SOLE-R1: Video-Language Reasoning as the Sole Reward for On-Robot Reinforcement Learning

arXiv · 2026-03-30 · High-Level Structured Planning / Multimodal Reasoning / Reward Reasoning