Jailbreak Attack on AdversarialBenchmark
0.512ASR (m)S-PIEL
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| S-PIELTarget Model Type=Mistral 7B2025.03 | 0.512 | 0.543 | |
| S-PIELTarget Model Type=Llama-2-7B2025.03 | 0.498 | 0.533 | |
| S-PIELTarget Model Type=Gemma-2-9B2025.03 | 0.498 | 0.506 | |
| S-PIELTarget Model Type=Llama-3.1-8B2025.03 | 0.38 | 0.402 | |
| S-PIELTarget Model Type=Llama-3.1-8B (DeepSeek-R1-Distilled)2025.03 | 0.354 | 0.368 | |
| GCGTarget Model Type=Mistral 7B2025.03 | 0.194 | 0.212 | |
| GCGTarget Model Type=Gemma-2-9B2025.03 | 0.146 | 0.155 | |
| GCGTarget Model Type=Llama-2-7B2025.03 | 0.12 | 0.16 | |
| GCGTarget Model Type=Llama-3.1-8B2025.03 | 0.056 | 0.067 | |
| Vanilla PromptTarget Model Type=Llama-2-7B2025.03 | 0 | 0 | |
| Vanilla PromptTarget Model Type=Llama-3.1-8B2025.03 | 0 | 0 | |
| Vanilla PromptTarget Model Type=Mistral 7B2025.03 | 0 | 0 | |
| Vanilla PromptTarget Model Type=Gemma-2-9B2025.03 | 0 | 0 | |
| Vanilla PromptTarget Model Type=Llama-3.1-8B (DeepSeek-R1-Distilled)2025.03 | 0 | 0 | |
| GCGTarget Model Type=Llama-3.1-8B (DeepSeek-R1-Distilled)2025.03 | 0 | 0 |