ResearchBenchmarksJailbreak Attack Evaluation on CuriosityFollow5Successful Attack CountPOLARIS4.0410.521723.48May 24, 2026Evaluation ResultsMethodMethodLinksSuccessful Attack CountPOLARISTarget Model=Qwen3-8B,...Target Model=Qwen3-8B, Evaluator=GPT-5-mini2026.055POLARISTarget Model=Qwen3-8B,...Target Model=Qwen3-8B, Evaluator=DeepSeek-R12026.0529