Jailbreak Defense Evaluation on WGT
2.595Strong-Reject Score (SR)vanilla
Evaluation Results
| Method | Links | |
|---|---|---|
| vanillaBase Model=Mistral-7b-U2026.02 | 2.595 | |
| vanillaBase Model=Llama2-7b-U2026.02 | 2.551 | |
| SCANSBase Model=Mistral-7b-U2026.02 | 2.519 | |
| SCANSBase Model=Llama2-7b-U2026.02 | 2.441 | |
| SAGEBase Model=Llama2-7b-U2026.02 | 2.425 | |
| SAGEBase Model=Mistral-7b-U2026.02 | 2.401 | |
| vanillaBase Model=Llama3.1-8b-D2026.02 | 2.219 | |
| SAGEBase Model=Llama3.1-8b-D2026.02 | 2.207 | |
| SCANSBase Model=Llama3.1-8b-D2026.02 | 2.164 | |
| ICDBase Model=Llama2-7b-U2026.02 | 2.163 | |
| SelfDefenDBase Model=Llama2-7b-U2026.02 | 2.158 | |
| InstructBase Model=Mistral-7b-U2026.02 | 2.148 | |
| SelfDefenDBase Model=Llama3.1-8b-D2026.02 | 1.993 | |
| SelfDefenDBase Model=Mistral-7b-U2026.02 | 1.813 | |
| ICDBase Model=Mistral-7b-U2026.02 | 1.749 | |
| ICDBase Model=Llama3.1-8b-D2026.02 | 1.744 | |
| AISABase Model=Llama2-7b-U2026.02 | 1.726 | |
| InstructBase Model=Llama3.1-8b-D2026.02 | 1.621 | |
| AISABase Model=Llama3.1-8b-D2026.02 | 1.414 | |
| AISABase Model=Mistral-7b-U2026.02 | 1.346 | |
| InstructBase Model=Llama2-7b-U2026.02 | 1.29 |