Jailbreak Attack on Llama 8B 3.1
96NR RateSWhisper
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| SWhisperSurrogate Model=LLaMA-3.1-8B2026.03 | 96 | 94.5 | |
| SWhisperSurrogate Model=Qwen2.5-7B2026.03 | 58 | 57.8 | |
| SWhisperSurrogate Model=Gemma-3-4B2026.03 | 54 | 52.5 | |
| SWhisperSurrogate Model=Mistral-7B2026.03 | 46 | 45.8 | |
| Auto-DANSurrogate Model=LLaMA-3.1-8B2026.03 | 30 | 29.8 | |
| Auto-DANSurrogate Model=Mistral-7B2026.03 | 28 | 27 | |
| Auto-DANSurrogate Model=Qwen2.5-7B2026.03 | 18 | 18 | |
| PAIRSurrogate Model=Mistral-7B2026.03 | 16 | 13.5 | |
| Auto-DANSurrogate Model=Gemma-3-4B2026.03 | 16 | 16 | |
| PAIRSurrogate Model=LLaMA-3.1-8B2026.03 | 14 | 10.5 | |
| PAIRSurrogate Model=Gemma-3-4B2026.03 | 12 | 10.5 | |
| PAIRSurrogate Model=Qwen2.5-7B2026.03 | 10 | 9.5 | |
| GCGSurrogate Model=LLaMA-3.1-8B2026.03 | 2 | 2 | |
| GCGSurrogate Model=Mistral-7B2026.03 | 2 | 2 | |
| IGCGSurrogate Model=Qwen2.5-7B2026.03 | 2 | 2 | |
| GCGSurrogate Model=Gemma-3-4B2026.03 | 0 | 0 | |
| GCGSurrogate Model=Qwen2.5-7B2026.03 | 0 | 0 | |
| IGCGSurrogate Model=LLaMA-3.1-8B2026.03 | 0 | 0 | |
| IGCGSurrogate Model=Gemma-3-4B2026.03 | 0 | 0 | |
| IGCGSurrogate Model=Mistral-7B2026.03 | 0 | 0 |