NeurIPS

Embracing Trustworthy Brain-Agent Collaboration as Paradigm Extension for Intelligent Assistive Technologies

Yankai Chen, Xinni Zhang, Yifei Zhang, Yangning Li, Henry Peng Zou, Chunyu Miao, Weizhi Zhang, Xue Liu, Philip S. Yu

机构 * University of Illinois Chicago（伊利诺伊大学芝加哥分校） ； MBZUAI（马克斯·普朗克人工智能研究所） ； McGill University（麦吉尔大学） ； The Chinese University of Hong Kong（香港中文大学） ； Nanyang Technological University（南洋理工大学） ； Tsinghua University（清华大学） 纠错

Comments Accepted by NeurIPS'25 Position Track

Differentially Private High-dimensional Variable Selection via Integer Programming

Petros Prastakos, Kayhan Behdin, Rahul Mazumder

机构 * Operations Research Center（运筹优化中心） ； MIT（麻省理工学院） ； Sloan School of Management（斯隆管理学院） 纠错

Comments NeurIPS 2025

VLM-SlideEval: Evaluating VLMs on Structured Comprehension and Perturbation Sensitivity in PPT

Hyeonsu Kang, Emily Bao, Anjan Goswami

Comments 39th Conference on Neural Information Processing Systems (NeurIPS 2025) Workshop: Evaluating the Evolving LLM Lifecycle - Benchmarks, Emergent Abilities, and Scaling

Fast Non-Log-Concave Sampling under Nonconvex Equality and Inequality Constraints with Landing

Kijung Jeon, Michael Muehlebach, Molei Tao

机构 * Georgia Institute of Technology（佐治亚理工学院） ； MPI-IS（马克斯·普朗克研究所） 纠错

Comments 62 pages

Journal ref NeurIPS 2025

Predictive Coding Enhances Meta-RL To Achieve Interpretable Bayes-Optimal Belief Representation Under Partial Observability

Po-Chen Kuo, Han Hou, Will Dabney, Edgar Y. Walker

机构 * University of Washington（华盛顿大学） ； Allen Institute for Neural Dynamics（神经动态研究院） ； Google DeepMind（谷歌DeepMind） 纠错

Comments Accepted to Annual Conference on Neural Information Processing Systems (NeurIPS) 2025

Online Optimization for Offline Safe Reinforcement Learning

Yassine Chemingui, Aryan Deshwal, Alan Fern, Thanh Nguyen-Tang, Janardhan Rao Doppa

机构 * Washington State University（华盛顿州立大学） ； University of Minnesota（明尼苏达大学） ； Oregon State University（俄勒冈州立大学） ； New Jersey Institute of Technology（新泽西理工学院） 纠错

Comments To appear in NeurIPS 2025 Conference

From Black-box to Causal-box: Towards Building More Interpretable Models

Inwoo Hwang, Yushu Pan, Elias Bareinboim

机构 * Causal Artificial Intelligence Lab（因果人工智能实验室） ； Columbia University（哥伦比亚大学） 纠错

Comments NeurIPS 2025

Deep Learning on Real-World Graphs

Emanuele Rossi

Comments The thesis was submitted for the degree of Doctor of Philosophy in Computing at Imperial College London (February 2024), under the supervision of Prof. Michael M. Bronstein. It includes work published at ICML, ICLR, NeurIPS, and the Learning on Graphs Conference

Two-Steps Diffusion Policy for Robotic Manipulation via Genetic Denoising

Mateo Clemente, Leo Brunswic, Rui Heng Yang, Xuan Zhao, Yasser Khalil, Haoyu Lei, Amir Rasouli, Yinchuan Li

机构 * Huawei Technologies Canada（华为技术加拿大） 纠错

Comments 16 pages, 11 figure, 2 tables, accepted at Neurips 2025

Revisiting Orbital Minimization Method for Neural Operator Decomposition

J. Jon Ryu, Samuel Zhou, Gregory W. Wornell

Comments 25 pages, 8 figures. To appear at NeurIPS 2025

Explaining and Mitigating Crosslingual Tokenizer Inequities

Catherine Arnett, Tyler A. Chang, Stella Biderman, Benjamin K. Bergen

机构 * EleutherAI ； UC San Diego（加州大学圣地亚哥分校） 纠错

Comments Accepted to NeurIPS 2025

Software Engineering Agents for Embodied Controller Generation : A Study in Minigrid Environments

Timothé Boulet, Xavier Hinaut, Clément Moulin-Frier

机构 * Inria Center of Bordeaux University（Inria 波尔多中心大学） ； Flowers AI & CogSci Lab（Flowers AI 与认知科学实验室） ； LaBRI, Bordeaux INP, CNRS UMR 5800（LaBRI，波尔多 INP，CNRS UMR 5800） ； IMN, CNRS UMR 5293（IMN，CNRS UMR 5293） ； Inria - Flowers AI & CogSci Lab - BioTiC（Inria - Flowers AI 与认知科学实验室 - BioTiC） 纠错

Comments 10 pages, 7 figures

Journal ref 39th Conference on Neural Information Processing Systems (NeurIPS 2025) Workshop: Efficient Reasoning

Restoring Pruned Large Language Models via Lost Component Compensation

Zijian Feng, Hanzhang Zhou, Zixiao Zhu, Tianjiao Li, Jia Jim Deryl Chua, Lee Onn Mak, Gee Wah Ng, Kezhi Mao

机构 * School of Electrical and Electronic Engineering, Nanyang Technological University, Singapore（电子与电气工程学院，南洋理工大学，新加坡） ； Home Team Science and Technology Agency (HTX), Singapore（科技局（HTX），新加坡） 纠错

Comments NeurIPS 2025 Spotlight

Avi: Action from Volumetric Inference

Harris Song, Long Le

机构 * University of California, Los Angeles（加州大学洛杉矶分校） ； University of pennsylvania（宾夕法尼亚大学） 纠错

Comments NeurIPS 2025 Workshop on Embodied World Models for Decision Making. URL: https://avi-3drobot.github.io/

Disentanglement Beyond Static vs. Dynamic: A Benchmark and Evaluation Framework for Multi-Factor Sequential Representations

Tal Barami, Nimrod Berman, Ilan Naiman, Amos H. Hason, Rotem Ezra, Omri Azencot

机构 * Faculty of Computer and Information Science（计算机与信息科学学院） ； Ben-Gurion University of the Negev（贝加尔-内盖夫大学） 纠错

Comments Accepted for the Thirty-ninth Annual Conference on Neural Information Processing Systems Datasets and Benchmarks Track

Holistic Order Prediction in Natural Scenes

Pierre Musacchio, Hyunmin Lee, Jaesik Park

机构 * Seoul National University（首尔国立大学） ； LG AI Research（LG人工智能研究） 纠错

Comments 24 pages, 11 figures, 6 tables

Journal ref The Thirty-Ninth Annual Conference on Neural Information Processing Systems (2025)

Decoder-Hybrid-Decoder Architecture for Efficient Reasoning with Long Generation

Liliang Ren, Congcong Chen, Haoran Xu, Young Jin Kim, Adam Atkinson, Zheng Zhan, Jiankai Sun, Baolin Peng, Liyuan Liu, Shuohang Wang, Hao Cheng, Jianfeng Gao, Weizhu Chen, Yelong Shen

机构 * Microsoft（微软公司） ； Stanford University（斯坦福大学） 纠错

Comments Accepted by NeurIPS 2025. Camera-ready Version

When Can Model-Free Reinforcement Learning be Enough for Thinking?

Josiah P. Hanna, Nicholas E. Corrado

机构 * Computer Sciences Department University of Wisconsin – Madison（威斯康星大学麦迪逊分校计算机科学系） 纠错

Comments 26 pages, 4 figures, Accepted to NeurIPS 2025

Vision Transformers Don't Need Trained Registers

Nick Jiang, Amil Dravid, Alexei Efros, Yossi Gandelsman

机构 * UC Berkeley（伯克利大学） 纠错

Comments Project page and code: https://avdravid.github.io/test-time-registers. Accepted to NeurIPS '25 (spotlight)

The Surprising Effectiveness of Negative Reinforcement in LLM Reasoning

Xinyu Zhu, Mengzhou Xia, Zhepei Wei, Wei-Lin Chen, Danqi Chen, Yu Meng

机构 * Computer Science Department, University of Virginia（弗吉尼亚大学计算机科学系） ； Princeton Language and Intelligence (PLI), Princeton University（普林斯顿大学语言与智能研究所） 纠错

Comments Accepted to NeurIPS 2025

On the Hardness of Approximating Distributions with Tractable Probabilistic Models

John Leland, YooJung Choi

机构 * School of Computing and Augmented Intelligence（计算与增强智能学院） ； Arizona State University（亚利桑那州立大学） 纠错

Comments Accepted to NeurIPS 2025. Proceedings forthcoming

$\texttt{AVROBUSTBENCH}$: Benchmarking the Robustness of Audio-Visual Recognition Models at Test-Time

Sarthak Kumar Maharana, Saksham Singh Kushwaha, Baoming Zhang, Adrian Rodriguez, Songtao Wei, Yapeng Tian, Yunhui Guo

机构 * The University of Texas at Dallas（德克萨斯大学达拉斯分校） 纠错

Comments 39th Conference on Neural Information Processing Systems (NeurIPS 2025) Track on Datasets and Benchmarks

Uni-MuMER: Unified Multi-Task Fine-Tuning of Vision-Language Model for Handwritten Mathematical Expression Recognition

Yu Li, Jin Jiang, Jianhua Zhu, Shuai Peng, Baole Wei, Yuxuan Zhou, Liangcai Gao

机构 * Wangxuan Institute of Computer Technology, Peking University（计算机技术研究院，北京大学） 纠错

Comments Accepted by NeurIPS 2025 as a spotlight

Statistical Inference under Performativity

Xiang Li, Yunai Li, Huiying Zhong, Lihua Lei, Zhun Deng

机构 * Northwestern University（西北大学） ； Massachusetts Institute of Technology（麻省理工学院） ； Stanford University（斯坦福大学） ； UNC at Chapel Hill（北卡罗来纳大学教堂山分校） 纠错

Comments Accepted by NeurIPS 2025

Raw2Drive: Reinforcement Learning with Aligned World Models for End-to-End Autonomous Driving (in CARLA v2)

Zhenjie Yang, Xiaosong Jia, Qifeng Li, Xue Yang, Maoqing Yao, Junchi Yan

机构 * Sch. of CS, Sch. of AIS, Sch. of AI, Shanghai Jiao Tong University（计算机科学学院、人工智能科学学院、人工智能学院，上海交通大学） ； Institute of Trustworthy Embodied AI, Fudan University（可信具身人工智能研究院，复旦大学） ； AgiBot 纠错

Comments Accepted by NeurIPS 2025

Physics-informed Reduced Order Modeling of Time-dependent PDEs via Differentiable Solvers

Nima Hosseini Dashtbayaz, Hesam Salehipour, Adrian Butscher, Nigel Morris

机构 * Department of Computer Science University of Western Ontario（计算机科学系大学西部安大略） 纠错

Comments Accepted at NeurIPS 2025

RLVR-World: Training World Models with Reinforcement Learning

Jialong Wu, Shaofeng Yin, Ningya Feng, Mingsheng Long

机构 * School of Software, BNRist, Tsinghua University（软件学院，北研所，清华大学） ； Zhili College, Tsinghua University（知礼学院，清华大学） 纠错

Comments NeurIPS 2025. Code is available at project website: https://thuml.github.io/RLVR-World/

Variational Regularized Unbalanced Optimal Transport: Single Network, Least Action

Yuhao Sun, Zhenyi Zhang, Zihan Wang, Tiejun Li, Peijie Zhou

机构 * Center for Machine Learning Research, Peking University（北京大学机器学习研究中心） ； LMAM and School of Mathematical Sciences, Peking University（北京大学LMAM与数学科学学院） ； Center for Quantitative Biology, Peking University（北京大学定量生物学中心） ； NELBDA, Peking University（北京大学NELBDA研究中心） ； AI for Science Institute, Beijing（北京人工智能科学研究院） 纠错

Comments Published as a conference paper at NeurIPS 2025

Minimizing False-Positive Attributions in Explanations of Non-Linear Models

Anders Gjølbye, Stefan Haufe, Lars Kai Hansen

机构 * Technical University of Denmark（丹麦技术大学） ； Technische Universität Berlin（柏林技术大学） ； Physikalisch-Technische Bundesanstalt, Berlin（柏林物理技术部） ； Charité – Universitätsmedizin Berlin（柏林查理医院大学医学中心） 纠错

Comments Accepted at NeurIPS 2025. Code: https://github.com/gjoelbye/PatternLocal

Modeling Cell Dynamics and Interactions with Unbalanced Mean Field Schrödinger Bridge

Zhenyi Zhang, Zihan Wang, Yuhao Sun, Tiejun Li, Peijie Zhou

机构 * LMAM and School of Mathematical Sciences, Peking University（北京大学理学院和数学科学学院） ； Center for Quantitative Biology, Peking University（北京大学定量生物学中心） ； Center for Machine Learning Research, Peking University（北京大学机器学习研究中心） ； NELBDA, Peking University（北京大学NELBDA研究中心） ； AI for Science Institute, Beijing（北京人工智能科学研究院） 纠错

Comments Published as a conference paper at NeurIPS 2025