Perception Before Reasoning: Two-Stage Reinforcement Learning for Visual Reasoning in Vision-Language Models

OffRLReLMLRMVLMRe-assign community