Yang, C.; Littman, M.; Carbin, M. Computably Continuous Reinforcement-Learning Objectives Are PAC-Learnable. AAAI 2023, 37, 10729-10736.