Sparse Autoencoders, Again?
Comments Accepted to the International Conference on Machine Learning (ICML) 2025
期刊&会议
ICML
International Conference on Machine Learning · 会议 · Machine Learning
Smoothed Preference Optimization via ReNoise Inversion for Aligning Diffusion Models with Varied Human Preferences
Comments Accepted by ICML 2025
HGOT: Self-supervised Heterogeneous Graph Neural Network with Optimal Transport
Comments The paper has 9 pages of text and 13 pages in total (including acknowledgments, impact statement, references, and appendix), with 6 figures and 2 tables. This paper has been accepted by ICML 2025 conference and this is a final version of the manuscript submitted to the conference
On the Importance of Gaussianizing Representations
Comments ICML 2025 Proceedings
Investigating Non-Transitivity in LLM-as-a-Judge
机构 * AI Centre, UCL（UCL人工智能中心）
Comments ICML 2025 (Spotlight)
Relational Conformal Prediction for Correlated Time Series
Comments ICML 2025
RoSTE: An Efficient Quantization-Aware Supervised Fine-Tuning Approach for Large Language Models
Comments accepted by ICML 2025
Toward Efficient Kernel-Based Solvers for Nonlinear PDEs
Journal ref Forty-Second International Conference on Machine Learning (ICML2025)
Exploring Representations and Interventions in Time Series Foundation Models
Comments Accepted at ICML'25
Longitudinal Targeted Minimum Loss-based Estimation with Temporal-Difference Heterogeneous Transformer
Comments Published in ICML 2024, PMLR 235
Journal ref Proceedings of the 41st International Conference on Machine Learning, PMLR 235:45097-45113, 2024
Truly Self-Improving Agents Require Intrinsic Metacognitive Learning
机构 * DAMTP, University of Cambridge, Cambridge, UK（剑桥大学 DAMTP 实验室）
Comments Published as a conference paper at ICML 2025
Privacy Amplification Through Synthetic Data: Insights from Linear Regression
Comments 26 pages, ICML 2025
UnHiPPO: Uncertainty-aware Initialization for State Space Models
机构 * Munich Data Science Institute, Technical University of Munich（慕尼黑技术大学数据科学研究所） ； Department of Computer Science, Technical University of Munich（慕尼黑技术大学计算机科学系）
Comments Published at ICML 2025
TIMING: Temporality-Aware Integrated Gradients for Time Series Explanation
Comments ICML 2025 Spotlight Presentation; Code is available at https://github.com/drumpt/TIMING
Identifying and Understanding Cross-Class Features in Adversarial Training
Comments ICML 2025
Learning Joint Interventional Effects from Single-Variable Interventions in Additive Models
机构 * Max Planck Institute for Intelligent Systems（马克斯·普朗克智能系统研究所） ； Amazon Research（亚马逊研究） ； Tübingen AI Center（图宾根人工智能中心） ； ELLIS Institute（ELLIS研究所）
Comments To be published at the International Conference on Machine Learning (ICML) 2025
A Reasoning-Based Approach to Cryptic Crossword Clue Solving
Comments 9 page paper plus Appendices. Accepted to ICML 2025
Rethinking the Stability-Plasticity Trade-off in Continual Learning from an Architectural Perspective
机构 * Aojun Lu（阿俊·卢） ； Hangjie Yuan（元航杰） ； Tao Feng（冯涛） ； Yanan Sun（孙燕安）
Comments Accepted to ICML 2025
Piloting Structure-Based Drug Design via Modality-Specific Optimal Schedule
Comments Accepted to ICML 2025
Connecting Thompson Sampling and UCB: Towards More Efficient Trade-offs Between Privacy and Regret
Comments Camera-ready Version for ICML 2025
LLM Social Simulations Are a Promising Research Method
Comments Published at ICML 2025
Stochastic Poisson Surface Reconstruction with One Solve using Geometric Gaussian Processes
机构 * cornell（康奈尔大学） ； mit（麻省理工学院） ； columbia（哥伦比亚大学）
Journal ref International Conference on Machine Learning, 2025
On the Power of Context-Enhanced Learning in LLMs
机构 * princeton university（普林斯顿大学）
Comments 77 pages, 18 figures; ICML 2025 Main Conference
Contrastive Visual Data Augmentation
Journal ref ICML 2025
Combinatorial Reinforcement Learning with Preference Feedback
机构 * Seoul National University, Seoul, Korea（首尔国立大学）
Comments Accepted at ICML 2025
EmbodiedBench: Comprehensive Benchmarking Multi-modal Large Language Models for Vision-Driven Embodied Agents
Comments Accepted to ICML 2025
CurvGAD: Leveraging Curvature for Enhanced Graph Anomaly Detection
Comments ICML 2025
$μ$nit Scaling: Simple and Scalable FP8 LLM Training
Comments ICML 2025; 17 pages, 13 figures
Fast Video Generation with Sliding Tile Attention
Comments Accepted by ICML 2025
A theoretical framework for overfitting in energy-based modeling
Comments 29 pages, 20 figures (including appendix). Accepted at Proceedings of the 42nd International Conference on Machine Learning, Vancouver, Canada. PMLR 267, 2025
