List your company

Adversarial Attack on LibrAI Do-Not-Answer and IndicAlign (145 prompts)

59ASR (%)GCG (Brute-Force)
31.9638.984653.02Jan 20, 2026

Evaluation results

MethodLinks
GCG (Brute-Force)Target Model=Vicuna 7B2026.01596.3
PEZTarget Model=Vicuna 7B2026.01397.3
GBDATarget Model=Vicuna 7B2026.01357.1
RECAPTarget Model=Vicuna 7B2026.01334