Gender bias evaluation on SEAT
0.181SEAT 6CCPA
Evaluation Results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| CCPABackbone=BERT, prompt_template=(1, 1, 1)2023.07 | 0.181 | -0.317 | 0.104 | 0.633 | 0.142 | 0.115 | 0.249 | |
| CCPABackbone=ELEATRA, prompt_template=(1, 1, 1)2023.07 | 0.251 | 0.012 | 0.647 | 0.437 | 0.72 | 0.46 | 0.421 | |
| Sent-DebiasBackbone=BERT2023.07 | 0.336 | -0.314 | -0.624 | 0.514 | 0.391 | 0.436 | 0.436 | |
| Auto-DebiasBackbone=BERT2023.07 | 0.373 | -0.056 | 0.745 | 1.175 | 0.856 | 0.823 | 0.671 | |
| DropoutBackbone=LLaMa3-8b2026.05 | 0.38 | 0.38 | 0.31 | 0.4 | 0.48 | 0.58 | — | |
| CCPABackbone=DistilBERT, prompt_template=(1, 1, 1)2023.07 | 0.409 | -0.024 | 0.138 | -0.029 | -0.029 | 0.283 | 0.152 | |
| DebiasRAGBackbone=LLaMa3-8b2026.05 | 0.42 | 0.36 | 0.25 | 0.25 | 0.35 | 0.64 | — | |
| originalBackbone=LLaMa3-8b2026.05 | 0.48 | 0.11 | 0.25 | 0.25 | 0.4 | 0.64 | — | |
| CDABackbone=BERT2023.07 | 0.596 | -0.103 | -0.236 | 0.8 | 0.394 | 0.734 | 0.477 | |
| INLPBackbone=BERT2023.07 | 0.619 | -0.226 | 0.326 | 0.591 | 0.43 | 0.549 | 0.457 | |
| MABELBackbone=BERT2023.07 | 0.664 | 0.167 | 0.479 | 0.647 | 0.465 | 0.57 | 0.499 | |
| FairFilBackbone=BERT2023.07 | 0.683 | -0.14 | -0.616 | 0.839 | 0.049 | -0.501 | 0.471 | |
| ELEATRABackbone=ELEATRA2023.07 | 0.82 | 0.036 | 1.18 | 1.007 | 0.782 | 0.958 | 0.797 | |
| DropoutBackbone=BERT2023.07 | 0.912 | 0.121 | 0.321 | 0.857 | 0.777 | 0.867 | 0.642 | |
| BERTBackbone=BERT2023.07 | 0.932 | 0.09 | -0.124 | 0.937 | 0.783 | 0.858 | 0.621 | |
| DistilBERTBackbone=DistilBERT2023.07 | 1.38 | 0.446 | -0.179 | 1.242 | 0.837 | 1.217 | 0.883 |