Natural Language Understanding on GLUE 2018 (test dev)
87.7MNLIRoBERTa
Evaluation results
| Method | Links | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| RoBERTaBackbone=RoBERTa-base2021.10 | 87.7 | 92.4 | 91.8 | 95.4 | 64.1 | 91.4 | 78.4 | 90.7 | 86.5 | |
| GEEPDescription=second-phase pre-training with gender equality prompt2021.10 | 87.7 | 92.4 | 91.7 | 95.4 | 50.5 | 89.8 | 68.7 | 89.9 | 83.3 | |
| SPPADescription=second-phase pre-training on gender-neutral data2021.10 | 87.2 | 92.4 | 91.3 | 94.7 | 38.9 | 88.8 | 60.2 | 88.3 | 80.2 |