List your company

Jailbreak Attack Evaluation on Curiosity

5Successful Attack CountPOLARIS
4.0410.521723.48May 24, 2026

Evaluation results

MethodLinks
POLARISTarget Model=Qwen3-8B, Evaluator=GPT-5-mini2026.055
POLARISTarget Model=Qwen3-8B, Evaluator=DeepSeek-R12026.0529