Jailbreak Attack on SafetyBench MiniGPT-4 Integrated from AdvBench (test) benchmark leaderboard · SOTA2 Research

Jailbreak Attack on SafetyBench MiniGPT-4 Integrated from AdvBench (test)

0.7024IA (Illegal Activity)MLAI

Evaluation Results

Method	Links
MLAI2024.11		0.7024	0.6657	0.5028	0.7202	0.7015	0.5067	0.6901	0.9674	0.7892	0.9516	1	0.9309	0.9785	0.7775
Liu et al.2024.11		0.1454	0.1192	0.1988	0.2451	0.0491	0.1856	0.2094	0.7911	0.105	0.8528	0.8812	0.9394	0.9175	0.4338
Qi et al.2024.11		0.1155	0.0397	0.1552	0.2343	0.068	0.1471	0.1911	0.7633	0.1297	0.8248	0.9093	0.9122	0.9125	0.4156
Plain Text2024.11		0.0192	0.0168	0.0332	0.0298	0.0568	0.0317	0.0414	0.6767	0.0897	0.7456	0.8433	0.765	0.9029	0.3271