ChA-MAEViT: Unifying Channel-Aware Masked Autoencoders and Multi-Channel Vision Transformers for Improved Cross-Channel Learning
机构 * Boston University（波士顿大学） ； Morgridge Institute, UW–Madison（莫里格德研究所，威斯康星大学麦迪逊分校）
Comments Accepted to NeurIPS 2025
期刊&会议
NeurIPS
Conference on Neural Information Processing Systems · 会议 · Machine Learning
The Lighthouse of Language: Enhancing LLM Agents via Critique-Guided Improvement
机构 * Fudan University（复旦大学） ； Tencent Hunyuan（腾讯文言）
Comments NeurIPS 2025
Technical Debt in In-Context Learning: Diminishing Efficiency in Long Context
机构 * Department of Industrial Engineering & Management Sciences（工业工程与管理科学系）
Comments NeurIPS 2025 Camera Ready
FocalCodec: Low-Bitrate Speech Coding via Focal Modulation Networks
机构 * Concordia University（康科德大学） ； Mila-Quebec AI Institute（魁北克AI研究院） ； Fondazione Bruno Kessler（布鲁诺·凯斯勒基金会） ； University of Trento（特伦托大学） ； Université Laval（拉瓦尔大学）
Comments Accepted at NeurIPS 2025
Dual-Flow: Transferable Multi-Target, Instance-Agnostic Attacks via In-the-wild Cascading Flow Optimization
机构 * Tsinghua University（清华大学） ； Ant Group（蚂蚁集团）
Comments Accepted at NeurIPS 2025
Efficient Adaptive Federated Optimization
机构 * Department of Statistics University of Chicago（芝加哥大学统计系） ； Department of Computer Science Columbia University（哥伦比亚大学计算机科学系） ； Google DeepMind（谷歌DeepMind） ； Department of Computer Science University of Chicago（芝加哥大学计算机科学系）
Comments Accepted to NeurIPS 2025
Equivariance by Contrast: Identifiable Equivariant Embeddings from Unlabeled Finite Group Actions
机构 * Institute of Computational Biology, Helmholtz Munich（赫尔姆霍茨慕尼黑计算生物学研究所） ； Munich Center for Machine Learning (MCML)（慕尼黑机器学习中心） ； Tübingen AI Center（图宾根人工智能中心）
Comments Accepted at NeurIPS 2025. The last two authors contributed equally. Code is available at https://github.com/dynamical-inference/ebc
On Thin Ice: Towards Explainable Conservation Monitoring via Attribution and Perturbations
机构 * Duke University（杜克大学） ； University of Agder（阿格德大学） ； University of Cambridge（剑桥大学） ； U.S. National Park Service Department of Interior（美国国家公园管理局） ； Alaska Spatial Science Fairbanks, AK（阿拉斯加空间科学研究所）
Comments NeurIPS Imageomics Workshop 2025
A Multimodal Benchmark for Framing of Oil & Gas Advertising and Potential Greenwashing Detection
机构 * Hitachi, Ltd.（日立公司） ； Stanford University（斯坦福大学） ； Centre for the Acceleration of Social Technology（社会技术加速中心） ； Princeton University（普林斯顿大学）
Comments Forthcoming in NeurIPS 2025 Datasets and Benchmarks Track
Few-Shot Knowledge Distillation of LLMs With Counterfactual Explanations
机构 * University of Maryland, College Park（马里兰大学 College Park 分校）
Comments NeurIPS 2025
Fisher meets Feynman: score-based variational inference with a product of experts
机构 * Flatiron Institute（Flatiron研究所） ； Columbia University（哥伦比亚大学）
Comments 27 pages, 11 figures. To appear in Advances in Neural Processing Information Systems (NeurIPS), 2025
HollowFlow: Efficient Sample Likelihood Evaluation using Hollow Message Passing
机构 * Department of Computer Science and Engineering, Chalmers University of Technology and University of Gothenburg（计算机科学与工程系，查尔姆斯理工大学和戈尔德堡大学）
Comments Accepted to NeurIPS 2025
FrameShield: Adversarially Robust Video Anomaly Detection
Comments 28 page, 5 figures
Journal ref 39th Conference on Neural Information Processing Systems (NeurIPS 2025)
EU-Agent-Bench: Measuring Illegal Behavior of LLM Agents Under EU Law
机构 * AI Safety Initiative Groningen（格罗宁根人工智能安全倡议）
Comments Accepted at the Workshop on Regulatable ML at the 39th Conference on Neural Information Processing Systems (NeurIPS 2025)
Brain-tuning Improves Generalizability and Efficiency of Brain Alignment in Speech Models
机构 * Max Planck Institute for Software Systems（马克斯·普朗克软件系统研究所）
Comments Published at the 39th Conference on Neural Information Processing Systems (NeurIPS 2025)
Towards a Golden Classifier-Free Guidance Path via Foresight Fixed Point Iterations
机构 * Department of Mathematics, The Hong Kong University of Science and Technology（香港科技大学数学系） ； Shenzhen-Hong Kong Collaborative Innovation Research Institute, HKUST（深圳-香港协同创新研究院）
Comments Accepted at NeurIPS 2025 (Spotlight)
MRO: Enhancing Reasoning in Diffusion Language Models via Multi-Reward Optimization
Comments Accepted by NeurIPS 2025
Finite-Time Analysis of Stochastic Nonconvex Nonsmooth Optimization on the Riemannian Manifolds
机构 * Northeastern University（东北大学） ； Tsinghua University（清华大学）
Comments To Appear in NeurIPS 2025
Risk Management for Mitigating Benchmark Failure Modes: BenchRisk
Comments 19 pages, 7 figures, to be published in the 39th Conference on Neural Information Processing Systems (NeurIPS 2025)
Multi-Task Vehicle Routing Solver via Mixture of Specialized Experts under State-Decomposable MDP
机构 * The Hong Kong University of Science and Technology（香港科学与技术大学） ； Singapore Management University（新加坡管理学院） ； The Hong Kong University of Science and Technology (Guangzhou)（香港科学与技术大学（广州）） ； Tencent AI Lab（腾讯AI实验室） ； University of Alberta（阿尔伯塔大学） ； Shanghai Jiao Tong University（上海交通大学）
Comments Accepted to NeurIPS 2025
MoniTor: Exploiting Large Language Models with Instruction for Online Video Anomaly Detection
机构 * College of Artificial Intelligence, Nanjing University of Aeronautics and Astronautics（人工智能学院，南京航空航天大学） ； Key Laboratory of Brain-Machine Intelligence Technology, Ministry of Education, China（脑机智能技术重点实验室，教育部，中国）
Comments Accepted to NeurIPS 2025. The first two authors hold equal contributions
Scalable Neural Incentive Design with Parameterized Mean-Field Approximation
Comments 52 pages, to appear at NeurIPS 2025
OpenHype: Hyperbolic Embeddings for Hierarchical Open-Vocabulary Radiance Fields
机构 * TU Wien（维也纳技术大学） ； Ulm University（乌尔姆大学） ； Fondazione Bruno Kessler（布鲁诺·凯塞研究所）
Journal ref NeurIPS 2025
AutoOpt: A Dataset and a Unified Framework for Automating Optimization Problem Solving
机构 * Brij Disa Centre for Data Science and AI（Brij Disa 数据科学与人工智能中心） ； Indian Institute of Management Ahmedabad（印度管理学院阿赫迈德亚德）
Comments NeurIPS 2025, 28 pages, 11 figures, 11 tables
Causality Meets Locality: Provably Generalizable and Scalable Policy Learning for Networked Systems
机构 * King’s College London（伦敦国王学院） ； Eindhoven University of Technology（埃因霍温理工大学） ； University of California, San Diego（加州大学圣地亚哥分校） ； The Alan Turing Institute（艾伦·图灵研究所）
Comments NeurIPS 2025 (Spotlight)
MUVR: A Multi-Modal Untrimmed Video Retrieval Benchmark with Multi-Level Visual Correspondence
机构 * MoE Key Laboratory of Brain-Machine Intelligence Technology, College of Artificial Intelligence, Nanjing University of Aeronautics and Astronautics（脑机智能技术MoE实验室，人工智能学院，南京航空航天大学） ； Nanjing University（南京大学） ； The Hong Kong Polytechnic University（香港理工大学）
Comments Accepted to NeurIPS 2025 D&B Track
Unveiling the Spatial-temporal Effective Receptive Fields of Spiking Neural Networks
Comments Acceped by 39th Conference on Neural Information Processing Systems (NeurIPS 2025)
Depth-Supervised Fusion Network for Seamless-Free Image Stitching
机构 * College of Information Science and Technology, Dalian Maritime University（大连海事大学信息科学与技术学院） ； School of Software Technology, Dalian University of Technology（大连理工大学软件学院）
Comments Accepted to Neurips 2025
Cost-Sensitive Freeze-thaw Bayesian Optimization for Efficient Hyperparameter Tuning
机构 * KAIST（韩国科学技术院） ； University of Freiburg（弗赖堡大学） ； Korea University（韩国大学）
Comments Published at NeurIPS 2025
FairImagen: Post-Processing for Bias Mitigation in Text-to-Image Models
机构 * The Chinese University of Hong Kong（香港中文大学） ； University of Oxford（牛津大学） ； University of Cambridge（剑桥大学） ； Trinity College Dublin（都柏林大学三一学院）
Comments Neurips 2025
