ICML

On Zero-Initialized Attention: Optimal Prompt and Gating Factor Estimation

Nghiem T. Diep, Huy Nguyen, Chau Nguyen, Minh Le, Duy M. H. Nguyen, Daniel Sonntag, Mathias Niepert, Nhat Ho

机构 * University of Science, VNU-HCM, Ho Chi Minh City, Vietnam（越南胡志明市VNU-HCM大学） ； Viet Nam National University, Ho Chi Minh City, Vietnam（越南国家大学） ； German Research Center for Artificial Intelligence (DFKI)（德国人工智能研究中心） ； Qualcomm AI Research, an initiative of Qualcomm Technologies, Inc.（高通人工智能研究） ； The University of Texas at Austin（德克萨斯大学奥斯汀分校） ； University of Stuttgart（斯图加特大学） ； Max Planck Research School for Intelligent Systems (IMPRS-IS)（马克斯·普朗克智能系统研究学校） ； Oldenburg University（奥尔登堡大学） 纠错

Comments Accepted at ICML 2025

REVOLVE: Optimizing AI Systems by Tracking Response Evolution in Textual Optimization

Peiyan Zhang, Haibo Jin, Leyang Hu, Xinnuo Li, Liying Kang, Man Luo, Yangqiu Song, Haohan Wang

Comments 20 pages, 2 figures, accepted by ICML 2025

Interchangeable Token Embeddings for Extendable Vocabulary and Alpha-Equivalence

İlker Işık, Ramazan Gokberk Cinbis, Ebru Aydin Gol

机构 * Department of Computer Engineering, Middle East Technical University, Ankara, Turkey（中欧技术大学计算机工程系） ； Microsoft, İstanbul, Turkey（微软公司） 纠错

Comments ICML 2025 Poster Paper, Camera Ready Version

Closed-Loop Long-Horizon Robotic Planning via Equilibrium Sequence Modeling

Jinghan Li, Zhicheng Sun, Yadong Mu

Comments ICML 2025

M3-JEPA: Multimodal Alignment via Multi-gate MoE based on the Joint-Embedding Predictive Architecture

Hongyang Lei, Xiaolong Cheng, Qi Qin, Dan Wang, Kun Fan, Huazhen Huang, Qingqing Gu, Yetao Wu, Zhonglin Jiang, Yong Chen, Luo Ji

机构 * Geely AI Lab（Geely人工智能实验室） ； Shenzhen Institutes of Advanced Technology（深圳先进技术研究院） ； Peking University（北京大学） 纠错

Comments 16 pages, 5 figures. ICML 2025

Self-Composing Policies for Scalable Continual Reinforcement Learning

Mikel Malagón, Josu Ceberio, Jose A. Lozano

机构 * Department of Computer Science and Artificial Intelligence, University of the Basque Country UPV/EHU, Donostia-San Sebastian, Spain（计算机科学与人工智能系，巴斯克国家大学UPV/EHU，圣地亚哥-桑塞班西斯，西班牙） ； Basque Center for Applied Mathematics (BCAM), Bilbao, Spain（应用数学巴斯克中心（BCAM），毕尔巴鄂，西班牙） 纠错

Comments ICML 2024 (oral)

A Variational Framework for Improving Naturalness in Generative Spoken Language Models

Li-Wei Chen, Takuya Higuchi, Zakaria Aldeneh, Ahmed Hussen Abdelaziz, Alexander Rudnicky

机构 * Language Technology Institute, Carnegie Mellon University（卡内基梅隆大学语言技术研究所） ； Apple（苹果公司） 纠错

Comments International Conference on Machine Learning (ICML) 2025

Adaptive Accompaniment with ReaLchords

Yusong Wu, Tim Cooijmans, Kyle Kastner, Adam Roberts, Ian Simon, Alexander Scarlatos, Chris Donahue, Cassie Tarakajian, Shayegan Omidshafiei, Aaron Courville, Pablo Samuel Castro, Natasha Jaques, Cheng-Zhi Anna Huang

机构 * Mila - Quebec AI Institute, Université de Montréal（魁北克AI研究所-蒙特利尔大学） ； Canada CIFAR AI Chair（加拿大CIFAR人工智能主席） ； Google DeepMind（谷歌DeepMind） ； Work done while at Google（在谷歌工作的期间） ； University of Massachusetts Amherst（马萨诸塞大学阿默斯特分校） ； Google（谷歌） ； University of Washington（华盛顿大学） ； Carnegie Mellon University（卡内基梅隆大学） 纠错

Comments Accepted by ICML 2024

Expressive Score-Based Priors for Distribution Matching with Geometry-Preserving Regularization

Ziyu Gong, Jim Lim, David I. Inouye

机构 * Elmore Family School of Electrical and Computer Engineering, Purdue University, West Lafayette, IN, USA（电子工程系，普渡大学，西拉法基，印第安纳州，美国） 纠错

Comments 32 pages, 20 figures. Accepted to ICML 2025

Journal ref Proceedings of the 42nd International Conference on Machine Learning, Vancouver, Canada, 2025

TGDPO: Harnessing Token-Level Reward Guidance for Enhancing Direct Preference Optimization

Mingkang Zhu, Xi Chen, Zhongdao Wang, Bei Yu, Hengshuang Zhao, Jiaya Jia

机构 * The Chinese University of Hong Kong（香港中文大学） ； The University of Hong Kong（香港大学） ； The Hong Kong University of Science（香港科学大学） ； Huawei（华为） 纠错

Comments ICML 2025

Train Once, Forget Precisely: Anchored Optimization for Efficient Post-Hoc Unlearning

Prabhav Sanga, Jaskaran Singh, Arun K. Dubey

机构 * University College London, London, United Kingdom（伦敦大学学院） ； University of Nottingham, Nottingham, United Kingdom（诺丁汉大学） ； Bharati Vidyapeeth College of Engineering, New Delhi, India（巴里蒂大学工程学院） 纠错

Comments Accepted at ICML MUGen'25

From Black Boxes to Transparent Minds: Evaluating and Enhancing the Theory of Mind in Multimodal Large Language Models

Xinyang Li, Siqi Liu, Bochao Zou, Jiansheng Chen, Huimin Ma

机构 * School of Computer and Communication Engineering, University of Science and Technology Beijing（计算机与通信工程学院，科学技术大学） 纠错

Comments 24 pages, 22 figures, accepted at ICML 2025, project page: see https://annaisavailable.github.io/GridToM/

Constant Stepsize Local GD for Logistic Regression: Acceleration by Instability

Michael Crawshaw, Blake Woodworth, Mingrui Liu

机构 * George Mason University（乔治·玛森大学） ； George Washington University（乔治·华盛顿大学） 纠错

Comments ICML 2025

DeepSeq: High-Throughput Single-Cell RNA Sequencing Data Labeling via Web Search-Augmented Agentic Generative AI Foundation Models

Saleem A. Al Dajani, Abel Sanchez, John R. Williams

机构 * Massachusetts Institute of Technology（麻省理工学院） 纠错

Comments 4 pages, 5 figures, Accepted by ICML 2025 FM4LS https://openreview.net/forum?id=zNjXOZxEYB . Workshop on Multi-modal Foundation Models and Large Language Models for Life Sciences (FM4LS)}, July 2025

Journal ref International Conference on Machine Learning (ICML). Workshop on Multi-modal Foundation Models and Large Language Models for Life Sciences (FM4LS), July 2025

ReFrame: Layer Caching for Accelerated Inference in Real-Time Rendering

Lufei Liu, Tor M. Aamodt

机构 * University of British Columbia（不列颠哥伦比亚大学） 纠错

Comments Published at ICML 2025

An Open-Source Software Toolkit & Benchmark Suite for the Evaluation and Adaptation of Multimodal Action Models

Pranav Guruprasad, Yangyue Wang, Sudipta Chowdhury, Jaewoo Song, Harshvardhan Sikka

机构 * Metarch ； Manifold Research ； Georgia Institute of Technology（佐治亚理工学院） 纠错

Comments ICML CodeML Workshop, 13 Pages, 6 Figures, 2 Tables

FSL-SAGE: Accelerating Federated Split Learning via Smashed Activation Gradient Estimation

Srijith Nair, Michael Lin, Peizhong Ju, Amirreza Talebi, Elizabeth Serena Bentley, Jia Liu

机构 * The Ohio State University（俄亥俄州立大学） ； Department of Electrical and Computer Engineering（电气与计算机工程系） ； Department of Industrial Engineering（工业工程系） ； University of Kentucky（肯塔基大学） ； Air Force Research Laboratory（空军研究实验室） 纠错

Comments Accepted for poster presentation at ICML 2025, 23 pages, 14 figures

Neural Genetic Search in Discrete Spaces

Hyeonah Kim, Sanghyeok Choi, Jiwoo Son, Jinkyoo Park, Changhyun Kwon

机构 * Mila - Quebec AI Institute（魁北克AI研究所） 纠错

Comments ICML 2025

Nonlinearly Preconditioned Gradient Methods under Generalized Smoothness

Konstantinos Oikonomidis, Jan Quan, Emanuel Laude, Panagiotis Patrinos

Comments ICML 2025 oral

Position: Editing Large Language Models Poses Serious Safety Risks

Paul Youssef, Zhixue Zhao, Daniel Braun, Jörg Schlötterer, Christin Seifert

机构 * Marburg University（马尔堡大学） ； University of Sheffield（谢菲尔德大学） ； University of Mannheim（曼海姆大学） 纠错

Comments Accepted at ICML 2025

Temperature-Annealed Boltzmann Generators

Henrik Schopmans, Pascal Friederich

机构 * Institute of Nanotechnology, Karlsruhe Institute of Technology（卡尔斯鲁厄理工学院纳米技术研究所） ； Institute of Theoretical Informatics, Karlsruhe Institute of Technology（卡尔斯鲁厄理工学院理论计算机科学研究所） 纠错

Journal ref Proceedings of the 42nd International Conference on Machine Learning (ICML), Vancouver, Canada. PMLR 267, 2025

Robust Multi-bit Text Watermark with LLM-based Paraphrasers

Xiaojun Xu, Jinghan Jia, Yuanshun Yao, Yang Liu, Hang Li

机构 * University of California, Santa Cruz（加州大学圣克鲁兹分校） 纠错

Comments Accepted by ICML 2025

DMM: Distributed Matrix Mechanism for Differentially-Private Federated Learning Based on Constant-Overhead Linear Secret Resharing

Alexander Bienstock, Ujjwal Kumar, Antigoni Polychroniadou

机构 * J.P. Morgan AI Research \& J.P. Morgan AlgoCRYPT CoE, New York, New York, USA 纠错

Comments International Conference on Machine Learning (ICML), 2025

Improved Off-policy Reinforcement Learning in Biological Sequence Design

Hyeonah Kim, Minsu Kim, Taeyoung Yun, Sanghyeok Choi, Emmanuel Bengio, Alex Hernández-García, Jinkyoo Park

机构 * Mila - Quebec AI Institute（魁北克人工智能研究所） ； Valence Labs（Valence实验室） 纠错

Comments ICML 2025

Learning Invariant Causal Mechanism from Vision-Language Models

Zeen Song, Siyu Zhao, Xingyu Zhang, Jiangmeng Li, Changwen Zheng, Wenwen Qiang

机构 * Institute of Software Chinese Academy of Sciences, Beijing, China（中国科学院软件研究所） ； University of the Chinese Academy of Sciences（中国科学院大学） 纠错

Comments Accepted to ICML 2025

Sketch-Plan-Generalize: Learning and Planning with Neuro-Symbolic Programmatic Representations for Inductive Spatial Concepts

Namasivayam Kalithasan, Sachit Sachdeva, Himanshu Gaurav Singh, Vishal Bindal, Arnav Tuli, Gurarmaan Singh Panjeta, Harsh Himanshu Vora, Divyanshu Aggarwal, Rohan Paul, Parag Singla

Comments Programmatic Representations for Agent Learning Worskop, ICML 2025

The Courage to Stop: Overcoming Sunk Cost Fallacy in Deep Reinforcement Learning

Jiashun Liu, Johan Obando-Ceron, Pablo Samuel Castro, Aaron Courville, Ling Pan

机构 * Hong Kong University of Science ； Mila - Qu\'ebec AI Institute 纠错

Comments Proceedings of the 42nd International Conference on Machine Learning (ICML 2025)

TensorSLM: Energy-efficient Embedding Compression of Sub-billion Parameter Language Models on Low-end Devices

Mingxue Xu, Yao Lei Xu, Danilo P. Mandic

机构 * Imperial College London（伦敦帝国理工学院） 纠错

Comments ICML 2025 Workshop on Tiny Titans: The next wave of On-Device Learning for Foundational Models (TTODLer-FM)

LapDDPM: A Conditional Graph Diffusion Model for scRNA-seq Generation with Spectral Adversarial Perturbations

Lorenzo Bini, Stephane Marchand-Maillet

机构 * Department of Computer Science, University of Geneva, Geneva, Switzerland（日内瓦大学计算机科学系） 纠错

Comments LapDDPM is a novel conditional graph diffusion model for scRNA-seq generation. Leveraging spectral adversarial perturbations, it ensures robustness and yields high-fidelity, biologically plausible, and cell-type-specific samples for complex data. Proceedings of the ICML 2025 GenBio Workshop: The 2nd Workshop on Generative AI and Biology, Vancouver, Canada, 2025

The Butterfly Effect: Neural Network Training Trajectories Are Highly Sensitive to Initial Conditions

Devin Kwok, Gül Sena Altıntaş, Colin Raffel, David Rolnick

机构 * School of Computer Science, McGill University, Montreal, Canada（麦吉尔大学计算机科学学院） ； Mila -- Quebec AI Institute, Montreal, Canada（魁北克人工智能研究所） ； University of Toronto, Vector Canada（多伦多大学） ； Vector Institute, Toronto, Canada（向量研究所） ； McGill University（麦吉尔大学） ； Mila -- Quebec AI Institute（魁北克人工智能研究所） ； University of Toronto（多伦多大学） ； Vector Institute（向量研究所） 纠错

Comments Published in ICML 2025. The first two authors contributed equally. 29 pages, 28 figures