Safety Defense on Q-LatHarmful
6.52ASRQwen-2.5-7B-Instruct
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Qwen-2.5-7B-InstructBackbone=Qwen-2.5-7B-Instruct2026.05 | 6.52 | 1.22 | |
| SPARDBackbone=Qwen-2.5-7B-Instruct2026.05 | 7.33 | 1.29 | |
| LisaBackbone=Qwen-2.5-7B-Instruct2026.05 | 8.55 | 1.25 | |
| SafeGradBackbone=Qwen-2.5-7B-Instruct2026.05 | 27 | 1.77 | |
| PTSTBackbone=Qwen-2.5-7B-Instruct2026.05 | 83.5 | 4.11 | |
| SafeInstrBackbone=Qwen-2.5-7B-Instruct2026.05 | 85.74 | 4.34 | |
| SFTBackbone=Qwen-2.5-7B-Instruct2026.05 | 96.74 | 4.8 |