This is an interesting paper on the application of ML / RL methods to the problem of reducing traffic congestion. The reviewers have come to a clear consensus that this paper is ready for acceptance at NeurIPS, and in discussion even the most critical reviewer has noted that this paper is ready for acceptance and that it would be a good contribution to the conference. I'd like to use this opportunity to ask the authors to ensure that they address the reviewer feedback fully in revision. In particular, there are several cases where the author response indicated "lack of space" as a reason why formal definitions, additional results, etc., were not included. Space is indeed difficult, but in cases like these preparing a supplement / appendix is a good way to go and can help the community learn as much as possible from the work. Please do address all feedback either in the main text or by creating a supplement or appendix. Many thanks!