A Vision-Language-Action-Critic Model for Robotic Real-World Reinforcement Learning
This addresses the challenge of reward engineering and exploration efficiency for robotic reinforcement learning in real-world settings, representing a significant advance rather than an incremental improvement.
The paper tackles the problem of robotic real-world reinforcement learning being bottlenecked by sparse rewards and inefficient exploration by introducing VLAC, a vision-language-action-critic model that provides dense progress rewards and supports one-shot transfer to unseen tasks. VLAC increased success rates from about 30% to about 90% within 200 real-world episodes, with human-in-the-loop interventions further improving sample efficiency by 50% and achieving up to 100% final success.
Robotic real-world reinforcement learning (RL) with vision-language-action (VLA) models is bottlenecked by sparse, handcrafted rewards and inefficient exploration. We introduce VLAC, a general process reward model built upon InternVL and trained on large scale heterogeneous datasets. Given pairwise observations and a language goal, it outputs dense progress delta and done signal, eliminating task-specific reward engineering, and supports one-shot in-context transfer to unseen tasks and environments. VLAC is trained on vision-language datasets to strengthen perception, dialogic and reasoning capabilities, together with robot and human trajectories data that ground action generation and progress estimation, and additionally strengthened to reject irrelevant prompts as well as detect regression or stagnation by constructing large numbers of negative and semantically mismatched samples. With prompt control, a single VLAC model alternately generating reward and action tokens, unifying critic and policy. Deployed inside an asynchronous real-world RL loop, we layer a graded human-in-the-loop protocol (offline demonstration replay, return and explore, human guided explore) that accelerates exploration and stabilizes early learning. Across four distinct real-world manipulation tasks, VLAC lifts success rates from about 30\% to about 90\% within 200 real-world interaction episodes; incorporating human-in-the-loop interventions yields a further 50% improvement in sample efficiency and achieves up to 100% final success.