OmniVLA: An Omni-Modal Vision-Language-Action Model for Robot Navigation
Comments 9 pages, 7 figures, 6 tables
作者
Sergey Levine
Robotics / Reinforcement Learning
Compute-Optimal Scaling for Value-Based Deep RL
机构 * University of California, Berkeley（加州大学伯克利分校） ； Carnegie Mellon University（卡内基梅隆大学）
Value-Based Deep RL Scales Predictably
机构 * University of Warsaw（华沙大学）
Comments ICML 2025
Adapt On-the-Go: Behavior Modulation for Single-Life Robot Deployment
机构 * Stanford University（斯坦福大学） ； UC Berkeley（加州大学伯克利分校）
Hi Robot: Open-Ended Instruction Following with Hierarchical Vision-Language-Action Models
机构 * Stanford University（斯坦福大学） ； University of California, Berkeley（加州大学伯克利分校）
Comments ICML 2025
Behavioral Exploration: Learning to Explore via In-Context Adaptation
机构 * Department of Electrical Engineering \& Computer Science, University of California, Berkeley
Leveraging Skills from Unlabeled Prior Data for Efficient Online Exploration
机构 * University of California, Berkeley（加州大学伯克利分校）
Comments 28 pages, 20 figures, 42nd International Conference on Machine Learning (ICML 2025)
Efficient Online Reinforcement Learning Fine-Tuning Need Not Retain Offline Data
机构 * UC Berkeley（加州大学伯克利分校） ； Carnegie Mellon University（卡内基梅隆大学）
Journal ref International Conference on Learning Representation (ICLR) 2025
Steering Your Diffusion Policy with Latent Space Reinforcement Learning
机构 * UC Berkeley（伯克利大学） ； University of Washington（华盛顿大学） ； Amazon（亚马逊公司）
One Step Diffusion via Shortcut Models
机构 * UC Berkeley（伯克利大学）
Visual Pre-Training on Unlabeled Images using Reinforcement Learning
Stabilizing Contrastive RL: Techniques for Robotic Goal Reaching from Offline Data
Comments ICLR 2024 Spotlight (< 5%). Website (https://chongyi-zheng.github.io/stable_contrastive_rl) and code (https://github.com/chongyi-zheng/stable_contrastive_rl)
Dynamic Search for Inference-Time Alignment in Diffusion Models
机构 * Texas A&M University（德克萨斯A&M大学） ； Genentech（基因泰克） ； UC Berkeley（加州大学伯克利分校）
Self-Challenging Language Model Agents
机构 * UC Berkeley（伯克利大学） ； FAIR at Meta（Meta 公司人工智能研究部）
Knowledge Insulating Vision-Language-Action Models: Train Fast, Run Fast, Generalize Better
机构 * Physical Intelligence
Diffusion Guidance Is a Controllable Policy Improvement Operator
机构 * UC Berkeley（加州大学伯克利分校）
SFT Memorizes, RL Generalizes: A Comparative Study of Foundation Model Post-training
Comments Website at https://tianzhechu.com/SFTvsRL
Flow Q-Learning
机构 * ucb（加州大学伯克利分校）
Comments ICML 2025
Inference via Interpolation: Contrastive Representations Provably Enable Planning and Inference
机构 * Princeton University（普林斯顿大学） ； UC Berkeley（加州大学伯克利分校） ； Carnegie Mellon University（卡内基梅隆大学）
Comments Code: https://github.com/vivekmyers/contrastive_planning
Journal ref Neural information processing systems (2024)
Training Strategies for Efficient Embodied Reasoning
机构 * UC Berkeley（伯克利大学） ； Stanford University（斯坦福大学） ； Physical Intelligence（物理智能）
Comments Updated figure layout, added project page link
Open X-Embodiment: Robotic Learning Datasets and RT-X Models
Comments Project website: https://robotics-transformer-x.github.io
Prioritized Generative Replay
机构 * Department of Electrical Engineering and Computer Science（电气工程与计算机科学系） ； University of California, Berkeley（加州大学伯克利分校）
Comments Project page available at: https://pgenreplay.github.io
$π_{0.5}$: a Vision-Language-Action Model with Open-World Generalization
机构 * Physical Intelligence
DROID: A Large-Scale In-The-Wild Robot Manipulation Dataset
Comments Project website: https://droid-dataset.github.io/
Reward-Guided Iterative Refinement in Diffusion Models at Test-Time with Applications to Protein and DNA Design
Comments Under review. If you have any suggestions/missing references, please let us know
AutoEval: Autonomous Evaluation of Generalist Robot Manipulation Policies in the Real World
ViVa: Video-Trained Value Functions for Guiding Online RL from Diverse Data
Precise and Dexterous Robotic Manipulation via Human-in-the-Loop Reinforcement Learning
SERL: A Software Suite for Sample-Efficient Robotic Reinforcement Learning
Comments ICRA 2024
SWEET-RL: Training Multi-Turn LLM Agents on Collaborative Reasoning Tasks
Comments 29 pages, 16 figures
