Conformalized Interactive Imitation Learning: Handling Expert Shift and Intermittent Feedback