Jailbreak Attack Evaluation on HarmfulQA
16ASRDeepSeek V3.1
Evaluation Results
| Method | Links | |
|---|---|---|
| DeepSeek V3.1Model=DeepSeek V3.12026.01 | 16 | |
| Grok 3 MiniModel=Grok 3 Mini2026.01 | 17.5 | |
| Gemini 2.5 FlashModel=Gemini 2.5 Flash2026.01 | 18 | |
| Qwen2.5 7BModel=Qwen2.5 7B2026.01 | 21 | |
| GPT-4o-miniModel=GPT-4o-mini2026.01 | 22.5 | |
| Mixtral 8×7BModel=Mixtral 8×7B2026.01 | 48.5 |