There was initially some concern that this paper didn't have highly significant results since it appears to be the case that your work is reducing the problem of equilibrium computation to convex optimization. However, after some consideration, the reviewers were swayed by your rebuttal and agreed that the results were not trivial, the topic was interesting, and this appears to be an interesting application of optimization methods. We decided to accept your paper for NeurIPS this year, but we would strongly suggest that you incorporate some of the comments in your rebuttal into the main document, to better highlight the novelty. We strongly encourage you to take into account Reviewer 4's comments, who carefully went through the paper.