Jailbreak Attack on Qwen2.5-7B
0.02Normalized Rate (NR)IGCG
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| IGCGSurrogate Model=Qwen2.5-7B2026.03 | 0.02 | 0.02 | |
| GCGSurrogate Model=Gemma-3-4B2026.03 | 0.04 | 0.038 | |
| GCGSurrogate Model=Mistral-7B2026.03 | 0.04 | 0.015 | |
| IGCGSurrogate Model=Mistral-7B2026.03 | 0.04 | 0.033 | |
| GCGSurrogate Model=LLaMA-3.1-8B2026.03 | 0.06 | 0.05 | |
| IGCGSurrogate Model=Gemma-3-4B2026.03 | 0.06 | 0.043 | |
| GCGSurrogate Model=Qwen2.5-7B2026.03 | 0.1 | 0.068 | |
| PAIRSurrogate Model=Gemma-3-4B2026.03 | 0.14 | 0.11 | |
| IGCGSurrogate Model=LLaMA-3.1-8B2026.03 | 0.18 | 0.136 | |
| PAIRSurrogate Model=LLaMA-3.1-8B2026.03 | 0.2 | 0.173 | |
| PAIRSurrogate Model=Mistral-7B2026.03 | 0.2 | 0.165 | |
| PAIRSurrogate Model=Qwen2.5-7B2026.03 | 0.26 | 0.236 | |
| Auto-DANSurrogate Model=LLaMA-3.1-8B2026.03 | 0.68 | 0.638 | |
| Auto-DANSurrogate Model=Gemma-3-4B2026.03 | 0.86 | 0.853 | |
| Auto-DANSurrogate Model=Qwen2.5-7B2026.03 | 0.88 | 0.873 | |
| Auto-DANSurrogate Model=Mistral-7B2026.03 | 0.88 | 0.86 | |
| SWhisperSurrogate Model=Qwen2.5-7B2026.03 | 0.88 | 0.68 | |
| SWhisperSurrogate Model=Mistral-7B2026.03 | 0.92 | 0.81 | |
| SWhisperSurrogate Model=Gemma-3-4B2026.03 | 0.96 | 0.86 | |
| SWhisperSurrogate Model=LLaMA-3.1-8B2026.03 | 1 | 0.935 |