List your company

Jailbreak Attack Evaluation on SORRY

28Attack Success CountPOLARIS
27.829.1530.531.85May 24, 2026

Evaluation results

MethodLinks
POLARISTarget Model=Qwen3-8B, Evaluator=DeepSeek-R12026.0528
POLARISTarget Model=Qwen3-8B, Evaluator=GPT-5-mini2026.0533