Safety Defense on LatHarmful
7.88ASRQwen-2.5-7B-Instruct
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Qwen-2.5-7B-InstructBackbone=Qwen-2.5-7B-Instruct2026.05 | 7.88 | 1.23 | |
| LisaBackbone=Qwen-2.5-7B-Instruct2026.05 | 7.88 | 1.26 | |
| SPARDBackbone=Qwen-2.5-7B-Instruct2026.05 | 7.88 | 1.31 | |
| SafeGradBackbone=Qwen-2.5-7B-Instruct2026.05 | 20 | 1.75 | |
| SafeInstrBackbone=Qwen-2.5-7B-Instruct2026.05 | 76.77 | 3.99 | |
| PTSTBackbone=Qwen-2.5-7B-Instruct2026.05 | 82.42 | 4.13 | |
| SFTBackbone=Qwen-2.5-7B-Instruct2026.05 | 91.92 | 4.6 |