Jailbreak Attack Evaluation on Do-Not-Answer
2.5ASRGemini 2.5 Flash
Evaluation Results
| Method | Links | |
|---|---|---|
| Gemini 2.5 FlashModel=Gemini 2.5 Flash2026.01 | 2.5 | |
| GPT-4o-miniModel=GPT-4o-mini2026.01 | 3 | |
| Qwen2.5 7BModel=Qwen2.5 7B2026.01 | 3 | |
| Grok 3 MiniModel=Grok 3 Mini2026.01 | 4 | |
| DeepSeek V3.1Model=DeepSeek V3.12026.01 | 4.5 | |
| Mixtral 8×7BModel=Mixtral 8×7B2026.01 | 14 |