Harmful Query on JailbreakB
0.67ASRNSPO
Evaluation Results
| Method | Links | |
|---|---|---|
| NSPOBackbone=Qwen2.5-7B-Instruct2025.12 | 0.67 | |
| NSPOBackbone=Llama3-8B-Instruct2025.12 | 1.33 | |
| BFPOBackbone=Qwen2.5-7B-Instruct2025.12 | 1.33 | |
| Llama3-8B-InstructBackbone=Llama3-8B-Instruct2025.12 | 2 | |
| MoCANBackbone=Qwen2.5-7B-Instruct2025.12 | 2 | |
| PeCANBackbone=Llama3-8B-Instruct2025.12 | 2.33 | |
| DPO-HBackbone=Qwen2.5-7B-Instruct2025.12 | 2.33 | |
| DPO-SBackbone=Llama3-8B-Instruct2025.12 | 2.67 | |
| MoCANBackbone=Llama3-8B-Instruct2025.12 | 2.67 | |
| BFPOBackbone=Llama3-8B-Instruct2025.12 | 2.67 | |
| DPO-SBackbone=Qwen2.5-7B-Instruct2025.12 | 2.67 | |
| Qwen2.5-7B-InstructBackbone=Qwen2.5-7B-Instruct2025.12 | 3 | |
| W-DOORBackbone=Qwen2.5-7B-Instruct2025.12 | 3.33 | |
| W-DOORBackbone=Llama3-8B-Instruct2025.12 | 4.33 | |
| DPO-MixBackbone=Qwen2.5-7B-Instruct2025.12 | 4.33 | |
| PeCANBackbone=Qwen2.5-7B-Instruct2025.12 | 5.67 | |
| DPO-MixBackbone=Llama3-8B-Instruct2025.12 | 6.67 | |
| SafeRLHFBackbone=Llama3-8B-Instruct2025.12 | 16.33 | |
| SafeRLHFBackbone=Qwen2.5-7B-Instruct2025.12 | 25 | |
| DPO-HBackbone=Llama3-8B-Instruct2025.12 | 47 |