Analyzing the Generalization Gap in Offline Reinforcement Learning

Offline learning algorithms struggle to generalize to new environments, highlighting the need for improved generalization capabilities.