PF3plat: Pose-Free Feed-Forward 3D Gaussian Splatting
机构 * Microsoft Research Asia（微软亚洲研究院）
Comments Accepted by ICML'25
期刊&会议
ICML
International Conference on Machine Learning · 会议 · Machine Learning
Weak-to-Strong Jailbreaking on Large Language Models
机构 * Sea AI Lab, Singapore（新加坡海智实验室） ； Carnegie Mellon University（卡内基梅隆大学）
Comments ICML 2025
Efficient and Robust Semantic Image Communication via Stable Cascade
Comments Accepted at ICML 2025 Workshop on Machine Learning for Wireless Communication and Networks (ML4Wireless)
Continual Generalized Category Discovery: Learning and Forgetting from a Bayesian Perspective
机构 * Department of Computer Science, UCL Centre for Artificial Intelligence, University College London, London, UK（计算机科学系，UCL人工智能中心，伦敦大学学院，伦敦，英国） ； University of Southampton, Southampton, UK（南安普顿大学，南安普顿，英国）
Comments 20 pages, 6 figures. Forty-second International Conference on Machine Learning. 2025
SToFM: a Multi-scale Foundation Model for Spatial Transcriptomics
机构 * Department of Computer Science and Tecnology, Tsinghua University（清华大学计算机科学与技术系） ； Institute for AI Industry Research (AIR), Tsinghua University（清华大学人工智能产业研究院） ； School of Mathematical Sciences, Peking University（北京大学数学科学学院） ； PharMolix Inc.（PharMolix公司）
Comments Accpeted by ICML 2025
SADA: Stability-guided Adaptive Diffusion Acceleration
机构 * Department of Electrical and Computer Engineering, Duke University, Durham, U.S.A（电子与计算机工程系，杜克大学，Durham，美国） ； Department of Statistical Science, Duke University, Durham, U.S.A（统计科学系，杜克大学，Durham，美国）
Comments Accepted and published by ICML 2025. Code is available at: https://github.com/Ting-Justin-Jiang/sada-icml
laplax -- Laplace Approximations with JAX
机构 * Tübingen AI center, University of Tübingen, Tübingen, Germany（图宾根人工智能中心，图宾根大学）
Comments Submission to the ICML 2025 Workshop on Championing Open-source Development in Machine Learning (CODEML '25)
Fast and Scalable Gene Embedding Search: A Comparative Study of FAISS and ScaNN
机构 * Department of Electrical and Computer Engineering, Drexel University, Philadelphia, USA（电气与计算机工程系，德雷塞尔大学，费城，美国）
Journal ref GenBio @ ICML 2025 Workshop, OpenReview, June 2025 GenBio @ ICML 2025 Workshop, OpenReview, June 2025
Koel-TTS: Enhancing LLM based Speech Generation with Preference Alignment and Classifier Free Guidance
机构 * NVIDIA Corporation（英伟达公司）
Journal ref ICML Workshop on Machine Learning for Audio, 2025
GenMol: A Drug Discovery Generalist with Discrete Diffusion
机构 * NVIDIA
Comments ICML 2025
Language model developers should report train-test overlap
机构 * Stanford University（斯坦福大学）
Comments ICML 2025 Spotlight; 23 pages
Mixture-of-Expert Variational Autoencoders for Cross-Modality Embedding of Type Ia Supernova Data
Comments Accepted to the Machine Learning for Astrophysics Workshop co-located at ICML 2025 and selected for a spotlight talk
LLM Data Selection and Utilization via Dynamic Bi-level Optimization
机构 * School of Artificial Intelligence, University of Chinese Academy of Sciences（中国科学院大学人工智能学院） ； The Key Laboratory of Cognition and Decision Intelligence for Complex Systems, Institute of Automation, Chinese Academy of Sciences（中国科学院自动化研究所认知与决策智能复杂系统重点实验室） ； Huawei Noah’s Ark Lab（华为诺亚实验室） ； College of Intelligence and Computing, Tianjin University（天津大学智能与计算学院） ； Nanyang Technological University（南洋理工大学）
Comments The 42nd International Conference on Machine Learning (ICML 2025)
MPO: An Efficient Post-Processing Framework for Mixing Diverse Preference Alignment
机构 * Department of Statistics, Rutgers University, New Brunswick, United States ； Department of Computer Science, Rutgers University, New Brunswick, United States ； College of Management of Technology, EPFL, Switzerland ； Department of Computer Science, ETH Zurich, Switzerland
Comments ICML 2025
TorchAO: PyTorch-Native Training-to-Serving Model Optimization
机构 * Meta Platforms Inc.（Meta公司） ； OpenTeams Inc.（OpenTeams公司）
Comments 5 pages, 3 figures, published in CODEML@ICML25
Journal ref ICML 2025 Workshop on Championing Open-source DEvelopment (CODEML 2025)
Beyond the ATE: Interpretable Modelling of Treatment Effects over Dose and Time
机构 * Department of Mathematics and Theoretical Physics, University of Cambridge, United Kingdom（数学与理论物理系，剑桥大学，英国）
Comments Presented at the Actionable Interpretability Workshop at ICML 2025
Global Convergence and Rich Feature Learning in $L$-Layer Infinite-Width Neural Networks under $μ$P Parametrization
机构 * Department of Computer Science, University of California, Los Angeles, USA（计算机科学系，加州大学洛杉矶分校）
Comments 28 pages, 17 figures, 2 tables. In ICML 2025
Practical Principles for AI Cost and Compute Accounting
机构 * MIT CSAIL ； Stanford University（斯坦福大学） ； Future of Life Institute（未来生命研究所）
Comments Accepted to the ICML 2025 TAIG Workshop
FlexiClip: Locality-Preserving Free-Form Character Animation
机构 * Search Technology Center India, Microsoft, IDC, Bengaluru, India(Bharat)（印度搜索技术中心、微软、IDC、班加罗尔、印度（巴哈拉特））
Comments 13 pages, 4 figures, 7 tables, Accepted in ICML 2025, https://openreview.net/forum?id=xtxCM4XZ82
Journal ref Proceedings of the 42nd International Conference on Machine Learning, Vancouver, Canada. PMLR 267, 2025
Closed-form Solutions: A New Perspective on Solving Differential Equations
机构 * AnnLab, Institute of Semiconductors, Chinese Academy of Sciences, Beijing, China（安 lab，半导体研究所，中国科学院，北京，中国） ； College of Materials Science and Opto-Electronic Technology, University of Chinese Academy of Sciences, Beijing, China（材料科学与光电技术学院，中国科学院大学，北京，中国） ； School of Integrated Circuits, University of Chinese Academy of Sciences, Beijing, China（集成电路学院，中国科学院大学，北京，中国） ； School of Industry-education Integration, University of Chinese Academy of Sciences, Beijing, China（产教融合学院，中国科学院大学，北京，中国）
Comments Accepted as a poster at the 42nd International Conference on Machine Learning (ICML 2025), Vancouver, Canada
Pruning Increases Orderedness in Recurrent Computation
机构 * Harvard College（哈佛学院）
Comments 8 pages, 11 figures, 2 tables, Workshop on Methods and Opportunities at Small Scale (MOSS), ICML 2025
Solo Connection: A Parameter Efficient Fine-Tuning Technique for Transformers
机构 * Department of Data Science, Worcester Polytechnic Institute, MA, United States（数据科学系，沃斯特理工学院，马萨诸塞州，美国） ； Department of Data Science, Mathematics and Computer Science, Worcester Polytechnic Institute, MA, United States（数据科学系，数学与计算机科学，沃斯特理工学院，马萨诸塞州，美国） ； Department of Data Science, Mathematics（数据科学系，数学） ； Computer Science, Worcester Polytechnic Institute, MA, United States（计算机科学，沃斯特理工学院，马萨诸塞州，美国）
Journal ref ICML SIM workshop 2025
ProofCompass: Enhancing Specialized Provers with LLM Guidance
机构 * Federal University of Rio de Janeiro（里约热内卢联邦大学） ； Harvard John A. Paulson School of Engineering and Applied Sciences（哈佛大学约翰·A·保罗森工程与应用科学学院） ； Stanford University（斯坦福大学）
Comments 19 pages, 7 figures. Accepted at the 2nd AI for MATH Workshop at the 42nd International Conference on Machine Learning (ICML 2025)
LaCache: Ladder-Shaped KV Caching for Efficient Long-Context Modeling of Large Language Models
机构 * NVIDIA
Comments ICML 2025. Code: https://github.com/GATECH-EIC/LaCache
Catalyst: a Novel Regularizer for Structured Pruning with Auxiliary Extension of Parameter Space
Comments ICML 2025 workshop HiLD 2025 (3rd workshop on High-dimensional Learning Dynamics)
DIVER-0 : A Fully Channel Equivariant EEG Foundation Model
机构 * snu（首尔国立大学） ； bnl（布鲁克海文国家实验室）
Comments 11 pages, 1 figures, ICML 2025 Workshop on GenBio
Interpretable Reward Modeling with Active Concept Bottlenecks
机构 * Department of Computer Science, ETH Zurich, Zurich, Switzerland（苏黎世联邦理工学院计算机科学系） ； Department of Applied Mathematics and Theoretical Physics, University of Cambridge, United Kingdom（剑桥大学应用数学与理论物理系）
Journal ref ICML 2025 Workshop on Programmatic Representations for Agent Learning
Beyond Atomic Geometry Representations in Materials Science: A Human-in-the-Loop Multimodal Framework
机构 * Computer Engineering, Texas A\&M University, College Station, TX 77843, USA ； College of Science ； Engineering, Hamad Bin Khalifa University, Doha, Qatar ； Dept. of Electrical \& Computer Engineering, Texas A\&M University at Qatar, Doha, Qatar ； Orthotics, Ankara University, Ankara, Turkey
Comments Presented at ICML 2025 Workshop on DataWorld
Grokking at the Edge of Linear Separability
机构 * Raymond and Beverly Sackler School of Physics and Astronomy, Tel Aviv University, Tel Aviv 69978, Israel（雷蒙德与贝弗利·萨克勒物理与天文学学院，特拉维夫大学） ； École Polytechnique Fédérale de Lausanne (EPFL), Lausanne, Switzerland（洛桑联邦理工学院（EPFL），洛桑，瑞士）
Comments Camera-ready version, ICML 2025
Sortformer: A Novel Approach for Permutation-Resolved Speaker Supervision in Speech-to-Text Systems
机构 * nvidia
Comments Published at ICML 2025
Journal ref Proceedings of the 42nd International Conference on Machine Learning (ICML), 2025
