Jailbreak Attack on Grok 4
82NR ScoreAuto-DAN
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Auto-DANSurrogate Model=Qwen2.5-7B2026.03 | 82 | 82 | |
| Auto-DANSurrogate Model=Mistral-7B2026.03 | 82 | 81 | |
| Auto-DANSurrogate Model=Gemma-3-4B2026.03 | 78 | 76.3 | |
| SWhisperSurrogate Model=LLaMA-3.1-8B2026.03 | 78 | 73.3 | |
| SWhisperSurrogate Model=Gemma-3-4B2026.03 | 78 | 69.5 | |
| SWhisperSurrogate Model=Qwen2.5-7B2026.03 | 78 | 74 | |
| SWhisperSurrogate Model=Mistral-7B2026.03 | 70 | 66 | |
| Auto-DANSurrogate Model=LLaMA-3.1-8B2026.03 | 66 | 65.8 | |
| PAIRSurrogate Model=Qwen2.5-7B2026.03 | 64 | 61.8 | |
| PAIRSurrogate Model=Gemma-3-4B2026.03 | 50 | 47.5 | |
| PAIRSurrogate Model=LLaMA-3.1-8B2026.03 | 46 | 43.8 | |
| PAIRSurrogate Model=Mistral-7B2026.03 | 46 | 43.3 | |
| GCGSurrogate Model=LLaMA-3.1-8B2026.03 | 22 | 20.8 | |
| GCGSurrogate Model=Gemma-3-4B2026.03 | 10 | 9.8 | |
| IGCGSurrogate Model=LLaMA-3.1-8B2026.03 | 10 | 9.3 | |
| IGCGSurrogate Model=Gemma-3-4B2026.03 | 8 | 7.8 | |
| GCGSurrogate Model=Qwen2.5-7B2026.03 | 6 | 6 | |
| GCGSurrogate Model=Mistral-7B2026.03 | 6 | 6 | |
| IGCGSurrogate Model=Qwen2.5-7B2026.03 | 6 | 6 | |
| IGCGSurrogate Model=Mistral-7B2026.03 | 2 | 2 |