Dataset Splits and Corresponding Code Mapping

[yejipark-m]

First, thank you for the outstanding work on this paper and its accompanying codebase—it’s truly impressive!

I have a question regarding the dataset splits and their connection to the reported results. Specifically:

How do the Adversarial (in Appendix) scores relate to the datasets in AAAAB_data.json and BAAAA_data.json?
Are these files directly tied to those scores, or is there an additional mapping or preprocessing step involved?

Your clarification would be greatly appreciated. Once again, thank you for your excellent work on this project!