Question Answering on SQuAD benign 1.0 (test)
93.8F1 ScoreFreeLB
Evaluation results
| Method | Links | ||
|---|---|---|---|
| FreeLBTraining protocol=Adversarial Training, Model backbone=RoBERTa_Large2020.10 | 93.8 | 87.3 | |
| InfoBERTTraining protocol=Adversarial Training, Model backbone=RoBERTa_Large2020.10 | 93.7 | 87 | |
| VanillaTraining protocol=Standard Training, Model backbone=RoBERTa_Large2020.10 | 93.5 | 86.9 | |
| InfoBERTTraining protocol=Standard Training, Model backbone=RoBERTa_Large2020.10 | 93.5 | 87 |