The reviewers agreed that this submission makes an interesting and novel contribution to NeurIPS. I strongly encourage the authors to address the remaining comments of reviewers 3 and 6, to clarify the test batch size, improve figure 2, and add citations.