xChemAgents: Agentic AI for Explainable Quantum Chemistry
Comments Accepted Paper at ICML 2025 Workshop on MAS
期刊&会议
ICML
International Conference on Machine Learning · 会议 · Machine Learning
Discovering Global False Negatives On the Fly for Self-supervised Contrastive Learning
Comments Accepted to ICML 2025
Counterfactual Influence as a Distributional Quantity
机构 * Imperial College London（帝国理工学院伦敦分校） ； Google DeepMind（谷歌DeepMind）
Comments Workshop on The Impact of Memorization on Trustworthy Foundation Models (MemFM) @ ICML 2025
Q-resafe: Assessing Safety Risks and Quantization-aware Safety Patching for Quantized Large Language Models
机构 * The State Key Laboratory of Blockchain and Data Security, Zhejiang University（区块链与数据安全国家重点实验室，浙江大学） ； Sun Yat-sen University（中山大学） ； Hangzhou High-Tech Zone (Binjiang) Institute of Blockchain and Data Security（杭州高新技术区（滨江）区块链与数据安全研究院）
Comments ICML 2025
ADDQ: Adaptive Distributional Double Q-Learning
机构 * Institute of Mathematics, University of Mannheim, Germany（曼海姆大学数学研究所） ； Department of Mathematics and Computer Science, Freie Universität Berlin, Germany（柏林自由大学数学与计算机科学系）
Journal ref ICML 2025
Morse: Dual-Sampling for Lossless Acceleration of Diffusion Models
机构 * Intel Labs China（英特尔中国实验室）
Comments Fixed a prompt typo in Figure 18 of the Appendix. This work is accepted to ICML 2025. The project page: https://github.com/deep-optimization/Morse
Evaluating Generalization and Representation Stability in Small LMs via Prompting, Fine-Tuning and Out-of-Distribution Prompts
机构 * Boston University, Boston, USA（波士顿大学） ； Carnegie Mellon University, Pittsburgh, USA（卡内基梅隆大学） ； Stanford University, Palo Alto, USA（斯坦福大学）
Comments Accepted at ICML
Angle Domain Guidance: Latent Diffusion Requires Rotation Rather Than Extrapolation
机构 * Department of Electronic Engineering, Tsinghua University, Beijing, China（清华大学电子工程系） ； Zhili College, Tsinghua University, Beijing, China（清华大学紫金学院）
Comments Accepted at ICML 2025
Provably Improving Generalization of Few-Shot Models with Synthetic Data
机构 * FPT Software AI Center（FPT软件人工智能中心） ； Hanoi University of Science（河内科学大学） ； VinUniversity（文大学） ； University of Warwick（沃里克大学）
Comments ICML 2025. Our code is released at https://github.com/Fsoft-AIC/ProtoAug
Adversarial Reasoning at Jailbreaking Time
机构 * University of Pennsylvania（宾夕法尼亚大学）
Comments Accepted to the 42nd International Conference on Machine Learning (ICML 2025)
Separating Tongue from Thought: Activation Patching Reveals Language-Agnostic Concept Representations in Transformers
机构 * ENS Paris-Saclay（巴黎萨克雷大学） ； Université Paris-Saclay（巴黎萨克雷大学） ； Northeastern（东北大学） ； Princeton（普林斯顿大学） ； Cornell（康奈尔大学） ； EPFL（洛桑联邦理工学院）
Comments 20 pages, 14 figures, previous version published under the title "How Do Llamas Process Multilingual Text? A Latent Exploration through Activation Patching" at the ICML 2024 mechanistic interpretability workshop at https://openreview.net/forum?id=0ku2hIm4BS
Can One Safety Loop Guard Them All? Agentic Guard Rails for Federated Computing
机构 * Division of Cancer Registry of Norway, Norwegian Institute of Public Health, Oslo, Norway（挪威癌症登记处，挪威公共卫生研究所，奥斯陆） ； Department of Physics and Technology, The Arctic University of Norway, Tromsø, Norway（挪威极地大学物理与技术系）
Comments Accepted at ICML 2025 Workshop on Collaborative and Federated Agentic Workflows (CFAgentic@ICML'25)
Protein Structure Tokenization: Benchmarking and New Recipe
机构 * Mila - Quebec AI Institute（魁北克AI研究所） ； Amazon（亚马逊） ； University of Montreal（蒙特利尔大学）
Comments Accepted at ICML 2025
Cross-regularization: Adaptive Model Complexity through Validation Gradients
机构 * NightCity Labs（夜城实验室）
Comments 21 pages, 13 figures. Accepted at ICML 2025
Maximal Update Parametrization and Zero-Shot Hyperparameter Transfer for Fourier Neural Operators
机构 * Carnegie Mellon University（卡内基梅隆大学） ； Brookhaven National Laboratory（布鲁克海文国家实验室）
Comments ICML 2025
Deep Electromagnetic Structure Design Under Limited Evaluation Budgets
机构 * School of Future Technologies, South China University of Technology（未来技术学院，华南理工大学） ； School of Software Engineering, South China University of Technology（软件工程学院，华南理工大学） ； School of Microelectronics, South China University of Technology（微电子学院，华南理工大学） ； Pazhou Laboratory（琶洲实验室） ； Peng Cheng Laboratory（鹏城实验室）
Comments ICML 2025 (accepted)
Private Model Personalization Revisited
机构 * Department of Computer Science & Engineering, The Ohio State University（计算机科学与工程系，俄亥俄州立大学） ； Department of Computer Science & Engineering and the Translational Data Analytics Institute (TDAI), The Ohio State University（计算机科学与工程系和转化数据分析研究所（TDAI），俄亥俄州立大学） ； Google Research, NY（谷歌研究，纽约）
Comments ICML 2025
Distilling Tool Knowledge into Language Models via Back-Translated Traces
机构 * University of Oxford（牛津大学） ； Amazon（亚马逊公司） ； University of Cambridge（剑桥大学） ； University of Los Angeles（洛杉矶大学） ； University of Chicago（芝加哥大学） ； Ludwig Maximilian University of Munich（慕尼黑莱布尼茨大学）
Comments Accepted in Workshop in Multi-Agent Systems in the Era of Foundation Models: Opportunities, Challenges and Futures, ICML 2025
When Diffusion Models Memorize: Inductive Biases in Probability Flow of Minimum-Norm Shallow Neural Nets
机构 * Marquette University（马quette大学）
Comments Accepted to the Forty-second International Conference on Machine Learning (ICML 2025)
eccDNAMamba: A Pre-Trained Model for Ultra-Long eccDNA Sequence Analysis
机构 * Department of Computer Science, Brown University, Providence, RI, USA（计算机科学系，布朗大学）
Comments Accepted by ICML 2025 Generative AI and Biology (GenBio) Workshop
MuseControlLite: Multifunctional Music Generation with Lightweight Conditioners
机构 * National Taiwan University（国立台湾大学） ； Massachusetts Institute of Technology（麻省理工学院）
Comments Accepted by the 42nd International Conference on Machine Learning (ICML 2025)
Privacy-Shielded Image Compression: Defending Against Exploitation from Vision-Language Pretrained Models
机构 * Pengcheng Laboratory（鹏城实验室） ； Guangdong Laboratory of Artificial Intelligence（广东省人工智能与数字经济实验室） ； Shenzhen University（深圳大学）
Comments 11 pages, 6 figures, publised to ICML 2025
Compositional Scene Understanding through Inverse Generative Modeling
机构 * Harvard University（哈佛大学）
Comments ICML 2025, Webpage: https://energy-based-model.github.io/compositional-inference
SASSHA: Sharpness-aware Adaptive Second-order Optimization with Stable Hessian Approximation
Comments ICML 2025
Flopping for FLOPs: Leveraging equivariance for computational efficiency
机构 * Chalmers University of Technology（楚德斯技术大学）
Comments ICML 2025
LAuReL: Learned Augmented Residual Layer
机构 * Google Research, Mountain View, CA.（谷歌研究，山景城，加利福尼亚州） ； Google Research, New York, NY.（谷歌研究，纽约，纽约州）
Comments Accepted at 42nd International Conference on Machine Learning (2025), Vancouver, Canada
Learning Decision Policies with Instrumental Variables through Double Machine Learning
机构 * Department of Computer Science, University of Oxford（牛津大学计算机科学系） ； Department of Electrical Engineering and Computer Science, Massachusetts Institute of Technology（麻省理工学院电子工程与计算机科学系）
Comments Accepted at ICML 2024
Journal ref Proceedings of the 41st International Conference on Machine Learning, PMLR 235:44489-44514, 2024
CommVQ: Commutative Vector Quantization for KV Cache Compression
机构 * University of Massachusetts Amherst（马萨诸塞大学阿默斯特分校） ； Massachusetts Institute of Technology（麻省理工学院） ； Princeton University（普林斯顿大学） ； Apple Inc（苹果公司）
Comments ICML 2025 poster
LIGHTHOUSE: Fast and precise distance to shoreline calculations from anywhere on earth
机构 * Allen Institute for AI (Ai2)（Allen人工智能研究所）
Comments 8 pages, 7 figures, 1 table, ICML 2025 ML4RS
ContinualFlow: Learning and Unlearning with Neural Flow Matching
机构 * Department of Computer Science, University of Pisa, Pisa, Italy（比萨大学计算机科学系） ； School of Public Health, Yale University, New Haven, U.S.A.（耶鲁大学公共卫生学院）
Comments Accepted at the ICML 2025 Workshop on Machine Unlearning for Generative AI (MUGen @ ICML25, Vancouver, July 2025)
