List your company

Jailbreak Attack Success Rate by Model on JailbreakBench

77Llama2 7B Attack Success RateLASH
-3.0817.7138.559.29May 20, 2026

Evaluation results

MethodLinks
LASHDefense=Perplexity2026.0577848177782870.83
LASHDefense=Smooth LLM2026.0547584333571141.5
TAPDefense=Perplexity2026.0531681751532440.67
TAPDefense=Smooth LLM2026.0522381124441024.83
LASHDefense=Llama Guard2026.0513171718261718
FlipAttackDefense=Perplexity2026.058513077171032.17
TAPDefense=Llama Guard2026.05795710118.17
AutoDAN-TDefense=Perplexity2026.057443037571131
AutoDANDefense=Llama Guard2026.05411———2
AutoDANDefense=Smooth LLM2026.053340———15.33
AutoDAN-TDefense=Smooth LLM2026.05344133249524.33
AutoDAN-TDefense=Llama Guard2026.0523921841918.5
PAIRDefense=Perplexity2026.05118494050126.5
PAIRDefense=Smooth LLM2026.0504532806.67
FlipAttackDefense=Smooth LLM2026.0504036006.67
PAIRDefense=Llama Guard2026.050109132119
FlipAttackDefense=Llama Guard2026.050424453.17
AutoDANDefense=Perplexity2026.050753———20