This paper was considered thoroughly executed and well evaluated. The idea of applying GAN to modeling tabular data was deemed mixed: on the one hand, it is a very important data type that has received relatively little attention in the NeurIPS community, but on the other hand many of the core ideas and developments were deemed to be potentially specific to tabular data and thus not highly generalizable/informative/engaging to the NeurIPS community at large. The rebuttal and discussion of the ablation study have favorably improved reviewers opinions.