Jailbreak Defense Evaluation on XST
3.03Strong-Reject Scorevanilla
Evaluation Results
| Method | Links | |
|---|---|---|
| vanillaBase Model=Mistral-7b-U2026.02 | 3.03 | |
| SAGEBase Model=Mistral-7b-U2026.02 | 2.905 | |
| SAGEBase Model=Llama2-7b-U2026.02 | 2.115 | |
| ICDBase Model=Llama3.1-8b-D2026.02 | 1.945 | |
| vanillaBase Model=Llama2-7b-U2026.02 | 1.8 | |
| SCANSBase Model=Mistral-7b-U2026.02 | 1.79 | |
| ICDBase Model=Llama2-7b-U2026.02 | 1.71 | |
| vanillaBase Model=Llama3.1-8b-D2026.02 | 1.56 | |
| AISABase Model=Llama2-7b-U2026.02 | 1.455 | |
| SCANSBase Model=Llama3.1-8b-D2026.02 | 1.39 | |
| SelfDefenDBase Model=Llama2-7b-U2026.02 | 1.345 | |
| SCANSBase Model=Llama2-7b-U2026.02 | 1.32 | |
| SelfDefenDBase Model=Llama3.1-8b-D2026.02 | 1.31 | |
| AISABase Model=Llama3.1-8b-D2026.02 | 1.285 | |
| AISABase Model=Mistral-7b-U2026.02 | 1.275 | |
| InstructBase Model=Mistral-7b-U2026.02 | 1.21 | |
| SelfDefenDBase Model=Mistral-7b-U2026.02 | 1.15 | |
| InstructBase Model=Llama3.1-8b-D2026.02 | 1.08 | |
| SAGEBase Model=Llama3.1-8b-D2026.02 | 1.07 | |
| ICDBase Model=Mistral-7b-U2026.02 | 1.065 | |
| InstructBase Model=Llama2-7b-U2026.02 | 1 |