Jailbreak Safety Evaluation on PAP
2ASRSafePath
Evaluation Results
| Method | Links | |
|---|---|---|
| SafePathBackbone Model=DeepSeek-R1-Distill-Llama-8B2026.06 | 2 | |
| ST-DBackbone Model=DeepSeek-R1-Distill-Llama-8B2026.06 | 2 | |
| ST-SBackbone Model=DeepSeek-R1-Distill-Llama-70B2026.06 | 2 | |
| ST-DBackbone Model=DeepSeek-R1-Distill-Llama-70B2026.06 | 2 | |
| ST-DBackbone Model=Qwen3-8B2026.06 | 4 | |
| ST-SBackbone Model=DeepSeek-R1-Distill-Llama-8B2026.06 | 4 | |
| ST-DBackbone Model=Qwen3-32B2026.06 | 4 | |
| ST-SBackbone Model=Qwen3-8B2026.06 | 10 | |
| Star1Backbone Model=Qwen3-32B2026.06 | 10 | |
| SafePathBackbone Model=Qwen3-32B2026.06 | 10 | |
| ST-SBackbone Model=Qwen3-32B2026.06 | 10 | |
| BaseBackbone Model=Qwen3-8B2026.06 | 12 | |
| Star1Backbone Model=Qwen3-8B2026.06 | 12 | |
| SafePathBackbone Model=Qwen3-8B2026.06 | 12 | |
| BaseBackbone Model=Qwen3-32B2026.06 | 12 | |
| SafePathBackbone Model=DeepSeek-R1-Distill-Llama-70B2026.06 | 12 | |
| Star1Backbone Model=DeepSeek-R1-Distill-Llama-8B2026.06 | 14 | |
| Star1Backbone Model=DeepSeek-R1-Distill-Llama-70B2026.06 | 14 | |
| BaseBackbone Model=DeepSeek-R1-Distill-Llama-70B2026.06 | 18 | |
| BaseBackbone Model=DeepSeek-R1-Distill-Llama-8B2026.06 | 20 |