ComPO: Preference Alignment via Comparison Oracles
机构 * Columbia University（哥伦比亚大学） ； Stern School of Business（斯特恩商学院） ； New York University（纽约大学） ； DAMO Academy, Alibaba Group US（阿里云达摩院）
Comments Accepted to NeurIPS 2025
期刊&会议
NeurIPS
Conference on Neural Information Processing Systems · 会议 · Machine Learning
VideoHallu: Evaluating and Mitigating Multi-modal Hallucinations on Synthetic Video Understanding
机构 * University of Maryland, College Park（马里兰大学学院公园分校）
Journal ref NeurIPS 2025
GVPO: Group Variance Policy Optimization for Large Language Model Post-Training
机构 * Thrust of Artificial Intelligence, Hong Kong University of Science and Technology (Guangzhou)（香港科技大学（广州）人工智能研究所） ； Zuoyebang Education Technology（佐业邦教育科技） ； Department of Computer Science and Engineering, HKUST（香港科技大学计算机科学与工程系）
Comments Accepted by NeurIPS 2025
Scaling Laws For Scalable Oversight
机构 * MIT（麻省理工学院）
Comments 33 pages, 17 figures; The first three authors contributed equally
Journal ref NeurIPS 2025 (Spotlight)
Adversarial Locomotion and Motion Imitation for Humanoid Policy Learning
机构 * Institute of Artificial Intelligence (TeleAI), China Telecom（人工智能研究院（TeleAI），中国电信） ； ShanghaiTech University（上海科技大学） ； University of Science and Technology of China（中国科学技术大学） ； Northwestern Polytechnical University（西北工业大学） ； Tsinghua University（清华大学）
Comments NeurIPS 2025. Code: https://github.com/TeleHuman/ALMI-Open, Dataset: https://huggingface.co/datasets/TeleEmbodied/ALMI-X
Better Estimation of the Kullback--Leibler Divergence Between Language Models
机构 * ETH Zürich（苏黎世联邦理工学院）
Comments NeurIPS 2025
Segment then Splat: Unified 3D Open-Vocabulary Segmentation via Gaussian Splatting
机构 * Case Western Reserve University（凯斯西储大学） ； Westlake University（西湖大学）
Comments NeurIPS 2025. Project page: https://vulab-ai.github.io/Segment-then-Splat/
Think or Not Think: A Study of Explicit Thinking in Rule-Based Visual Reinforcement Fine-Tuning
机构 * Shanghai AI Laboratory（上海人工智能实验室） ； University of Southern California（南加州大学） ； Emory University（埃默里大学） ； Chinese University of Hong Kong（香港中文大学）
Comments Neurips 2025 Spotlight
A Frustratingly Simple Yet Highly Effective Attack Baseline: Over 90% Success Rate Against the Strong Black-box Models of GPT-4.5/4o/o1
机构 * VILA Lab, Department of Machine Learning, MBZUAI（VILA实验室，机器学习系，MBZUAI）
Comments NeurIPS 2025. Code at: https://github.com/VILA-Lab/M-Attack
Reasoning is Periodicity? Improving Large Language Models Through Effective Periodicity Modeling
机构 * School of Computer Science, Peking University（北京大学计算机学院） ； aiXcoder ； The Hong Kong University of Science and Technology（香港科学与技术大学） ； Advanced Institute of Big Data（大数据高级研究院）
Comments Accepted to NeurIPS'25
KL Penalty Control via Perturbation for Direct Preference Optimization
Comments Published as a main conference track paper at NeurIPS 2025
Provably Efficient Online RLHF with One-Pass Reward Modeling
机构 * National Key Laboratory for Novel Software Technology, Nanjing University, China（国家新型软件技术实验室，南京大学） ； School of Artificial Intelligence, Nanjing University, China（人工智能学院，南京大学）
Comments NeurIPS 2025; The first two authors contributed equally
Efficient Randomized Experiments Using Foundation Models
Comments Accepted for presentation at the Conference on Neural Information Processing Systems (NeurIPS) 2025
AttentionPredictor: Temporal Patterns Matter for KV Cache Compression
机构 * MoE Key Laboratory of Brain-inspired Intelligent Perception and Cognition, University of Science and Technology of China（脑启发智能感知与认知联合实验室，中国科学技术大学） ； Huawei Noah’s Ark Lab（华为诺亚实验室） ； College of Intelligence and Computing, Tianjin University（智能与计算学院，天津大学）
Comments NeurIPS 2025
Diffusion Generative Modeling on Lie Group Representations
机构 * Pfizer Worldwide Research and Development（辉瑞全球研发）
Comments 29 pages. Accepted as a spotlight paper at NeurIPS 2025
MIBP-Cert: Certified Training against Data Perturbations with Mixed-Integer Bilinear Programs
机构 * CISPA Helmholtz Center for Information Security（CISPA信息安全研究中心） ； Department of Computer Science University of Oxford（计算机科学系牛津大学）
Journal ref NeurIPS 2025
Last Iterate Convergence in Monotone Mean Field Games
机构 * RIKEN AIP Tokyo（日本理化学研究所AIP东京）
Comments Accepted by NeurIPS 2025, 27 pages, 2 figures, 1 table
Learning to Better Search with Language Models via Guided Reinforced Self-Training
机构 * Seoul National University（首尔国立大学） ； KRAFTON
Comments Accepted at NeurIPS 2025
Secure and Confidential Certificates of Online Fairness
Journal ref 39th Conference on Neural Information Processing Systems (NeurIPS 2025)
FAN: Fourier Analysis Networks
机构 * School of Computer Science, Peking University（北京大学计算机学院） ； The Hong Kong University of Science and Technology（香港科学与技术大学） ； ByteDance（字节跳动）
Comments Accepted to NeurIPS'25
Efficient Federated Learning against Byzantine Attacks and Data Heterogeneity via Aggregating Normalized Gradients
Comments Accepted by NeurIPS 2025
Score-based Generative Neural Networks for Large-Scale Optimal Transport
机构 * Northeastern University（东北大学） ； Brandeis University（布兰迪斯大学） ； MIT（麻省理工学院）
Comments Correction to NeurIPS 2021 Paper 4675 fixing an oversight in the statement of Assn 4.1; thanks ZY for pointing this out!
Long-Tailed Classification by Keeping the Good and Removing the Bad Momentum Causal Effect
机构 * Nanyang Technological University（南洋理工大学） ； Alibaba Group（阿里巴巴集团）
Comments This paper is accepted by NeurIPS 2020. The code is available on GitHub: https://github.com/KaihuaTang/Long-Tailed-Recognition.pytorch
I2-NeRF: Learning Neural Radiance Fields Under Physically-Grounded Media Interactions
机构 * The University of Tokyo（东京大学） ； RIKEN（日本研究机构）
Journal ref Advances in Neural Information Processing Systems, 2025
Solving Continuous Mean Field Games: Deep Reinforcement Learning for Non-Stationary Dynamics
机构 * University of Cambridge（剑桥大学） ； KTH Royal Institute of Technology（皇家理工学院） ； University of California, Los Angeles（加州大学洛杉矶分校） ； NYU Center for Data Science（纽约大学数据科学中心） ； NYU Shanghai（纽约大学上海分校）
Comments Neurips 2025
Edit Less, Achieve More: Dynamic Sparse Neuron Masking for Lifelong Knowledge Editing in LLMs
机构 * Key Lab of Intell. Info. Process., Inst. of Comput. Tech., CAS（智能信息处理重点实验室，计算技术研究所，中国科学院） ； University of Chinese Academy of Sciences（中国科学院大学） ； Agriculture Information Institute, CAAS（农业信息研究所，中国农业科学院）
Comments 19 pages, 11 figures, Accepted by NeurIPS 2025
Probing Neural Combinatorial Optimization Models
机构 * Singapore Management University（新加坡国立管理学院） ； Massachusetts Institute of Technology（麻省理工学院）
Comments 39 pages, 16 figures. Accepted as Spotlight at NeurIPS 2025
Mint: A Simple Test-Time Adaptation of Vision-Language Models against Common Corruptions
机构 * University of Illinois Urbana-Champaign（伊利诺伊大学厄巴纳-香槟分校）
Comments Accepted by NeurIPS 2025
Embracing Trustworthy Brain-Agent Collaboration as Paradigm Extension for Intelligent Assistive Technologies
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）
Comments Accepted by NeurIPS'25 Position Track
Differentially Private High-dimensional Variable Selection via Integer Programming
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）
Comments NeurIPS 2025
