R1 thought the paper was good but perhaps too narrow in scope, though R1 was positively moved by the rebuttal. R2 agreed that motivation was the biggest weakness and otherwise was very positive. R3, agreeing on the technical points, thought there was enough applicability to NeurIPS. R4 significantly raised their score after the rebuttal clarified some issues, and agreed with R3 that there was enough applicability. Overall, these were nice results, and the reviewers were quite positive.