NeurIPS

Fast Monte Carlo Tree Diffusion: 100x Speedup via Parallel Sparse Planning

Jaesik Yoon, Hyeonseo Cho, Yoshua Bengio, Sungjin Ahn

Comments 20 pages, 6 figures, NeurIPS 25 Spotlight

Router-R1: Teaching LLMs Multi-Round Routing and Aggregation via Reinforcement Learning

Haozhen Zhang, Tao Feng, Jiaxuan You

机构 * University of Illinois at Urbana-Champaign（伊利诺伊大学厄巴纳-香槟分校） 纠错

Comments Accepted by NeurIPS 2025. Code is available at https://github.com/ulab-uiuc/Router-R1. Models and Datasets are available at https://huggingface.co/collections/ulab-ai/router-r1-6851bbe099c7a56914b5db03

Sharper Convergence Rates for Nonconvex Optimisation via Reduction Mappings

Evan Markou, Thalaiyasingam Ajanthan, Stephen Gould

Comments Accepted to NeurIPS 2025 (Spotlight). 43 pages, 12 figures

Learning to Focus: Causal Attention Distillation via Gradient-Guided Token Pruning

Yiju Guo, Wenkai Yang, Zexu Sun, Ning Ding, Zhiyuan Liu, Yankai Lin

Comments Accepted at NeurIPS 2025, camera-ready version

RSafe: Incentivizing proactive reasoning to build robust and adaptive LLM safeguards

Jingnan Zheng, Xiangtian Ji, Yijun Lu, Chenhang Cui, Weixiang Zhao, Gelei Deng, Zhenkai Liang, An Zhang, Tat-Seng Chua

Journal ref 39th Conference on Neural Information Processing Systems (NeurIPS 2025). 39th Conference on Neural Information Processing Systems (NeurIPS 2025). 39th Conference on Neural Information Processing Systems (NeurIPS 2025)

Simulating Society Requires Simulating Thought

Chance Jiajie Li, Jiayi Wu, Zhenze Mo, Ao Qu, Yuhan Tang, Kaiya Ivy Zhao, Yulu Gan, Jie Fan, Jiangbo Yu, Jinhua Zhao, Paul Liang, Luis Alonso, Kent Larson

机构 * MIT Media Lab（麻省理工学院媒体实验室） ； MIT EECS（麻省理工学院电子工程与计算机科学系） ； MIT BCS（麻省理工学院生物工程与计算机科学系） ； MIT IDSS（麻省理工学院国际设计与研究系统） ； MIT CEE（麻省理工学院土木与环境工程系） ； MIT DUSP（麻省理工学院设计与科学计划） ； MIT Architecture（麻省理工学院建筑系） ； Northeastern University（东北大学） ； Brown University（布朗大学） ； McGill University（麦吉尔大学） 纠错

Comments NeurIPS 2025 (Position Paper Track)

BEAST: Efficient Tokenization of B-Splines Encoded Action Sequences for Imitation Learning

Hongyi Zhou, Weiran Liao, Xi Huang, Yucheng Tang, Fabian Otto, Xiaogang Jia, Xinkai Jiang, Simon Hilber, Ge Li, Qian Wang, Ömer Erdinç Yağmurlu, Nils Blank, Moritz Reuss, Rudolf Lioutikov

机构 * Karlsruhe Institute of Technology（卡尔斯鲁厄理工大学） ； Robotics Institute Germany（德国机器人研究所） ； Microsoft Research（微软研究院） 纠错

Comments Accepted by NeurIPS 2025

Rectified Point Flow: Generic Point Cloud Pose Estimation

Tao Sun, Liyuan Zhu, Shengyu Huang, Shuran Song, Iro Armeni

机构 * Stanford University（斯坦福大学） ； NVIDIA Research（NVIDIA研究） 纠错

Comments NeurIPS 2025 Camera-ready. Project page: https://rectified-pointflow.github.io/

Beyond Accuracy: Dissecting Mathematical Reasoning for LLMs Under Reinforcement Learning

Jiayu Wang, Yifei Ming, Zixuan Ke, Caiming Xiong, Shafiq Joty, Aws Albarghouthi, Frederic Sala

机构 * University of Wisconsin-Madison（威斯康星大学麦迪逊分校） ； Salesforce AI Research（Salesforce AI研究） 纠错

Comments Accepted to NeurIPS 2025

Grasp2Grasp: Vision-Based Dexterous Grasp Translation via Schrödinger Bridges

Tao Zhong, Jonah Buchanan, Christine Allen-Blanchette

机构 * Princeton University（普林斯顿大学） ； San Jose State University（圣何塞州立大学） ； Lockheed Martin Corporation（洛克希德·马丁公司） 纠错

Comments Accepted at NeurIPS 2025

Principled Data Augmentation for Learning to Solve Quadratic Programming Problems

Chendi Qian, Christopher Morris

机构 * Department of Computer Science（计算机科学系） ； RWTH Aachen University（亚琛工业大学） 纠错

Comments Accepted at NeurIPS 2025 as spotlight

zip2zip: Inference-Time Adaptive Tokenization via Online Compression

Saibo Geng, Nathan Ranchin, Yunzhen yao, Maxime Peyrard, Chris Wendler, Michael Gastpar, Robert West

机构 * EPFL（瑞士联邦理工学院） ； Northeastern University（东北大学） ； Université Grenoble Alpes（格勒诺布尔阿尔卑斯大学） ； CNRS（法国国家科学研究中心） ； Grenoble INP（格勒诺布尔INP） ； LIG（图灵研究所） 纠错

Comments NeurIPS 2025

FSNet: Feasibility-Seeking Neural Network for Constrained Optimization with Guarantees

Hoang T. Nguyen, Priya L. Donti

机构 * Massachusetts Institute of Technology（麻省理工学院） 纠错

Comments In NeurIPS 2025. Code available at https://github.com/MOSSLab-MIT/FSNet

The Rich and the Simple: On the Implicit Bias of Adam and SGD

Bhavya Vasudeva, Jung Whan Lee, Vatsal Sharan, Mahdi Soltanolkotabi

机构 * Department of Computer Science University of Southern California（计算机科学系 美国南加州大学） 纠错

Comments Accepted at NeurIPS 2025; 30 pages, 12 figures, 17 tables

LayerIF: Estimating Layer Quality for Large Language Models using Influence Functions

Hadi Askari, Shivanshu Gupta, Fei Wang, Anshuman Chhabra, Muhao Chen

机构 * University of California, Davis（加州大学戴维斯分校） ； University of California, Irvine（加州大学伊市分校） ； University of Southern California（南加州大学） ； University of South Florida（佛罗里达州立大学） 纠错

Comments Neurips 2025

RiverMamba: A State Space Model for Global River Discharge and Flood Forecasting

Mohamad Hakam Shams Eddin, Yikui Zhang, Stefan Kollet, Juergen Gall

机构 * Institute of Computer Science, University of Bonn（波恩大学计算机科学研究所） ； Lamarr Institute for Machine Learning and Artificial Intelligence（拉马尔人工智能与机器学习研究所） ； Institute of Bio- and Geosciences Agrosphere (IBG-3), Research Centre Jülich（茹里奇研究中心生物与地质科学农业球面研究所） ； Centre for High-Performance Scientific Computing in Terrestrial Systems, Geoverbund ABC/J, Jülich（地球系统高性能科学计算中心，Geoverbund ABC/J，茹里奇） 纠错

Comments Accepted at the 39th Conference on Neural Information Processing Systems (NeurIPS 2025). Main paper 10 pages, Appendix 54 pages

Towards Comprehensive Scene Understanding: Integrating First and Third-Person Views for LVLMs

Insu Lee, Wooje Park, Jaeyun Jang, Minyoung Noh, Kyuhong Shim, Byonghyo Shim

Comments Accepted to NeurIPS 2025 (Spotlight)

Revisiting Bi-Linear State Transitions in Recurrent Neural Networks

M. Reza Ebrahimi, Roland Memisevic

机构 * Qualcomm AI Research（高通人工智能研究） 纠错

Comments Accepted to NeurIPS 2025

Revisiting Multi-Agent World Modeling from a Diffusion-Inspired Perspective

Yang Zhang, Xinran Li, Jianing Ye, Shuang Qiu, Delin Qu, Xiu Li, Chongjie Zhang, Chenjia Bai

机构 * Tsinghua University（清华大学） ； The Hong Kong University of Science and Technology（香港科技大学） ； Washington University in St. Louis（圣路易斯华盛顿大学） ； City University of Hong Kong（香港城市大学） ； Fudan University（复旦大学） ； Institute of Artificial Intelligence (TeleAI)（人工智能研究所） ； China Telecom（中国电信） ； Shenzhen Research Institute of Northwestern Polytechnical University（西北工业大学深圳研究院） 纠错

Comments Accepted at NIPS'25

Can Agents Fix Agent Issues?

Alfin Wijaya Rahardja, Junwei Liu, Weitong Chen, Zhenpeng Chen, Yiling Lou

Comments Accepted by the 39th Annual Conference on Neural Information Processing Systems (NeurIPS 2025)

Lorentz Local Canonicalization: How to Make Any Network Lorentz-Equivariant

Jonas Spinner, Luigi Favaro, Peter Lippmann, Sebastian Pitz, Gerrit Gerhartz, Tilman Plehn, Fred A. Hamprecht

机构 * ITP, Heidelberg University, Germany（海德堡大学ITP, 德国） ； CP3, UCLouvain, Belgium（列日大学CP3, 比利时） ； IWR, Heidelberg University, Germany（海德堡大学IWR, 德国） 纠错

Comments 22 pages, 6 figures, 6 tables. v2: NeurIPS camera-ready version, 23 pages, 6 figures, 7 tables

Dependency Parsing is More Parameter-Efficient with Normalization

Paolo Gajo, Domenic Rosati, Hassan Sajjad, Alberto Barrón-Cedeño

机构 * University of Bologna（博洛尼亚大学） ； Dalhousie University（达尔豪斯大学） 纠错

Comments Accepted at NeurIPS 2025

SAMA: Towards Multi-Turn Referential Grounded Video Chat with Large Language Models

Ye Sun, Hao Zhang, Henghui Ding, Tiehua Zhang, Xingjun Ma, Yu-Gang Jiang

Comments NeurIPS 2025

How Does Sequence Modeling Architecture Influence Base Capabilities of Pre-trained Language Models? Exploring Key Architecture Design Principles to Avoid Base Capabilities Degradation

Xin Lu, Yanyan Zhao, Si Wei, Shijin Wang, Bing Qin, Ting Liu

机构 * Research Center for Social Computing and Interactive Robotics（社会计算与交互机器人研究中心） ； Harbin Institute of Technology（哈尔滨工业大学） ； iFLYTEK Co., Ltd（iFLYTEK公司） 纠错

Comments Accepted by NeurIPS 2025

AcuRank: Uncertainty-Aware Adaptive Computation for Listwise Reranking

Soyoung Yoon, Gyuwan Kim, Gyu-Hwung Cho, Seung-won Hwang

Comments Accepted at NeurIPS 2025. The first two authors contributed equally. Author order is randomly determined via coin toss

Scalable Valuation of Human Feedback through Provably Robust Model Alignment

Masahiro Fujisawa, Masaki Adachi, Michael A. Osborne

机构 * The University of Osaka（大阪大学） ； Lattice Lab, Toyota Motor Corporation（丰田公司Lattice实验室） ； Machine Learning Research Group, University of Oxford（牛津大学机器学习研究组） ； RIKEN AIP（理化学研究所AIP） 纠错

Comments Accepted by the 39th Conference on Neural Information Processing Systems (NeurIPS2025), 49 pages, 7 figures

Structured Linear CDEs: Maximally Expressive and Parallel-in-Time Sequence Models

Benjamin Walker, Lingyi Yang, Nicola Muca Cirone, Cristopher Salvi, Terry Lyons

机构 * Mathematical Institute, University of Oxford（牛津大学数学院） ； Department of Mathematics, Imperial College London（伦敦帝国理工学院数学系） 纠错

Comments 32 pages, 5 figures

Journal ref Proceedings of the Thirty-Ninth Annual Conference on Neural Information Processing Systems, 2025

Mind the GAP! The Challenges of Scale in Pixel-based Deep Reinforcement Learning

Ghada Sokar, Pablo Samuel Castro

机构 * Google DeepMind（谷歌DeepMind） 纠错

Comments NeurIPS 2025

Incremental Sequence Classification with Temporal Consistency

Lucas Maystre, Gabriel Barello, Tudor Berariu, Aleix Cambray, Rares Dolga, Alvaro Ortega Gonzalez, Andrei Nica, David Barber

机构 * UiPath（UiPath公司） ； UCL（大学学院） 纠错

Comments To be presented as a spotlight paper at NeurIPS 2025

LCDB 1.1: A Database Illustrating Learning Curves Are More Ill-Behaved Than Previously Thought

Cheng Yan, Felix Mohr, Tom Viering

机构 * Delft University of Technology（代尔夫特理工大学） ； Universidad de La Sabana（拉萨芭纳大学） 纠错

Comments Accepted at NeurIPS 2025 Datasets & Benchmarks Track