List your company

Jailbreak Detection on GCG_adapt jailbreak prompts (Llama2-7B) (test)

81ASRHiddenDetect
56.0462.526975.48Jun 5, 2026

Evaluation results

MethodLinks
HiddenDetectAdaptive loss coefficient (λ)=0.1, Target Model=Llama2-7B2026.06814978
MTKAdaptive loss coefficient (λ)=0.1, Adaptive loss=J3, Target Model=Llama2-7B2026.067276517
HSFAdaptive loss coefficient (λ)=0.1, Target Model=Llama2-7B2026.066728448
SaPAdaptive loss coefficient (λ)=0.1, Target Model=Llama2-7B2026.06620662
GradSafeAdaptive loss coefficient (λ)=0.05, Target Model=Llama2-7B2026.065712350