NeurIPS

Ultrametric Cluster Hierarchies: I Want 'em All!

Andrew Draganov, Pascal Weber, Rasmus Skibdahl Melanchton Jørgensen, Anna Beer, Claudia Plant, Ira Assent

Comments Published in the 39th Conference on Neural Information Processing Systems (NeurIPS 2025)

The Path Not Taken: RLVR Provably Learns Off the Principals

Hanqing Zhu, Zhenyu Zhang, Hanxian Huang, DiJia Su, Zechun Liu, Jiawei Zhao, Igor Fedorov, Hamed Pirsiavash, Zhizhou Sha, Jinwon Lee, David Z. Pan, Zhangyang Wang, Yuandong Tian, Kai Sheng Tai

机构 * Meta AI ； The University of Texas at Austin（德克萨斯大学奥斯汀分校） 纠错

Comments Preliminary version accepted as a spotlight in NeurIPS 2025 Workshop on Efficient Reasoning

CleverBirds: A Multiple-Choice Benchmark for Fine-grained Human Knowledge Tracing

Leonie Bossemeyer, Samuel Heinrich, Grant Van Horn, Oisin Mac Aodha

机构 * University of Edinburgh（爱丁堡大学） ； Cornell University（康奈尔大学） ； UMass Amherst（马萨诸塞大学阿姆赫斯特分校） 纠错

Comments To appear at NeurIPS 2025 - Datasets and Benchmarks Track

Galactification: painting galaxies onto dark matter only simulations using a transformer-based model

Shivam Pandey, Christopher C. Lovell, Chirag Modi, Benjamin D. Wandelt

机构 * Department of Physics and Astronomy, Johns Hopkins University（约翰霍普金斯大学物理与天文学系） ； Kavli Institute for Cosmology, University of Cambridge（剑桥大学卡弗利宇宙研究所） ； Center for Cosmology and Particle Physics, New York University（纽约大学宇宙与粒子物理中心） 纠错

Comments 8 pages, 4 figures. , accepted at Machine Learning and the Physical Sciences Workshop at NeurIPS 2025

RAPTR: Radar-based 3D Pose Estimation using Transformer

Sorachi Kato, Ryoma Yataka, Pu Perry Wang, Pedro Miraldo, Takuya Fujihashi, Petros Boufounos

机构 * Mitsubishi Electric Research Laboratories (MERL)（三菱电机研究实验室） ； The University of Osaka（大阪大学） ； Information Technology R&D Center (ITC)（信息科技研发中心） 纠错

Comments 26 pages, Accepted to NeurIPS 2025

DiagramIR: An Automatic Pipeline for Educational Math Diagram Evaluation

Vishal Kumar, Shubhra Mishra, Rebecca Hao, Rizwaan Malik, David Broman, Dorottya Demszky

机构 * Stanford University（斯坦福大学） ； KTH Royal Institute of Technology（皇家理工学院） 纠错

Comments Published at the Math-AI Workshop at NeurIPS 2025

Prioritizing Perception-Guided Self-Supervision: A New Paradigm for Causal Modeling in End-to-End Autonomous Driving

Yi Huang, Zhan Qu, Lihui Jiang, Bingbing Liu, Hongbo Zhang

机构 * The Chinese University of Hong Kong, Shenzhen（香港中文大学（深圳）） ； Huawei Noah’s Ark Lab（华为诺亚实验室） 纠错

Comments Accepted at NeurIPS 2025

UniFormer: Unified and Efficient Transformer for Reasoning Across General and Custom Computing

Zhuoheng Ran, Chong Wu, Renjie Xu, Maolin Che, Hong Yan

Comments Accepted on 24 September 2025 at NeurIPS 2025 Efficient Reasoning Workshop

DynaAct: Large Language Model Reasoning with Dynamic Action Spaces

Xueliang Zhao, Wei Wu, Jian Guan, Qintong Li, Lingpeng Kong

机构 * The University of Hong Kong（香港大学） ； Ant Group（蚂蚁集团） 纠错

Comments Accepted to NeurIPS 2025

Generalizable Insights for Graph Transformers in Theory and Practice

Timo Stoll, Luis Müller, Christopher Morris

Comments Accepted at NeurIPS 2025 as spotlight

The One Where They Brain-Tune for Social Cognition: Multi-Modal Brain-Tuning on Friends

Nico Policzer, Cameron Braunstein, Mariya Toneva

机构 * Saarland University（萨尔兰大学） ； MPI for Software Systems（软件系统Max Planck研究所） ； University of British Columbia（不列颠哥伦比亚大学） 纠错

Comments 20 pages, 7 figures. Appearing at the NeurIPS 2025 Workshop on Interpreting Cognition in Deep Learning Models

Diffusion Guided Adversarial State Perturbations in Reinforcement Learning

Xiaolin Sun, Feidi Liu, Zhengming Ding, ZiZhan Zheng

机构 * Department of Computer Science, Tulane University（Tulane大学计算机科学系） ； Shanghai Center for Mathematical Science, Fudan University（复旦大学上海数学科学中心） 纠错

Journal ref NeurIPS 2025 Poster

DRAGON: Guard LLM Unlearning in Context via Negative Detection and Reasoning

Yaxuan Wang, Chris Yuhao Liu, Quan Liu, Jinglong Pang, Wei Wei, Yujia Bao, Yang Liu

机构 * University of California, Santa Cruz（加州大学圣克ruz分校） ； Center for Advanced AI, Accenture（Accenture高级人工智能中心） 纠错

Comments Please refer to the NeurIPS 2025 submission: https://openreview.net/forum?id=FNuul0hlin The paper has been accepted to the ICML 2025 MUGen Workshop: https://openreview.net/forum?id=ET24oKP23c

Reliable Parameter Inference for the Epoch of Reionization using Balanced Neural Ratio Estimation

Diego González-Hernández, Molly Wolfson, Joseph F. Hennawi

Comments Machine Learning and the Physical Sciences Workshop, NeurIPS 2025

CryptoMoE: Privacy-Preserving and Scalable Mixture of Experts Inference via Balanced Expert Routing

Yifan Zhou, Tianshi Xu, Jue Hong, Ye Wu, Meng Li

Comments NeurIPS 2025

Normalization in Attention Dynamics

Nikita Karagodin, Shu Ge, Yury Polyanskiy, Philippe Rigollet

机构 * Department of EECS, MIT（电子工程与计算机科学系，麻省理工学院） ； Department of Mathematics, MIT（数学系，麻省理工学院） 纠错

Comments 39th Conference on Neural Information Processing Systems (NeurIPS 2025), 23 pages

Disentangled Representation Learning via Modular Compositional Bias

Whie Jung, Dong Hoon Lee, Seunghoon Hong

机构 * KAIST（韩国科学技术院） 纠错

Journal ref Advances in Neural Information Processing Systems (NeurIPS), 2025

S$^2$M-Former: Spiking Symmetric Mixing Branchformer for Brain Auditory Attention Detection

Jiaqi Wang, Zhengyu Ma, Xiongri Shen, Chenlin Zhou, Leilei Zhao, Han Zhang, Yi Zhong, Siqi Cai, Zhenxi Song, Zhiguo Zhang

机构 * Harbin Institute of Technology, Shenzhen（哈尔滨工业大学（深圳）） ； Pengcheng Laboratory（鹏城实验室） ； Peking University（北京大学） ； Harbin Institute of Technology（哈尔滨工业大学） ； Great Bay University（大湾区大学） 纠错

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

ElasticMM: Efficient Multimodal LLMs Serving with Elastic Multimodal Parallelism

Zedong Liu, Shenggan Cheng, Guangming Tan, Yang You, Dingwen Tao

机构 * Institute of Computing Technology, Chinese Academy of Sciences（中国科学院计算技术研究所） ； University of Electronic Science and Technology of China（电子科技大学） ； National University of Singapore（新加坡国立大学） 纠错

Comments Accepted at NeurIPS 2025 Oral (Thirty-Ninth Conference on Neural Information Processing Systems)

Generalizing Verifiable Instruction Following

Valentina Pyatkin, Saumya Malik, Victoria Graf, Hamish Ivison, Shengyi Huang, Pradeep Dasigi, Nathan Lambert, Hannaneh Hajishirzi

机构 * Allen Institute for Artificial Intelligence（艾伦人工智能研究所） ； University of Washington（华盛顿大学） 纠错

Comments 11 pages, accepted to NeurIPS 2025, Datasets & Benchmarks

Rethinking Losses for Diffusion Bridge Samplers

Sebastian Sanokowski, Lukas Gruber, Christoph Bartmann, Sepp Hochreiter, Sebastian Lehner

机构 * Technical University Munich, Chair of Robotics, Artificial Intelligence and Embedded Systems（慕尼黑技术大学机器人、人工智能与嵌入式系统教研室） ； ELLIS Unit Linz, LIT AI Lab, Johannes Kepler University Linz, Austria（林茨ELLIS单位，LIT人工智能实验室，林茨约翰内斯·开普勒大学，奥地利） ； NXAI Lab & NXAI GmbH, Linz, Austria（NXAI实验室及NXAI GmbH，林茨，奥地利） 纠错

Comments Accepted at NeurIPS 2025 as a Conference Paper

Zeroth-Order Optimization Finds Flat Minima

Liang Zhang, Bingcong Li, Kiran Koshy Thekumparampil, Sewoong Oh, Michael Muehlebach, Niao He

机构 * Department of Computer Science, ETH Zurich（苏黎世联邦理工学院计算机科学系） ； Max Planck Institute for Intelligent Systems（智能系统马克斯·普朗克研究所） ； Amazon（亚马逊公司） ； Paul G. Allen School of Computer Science and Engineering, University of Washington（华盛顿大学保罗·G·阿伦计算机科学与工程学校） 纠错

Comments NeurIPS 2025

UniSite: The First Cross-Structure Dataset and Learning Framework for End-to-End Ligand Binding Site Detection

Jigang Fan, Quanlin Wu, Shengjie Luo, Liwei Wang

机构 * Center for Data Science, Peking University（数据科学中心，北京大学） ； State Key Laboratory of General Artificial Intelligence, Peking University（通用人工智能国家重点实验室，北京大学） ； Center for Machine Learning Research, Peking University（机器学习研究中心，北京大学） 纠错

Comments Accepted by NeurIPS 2025 as a Spotlight paper

Journal ref NeurIPS 2025 (Spotlight)

FutureSightDrive: Thinking Visually with Spatio-Temporal CoT for Autonomous Driving

Shuang Zeng, Xinyuan Chang, Mengwei Xie, Xinran Liu, Yifan Bai, Zheng Pan, Mu Xu, Xing Wei, Ning Guo

机构 * Xi’an Jiaotong University（西安交通大学） ； Amap, Alibaba Group（阿里巴巴集团） ； DAMO Academy, Alibaba Group（阿里巴巴达摩院） 纠错

Comments Accepted to NeurIPS 2025 as Spotlight Presentation. Code: https://github.com/MIV-XJTU/FSDrive

UniEdit: A Unified Knowledge Editing Benchmark for Large Language Models

Qizhou Chen, Dakan Wang, Taolin Zhang, Zaoming Yan, Chengsong You, Chengyu Wang, Xiaofeng He

Comments NeurIPS 2025 Track on Datasets and Benchmarks. Dataset Download: https://huggingface.co/datasets/qizhou/UniEdit Code Download: https://github.com/qizhou000/UniEdit

Continuous Subspace Optimization for Continual Learning

Quan Cheng, Yuanyu Wan, Lingyu Wu, Chenping Hou, Lijun Zhang

机构 * National Key Laboratory for Novel Software Technology, Nanjing University（新型软件技术国家实验室，南京大学） ； School of Artificial Intelligence, Nanjing University（人工智能学院，南京大学） ； School of Software Technology, Zhejiang University（软件学院，浙江大学） ； Hangzhou High-Tech Zone (Binjiang) Institute of Blockchain and Data Security, Hangzhou, China（杭州高新技术区（滨江）区块链与数据安全研究院，杭州，中国） ； College of Science, National University of Defense Technology（科学学院，国防科技大学） 纠错

Comments Accepted to NeurIPS 2025

CATransformers: Carbon Aware Transformers Through Joint Model-Hardware Optimization

Irene Wang, Newsha Ardalani, Mostafa Elhoushi, Daniel Jiang, Samuel Hsia, Ekin Sumbul, Divya Mahajan, Carole-Jean Wu, Bilge Acun

机构 * Georgia Institute of Technology（佐治亚理工学院） ； FAIR at Meta（Meta的FAIR部门） ； Reality Labs at Meta（Meta的Reality Labs） ； Meta 纠错

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

Scalable Signature Kernel Computations for Long Time Series via Local Neumann Series Expansions

Matthew Tamayo-Rios, Alexander Schell, Rima Alaifari

机构 * Seminar for Applied Mathematics & AI Center ETH Zurich（应用数学与人工智能中心，苏黎世联邦理工学院） ； Department of Mathematics Technical University of Munich（数学系，慕尼黑技术大学） ； Department of Mathematics RWTH Aachen University（数学系，亚琛工业大学） 纠错

Comments NeurIPS 2025 (to appear), 32 pages, 9 figures

Strategizing against No-regret Learners

Yuan Deng, Jon Schneider, Balusubramanian Sivan

机构 * Duke University（杜克大学） ； Google Research（谷歌研究） 纠错

Comments 11 pages. NeurIPS 2019

LLMs vs. Traditional Sentiment Tools in Psychology: An Evaluation on Belgian-Dutch Narratives

Ratna Kandala, Katie Hoemann

Comments NeurIPS 2025 Workshop on Evaluating the Evolving LLM Lifecycle: Benchmarks, Emergent Abilities, and Scaling