Malicious Query Safety Evaluation on JailBench
100Refusal RatioSSD
Evaluation Results
| Method | Links | |
|---|---|---|
| SSDBase Model=Llama32026.04 | 100 | |
| AdaCDBase Model=Llama32026.04 | 100 | |
| DefaultBase Model=Gemma22026.04 | 100 | |
| SSDBase Model=Gemma22026.04 | 100 | |
| AdaCDBase Model=Gemma22026.04 | 100 | |
| PromptBase Model=Qwen32026.04 | 100 | |
| SSDBase Model=Qwen32026.04 | 100 | |
| AdaCDBase Model=Qwen32026.04 | 100 | |
| DefaultBase Model=Llama32026.04 | 99 | |
| PromptBase Model=Llama32026.04 | 99 | |
| PromptBase Model=Gemma22026.04 | 99 | |
| DefaultBase Model=Qwen32026.04 | 99 | |
| SurgicalBase Model=Qwen32026.04 | 97 | |
| SurgicalBase Model=Llama32026.04 | 95 | |
| SurgicalBase Model=Gemma22026.04 | 95 | |
| SelfCDBase Model=Gemma22026.04 | 90 | |
| SelfCDBase Model=Qwen32026.04 | 88 | |
| SelfCDBase Model=Llama32026.04 | 85 |