Review for NeurIPS paper: Policy Improvement via Imitation of Multiple Oracles