Learning Parametric Distributions from Samples and Preferences
Comments 28 pages, 8 figures. To be published in the Forty-Second International Conference on Machine Learning
期刊&会议
ICML
International Conference on Machine Learning · 会议 · Machine Learning
TimePoint: Accelerated Time Series Alignment via Self-Supervised Keypoint and Descriptor Learning
Comments ICML 2025
Improved Learning via k-DTW: A Novel Dissimilarity Measure for Curves
机构 * TU Dortmund（杜伊斯堡-艾森大学） ； University of Cologne（科隆大学） ； Université Côte d’Azur（蔚蓝海岸大学） ； CNRS（法国国家科学研究中心） ； Inria（法国国家信息与自动化技术研究所） ； Aarhus University（阿arhus大学）
Comments ICML 2025
Discriminative Policy Optimization for Token-Level Reward Models
机构 * Hongzhan Chen（洪湛 Chen） ； Tao Yang（杨 Tao） ； Shiping Gao（高世平） ； Ruijun Chen（陈瑞俊） ； Xiaojun Quan（钱小军） ； Hongtao Tian（田洪涛） ； Ting Yao（姚婷）
Comments ICML 2025
LADA: Scalable Label-Specific CLIP Adapter for Continual Learning
Comments Accepted at ICML 2025
Improving the Effective Receptive Field of Message-Passing Neural Networks
Comments Accepted to ICML 2025
Provable Benefit of Random Permutations over Uniform Sampling in Stochastic Coordinate Descent
Comments Accepted to ICML 2025. 68 pages, 15 figures
Improved Last-Iterate Convergence of Shuffling Gradient Methods for Nonsmooth Convex Optimization
机构 * Zijian Liu（刘子健） ； Zhengyuan Zhou（周正元）
Comments ICML 2025
Diverse Prototypical Ensembles Improve Robustness to Subpopulation Shift
Comments ICML 2025 Paper
Hyperbolic-PDE GNN: Spectral Graph Neural Networks in the Perspective of A System of Hyperbolic Partial Differential Equations
Comments 18 pages, 2 figures, published to ICML 2025
Journal ref International Conference on Machine Learning 2025
Directed Graph Grammars for Sequence-based Learning
Comments ICML 2025
Foundation Molecular Grammar: Multi-Modal Foundation Models Induce Interpretable Molecular Graph Languages
机构 * MIT CSAIL（麻省理工学院计算机科学与人工智能实验室） ； MIT Chemistry（麻省理工学院化学系） ； MIT-IBM Watson AI Lab, IBM Research（麻省理工-IBM Watson人工智能实验室，IBM研究院） ； University of Notre Dame（诺埃伯大学）
Comments ICML 2025
Generative Social Choice: The Next Generation
机构 * Hasso Plattner Institute（哈索普拉特纳研究所） ； Harvard University（哈佛大学）
Comments Accepted to ICML 2025
CFP-Gen: Combinatorial Functional Protein Generation via Diffusion Language Models
Comments Accepted at ICML 2025. Code is available at https://github.com/yinjunbo/cfpgen
BinauralFlow: A Causal and Streamable Approach for High-Quality Binaural Speech Synthesis with Flow Matching Models
Comments ICML 2025, 18 pages
Machine Learning Models Have a Supply Chain Problem
Comments 12 pages, four figures and one table
Journal ref ICML 2025
Update Your Transformer to the Latest Release: Re-Basin of Task Vectors
Comments Accepted at ICML 2025
SlimLLM: Accurate Structured Pruning for Large Language Models
Comments ICML 2025
Pessimism Principle Can Be Effective: Towards a Framework for Zero-Shot Transfer Reinforcement Learning
Comments Accepted to ICML 2025
Improved and Oracle-Efficient Online $\ell_1$-Multicalibration
机构 * H. Milton Stewart School of Industrial and Systems Engineering / Algorithms and Randomness Center, Georgia Institute of Technology, Atlanta, USA（赫尔曼·米尔顿·斯图尔特工业与系统工程学院/算法与随机性中心，佐治亚理工学院，亚特兰大，美国） ； School of Electrical and Computer Engineering/H. Milton Stewart School of Industrial and Systems Engineering, Georgia Institute of Technology, Atlanta, USA（电气与计算机工程学院/赫尔曼·米尔顿·斯图尔特工业与系统工程学院，佐治亚理工学院，亚特兰大，美国） ； School of Computer Science, Georgia Institute of Technology, Atlanta, GA, USA（计算机科学学院，佐治亚理工学院，亚特兰大，佐治亚州，美国）
Comments Accepted to ICML 2025
How Transformers Learn Regular Language Recognition: A Theoretical Study on Training Dynamics and Implicit Bias
Comments accepted by ICML 2025
BECAME: BayEsian Continual Learning with Adaptive Model MErging
Comments Accepted by ICML 2025
Hyperspherical Normalization for Scalable Deep Reinforcement Learning
Comments 50 pages. ICML'25 (spotlight)
From Individual Experience to Collective Evidence: A Reporting-Based Framework for Identifying Systemic Harms
机构 * University of California, Berkeley（加州大学伯克利分校）
Comments Updated terminology from "incident database" to "reporting database." To be presented at ICML 2025
Behavior-Regularized Diffusion Policy Optimization for Offline Reinforcement Learning
Comments Accepted by ICML 2025
Position: Scaling LLM Agents Requires Asymptotic Analysis with LLM Primitives
机构 * Elliot Meyerson ； Xin Qiu
Comments In Proceedings of the 42nd International Conference on Machine Learning (ICML 2025); 13 pages including references
CodeSteer: Symbolic-Augmented Language Models via Code/Text Guidance
机构 * Massachusetts Institute of Technology, Boston, MA, USA（麻省理工学院） ； Harvard University, Boston, MA, USA（哈佛大学） ； MIT-IBM Watson AI Lab, Boston, MA, USA（MIT-IBM Watson AI实验室） ； University of Illinois Urbana-Champaign, Urbana, IL, USA（伊利诺伊大学厄巴纳-香槟分校）
Comments 28 pages, 12 figures
Journal ref International Conference on Machine Learning (ICML'2025)
SPRI: Aligning Large Language Models with Context-Situated Principles
Comments Forty-Second International Conference on Machine Learning (ICML 2025) Camera-Ready Version
Joint Localization and Activation Editing for Low-Resource Fine-Tuning
Comments Accepted by ICML 2025 (camera-ready version). The code is released at https://github.com/wenlai-lavine/jola
Learning from Suboptimal Data in Continuous Control via Auto-Regressive Soft Q-Network
Comments Accepted by ICML 2025
