Safety Evaluation on CatQA-en (test)
1.66Harmful Content RateAdvBench-R2J Alignment
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| AdvBench-R2J AlignmentAlignment Dataset=AdvBench-R2J2025.02 | 1.66 | 15.27 | |
| AdvBench AlignmentAlignment Dataset=AdvBench2025.02 | 1.84 | 23.09 | |
| w/o SFTAlignment Dataset=None2025.02 | 3.82 | 64 |