ResearchBenchmarksJailbreaking on Jailbreaking promptsJailbreaking on Jailbreaking promptsFollow24Overall Jailbreak Success RateAttack13.616.31921.7Jun 4, 2025Evaluation resultsMethodMethodLinksOverall Jailbreak Success RateJailbreak Success Rate (Identity)Jailbreak Success Rate (Disinformation)Jailbreak Success Rate (Crime)Jailbreak Success Rate (X-Risk)AttackModel=MiniGPT 4-13BModel=MiniGPT 4-13B2025.062467122CleanModel=MiniGPT 4-13BModel=MiniGPT 4-13B2025.06163661DiffCAPModel=MiniGPT 4-13BModel=MiniGPT 4-13B2025.06143551