The "Extreme Toxic Human Medium Dataset" is a curated collection of prompt-response pairs designed to represent harmful, toxic, unethical, and problematic human-generated text. The prompts are designed to elicit undesirable responses from language models, and the accompanying responses exemplify such harmful outputs.
Warning: This dataset contains extremely offensive, toxic, disturbing, and potentially illegal content, including… See the full description on the dataset page:
https://huggingface.co/datasets/deepkaria/extreme-toxic-human-medium-dataset.
