Regard Evaluation on BOLD
0.0663GenderFair-GCG
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| Fair-GCGBackbone=Qwen 2.5 7B, Setting=Fair-GCG2026.06 | 0.0663 | 0.0527 | 0.4098 | 0.3243 | 0.2137 | |
| VanillaBackbone=Qwen 2.5 7B, Setting=Vanilla2026.06 | 0.0641 | 0.0389 | 0.4063 | 0.3388 | 0.2324 | |
| Fair-GCGBackbone=Llama 3.1 8B, Setting=Fair-GCG2026.06 | 0.0342 | 0.0607 | 0.4039 | 0.3261 | 0.2021 | |
| VanillaBackbone=Llama 3.1 8B, Setting=Vanilla2026.06 | 0.0251 | 0.0721 | 0.3812 | 0.3277 | 0.2122 |