Reliable Decision Making via Calibration Oriented Retrieval Augmented Generation
机构 * KAIST（韩国科学技术院） ； Kookmin University（韩国釜山大学）
Comments Accepted by NeurIPS 2025
期刊&会议
NeurIPS
Conference on Neural Information Processing Systems · 会议 · Machine Learning
Can DPO Learn Diverse Human Values? A Theoretical Scaling Law
机构 * Department of Computer Sciences University of Wisconsin-Madison（计算机科学系威斯康星大学麦迪逊分校）
Comments NeurIPS 2025
When to Trust Your Simulator: Dynamics-Aware Hybrid Offline-and-Online Reinforcement Learning
机构 * Tsinghua University（清华大学） ； Indian Institute of Technology（印度理工学院） ； Shanghai Jiaotong University（上海交通大学） ； Beijing National Research Center for Information Science and Technology（北京信息科学与技术国家研究中心） ； Shanghai AI Laboratory（上海人工智能实验室）
Comments NeurIPS 2022 Spotlight
Convergence and sample complexity of natural policy gradient primal-dual methods for constrained MDPs
机构 * University of Tennessee, Knoxville（田纳西大学） ； University of Maryland, College Park（马里兰大学） ； Apple Inc.（苹果公司） ； University of Illinois Urbana-Champaign（伊利诺伊大学厄巴纳-香槟分校） ； University of Southern California（南加州大学）
Comments 76 pages, 4 figures, 2 tables; Journal version of the NeurIPS 2020 paper; Accepted to JMLR
Improving Decision Trees through the Lens of Parameterized Local Search
Comments Accepted at NeurIPS 2025
Structure-Aware Spectral Sparsification via Uniform Edge Sampling
机构 * Department of Computer Science（计算机科学系） ； Purdue University（普渡大学）
Comments 19 pages, 4 figures, NeurIPS 2025
One Prompt Fits All: Universal Graph Adaptation for Pretrained Models
机构 * College of Intelligence and Computing, Tianjin University（天津大学智能与计算学院） ； School of Economics and Management, Beijing University of Posts and Telecommunications（北京邮电大学经济管理学院） ； Department of Data Science, George Washington University（乔治华盛顿大学数据科学系）
Comments accepted by NeurIPS 2025 main conference
OmniDraft: A Cross-vocabulary, Online Adaptive Drafter for On-device Speculative Decoding
机构 * Qualcomm AI Research（高通人工智能研究）
Comments Accepted to NeurIPS 2025
PubSub-VFL: Towards Efficient Two-Party Split Learning in Heterogeneous Environments via Publisher/Subscriber Architecture
机构 * Department of Computer Science, City University of Hong Kong（香港城市大学计算机科学系） ； ByteDance Inc.（字节跳动公司）
Comments Accepted at NeurIPS 2025
CrossAD: Time Series Anomaly Detection with Cross-scale Associations and Cross-window Modeling
机构 * East China Normal University（东华大学） ； Shandong Inspur Database Technology Co., Ltd.（山东 Inspur 数据库技术有限公司）
Comments Accepted by the thirty-ninth annual conference on Neural Information Processing Systems
VideoLucy: Deep Memory Backtracking for Long Video Understanding
机构 * National Key Laboratory of Multispectral Information Intelligent Processing Technology, School of Artificial Intelligence and Automation, Huazhong University of Science and Technology（多谱段信息智能处理国家级重点实验室，人工智能与自动化学院，华中科技大学） ； NUS（国立大学新加坡） ； S-Lab, NTU（NTU的S实验室） ； Shanghai AI Lab（上海人工智能实验室）
Comments NeurIPS-2025 Accepted Paper
Shallow Robustness, Deep Vulnerabilities: Multi-Turn Evaluation of Medical LLMs
机构 * Dynamo AI
Comments Dataset and code: https://huggingface.co/datasets/dynamoai-ml/MedQA-USMLE-4-MultiTurnRobust ; https://github.com/bmanczak/MedQA-MultiTurnRobustness Accepted as a poster at NeurIPS 2025 Workshop on GenAI for Health: Potential, Trust, and Policy Compliance
DSAS: A Universal Plug-and-Play Framework for Attention Optimization in Multi-Document Question Answering
机构 * Institute of Intelligent Computing, University of Electronic Science and Technology of China（电子科技大学智能计算研究院） ； School of Information and Software Engineering, University of Electronic Science and Technology of China（电子科技大学信息与软件工程学院）
Comments 27 pages, has been accepted by NeurIPS 2025
Hierarchical Koopman Diffusion: Fast Generation with Interpretable Diffusion Trajectory
机构 * Fudan University（复旦大学） ； The University of Hong Kong（香港大学）
Comments NeurIPS 2025
ImageSentinel: Protecting Visual Datasets from Unauthorized Retrieval-Augmented Image Generation
机构 * Department of Computer Science, Hong Kong Baptist University（香港 Baptist 大学计算机科学系） ； NVIDIA AI Technology Center, NVIDIA（NVIDIA 人工智能技术中心）
Comments Accepted at NeurIPS 2025
APCE: Adaptive Progressive Context Expansion for Long Context Processing
机构 * LG Electronics USA（LG电子美国公司）
Comments NeurIPS 2025 Workshop: ML For Systems
Statistical Guarantees for High-Dimensional Stochastic Gradient Descent
机构 * Department of Statistics University of Chicago（统计学系芝加哥大学） ； Department of Mathematics University of California, San Diego（数学系加州大学圣地亚哥分校） ； Department of Applied Mathematics University of Twente（应用数学系埃因霍温大学）
Comments Accepted to NeurIPS 2025
Semi-infinite Nonconvex Constrained Min-Max Optimization
Comments Accepted at NeurIPS 2025
CGBench: Benchmarking Language Model Scientific Reasoning for Clinical Genetics Research
机构 * Stanford University（斯坦福大学）
Comments Accepted at NeurIPS 2025
Efficient Restarts in Non-Stationary Model-Free Reinforcement Learning
机构 * Soka University of America（美国早稻田大学） ； University of Maryland, College Park（马里兰大学学院市分校）
Comments This paper contains 19 pages and 3 figures. To be presented at the 2nd Workshop on Aligning Reinforcement Learning Experimentalists and Theorists (ARLET 2025) at NeurIPS 2025
Robust Adversarial Reinforcement Learning in Stochastic Games via Sequence Modeling
机构 * University College London（伦敦大学学院）
Comments Accepted by Reliable ML Workshop @ NeurIPS 2025
Balancing Synthetic Data and Replay for Enhancing Task-Specific Capabilities
机构 * University of Freiburg（弗赖堡大学） ； ELLIS Institute Tübingen（图宾根ELLIS研究所） ； Open-Sci Collective（开放科学集体） ； LAION ； Prior Labs（Prior实验室）
Comments Presented at 39th Conference on Neural Information Processing Systems (NeurIPS 2025) Workshop on Continual and Compatible Foundation Model Updates (CCFM)
Zero-Shot Large Language Model Agents for Fully Automated Radiotherapy Treatment Planning
机构 * Department of Radiation Oncology, Duke University（杜克大学放射肿瘤科） ； Department of Radiation Oncology, University of California Irvine（加州大学尔湾分校放射肿瘤科）
Comments Accepted for poster presentation at the NeurIPS 2025 Workshop on GenAI for Health: Potential, Trust, and Policy Compliance
AGENTIQL: An Agent-Inspired Multi-Expert Framework for Text-to-SQL Generation
机构 * Concordia University（康科德大学）
Comments Accepted at NeurIPS 2025, ER "Efficient Reasoning" workshop
Adaptive Dual Reasoner: Large Reasoning Models Can Think Efficiently by Hybrid Reasoning
机构 * Tencent Youtu Lab（腾讯优图实验室）
Comments Accepted to NeurIPS 2025 Workshop on Efficient Reasoning
Scalable Policy-Based RL Algorithms for POMDPs
机构 * UIUC（伊利诺伊大学香槟分校）
Comments 36 pages, 3 Figures, Accepted at NeurIPS 2025
Limited Preference Data? Learning Better Reward Model with Latent Space Synthesis
机构 * Department of Computer Sciences, University of Wisconsin-Madison（威斯康星大学麦迪逊分校计算机科学系） ； College of Computing and Data Science, Nanyang Technological University（南洋理工大学计算机科学与数据科学学院）
Comments Accepted by NeurIPS 2025
Unspoken Hints: Accuracy Without Acknowledgement in LLM Reasoning
Comments 5 Pages, 4 Figures, 4 Tables
Journal ref 39th Conference on Neural Information Processing Systems, 2025, Workshop: Reliable ML from Unreliable Data
Clean First, Align Later: Benchmarking Preference Data Cleaning for Reliable LLM Alignment
Comments NeurIPS 2025
SCAN: Self-Denoising Monte Carlo Annotation for Robust Process Reward Learning
机构 * Soochow University（苏州大学） ； Tencent（腾讯）
Comments NeurIPS 2025. Project page: https://scan-prm.github.io/
