Out-of-Distribution Detection Methods Answer the Wrong Questions
Comments Extended version of ICML 2025 paper
期刊&会议
ICML
International Conference on Machine Learning · 会议 · Machine Learning
SPARSE-PIVOT: Dynamic correlation clustering for node insertions
Comments ICML 2025
LoRA Fine-Tuning Without GPUs: A CPU-Efficient Meta-Generation Framework for LLMs
机构 * Department of Mathematics, McMaster University, Hamilton, Canada（麦斯特大学数学系） ； Vector Institute, Toronto, Canada（向量研究所） ； University of Oxford, Oxford, United Kingdom（牛津大学）
Comments 5-page main paper (excluding references) + 11-page appendix, 3 tables, 1 figure. Accepted to ICML 2025 Workshop on Efficient Systems for Foundation Models
SE(3)-Equivariant Diffusion Policy in Spherical Fourier Space
机构 * Amazon Robotics（亚马逊机器人技术）
Comments Accepted at ICML 2025
Dynamic Similarity Graph Construction with Kernel Density Estimation
机构 * University of Edinburgh, United Kingdom（爱丁堡大学） ； University of St Andrews, United Kingdom（圣安德鲁大学）
Comments ICML'25
ICLShield: Exploring and Mitigating In-Context Learning Backdoor Attacks
机构 * Nanyang Technological University（南洋理工大学） ； Beihang University（北航大学）
Comments ICML 2025
ConceptAttention: Diffusion Transformers Learn Highly Interpretable Features
机构 * ibm（IBM研究院）
Comments Oral Presentation at ICML 2025, Best Paper Award at CVPR Workshop on Visual Concepts
NegMerge: Sign-Consensual Weight Merging for Machine Unlearning
机构 * Sogang University（ソガン大学） ； NAVER AI Lab（NAVER AI实验室）
Comments Accepted to ICML 2025
SafeMap: Robust HD Map Construction from Incomplete Observations
机构 * Beijing Academy of Artificial Intelligence（北京人工智能研究院） ； School of Computer Science, Hefei University of Technology（合肥工业大学计算机学院） ； Institute of Information Engineering, Chinese Academy of Sciences（中国科学院信息工程研究所） ； School of Computer Science, Wuhan University（武汉大学计算机学院） ； National University of Singapore（新加坡国立大学） ； Intelligent Interconnected Systems Laboratory of Anhui Province (Hefei University of Technology)（安徽省智能互联系统实验室（合肥工业大学））
Comments Accepted by ICML 2025
Neural Augmented Kalman Filters for Road Network assisted GNSS positioning
机构 * Department of Electrical Engineering, Eindhoven University of Technology（埃因霍温理工大学电子工程系） ； Qualcomm AI Research（高通人工智能研究）
Comments Accepted to ICML 2025 workshop ML4Wireless
Foundation Models for Clinical Records at Health System Scale
机构 * new york university（纽约大学）
Comments Accepted to ICML 2025 Workshop on Foundation Models for Structured Data
Exploring Large Action Sets with Hyperspherical Embeddings using von Mises-Fisher Sampling
机构 * Deezer Research（Deep Research（Deezer）） ； SPEIT, Shanghai Jiao Tong University（上海交通大学 SPEIT 实验室）
Comments 42nd International Conference on Machine Learning (ICML 2025)
Simulation-Efficient Cosmological Inference with Multi-Fidelity SBI
机构 * Kavli IPMU (WPI), UTIAS, The University of Tokyo, Japan（Kavli IPMU（WPI）、UTIAS、东京大学） ； Center for Data-Driven Discovery, Kavli IPMU (WPI), Japan（数据驱动发现中心、Kavli IPMU（WPI）、日本） ； Center for Computational Astrophysics, Flatiron Institute, New York, USA（计算天体物理学中心、Flatiron研究所、美国纽约） ； Department of Astrophysical Sciences, Princeton University, USA（天体物理学系、普林斯顿大学、美国） ； RCAST, The University of Tokyo, Japan（RCAST、东京大学、日本）
Comments 5 pages, 4 figures; accepted at ICML-colocated ML4Astro 2025 workshop
A Recipe for Causal Graph Regression: Confounding Effects Revisited
机构 * Hong Kong Baptist University（香港 Baptist 大学） ； Hong Kong University of Science（香港科学大学） ； Zhejiang University（浙江大学）
Comments ICML 2025 accepted
Beyond Sensor Data: Foundation Models of Behavioral Data from Wearables Improve Health Predictions
机构 * USC（美国大学） ； Apple Inc（苹果公司）
Comments Accepted to ICML 2025
Causal Discovery of Latent Variables in Galactic Archaeology
Comments Accepted to ML4Astro 2025 (Machine Learning for Astrophysics workshop at ICML 2025)
Accelerating Quantum Reinforcement Learning with a Quantum Natural Policy Gradient Based Approach
机构 * Purdue University, West Lafayette, IN, USA（普渡大学）
Comments Proceedings of the 42nd International Conference on Machine Learning
ETTA: Elucidating the Design Space of Text-to-Audio Models
机构 * NVIDIA
Comments ICML 2025. Demo: https://research.nvidia.com/labs/adlr/ETTA/ Code: https://github.com/NVIDIA/elucidated-text-to-audio
EvoPress: Accurate Dynamic Model Compression via Evolutionary Search
机构 * Yandex Research（Yandex研究）
Comments ICML camera-ready
Outlier Weighed Layerwise Sparsity (OWL): A Missing Secret Sauce for Pruning LLMs to High Sparsity
机构 * Eindhoven University of Technology, the Netherlands（埃因霍温理工大学） ； University of Surrey（萨里大学） ； Google Research, USA（谷歌研究院） ； University of Texas at Austin, USA（德克萨斯大学奥斯汀分校） ； Clemson University, USA（克莱姆森大学） ； University of Washington, USA（华盛顿大学） ； University of Oxford, UK（牛津大学）
Comments Published at ICML 2024
LibVulnWatch: A Deep Assessment Agent System and Leaderboard for Uncovering Hidden Vulnerabilities in Open-Source AI Libraries
机构 * Holistic AI ； University College London（伦敦大学学院）
Comments ACL 2025 Student Research Workshop and ICML 2025 TAIG Workshop
Simultaneous Multi-Robot Motion Planning with Projected Diffusion Models
机构 * University of Virginia（弗吉尼亚大学） ； University of California（加州大学）
Comments Published at the Forty-Second International Conference on Machine Learning (ICML 2025)
Attestable Audits: Verifiable AI Safety Benchmarks Using Trusted Execution Environments
机构 * Department of Computer Science and Technology, University of Cambridge（计算机科学与技术系，剑桥大学）
Comments ICML 2024 Workshop TAIG
When Will It Fail?: Anomaly to Prompt for Forecasting Future Anomalies in Time Series
机构 * Department of Artificial Intelligence, Kyung Hee University, Yongin, Republic of Korea（韩国成均馆大学人工智能系） ； Department of Artificial Intelligence, Korea University, Seoul, Republic of Korea（韩国大学人工智能系）
Comments 18 pages, 10 figures, 12 tables, ICML 2025
Accurate Parameter-Efficient Test-Time Adaptation for Time Series Forecasting
Comments Second Workshop on Test-Time Adaptation: Putting Updates to the Test! at ICML 2025, Vancouver, Canada. 2025
TuCo: Measuring the Contribution of Fine-Tuning to Individual Responses of LLMs
机构 * University of Oxford（牛津大学）
Comments ICML 2025
Not All Explanations for Deep Learning Phenomena Are Equally Valuable
机构 * Department of Applied Mathematics and Theoretical Physics, University of Cambridge（应用数学与理论物理系，剑桥大学）
Comments Accepted at ICML 2025 for oral presentation
MoMa: Modulating Mamba for Adapting Image Foundation Models to Video Recognition
机构 * Cooperative Medianet Innovation Center, Shanghai Jiao Tong University（上海交通大学协同中位创新中心） ； School of Artificial Intelligence, Shanghai Jiao Tong University（上海交通大学人工智能学院） ； Shanghai Artificial Intelligence Laboratory（上海人工智能实验室）
Comments ICML 2025 paper
BEST-Route: Adaptive LLM Routing with Test-Time Optimal Compute
机构 * The University of British Columbia（不列颠哥伦比亚大学） ； Microsoft（微软） ； Pennsylvania State University（宾夕法尼亚州立大学） ； AG2AI, Inc.（AG2AI公司）
Comments Accepted to ICML 2025 (main conference)
Residual Matrix Transformers: Scaling the Size of the Residual Stream
机构 * University of California Santa Cruz（加州大学圣克鲁兹分校）
Comments Accepted to ICML 2025
