Review for NeurIPS paper: Learning Guidance Rewards with Trajectory-space Smoothing