All reviewers agree that the paper is of high quality and will be of interest to the NeurIPS community. I strongly encourage the authors to include results on the Maze dataset as it has been brought up by two of the reviewers. I look forward to reading the camera ready version of the manuscript.