michaelcohen comments on Delegative Reinforcement Learning with a Merely Sane Advisor

[michaelcohen]

(as opposed to standard regret bounds in RL which are only applicable in the episodic setting)

??