Natural Language Inference on ANLI (dev)
76.4R1 ScoreInfoBERT
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| InfoBERTTraining=Adversarial Training, Model=RoBERTa2020.10 | 76.4 | 51.7 | 48.6 | 58.3 | |
| InfoBERTTraining=Standard Training, Model=RoBERTa2020.10 | 75.2 | 49.6 | 47.8 | 56.9 | |
| FreeLBTraining=Adversarial Training, Model=RoBERTa2020.10 | 75.2 | 47.4 | 45.3 | 55.3 | |
| SMART_RoBERTa-LARGETraining Data=MNLI + SNLI + ANLI + FEVER2019.11 | 74.5 | 50.9 | 47.6 | 57.1 | |
| SMARTTraining=Adversarial Training, Model=RoBERTa2020.10 | 74.5 | 50.9 | 47.6 | 57.1 | |
| SMART_RoBERTa-LARGETraining Data=ANLI2019.11 | 74.2 | 49.5 | 49.2 | 57.1 | |
| VanillaTraining=Standard Training, Model=RoBERTa2020.10 | 74.1 | 50.8 | 43.9 | 55.5 | |
| ALUMTraining=Adversarial Training, Model=RoBERTa2020.10 | 73.3 | 53.4 | 48.2 | 57.7 | |
| ALUMTraining=Adversarial Training, Model=BERT2020.10 | 62 | 48.6 | 48.1 | 52.6 | |
| InfoBERTTraining=Adversarial Training, Model=BERT2020.10 | 60.8 | 48.7 | 45.9 | 51.4 | |
| FreeLBTraining=Adversarial Training, Model=BERT2020.10 | 60.3 | 47.1 | 46.3 | 50.9 | |
| InfoBERTTraining=Standard Training, Model=BERT2020.10 | 59.3 | 48.9 | 45.5 | 50.9 | |
| VanillaTraining=Standard Training, Model=BERT2020.10 | 58.5 | 46.1 | 45.5 | 49.8 |