AdvPromptSet

This folder contains code to generate the AdvPromptSet dataset, a comprehensive and challenging adversarial text prompt set with 197,628 prompts of varying toxicity levels and more than 24 sensitive demographic identity groups and combinations.

Link: https://github.com/facebookresearch/ResponsibleNLP/blob/main/AdvPromptSet/README.md