This dataset is an augmented version of PKU-Alignment/BeaverTails-V.
The release is organized into 19 categories, and each category contains train and validation splits:
animal_abuse
dangerous_behavior
deception_in_personal_relationships
discriminatory_depictions
environmental_damage… See the full description on the dataset page:
https://huggingface.co/datasets/SII-fleeeecer/PragmaSafe-Beavertails.