Review for NeurIPS paper: Improving Generalization in Reinforcement Learning with Mixture Regularization