Improving Your Model Ranking on Chatbot Arena by Vote Rigging
机构 * Sea AI Lab（Sea AI实验室） ； Hong Kong University of Science and Technology（香港科学大学） ； Pennsylvania State University（宾夕法尼亚州立大学）
Comments ICML 2025
期刊&会议
ICML
International Conference on Machine Learning · 会议 · Machine Learning
FlatQuant: Flatness Matters for LLM Quantization
机构 * Huawei Noah's Ark Lab（华为诺亚实验室） ； Shenzhen International Graduate School, Tsinghua University（清华大学深圳国际研究生院） ； The Chinese University of Hong Kong（香港中文大学）
Comments 27 pages, accepted to ICML 2025
Rejecting Hallucinated State Targets during Planning
机构 * McGill University（麦吉尔大学） ； Mila (Quebec AI Institute)（蒙特利尔AI研究所） ； Google Deepmind（谷歌DeepMind）
Comments [20250810]: ICML 2025 Camera Ready, https://github.com/mila-iqia/delusions
EEG-Language Pretraining for Highly Label-Efficient Clinical Phenotyping
机构 * Charité – Universitätsmedizin Berlin, Department of Psychiatry and Psychotherapy, Berlin, Germany（柏林查理医院医学大学精神病与心理治疗系） ； Hertie Institute for AI in Brain Health, University of Tübingen, Germany（图宾根大学健康人工智能研究所）
Comments Accepted to ICML 2025
RNA-FrameFlow: Flow Matching for de novo 3D RNA Backbone Design
机构 * Yale University（耶鲁大学） ； University of Cambridge (UK)（剑桥大学） ； Lawrence Berkeley National Laboratory（伯克利国家实验室） ； Prescient Design, Genentech, Roche（Prescient Design、基因泰克、罗氏） ； University of Oxford (UK)（牛津大学） ； National University of Singapore（新加坡国立大学）
Comments Published in Transactions on Machine Learning Research (https://openreview.net/forum?id=wOc1Yx5s09). Also presented as an Oral at Machine Learning in Computational Biology 2024, ICML 2024 Structured Probabilistic Inference & Generative Modeling Workshop, and a Spotlight at ICML 2024 AI4Science Workshop
A Square Peg in a Square Hole: Meta-Expert for Long-Tailed Semi-Supervised Learning
机构 * Key Laboratory of New Generation Artificial Intelligence Technology and Its Interdisciplinary Applications (Southeast University), Ministry of Education, Nanjing, China（新一代人工智能技术及其交叉应用重点实验室（东南大学）） ； School of Computer Science and Engineering, Southeast University, Nanjing, China（计算机科学与工程学院（东南大学））
Comments The paper is accepted by ICML 2025
On the Emergence of Position Bias in Transformers
机构 * MIT IDSS \& LIDS ； MIT CSAIL
Comments ICML 2025 camera-ready
Rethinking the Bias of Foundation Model under Long-tailed Distribution
机构 * Gaoling School of Artificial Intelligence, Renmin University of China（中国人民大学耿丽人工智能学院） ； Beijing Key Laboratory of Big Data Management and Analysis Methods（北京大数据管理与分析方法重点实验室） ； Institute of Software Chinese Academy of Sciences（中国科学院软件研究所） ； Electrical and Computer Engineering, Carnegie Mellon University（卡内基梅隆大学电气与计算机工程学院） ； University of Chinese Academy of Sciences（中国科学院大学）
Comments Published as a conference paper in ICML 2025
Position: Lifetime tuning is incompatible with continual reinforcement learning
机构 * Department of Computing Science, University of Alberta, Edmonton, Canada（阿尔伯塔大学计算机科学系） ； Alberta Machine Intelligence Institute (Amii)（阿尔伯塔机器智能研究所） ； Canada CIFAR AI Chair（加拿大 CIFAR 人工智能主席）
Comments ICML 2025, position track: https://icml.cc/virtual/2025/poster/40153
Direct Robot Configuration Space Construction using Convolutional Encoder-Decoders
机构 * Columbia University Department of Computer Science（哥伦比亚大学计算机科学系） ； Columbia University Data Science Institute（哥伦比亚大学数据科学研究院）
Comments 8 pages, 7 figures, 4 tables; Appeared at the ICML 2025 Workshop on Building Physically Plausible World Models
Quantum Algorithms for Finite-horizon Markov Decision Processes
Comments Accepted in 42nd International Conference on Machine Learning (ICML 2025)
Online Sparsification of Bipartite-Like Clusters in Graphs
机构 * School of Informatics, University of Edinburgh, Edinburgh, United Kingdom（信息学院，爱丁堡大学，爱丁堡，英国）
Comments ICML 2025
RIFLEx: A Free Lunch for Length Extrapolation in Video Diffusion Transformers
机构 * Dept. of Comp. Sci. \& Tech., BNRist Center, THU-Bosch ML Center, Tsinghua University. ； The University of Texas at Austin. ； Gaoling School of Artificial Intelligence Renmin University of China Beijing, China. ； Beijing Key Laboratory of Research on Large Models ； Engineering Research Center of Next-Generation Intelligent Search ； Pazhou Laboratory (Huangpu)
Comments ICML 2025. Project page: https://riflex-video.github.io/
RLTHF: Targeted Human Feedback for LLM Alignment
机构 * Microsoft（微软公司） ； University of California, Los Angeles（加州大学洛杉矶分校）
Comments Presented at ICML 2025
Can Transformers Learn Full Bayesian Inference in Context?
机构 * Department of Statistics, LMU Munich（慕尼黑大学统计系） ； Munich Center for Machine Learning (MCML)（慕尼黑机器学习中心） ； Center for Data Science, New York University（纽约大学数据科学中心） ； Department of Computer Science, Technical University of Munich（慕尼黑技术大学计算机科学系）
Comments Published at ICML 2025 https://openreview.net/forum?id=9Ip6fihKbc&referrer=%5BAuthor%20Console%5D(%2Fgroup%3Fid%3DICML.cc%2F2025%2FConference%2FAuthors%23your-submissions
A solvable generative model with a linear, one-step denoiser
机构 * Harvard University（哈佛大学）
Comments Published at International Conference on Machine Learning 2025
Stochastic Encodings for Active Feature Acquisition
机构 * Department of Computer Science and Technology, University of Cambridge, Cambridge, United Kingdom（计算机科学与技术系，剑桥大学，剑桥，英国） ； Department of Applied Mathematics and Theoretical Physics, University of Cambridge, Cambridge, United Kingdom（应用数学与理论物理系，剑桥大学，剑桥，英国） ； Department of Statistics, University of Oxford, Oxford, United Kingdom（统计学系，牛津大学，牛津，英国） ； Department of Artificial Intelligence, Korea University, Seoul, Korea（人工智能系，韩国大学，首尔，韩国）
Comments 31 pages, 15 figures, 17 tables, published at ICML 2025
Adversarial Cooperative Rationalization: The Risk of Spurious Correlations in Even Clean Datasets
机构 * School of Computer Science（计算机科学学院） ； Technology, Huazhong University of Science（科学技术大学技术学院） ； Faculty of Artificial Intelligence in Education, Central China Normal University（教育人工智能学院，中国中部师范大学）
Comments ICML 2025
Tool Unlearning for Tool-Augmented LLMs
机构 * University of Massachusetts Lowell（马萨诸塞大学洛厄尔分校）
Comments ICML 2025 https://clu-uml.github.io/MU-Bench-Project-Page/
Avoiding Catastrophe in Online Learning by Asking for Help
机构 * University of California, Berkeley（加州大学伯克利分校）
Comments Accepted to ICML 2025
$α$-Mutual Information: A Tunable Privacy Measure for Privacy Protection in Data Sharing
Comments 2023 22nd IEEE International Conference on Machine Learning and Applications (ICMLA)
Enhancing Certified Robustness via Block Reflector Orthogonal Layers and Logit Annealing Loss
机构 * Department of Computer Science（计算机科学系） ； Information Engineering, National Taiwan University, Taiwan（信息工程，台湾国立台湾大学）
Comments ICML 2025 Spotlight
Evaluating Deepfake Detectors in the Wild
Comments Accepted to the ICML 2025 Workshop 'DataWorld: Unifying Data Curation Frameworks Across Domains'
Avoiding Leakage Poisoning: Concept Interventions Under Distribution Shifts
机构 * University of Cambridge（剑桥大学） ； Leap Laboratories Inc.（Leap实验室公司） ； IBM Research（IBM研究院）
Comments Presented at the Forty-Second International Conference on Machine Learning (ICML 2025). Post-conference manuscript
EQ-VAE: Equivariance Regularized Latent Space for Improved Generative Image Modeling
机构 * Archimedes,Athena Reaserch Center, Greece（阿基米德、阿泰纳研究中心） ； National Technical University of Athens, Greece（雅典技术大学） ； University of Crete, Greece（克里特大学）
Comments ICML 2025
Privacy Amplification by Structured Subsampling for Deep Differentially Private Time Series Forecasting
机构 * Technical University of Munich \& Munich Data Science Institute ； Machine Learning Research, Morgan Stanley
Comments Accepted as ICML 2025 Spotlight
Enhancing Math Reasoning in Small-sized LLMs via Preview Difficulty-Aware Intervention
机构 * Independent, Researcher, China（中国独立研究者）
Comments 7 pages, 1 table, accepted by SIM ICML@2025 Workshop
Learning to Incentivize in Repeated Principal-Agent Problems with Adversarial Agent Arrivals
机构 * Paul G. Allen School of Computer Science \& Engineering, University of Washington, Seattle, USA ； Department of Electrical ； Computer Engineering, University of Washington, Seattle, USA
Comments To appear at ICML 2025
Plan-and-Act: Improving Planning of Agents for Long-Horizon Tasks
机构 * University of Tokyo（东京大学）
Journal ref ICML 2025
Predictive Data Selection: The Data That Predicts Is the Data That Teaches
机构 * Vivo AI Lab（Vivo人工智能实验室）
Comments ICML 2025
