Jailbreak Attack on AdvBench-50 + Malicious Instruct
100ASRStegoAttack
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| StegoAttackTarget Model=Gemini-32025.05 | 100 | 4.97 | 4.71 | |
| StegoAttackTarget Model=DS-V3.22025.05 | 100 | 4.9 | 4.89 | |
| StegoAttackTarget Model=Qwen32025.05 | 99.33 | 4.99 | 4.44 | |
| DRATarget Model=DS-V3.22025.05 | 98 | 4.91 | 4.84 | |
| StegoAttackTarget Model=GPT-52025.05 | 82.67 | 3.98 | 2.44 | |
| DRATarget Model=Qwen32025.05 | 82 | 4.12 | 3.92 | |
| JailbrokenTarget Model=DS-V3.22025.05 | 72.67 | 3.33 | 3.72 | |
| DrAttackTarget Model=DS-V3.22025.05 | 56.67 | 2.62 | 3.69 | |
| DrAttackTarget Model=Gemini-32025.05 | 39.33 | 2.89 | 1.81 | |
| FFATarget Model=Qwen32025.05 | 34.67 | 1.53 | 3.4 | |
| FFATarget Model=DS-V3.22025.05 | 30.67 | 1.47 | 3.77 | |
| AutoDANTarget Model=Qwen32025.05 | 29.33 | 1.75 | 3.57 | |
| DrAttackTarget Model=Qwen32025.05 | 27.33 | 1.59 | 3.1 | |
| DRATarget Model=Gemini-32025.05 | 20 | 1.62 | 3.27 | |
| JailbrokenTarget Model=Gemini-32025.05 | 14 | 1.35 | 2.85 | |
| ArtPromptTarget Model=DS-V3.22025.05 | 13.33 | 1.79 | 1.75 | |
| ArtPromptTarget Model=Gemini-32025.05 | 12 | 1.27 | 2.41 | |
| CipherTarget Model=DS-V3.22025.05 | 12 | 1.39 | 2.2 | |
| FFATarget Model=Gemini-32025.05 | 10.67 | 1.25 | 3.25 | |
| JailbrokenTarget Model=Qwen32025.05 | 10.67 | 1.24 | 2.69 | |
| CipherTarget Model=Qwen32025.05 | 7.33 | 1.23 | 2.89 | |
| ArtPromptTarget Model=Qwen32025.05 | 4.67 | 1.13 | 3.03 | |
| VanillaTarget Model=Gemini-32025.05 | 4 | 1.05 | 3.88 | |
| GCGTarget Model=Gemini-32025.05 | 3.33 | 1.05 | 3.36 | |
| GCGTarget Model=Qwen32025.05 | 2.67 | 1.05 | 3.52 | |
| DrAttackTarget Model=GPT-52025.05 | 2.67 | 1.05 | 4.03 | |
| GCGTarget Model=DS-V3.22025.05 | 2 | 1.03 | 3.6 | |
| AutoDANTarget Model=Gemini-32025.05 | 2 | 1.03 | 3.57 | |
| DRATarget Model=GPT-52025.05 | 2 | 1.13 | 4.09 | |
| CipherTarget Model=Gemini-32025.05 | 1.33 | 1.06 | 1.87 | |
| VanillaTarget Model=GPT-52025.05 | 0.67 | 1.01 | 4.4 | |
| JailbrokenTarget Model=GPT-52025.05 | 0.67 | 1.01 | 4.04 | |
| VanillaTarget Model=DS-V3.22025.05 | 0 | 1 | 3.56 | |
| VanillaTarget Model=Qwen32025.05 | 0 | 1 | 3.72 | |
| GCGTarget Model=GPT-52025.05 | 0 | 1.49 | 3.89 | |
| FFATarget Model=GPT-52025.05 | 0 | 1 | 2.97 | |
| AutoDANTarget Model=GPT-52025.05 | 0 | 1 | 4 | |
| AutoDANTarget Model=DS-V3.22025.05 | 0 | 1 | 3.41 | |
| ArtPromptTarget Model=GPT-52025.05 | 0 | 1 | 2.92 | |
| CipherTarget Model=GPT-52025.05 | 0 | 1 | 3.17 |