ResearchBenchmarksJailbreak Attack Success Rate on JBShield (test)Jailbreak Attack Success Rate on JBShield (test)Follow95Attack Success Rate (ASR)JB-GCG90.2592.6259597.375May 4, 2026Evaluation resultsMethodMethodLinksAttack Success Rate (ASR)Toxic Detection RateJailbreak Detection RateASR (Direct)ASR (Masked)JB-GCGl_dir=18, l_opt=18l_dir=18, l_opt=182026.0595010048.426.3