[2302.02061] Reinforcement Learning with History-Dependent Dynamic Contexts