Jailbreak Attack on DeepSeek
0NR ScoreGCG
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| GCGSurrogate Model=Qwen2.5-7B2026.03 | 0 | 0 | |
| IGCGSurrogate Model=Mistral-7B2026.03 | 0 | 0 | |
| IGCGSurrogate Model=Gemma-3-4B2026.03 | 2 | 2 | |
| IGCGSurrogate Model=Qwen2.5-7B2026.03 | 2 | 1.3 | |
| GCGSurrogate Model=Mistral-7B2026.03 | 4 | 1.8 | |
| GCGSurrogate Model=Gemma-3-4B2026.03 | 6 | 6 | |
| IGCGSurrogate Model=LLaMA-3.1-8B2026.03 | 8 | 7.3 | |
| GCGSurrogate Model=LLaMA-3.1-8B2026.03 | 12 | 11.5 | |
| PAIRSurrogate Model=Gemma-3-4B2026.03 | 12 | 11.3 | |
| PAIRSurrogate Model=Mistral-7B2026.03 | 28 | 26.3 | |
| PAIRSurrogate Model=LLaMA-3.1-8B2026.03 | 30 | 28 | |
| PAIRSurrogate Model=Qwen2.5-7B2026.03 | 36 | 31.8 | |
| SWhisperSurrogate Model=Mistral-7B2026.03 | 56 | 53.8 | |
| Auto-DANSurrogate Model=LLaMA-3.1-8B2026.03 | 68 | 63.8 | |
| Auto-DANSurrogate Model=Mistral-7B2026.03 | 68 | 67.8 | |
| Auto-DANSurrogate Model=Qwen2.5-7B2026.03 | 72 | 70.8 | |
| SWhisperSurrogate Model=Qwen2.5-7B2026.03 | 72 | 71.3 | |
| SWhisperSurrogate Model=Gemma-3-4B2026.03 | 74 | 71.5 | |
| SWhisperSurrogate Model=LLaMA-3.1-8B2026.03 | 78 | 74.5 | |
| Auto-DANSurrogate Model=Gemma-3-4B2026.03 | 82 | 82 |