论文 Vision-Language Models are Zero-Shot Reward Models for Reinforcement Learning 的阅读。本文提出了...