Abstract: Offline reinforcement learning strives to enable agents to effectively utilize pre-collected offline datasets for learning. Such an offline setup tremendously mitigates the problems of ...