Poster
in
Workshop: Agent Learning in Open-Endedness Workshop
Vision-Language Models as a Source of Rewards
Harris Chan · Volodymyr Mnih · Feryal Behbahani · Michael Laskin · Luyu Wang · Fabio Pardo · Maxime Gazeau · Himanshu Sahni · Daniel Horgan · Kate Baumli · Yannick Schroecker · Stephen Spencer · Richie Steigerwald · John Quan · Gheorghe Comanici · Sebastian Flennerhag · Alexander Neitz · Lei Zhang · Tom Schaul · Satinder Singh · Clare Lyle · Tim Rocktäschel · Jack Parker-Holder · Kristian Holsheimer
Keywords: [ reward modeling ] [ generalist agents ] [ goal-conditioned reinforcement learning ]
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.