Provably Feedback-Efficient Reinforcement Learning via Active Reward Learning

An appropriate reward function is of paramount importance in specifying a task in reinforcement learning (RL).