Dataset for SafeAligner: Safety Alignment against Jailbreak Attacks via Response Disparity Guidance

from datasets import load_dataset dataset = load_dataset("csHuang/SafeAligner")

	Data Category




	

Scenario Num

Adult Content 34 12.2 19.6 272.3

Economic Harm 38 14.8 17.8 218.8

Fraud Deception 72 15.1 20.4 241.1

Illegal Activity 144 14.6 21.4 206.5

Hate/Harass/Violence 130 15.7 17.3 183.8

Malware 130 17.0 20.1 249.3

Physical Harm… See the full description on the dataset page: https://huggingface.co/datasets/csHuang/SafeAligner.