Taywon/

HH_full_parsed

This is the hh-rlhf dataset, with only the helpful split merged. And the format is parsed so that chosen and rejected are not strings but lists of dictionaries, where each dictionary is the conversation (following the more standard format).