Automated Reward Generation in Reinforcement Learning Using Vision Language Models

RL-VLM-F automates reward function generation using vision language models, outperforming prior methods and enabling effective policy learning.