List your company

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

92ASRSaP
80.5683.5386.589.47Jun 5, 2026

Evaluation results

MethodLinks
SaPAdaptive loss coefficient (λ)=0.1, Target Model=Vicuna-7B2026.06920892
HiddenDetectAdaptive loss coefficient (λ)=0.1, Target Model=Vicuna-7B2026.06900590
HSFAdaptive loss coefficient (λ)=0.1, Target Model=Vicuna-7B2026.06900790
MTKAdaptive loss coefficient (λ)=0.1, Adaptive loss=J3, Target Model=Vicuna-7B2026.068885413
GradSafeAdaptive loss coefficient (λ)=0.05, Target Model=Vicuna-7B2026.06810481