Vision-Language Models as a Source of Rewards
Keywords: reward modeling, goal-conditioned reinforcement learning, generalist agents
TL;DR: We investigate using frozen vision-language models from the CLIP family as a source of binary rewards to train language goal-conditioned agents in several visual environments (Playhouse, AndroidEnv)
Abstract: Building generalist agents that can accomplish many goals in rich open-ended environments is one of the research frontiers for reinforcement learning. A key limiting factor for building generalist agents with RL has been the need for a large number of reward functions for achieving different goals. We investigate the feasibility of using off-the-shelf vision-language models, or VLMs, as sources of rewards for reinforcement learning agents. We show how rewards for visual achievement of a variety of language goals can be derived from the CLIP family of models, and used to train RL agents that can achieve a variety of language goals. We showcase this approach in two distinct visual domains and present a scaling trend showing how larger VLMs lead to more accurate rewards for visual goal achievement, which in turn produces more capable RL agents.
Submission Number: 48