The reviewers generally found the paper's main idea original and interesting, and the discourse easy to follow. The rebuttal satisfied most of the reviewers' questions, and they thought the proposed changes would make the paper acceptable at NeurIPS. A few remaining points that should be addressed in the final draft: - still unclear how limiting the assumption of spherical embedding is, and the provided reference in the rebuttal doesn't really provide much insight - it would be good to include more compelling experimental results (at a minimum, a small case study on a real world graph)