Stereotype Fairness Identification on WinoBias cloze-style (test)
43.18P_stereoIG2
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| IG2Model=Llama-3.22026.02 | 43.18 | 43.43 | 13.39 | 0.25 | |
| BBAModel=Llama-3.12026.02 | 50.51 | 49.49 | 0 | 1.02 | |
| BBAModel=Llama-3.22026.02 | 51.52 | 48.48 | 0 | 3.04 | |
| FBAModel=Llama-3.12026.02 | 52.53 | 47.47 | 0 | 5.06 | |
| IG2Model=Llama-3.12026.02 | 57.32 | 42.68 | 0 | 14.64 | |
| Prefix PromptingModel=Llama-3.12026.02 | 58.08 | 41.92 | 0 | 16.16 | |
| FBAModel=Llama-3.22026.02 | 59.34 | 40.66 | 0 | 18.68 | |
| Auto-DebiasModel=Llama-3.12026.02 | 61.49 | 38.51 | 0 | 22.98 | |
| BaseModel=Llama-3.12026.02 | 62.63 | 37.37 | 0 | 25.26 | |
| Self-DebiasingModel=Llama-3.12026.02 | 85.61 | 14.39 | 0 | 71.22 | |
| Self-DebiasingModel=Llama-3.22026.02 | 88.64 | 11.11 | 0.25 | 77.53 | |
| BaseModel=Llama-3.22026.02 | 95.71 | 4.29 | 0 | 91.42 | |
| Prefix PromptingModel=Llama-3.22026.02 | 96.21 | 3.79 | 0 | 92.42 | |
| Auto-DebiasModel=Llama-3.22026.02 | 96.46 | 3.54 | 0 | 92.92 |