This table contains 25,029 rows and 12 columns of data related to problematic content dialogue, including context, response, safety labels, annotations, reasons, and source. It can be used to analyze and understand patterns in prosocial dialogues and identify potential safety concerns. (AI-generated)
context, response, rots, safety_label, safety_annotations, safety_annotation_reasons, source, etc, dialogue_id, response_id, episode_done