pyrit.datasets.fetch_llm_latent_adversarial_training_harmful_dataset#
- fetch_llm_latent_adversarial_training_harmful_dataset() SeedDataset[source]#
Fetch the LLM-LAT harmful dataset from Hugging Face.
This dataset contains prompts used to assess and analyze harmful behaviors in large language models. It is sourced from the LLM-LAT/harmful-dataset repository.
- Returns:
A dataset containing harmful prompts for LLM testing.
- Return type: