Unleashing Uncertainty: Efficient Machine Unlearning for Generative AI
机构 * ITI, Centre for Research \& Technology Hellas, Greece ； University of Amsterdam, Netherlands
Comments ICML 2025 workshop on Machine Unlearning for Generative AI
期刊&会议
ICML
International Conference on Machine Learning · 会议 · Machine Learning
MERIT: Maximum-normalized Element-wise Ratio for Language Model Large-batch Training
机构 * School of Computing, National University of Singapore, Singapore（新加坡国立大学计算机学院）
Comments ICML 2025
Diagonal Symmetrization of Neural Network Solvers for the Many-Electron Schrödinger Equation
机构 * Gatsby Unit, University College London, London, UK（伦敦大学学院Gatsby单元） ； Department of Computer Science, Princeton University, Princeton, NJ, USA（普林斯顿大学计算机科学系） ； Department of Mechanical Science and Engineering, University of Illinois at Urbana-Champaign, Champaign, IL, USA（伊利诺伊大学厄巴纳-香槟分校机械科学与工程系）
Comments ICML 2025 camera-ready version
Random Feature Representation Boosting
机构 * Department of Mathematics, Imperial College London, UK（帝国理工学院伦敦校区数学系） ； School of Computer Science, University of St. Gallen, Switzerland（圣加尔登大学计算机科学学院）
Comments To appear in ICML 2025
Tracking World States with Language Models: State-Based Evaluation Using Chess
机构 * University of Tokyo（东京大学）
Comments Spotlight presentation at ICML 2025 Workshop on Assessing World Models
Learning Game-Playing Agents with Generative Code Optimization
机构 * stanford（斯坦福大学）
Comments ICML 2025 Workshop on Programmatic Representations for Agent Learning, Vancouver, Canada
Putnam-AXIOM: A Functional and Static Benchmark for Measuring Higher Level Mathematical Reasoning in LLMs
机构 * Department of Computer Science, Stanford University, Stanford, USA（计算机科学系，斯坦福大学，斯坦福，美国）
Comments 27 pages total (10-page main paper + 17-page appendix), 12 figures, 6 tables. Submitted to ICML 2025 (under review)
Journal ref ICML 2025
Energy-Based Flow Matching for Generating 3D Molecular Structure
机构 * KTH Royal Institute of Technology, Stockholm, Sweden（皇家理工学院） ； Science for Life Laboratory, Sweden（生命科学实验室） ； DBB at Stockholm University, Sweden（斯德哥尔摩大学DBB实验室） ； Max Planck Institute for Polymer Research, Mainz, Germany（马克斯·普朗克聚合物研究所） ； The Alan Turing Institute, London, United Kingdom（艾伦·图灵研究所）
Comments Accepted to the International Conference on Machine Learning (2025)
Flexibility-Conditioned Protein Structure Design with Flow Matching
机构 * Heidelberg Institute for Theoretical Studies, Heidelberg, Germany（海德堡理论研究所） ； IWR, Heidelberg University, Heidelberg, Germany（海德堡大学生物医学研究所） ； Max Planck Institute for Polymer Research, Mainz, Germany（马克斯·普朗克聚合物研究所） ； Biophysics at Stockholm University（斯德哥尔摩大学生物物理学系） ； Science for Life Laboratory, Stockholm, Sweden（生命科学实验室） ； IAR, Karlsruhe Institute of Technology, Karlsruhe, Germany（卡尔斯鲁厄理工学院）
Comments ICML 2025
BiMark: Unbiased Multilayer Watermarking for Large Language Models
机构 * Griffith University, Brisbane（格里菲斯大学） ； RMIT University, Melbourne（皇家墨尔本理工大学） ； University of Technology Sydney, Sydney（悉尼大学）
Comments This paper is accepted by International Conference on Machine Learning (ICML) 2025
Constrained Diffusion Models for Synthesizing Representative Power Flow Datasets
机构 * Department of Electrical Engineering and Computer Science, University of Michigan（电气工程与计算机科学系，密歇根大学）
Comments This paper is the extended journal version of our paper at ICML 2025 Workshop "DataWorld: Unifying Data Curation Frameworks Across Domains"
Adversarial Robustness in Two-Stage Learning-to-Defer: Algorithms and Guarantees
机构 * School of Computing, National University of Singapore, Singapore（新加坡国立大学计算机学院） ； IRIT, Université de Toulouse, CNRS, Toulouse INP, Toulouse, France（图卢兹大学IRIT研究所、法国国家科学研究中心、图卢兹INP） ； Institute for Infocomm Research, Agency for Science, Technology and Research, Singapore（新加坡科技研究局信息与通信研究所）
Comments Accepted at the 42nd International Conference on Machine Learning (ICML 2025)
Visual Generation Without Guidance
机构 * Department of Computer Science \& Technology, Tsinghua University
Comments Accepted to ICML 2025
GPU Kernel Scientist: An LLM-Driven Framework for Iterative Kernel Optimization
机构 * LLM-Driven Framework for Iterative Kernel Optimization（LLM驱动的迭代核优化框架）
Comments 4+1 page paper plus Appendices and Supplementary zip file. Presented at the ES-FoMo "Efficient Systems for Foundation Models" workshop at ICML 2025
Mirror, Mirror of the Flow: How Does Regularization Shape Implicit Bias?
机构 * CISPA Helmholtz Center for Information Security（信息安全霍普夫研究所）
Comments 26 pages, 16 figures
Journal ref Forty-second International Conference on Machine Learning, 2025
One Example Shown, Many Concepts Known! Counterexample-Driven Conceptual Reasoning in Mathematical LLMs
机构 * Tsinghua University（清华大学） ； Sun-Yat Sen University（孙中山大学） ； School of Mathematical Science, Fudan University（复旦大学数学学院） ； Bytedance Inc.（字节跳动公司） ； INFLY TECH (Shanghai) Co., Ltd.（INFLY TECH（上海）有限公司） ； ARC Lab, Arizona State University（ARC实验室，亚利桑那州立大学） ； University of Illinois Chicago（伊利诺伊大学芝加哥分校） ； Peng Cheng Laboratory（鹏城实验室） ； Guangdong Provincial Key Laboratory of Fire Science（广东省消防科学与智能应急技术重点实验室）
Comments ICML 2025
A Diffusion Model Framework for Unsupervised Neural Combinatorial Optimization
机构 * Institute for Machine Learning, Johannes Kepler University, Linz, Austria（机器学习研究所，约翰尼斯·凯普大学，林茨，奥地利） ； NXAI GmbH（NXAI公司）
Comments Accepted at ICML 2024
Journal ref Proceedings of the 41st International Conference on Machine Learning, PMLR 235:43346-43367, 2024
Don't Think Twice! Over-Reasoning Impairs Confidence Calibration
机构 * Stanford University（斯坦福大学）
Comments Published at ICML 2025 Workshop on Reliable and Responsible Foundation Models
Generation of structure-guided pMHC-I libraries using Diffusion Models
机构 * Center for Computational Biology, University of California, Berkeley, Berkeley, California, USA（计算生物学中心，加州大学伯克利分校） ； Department of Electrical Engineering and Computer Sciences, University of California, Berkeley, Berkeley, California, USA（电气工程与计算机科学系，加州大学伯克利分校）
Comments Accepted to the The 2nd Workshop on Generative AI and Biology ICML Workshop 2025
Rethinking Addressing in Language Models via Contexualized Equivariant Positional Encoding
机构 * University of Texas at Austin（德克萨斯大学奥斯汀分校） ； Zhejiang University（浙江大学） ； Princeton University（普林斯顿大学）
Comments ICML 2025
Democratic AI is Possible. The Democracy Levels Framework Shows How It Might Work
Comments 31 pages. Accepted to the position paper track at ICML 2025. A previous version was presented at the Pluralistic Alignment Workshop at NeurIPS 2024. For ongoing work, see: https://democracylevels.org
Towards Understanding Gradient Dynamics of the Sliced-Wasserstein Distance via Critical Point Analysis
机构 * Laboratoire de Mathématiques d'Orsay, Université Paris-Saclay, Gif-sur-Yvette, France（巴黎-萨克雷大学奥赛数学实验室） ； Centre de recherche en économie et statistique, ENSAE, Palaiseau, France（经济与统计研究中心）
Comments 32p
Journal ref Forty-second International Conference on Machine Learning, 2025
No Metric to Rule Them All: Toward Principled Evaluations of Graph-Learning Datasets
机构 * University of Fribourg, Switzerland（弗里堡大学） ； Aalto University, Finland（阿尔托大学） ； Max Planck Institute for Informatics, Germany（马克斯·普朗克信息研究所）
Comments Accepted at ICML 2025
Towards the Use of Saliency Maps for Explaining Low-Quality Electrocardiograms to End Users
机构 * University of Amsterdam, Netherlands（阿姆斯特丹大学） ； Google, USA（谷歌） ； Microsoft Research, Canada（微软研究院） ； University of Cambridge, United Kingdom（剑桥大学）
Comments Accepted to ICML 2022 Workshop on Interpretable ML in Healthcare
Action-Constrained Imitation Learning
机构 * National Yang Ming Chiao Tung University, Hsinchu, Taiwan（National Yang Ming Chiao Tung University） ； University of Illinois at Urbana-Champaign, Illinois, United States（University of Illinois at Urbana-Champaign） ； National Taiwan University, Taipei, Taiwan（National Taiwan University）
Comments Published in ICML 2025
SketchDNN: Joint Continuous-Discrete Diffusion for CAD Sketch Generation
机构 * Department of Computer Science, Miami-Oxford University, Oxford OH, USA（计算机科学系，迈阿密-牛津大学，牛津 OH，美国）
Comments 17 pages, 63 figures, Proceedings of the 42nd International Conference on Machine Learning (ICML2025)
Efficient Network Automatic Relevance Determination
机构 * School of Data Science, Fudan University, Shanghai, China（复旦大学数据科学学院） ； Incubation Institute, Fudan University, Shanghai, China（复旦大学孵化院） ； Shanghai Academy of Artificial Intelligence for Science, Shanghai, China（上海人工智能科学研究院） ； Eberly College of Science, Pennsylvania State University, PA, United State（宾夕法尼亚州立大学科学学院） ； Department of Applied Mathematics and Theoretical Physics, University of Cambridge, Cambridge, United Kingdom（剑桥大学应用数学与理论物理系） ； Zhongshan Hospital, Fudan University, Shanghai, China（复旦大学中山医院）
Comments ICML 2025
Position: We Need Responsible, Application-Driven (RAD) AI Research
机构 * Tech \& Policy Lab, Law School, University of Western Australia, Australia ； College of Systems ； Society, Australian National University, Australia ； Centre for Business Research, University of Cambridge, United Kingdom
Comments 12 pages, 1 figure, Final Camera Ready version for proceedings with updated figure color and reference, Accepted to Proceedings of the 41 st International Conference on Machine Learning, Vancouver, Canada. PMLR 267, 2025
POPri: Private Federated Learning using Preference-Optimized Synthetic Data
机构 * Pittsburgh Supercomputing Center, Pittsburgh, USA（匹兹堡超级计算中心） ； Department of ECE, Carnegie Mellon University, Pittsburgh, PA（电子工程系，卡内基梅隆大学）
Comments ICML 2025 camera-ready
Recommendations with Sparse Comparison Data: Provably Fast Convergence for Nonconvex Matrix Factorization
机构 * EPFL（瑞士联邦理工学院） ； TUM（慕尼黑工业大学）
Comments Accepted for publication in ICML 2025
