When One Modality Sabotages the Others: A Diagnostic Lens on Multimodal Reasoning
机构 * Harvard University（哈佛大学） ； MIT Media Lab（麻省理工学院媒体实验室）
Comments Accepted at the Multimodal Algorithmic Reasoning (MAR) Workshop, NeurIPS 2025
期刊&会议
NeurIPS
Conference on Neural Information Processing Systems · 会议 · Machine Learning
Adam Reduces a Unique Form of Sharpness: Theoretical Insights Near the Minimizer Manifold
机构 * Tsinghua University（清华大学） ； Institute for Interdisciplinary Information Sciences（交叉信息研究院）
Journal ref The Thirty-Ninth Annual Conference on Neural Information Processing Systems (NeurIPS 2025)
VidEmo: Affective-Tree Reasoning for Emotion-Centric Video Foundation Models
机构 * Nankai University（南开大学） ； Pengcheng Laboratory（鹏城实验室） ； Kuaishou Technology（快手科技） ； Nankai International Advanced Research Institute (SHENZHEN·FUTIAN)（南开国际先进研究院（深圳·福田））
Comments 41 pages, 26 figures
Journal ref NeurIPS 2025
Curriculum Design for Trajectory-Constrained Agent: Compressing Chain-of-Thought Tokens in LLMs
机构 * MPI-SWS（马克斯·普朗克所际研究所）
Comments NeurIPS'25 paper
Differentiable Hierarchical Visual Tokenization
Comments NeurIPS 2025 Spotlight
Adaptive Neighborhood-Constrained Q Learning for Offline Reinforcement Learning
机构 * Department of Automation, Tsinghua University（自动化系，清华大学）
Comments Accepted to NeurIPS 2025 (Spotlight)
DetectiumFire: A Comprehensive Multi-modal Dataset Bridging Vision and Language for Fire Understanding
机构 * Department of Computer Science（计算机科学系） ； Tulane University（Tulane 大学） ； Department of Industrial Engineering and Management（工业工程与管理系） ； Aalto University（Aalto 大学）
Comments Advances in Neural Information Processing Systems 2025 (NeurIPS 2025), Poster, https://neurips.cc/virtual/2025/loc/san-diego/poster/121400
Human-Machine Ritual: Synergic Performance through Real-Time Motion Recognition
机构 * Department of Digital Arts and Experimental Media (DXARTS), University of Washington（数字艺术与实验媒体系（DXARTS）、华盛顿大学）
Comments 8 pages, 5 figures. Camera-ready manuscript for the Creative AI Track of NeurIPS 2025
LTD-Bench: Evaluating Large Language Models by Letting Them Draw
机构 * Youtu-Agent Team（优图智能团队）
Comments Accepted by NeurIPS 2025
Cycle-Sync: Robust Global Camera Pose Estimation through Enhanced Cycle-Consistent Synchronization
机构 * School of Mathematics, University of Minnesota（明尼苏达大学数学系） ； Department of Mathematics, University of California, Davis（加州大学戴维斯分校数学系）
Comments NeurIPS 2025 spotlight paper
Gradient-Variation Online Adaptivity for Accelerated Optimization with Hölder Smoothness
Comments NeurIPS 2025
QiMeng-NeuComBack: Self-Evolving Translation from IR to Assembly Code
机构 * State Key Lab of Processors, Institute of Computing Technology, Chinese Academy of Sciences, Beijing, China（中国科学院计算技术研究所处理器重点实验室，北京，中国） ； University of Chinese Academy of Sciences, Beijing, China（中国科学院大学，北京，中国）
Comments Accepted at NeurIPS 2025
GRAM-DTI: adaptive multimodal representation learning for drug target interaction prediction
机构 * University of Texas at Arlington（德克萨斯理工大学） ； Johnson & Johnson Innovative Medicine（强生创新医学）
Journal ref NeurIPS 2025 2nd Workshop on Multi-modal Foundation Models and Large Language Models for Life Sciences
Generative World Models of Tasks: LLM-Driven Hierarchical Scaffolding for Embodied Agents
机构 * Department of Computer Science University of Wisconsin-Madison（计算机科学系威斯康星大学麦迪逊分校）
Comments In the 39th Conference on Neural Information Processing Systems (NeurIPS 2025) Workshop: Embodied World Models for Decision Making (EWM)
The Physical Basis of Prediction: World Model Formation in Neural Organoids via an LLM-Generated Curriculum
机构 * Department of Computer Science University of Wisconsin-Madison（计算机科学系 威斯康星大学麦迪逊分校）
Comments Published in the proceedings of the 39th Conference on Neural Information Processing Systems (NeurIPS 2025) Workshop: Scaling Environments for Agents (SEA). Additionally accepted for presentation in NeurIPS 2025 Workshop: Embodied World Models for Decision Making
Co-Evolving Complexity: An Adversarial Framework for Automatic MARL Curricula
机构 * Department of Computer Science University of Wisconsin-Madison（计算机科学系 明尼苏达大学）
Comments Published in the proceedings of the 39th Conference on Neural Information Processing Systems (NeurIPS 2025) Workshop: Scaling Environments for Agents (SEA)
Prior-Guided Flow Matching for Target-Aware Molecule Design with Learnable Atom Number
机构 * School of Computer Science, Shanghai Jiao Tong University（上海交通大学计算机科学学院）
Comments Accepted by NeurIPS 2025
DIsoN: Decentralized Isolation Networks for Out-of-Distribution Detection in Medical Imaging
机构 * Department of Engineering Science, University of Oxford（工程科学系，牛津大学）
Comments Accepted at NeurIPS 2025
Reliably Detecting Model Failures in Deployment Without Labels
Comments 44 pages, 9 figures, 12 tables. Accepted at NeurIPS 2025. Code available at https://github.com/teivng/d3m
From Flat to Hierarchical: Extracting Sparse Representations with Matching Pursuit
机构 * EPFL（瑞士联邦理工学院） ； Kempner Institute, Harvard University（哈佛大学凯普纳研究所） ； CBS-NTT Program in Physics of Intelligence, Harvard University（哈佛大学物理智能CBSTNTT项目） ； Physics of Artificial Intelligence Group, NTT Research, Inc., Sunnyvale, CA, USA（NTT研究公司人工智能物理研究组） ； University of Alberta（阿尔伯塔大学） ； Alberta Machine Intelligence Institute (Amii)（阿尔伯塔人工智能研究所） ； SEAS, Harvard University（哈佛大学工程与应用科学学院）
Comments 39th Conference on Neural Information Processing Systems (NeurIPS 2025)
MTL-KD: Multi-Task Learning Via Knowledge Distillation for Generalizable Neural Vehicle Routing Solver
机构 * College of Computer Science and Software Engineering, Shenzhen University（深圳大学计算机科学与软件工程学院） ； School of Automation and Intelligent Manufacturing, Southern University of Science and Technology（南方科技大学自动化与智能制造学院） ； Guangdong Provincial Key Laboratory of Fully Actuated System Control Theory and Technology, Southern University of Science and Technology（广东省全驱动系统控制理论与技术重点实验室，南方科技大学） ； Department of Industrial Engineering and Innovation Sciences, Eindhoven University of Technology（埃因霍温理工大学工业工程与创新科学系）
Comments Accepted at NeurIPS 2025
SWE-rebench: An Automated Pipeline for Task Collection and Decontaminated Evaluation of Software Engineering Agents
机构 * Nebius
Comments Dataset: https://huggingface.co/datasets/nebius/SWE-rebench, SWE-rebench leaderboard https://swe-rebench.com NeurIPS 2025
A Unified Framework for Variable Selection in Model-Based Clustering with Missing Not at Random
机构 * Faculty of Mathematics and Computer Science, University of Science, Ho Chi Minh City, Vietnam（越南科技大学数学与计算机科学学院） ； Vietnam National University Ho Chi Minh City, Vietnam（越南胡志明市国家大学） ； School of Information and Communications Technology, Hanoi University of Science and Technology, Ha Noi, Vietnam（河内科学技术大学信息与通信技术学院） ； ARC Centre of Excellence for the Mathematical Analysis of Cellular Systems（细胞系统数学分析卓越中心） ； School of Mathematical Sciences, Queensland University of Technology, Brisbane City, Australia（昆士兰科技大学数学科学学院）
Comments Binh H. Ho, Long Nguyen Chi, and TrungTin Nguyen are co-first authors. Correct final typos for the NeurIPS 2025 camera-ready version
Journal ref 39th Conference on Neural Information Processing Systems (NeurIPS 2025)
Dense Backpropagation Improves Training for Sparse Mixture-of-Experts
机构 * University of Maryland（马里兰大学） ； Capital One ； University of Chicago（芝加哥大学） ； Mila – Quebec AI Institute（魁北克人工智能研究所）
Comments NeurIPS 2025
RoMA: Scaling up Mamba-based Foundation Models for Remote Sensing
机构 * College of Computer Science and Technology, National University of Defense Technology, China（国防科技大学计算机科学与技术学院） ； Tsinghua University, China（清华大学） ； Beijing University of Posts and Telecommunications, China（北京邮电大学） ； School of Computer Science, Wuhan University, China（武汉大学计算机学院） ； Zhongguancun Academy, China（中关村学院）
Comments NeurIPS 2025
Position: Bridge the Gaps between Machine Unlearning and AI Regulation
机构 * University of Cambridge（剑桥大学）
Comments NeurIPS 2025 Position Paper Track Oral, https://openreview.net/forum?id=0ngi2StMwC
Image Super-Resolution with Guarantees via Conformalized Generative Models
机构 * School of Applied Mathematics（应用数学学院） ； Getulio Vargas Foundation（格尔维斯基金会）
Comments To appear at NeurIPS 2025. 17 pages, 7 figures
Twilight: Adaptive Attention Sparsity with Hierarchical Top-$p$ Pruning
机构 * Tsinghua University（清华大学） ； Massachusetts Institute of Technology（麻省理工学院） ； University of California, Berkeley（加州大学伯克利分校）
Comments To appear on NeurIPS 2025 (spotlight)
On Extending Direct Preference Optimization to Accommodate Ties
机构 * Department of Engineering University of Cambridge（工程系剑桥大学）
Comments 24 pages, NeurIPS 2025
Learning Interactive World Model for Object-Centric Reinforcement Learning
机构 * University of California San Diego（加州大学圣地亚哥分校） ； Mohamed bin Zayed University of Artificial Intelligence（马尔代夫穆罕默德·本·扎耶德人工智能大学） ； University of Amsterdam（阿姆斯特丹大学）
Comments NeurIPS 2025
