4 datasets found

The dataset used in the SafeDecoding paper, which contains 32 harmful queries spanning 16 harmful categories.

The Comprehensive Assessment of Jailbreak Attacks against LLMs dataset is used to evaluate the effectiveness of jailbreak attacks on language models.

The dataset used to evaluate the effectiveness of different jailbreak attack methods against LLMs. The dataset contains 160 forbidden questions with high diversity.

The dataset used in the paper to evaluate the effectiveness of different jailbreak attack methods against Large Language Models (LLMs).