Review for NeurIPS paper: Preference-based Reinforcement Learning with Finite-Time Guarantees