List your company

Jailbreak Defense on Common Jailbreak Attacks Suite (Qwen3.5-9B)

0IJP ASR (Manual)Qwen3.5-9B (No Defense)
-0.001-0.000500.0005Apr 1, 2026

Evaluation results

MethodLinks
Qwen3.5-9B (No Defense)Defense=None2026.040—0—0—0—32—33—11—9.5—
Perplexity Filter2026.0401000620100010032100331001193.670.113,073.1
GradientCuff2026.04010001000100010032100331001010073.691,427.98
Prompt Guard2026.040008042094321003397.551056.927.771,174.32
Llama GuardEvaluation Stage=Pre-processing2026.04056.10390500843295.243398.671170.50.413,547.56
Llama GuardEvaluation Stage=Post-processing2026.04054.9034045075069.213391.331161.570.5213,694.89
SelfDefendMethod Configuration=Direct2026.04025.8080150583270.483360.51139.630.6813,086.41
SelfDefendMethod Configuration=Intent2026.04029.40801101432603348.831128.541.9913,091.62
WildGuardEvaluation Stage=Pre-processing2026.0403.4020205032803393.171138.431.0513,921.18
WildGuardEvaluation Stage=Post-processing2026.04074.8078072064087.623398.671179.180.9813,979.57
GuardReasonerEvaluation Stage=Pre-processing2026.0400.90001036036.19——014.82——
GuardReasonerEvaluation Stage=Post-processing2026.04082.7073073080085.71——078.88——
SelfGrader2026.04000000000000001.25355.97