Ad-Hoc Human-AI Coordination Challenge
机构 * FLAIR, University of Oxford, Oxford, UK（FLAIR，牛津大学，牛津，英国） ； Department of Computer Science, University of Oxford, UK（计算机科学系，牛津大学，英国）
Comments Published at ICML 2025
期刊&会议
ICML
International Conference on Machine Learning · 会议 · Machine Learning
Early Stopping Tabular In-Context Learning
机构 * Department of Computer Science, University of Freiburg, Freiburg, Germany（弗赖堡大学计算机科学系） ； Zuse School ELIZA, Darmstadt, Germany（泽乌斯学校ELIZA） ； ELLIS Institute, Tübingen, Germany（ELLIS研究所） ； Prior Labs, Freiburg, Germany（Prior实验室）
Comments ICML Workshop Paper
Training Flexible Models of Genetic Variant Effects from Functional Annotations using Accelerated Linear Algebra
机构 * New York University（纽约大学）
Comments For example: ICML 2025. Code available at: https://github.com/AlanNawzadAmin/DeepWAS
Learning Time-Aware Causal Representation for Model Generalization in Evolving Domains
机构 * Independent Researcher, China（独立研究者）
Comments ICML 2025
Scalable Non-Equivariant 3D Molecule Generation via Rotational Alignment
机构 * ETH Zurich（苏黎世联邦理工学院）
Comments ICML 2025; added conditional generation results
BIMgent: Towards Autonomous Building Modeling via Computer-use Agents
机构 * Chair of Computing in Civil and Building Engineering, Technical University of Munich, Germany（土木与建筑工程计算系，慕尼黑技术大学） ； TUM Georg Nemetschek Institute, Munich, Germany（慕尼黑技术大学Georg Nemetschek研究所）
Comments ICML 2025 Workshop on Computer Use Agents
Gumiho: A Hybrid Architecture to Prioritize Early Tokens in Speculative Decoding
机构 * Advanced Micro Devices, Inc., Beijing, China（Advanced Micro Devices公司，北京） ； Department of Electrical and Electronic Engineering, The University of Hong Kong（香港大学电子与电气工程系） ； Institute of Artificial Intelligence and Robotics, Xi’an Jiaotong University（西安交通大学人工智能与机器人研究所）
Comments Accepted to the 42nd International Conference on Machine Learning (ICML 2025). Code: https://github.com/AMD-AIG-AIMA/Gumiho
What can large language models do for sustainable food?
机构 * Stanford University（斯坦福大学）
Comments ICML camera ready version
How to Move Your Dragon: Text-to-Motion Synthesis for Large-Vocabulary Objects
机构 * Seoul National University（首尔国立大学） ； NVIDIA（英伟达）
Comments Accepted to ICML 2025
Benefits of Early Stopping in Gradient Descent for Overparameterized Logistic Regression
机构 * University of California, Berkeley（加州大学伯克利分校） ； Google DeepMind（谷歌DeepMind） ； New York University（纽约大学）
Comments ICML 2025 Camera Ready
Multimodal Medical Code Tokenizer
机构 * Department of Biomedical Informatics, Harvard Medical School, Boston, MA, USA（生物医学信息学系，哈佛医学院，波士顿，马萨诸塞州，美国）
Comments ICML'25
Efficiently Serving Large Multimodal Models Using EPD Disaggregation
机构 * Huawei Technologies Canada, BC, Canada（华为加拿大技术有限公司） ； Simon Fraser University, BC, Canada（西蒙弗雷泽大学） ； Huawei Cloud, China（华为云）
Comments 17 pages, 12 figures, 9 tables
Journal ref International Conference on Machine Proceedings of the 42nd International Conference on Machine Learning, Vancouver, Canada. PMLR 267, 2025
On Fine-Grained Distinct Element Estimation
Comments ICML 2025
Continual Learning for Wireless Channel Prediction
Comments Accepted at ICML Workshop on ML4Wireless
Demystifying Singular Defects in Large Language Models
机构 * EPFL（苏黎世联邦理工学院） ； School of Computer and Communication Sciences（计算机与通信科学学院） ； Swiss Data Science Center（瑞士数据科学中心） ； University of Chinese Academy of Sciences（中国科学院大学）
Comments ICML 2025
Universal Retrieval for Multimodal Trajectory Modeling
Comments 18 pages, 3 figures, accepted by Workshop on Computer-use Agents @ ICML 2025
From Token to Rhythm: A Multi-Scale Approach for ECG-Language Pretraining
机构 * School of Computing and Data Science, The University of Hong Kong, Hong Kong SAR, China（计算与数据科学学院，香港大学，香港特别行政区，中国）
Comments ICML 2025
Foundation Model Insights and a Multi-Model Approach for Superior Fine-Grained One-shot Subset Selection
机构 * National Engineering Research Center for Multimedia Software, Institute of Artificial Intelligence, School of Computer Science, Wuhan University, China（国家多媒体软件工程研究中心、人工智能研究院、计算机科学学院、武汉大学，中国） ； Hubei Key Laboratory of Multimedia（湖北省多媒体关键实验室） ； The University of Tokyo, Japan（东京大学，日本） ； National Institute of Informatics, Japan（日本信息处理学会） ； School of Computer Science（计算机科学学院） ； Technology, Wuhan University of Science（科技，武汉科技大学）
Comments 18 pages, 10 figures, accepted by ICML 2025
$\text{C}^{2}\text{BNVAE}$: Dual-Conditional Deep Generation of Network Traffic Data for Network Intrusion Detection System Balancing
Comments In ICML 2025 Workshop NewInML
SENSEI: Semantic Exploration Guided by Foundation Models to Learn Versatile World Models
机构 * Autonomous Learning, University of Tübingen（图宾根大学自主学习研究所） ； Empirical Inference, Max Planck Institute for Intelligent Systems（马克斯·普朗克智能系统研究所） ； Neuro-Cognitive Modeling, University of Tübingen（图宾根大学神经认知建模研究所） ； VISLab, University of Amsterdam（阿姆斯特丹大学视觉实验室）
Comments ICML 2025 camera-ready version. Project webpage at https://sites.google.com/view/sensei-paper
Generative Data Mining with Longtail-Guided Diffusion
机构 * OpenAI ； Colombia University（哥伦比亚大学） ； Meta
Comments 20 pages
Journal ref Proceedings of the 42nd International Conference on Machine Learning (ICML), 2025
Leveraging Online Olympiad-Level Math Problems for LLMs Training and Contamination-Resistant Evaluation
机构 * University of British Columbia（不列颠哥伦比亚大学） ； Vector Institute for AI（向量人工智能研究所） ； Canada CIFAR AI Chair（加拿大CIFAR人工智能主席）
Comments ICML 2025 Camera Ready
Metadata Conditioning Accelerates Language Model Pre-training
机构 * Princeton Language and Intelligence（普林斯顿语言与智能）
Comments Accepted to ICML 2025. Code available at https://github.com/princeton-pli/MeCo
QT-DoG: Quantization-aware Training for Domain Generalization
机构 * CVLab, EPFL, Switzerland（EPFL计算机视觉实验室） ； Swiss Data Science Center, Switzerland（瑞士数据科学中心）
Comments Accepted at International Conference on Machine Learning (ICML) 2025. Project website: https://saqibjaved1.github.io/QT_DoG/
MimicMotion: High-Quality Human Motion Video Generation with Confidence-aware Pose Guidance
机构 * Tencent（腾讯） ； Shanghai Jiao Tong University（上海交通大学）
Comments ICML 2025
Aligning Spoken Dialogue Models from User Interactions
机构 * Department of Computer Science, Cornell University. Work done at Kyutai.（计算机科学系，康奈尔大学。在Kyutai工作）
Comments Accepted at ICML 2025
Towards an Optimal Control Perspective of ResNet Training
机构 * Institute of Energy Systems, Energy Efficiency and Energy Economics, TU Dortmund, Germany（能源系统、能效与能源经济学研究所，杜伊斯堡-艾森大学） ； Department of Computer Science, Ruhr University Bochum, Germany（计算机科学系，波鸿鲁尔大学） ； Institute of Control Systems, Hamburg University of Technology, Germany（控制系统研究所，汉堡技术大学）
Comments Accepted for presentation at the High-dimensional Learning Dynamics (HiLD) workshop at ICML 2025
Action-Minimization Meets Generative Modeling: Efficient Transition Path Sampling with the Onsager-Machlup Functional
机构 * Computational Research, LBNL（劳伦斯伯克利国家实验室计算研究部） ； Faculty of Mathematics（数学系） ； Physics, Charles University（查尔斯大学物理系） ； Done prior to joining Amazon（加入亚马逊之前）
Comments ICML 2025
Transformer-Based Spatial-Temporal Counterfactual Outcomes Estimation
机构 * College of Computer Science and Technology, National University of Defense Technology, Changsha, China（计算机科学与技术学院，国防科技大学，长沙，中国） ； Intelligent Game and Decision Lab, Academy of Military Science, Beijing, China（智能游戏与决策实验室，军事科学院，北京，中国）
Comments 24 pages, accepted at ICML 2025
GoIRL: Graph-Oriented Inverse Reinforcement Learning for Multimodal Trajectory Prediction
机构 * Department of Electronic and Computer Engineering, Hong Kong University of Science and Technology, Hong Kong, China（香港理工大学电子与计算机工程系） ； Voyager Research, Didi Chuxing, China（维杰研究，滴滴出行，中国） ； Zhuoyu Technology Co., Ltd., Shenzhen, China（珠宇科技有限公司，深圳，中国）
Comments Accepted by ICML 2025
