve this dilemma by integrating the strengths of both paradigms. |
Wang Chao; Bei Li; Jiaqi Zhang; Xinyu Liu; Yuchun Fan; Linkun Lyu; Xin Chen; Jingang Wang; Tong Xiao; Peng Pei; Xunliang Cai; |
| 499 | An Algebraic View of The Expressivity of Recurrent Language Models Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: Moreover, many proofs are highly architecture-specific and hard to transfer across closely related models. We address these issues with a unifying algebraic framework for a broad class of RNN language models, formally translating them to wreath products of transformation semigroups. |
Franz Nowak; Reda Boumasmoud; Ryan Cotterell; |
| 500 | Edit-Based Refinement for Parallel Masked Diffusion Language Models Related Papers Related Patents Related Grants Related Venues Related Experts View Save Highlight: In this paper, we propose ME-DLM, an edit-based refinement framework that augments diffusion generation with a lightweight post-generation editing step. |
Houxing Ren; Mingjie Zhan; Zimu Lu; Ke Wang; Yunqiao Yang; Haotian Hou; Junting Pan; Hongsheng Li; |
This table only includes 500 papers selected by our daily digest algorithm. To continue with the full list (~6,500 papers), please visit Paper Digest: ICML-2026 (Full List).