ICML

SAEBench: A Comprehensive Benchmark for Sparse Autoencoders in Language Model Interpretability

Adam Karvonen, Can Rager, Johnny Lin, Curt Tigges, Joseph Bloom, David Chanin, Yeu-Tong Lau, Eoin Farrell, Callum McDougall, Kola Ayonrinde, Demian Till, Matthew Wearden, Arthur Conmy, Samuel Marks, Neel Nanda

Comments Accepted to ICML 2025 main conference

Learning Cascade Ranking as One Network

Yunli Wang, Zhen Zhang, Zhiqiang Wang, Zixuan Yang, Yu Li, Jian Yang, Shiyang Wen, Peng Jiang, Kun Gai

Comments Accepted by ICML 2025

Dynamic Benchmarking of Reasoning Capabilities in Code Large Language Models Under Data Contamination

Simin Chen, Pranav Pusarla, Baishakhi Ray

Comments This paper is accepted to ICML 2025. Website: https://codekaleidoscope.github.io/dycodeeval.html

Generalization in Federated Learning: A Conditional Mutual Information Framework

Ziqiao Wang, Cheng Long, Yongyi Mao

机构 * Ziqiao Wang（瓦氏） ； Cheng Long（龙成） ； Yongyi Mao（茅永毅） 纠错

Comments Accepted to ICML 2025

M3HF: Multi-agent Reinforcement Learning from Multi-phase Human Feedback of Mixed Quality

Ziyan Wang, Zhicheng Zhang, Fei Fang, Yali Du

机构 * Department of Informatics, King’s College London, London, United Kingdom（伦敦大学国王学院信息学系） ； Societal Systems Department, Carnegie Mellon University, Pittsburgh, PA, USA（卡内基梅隆大学社会系统系） 纠错

Comments Accepted to ICML 2025

Dynamical phases of short-term memory mechanisms in RNNs

Bariscan Kurtkaya, Fatih Dinc, Mert Yuksekgonul, Marta Blanco-Pozo, Ege Cirakman, Mark Schnitzer, Yucel Yemez, Hidenori Tanaka, Peng Yuan, Nina Miolane

Comments published in ICML 2025

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

FlexTok: Resampling Images into 1D Token Sequences of Flexible Length

Roman Bachmann, Jesse Allardice, David Mizrahi, Enrico Fini, Oğuzhan Fatih Kar, Elmira Amirloo, Alaaeldin El-Nouby, Amir Zamir, Afshin Dehghan

机构 * Apple（苹果公司） ； Swiss Federal Institute of Technology Lausanne (EPFL)（瑞士联邦理工学院洛桑分校） 纠错

Comments ICML 2025. Project page at https://flextok.epfl.ch/

How Compositional Generalization and Creativity Improve as Diffusion Models are Trained

Alessandro Favero, Antonio Sclocchi, Francesco Cagnetta, Pascal Frossard, Matthieu Wyart

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

CellFlux: Simulating Cellular Morphology Changes via Flow Matching

Yuhui Zhang, Yuchang Su, Chenyu Wang, Tianhong Li, Zoe Wefers, Jeffrey Nirschl, James Burgess, Daisy Ding, Alejandro Lozano, Emma Lundberg, Serena Yeung-Levy

Comments Published at ICML 2025

Visual Attention Never Fades: Selective Progressive Attention ReCalibration for Detailed Image Captioning in Multimodal Large Language Models

Mingi Jung, Saehyung Lee, Eunji Kim, Sungroh Yoon

Comments ICML 2025

An Adaptive Orthogonal Convolution Scheme for Efficient and Flexible CNN Architectures

Thibaut Boissin, Franck Mamalet, Thomas Fel, Agustin Martin Picard, Thomas Massena, Mathieu Serrurier

机构 * Innovation \& Research Division, SNCF, France ； Kempner Institute, Harvard University, USA 纠错

Journal ref International Conference on Machine Learning, 2025, Vancouver, Canada

HashAttention: Semantic Sparsity for Faster Inference

Aditya Desai, Shuo Yang, Alejandro Cuadron, Matei Zaharia, Joseph E. Gonzalez, Ion Stoica

Comments Accepted at ICML'2025

MMedPO: Aligning Medical Vision-Language Models with Clinical-Aware Multimodal Preference Optimization

Kangyu Zhu, Peng Xia, Yun Li, Hongtu Zhu, Sheng Wang, Huaxiu Yao

机构 * UNC Chapel-Hill（UNC夏洛特-希尔分校） ； Brown University（布朗大学） ； University of Washington（华盛顿大学） 纠错

Comments ICML 2025

Diffusion-VLA: Generalizable and Interpretable Robot Foundation Model via Self-Generated Reasoning

Junjie Wen, Minjie Zhu, Yichen Zhu, Zhibin Tang, Jinming Li, Zhongyi Zhou, Chengmeng Li, Xiaoyu Liu, Yaxin Peng, Chaomin Shen, Feifei Feng

Comments Accepted by ICML 2025. The project page is available at: http://diffusion-vla.github.io

Subspace Optimization for Large Language Models with Convergence Guarantees

Yutong He, Pengrui Li, Yipeng Hu, Chuyan Chen, Kun Yuan

机构 * Peking University（北京大学） ； Zhongguancun Academy（中关村学院） ； Beihang University（北航大学） ； AI for Science Institute, Beijing, China（人工智能科学研究院，北京，中国） ； National Engineering Laboratory for Big Data Analytics（大数据分析与应用国家工程实验室） 纠错

Comments Accepted by ICML 2025

VinePPO: Refining Credit Assignment in RL Training of LLMs

Amirhossein Kazemnejad, Milad Aghajohari, Eva Portelance, Alessandro Sordoni, Siva Reddy, Aaron Courville, Nicolas Le Roux

Comments Accepted at ICML 2025; 12 pages and 22 pages Appendix

EasyInv: Toward Fast and Better DDIM Inversion

Ziyue Zhang, Mingbao Lin, Shuicheng Yan, Rongrong Ji

Comments Accepted by ICML 2025

Data-Juicer Sandbox: A Feedback-Driven Suite for Multimodal Data-Model Co-development

Daoyuan Chen, Haibin Wang, Yilun Huang, Ce Ge, Yaliang Li, Bolin Ding, Jingren Zhou

Comments Accepted by ICML 2025 (Spotlight). 33 pages, 16 tables, 14 figures

Quantifying Prediction Consistency Under Fine-Tuning Multiplicity in Tabular LLMs

Faisal Hamman, Pasan Dissanayake, Saumitra Mishra, Freddy Lecue, Sanghamitra Dutta

Comments International Conference on Machine Learning (ICML), 2025

Craftium: Bridging Flexibility and Efficiency for Rich 3D Single- and Multi-Agent Environments

Mikel Malagón, Josu Ceberio, Jose A. Lozano

机构 * Department of Computer Science and Artificial Intelligence, University of the Basque Country UPV/EHU（计算机科学与人工智能系，巴斯克国家大学UPV/EHU） ； Basque Center for Applied Mathematics (BCAM)（巴斯克应用数学中心） 纠错

Comments ICML 2025. Project's website: https://github.com/mikelma/craftium/

LDMol: A Text-to-Molecule Diffusion Model with Structurally Informative Latent Space Surpasses AR Models

Jinho Chang, Jong Chul Ye

Comments Poster in ICML 2025; 19 pages, 13 figures

Representation Surgery: Theory and Practice of Affine Steering

Shashwat Singh, Shauli Ravfogel, Jonathan Herzig, Roee Aharoni, Ryan Cotterell, Ponnurangam Kumaraguru

Comments Accepted in ICML 2024

Understanding the Training Speedup from Sampling with Approximate Losses

Rudrajit Das, Xi Chen, Bertram Ieong, Parikshit Bansal, Sujay Sanghavi

Comments Updated version of our paper published in ICML 2024 (https://proceedings.mlr.press/v235/das24b.html)

$\texttt{BTSbot}$: A Multi-input Convolutional Neural Network to Automate and Expedite Bright Transient Identification for the Zwicky Transient Facility

Nabeel Rehemtulla, Adam A. Miller, Michael W. Coughlin, Theophile Jegou du Laz

Comments Accepted at the ICML 2023 Workshop on ML for Astrophysics; see Rehemtulla et al. 2024, ApJ, 972, 7R for the full BTSbot publication

AUTOCIRCUIT-RL: Reinforcement Learning-Driven LLM for Automated Circuit Topology Generation

Prashanth Vijayaraghavan, Luyao Shi, Ehsan Degan, Vandana Mukherjee, Xin Zhang

机构 * Prashanth Vijayaraghavan（独立研究者） ； Luyao Shi（独立研究者） ； Ehsan Degan（独立研究者） ； Vandana Mukherjee（独立研究者） ； Xin Zhang（独立研究者） 纠错

Comments 9 Pages (Content), 4 Pages (Appendix), 7 figures, ICML'2025

Revisiting Continuity of Image Tokens for Cross-domain Few-shot Learning

Shuai Yi, Yixiong Zou, Yuhua Li, Ruixuan Li

Comments Accepted by ICML 2025(spotlight)

Corrigibility as a Singular Target: A Vision for Inherently Reliable Foundation Models

Ram Potham, Max Harms

机构 * Independent Researcher（独立研究者） ； Machine Intelligence Research Institute（机器智能研究院） 纠错

Comments Preprint. This work has been submitted to the Reliable and Responsible Foundation Models Workshop at ICML 2025 for review

PAC Learning with Improvements

Idan Attias, Avrim Blum, Keziah Naggita, Donya Saless, Dravyansh Sharma, Matthew Walter

机构 * University of Illinois at Chicago（伊利诺伊大学香槟分校） ； Toyota Technological Institute at Chicago（芝加哥丰田技术研究所） ； Northwestern University（西北大学） 纠错

Comments 41 pages, 13 figures, ICML 2025

Graph Generative Pre-trained Transformer

Xiaohui Chen, Yinkai Wang, Jiaxing He, Yuanqi Du, Soha Hassoun, Xiaolin Xu, Li-Ping Liu

Comments ICML 2025

Random Registers for Cross-Domain Few-Shot Learning

Shuai Yi, Yixiong Zou, Yuhua Li, Ruixuan Li

Comments Accepted by ICML 2025