List your company

Jailbreak Defense on Jailbreak Attacks

28.1ES ASRAdL*
25.41243.55661.779.844Apr 12, 2026

Evaluation results

MethodLinks
AdL*Backbone=LLaMA2026.0428.1—0
LIRA*Backbone=LLaMA2026.0440.6—0
AdL*Backbone=Gemma2026.0449.21.50
GDBackbone=LLaMA2026.0457.8—0.7
CBBackbone=LLaMA2026.0465.6—3.9
LIRA*Backbone=Gemma2026.0475190.8
CBBackbone=Gemma2026.0493.722.22.3
ITBackbone=LLaMA2026.0493.8—0
GDBackbone=Gemma2026.0493.944.50.8
ITBackbone=Gemma2026.0495.357.10