Reading Comprehension on SQuAD (dev)
0.922F1 ScoreSpanBERT-base
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| SpanBERT-baseQuery dependency=Query-Dependent, Model scale=base2020.12 | 0.922 | 0.857 | |
| BERT-baseQuery dependency=Query-Dependent, Model scale=base2020.12 | 0.885 | 0.808 | |
| DensePhrasesQuery dependency=Query-Agnostic2020.12 | 0.863 | 0.783 | |
| DenSPI + SparcQuery dependency=Query-Agnostic, Model scale=BERT-large2020.12 | 0.848 | 0.764 | |
| QANetmodel_type=single model2018.11 | 0.827 | 0.736 | |
| DenSPIQuery dependency=Query-Agnostic2020.12 | 0.817 | 0.736 | |
| DECAPROPmodel_type=single model2018.11 | 0.814 | 0.729 | |
| Clark and GardnerModel=(1)2017.06 | 0.808 | 0.716 | |
| R-NET (Wang et al.)model_type=single model2018.11 | 0.806 | 0.723 | |
| DCN + CoVemodel_type=single model2018.11 | 0.799 | 0.713 | |
| Embedding RefinementModel=BiLSTM+p+q+A+W2017.06 | 0.797 | 0.695 | |
| R-NET (Our re-implementation)model_type=single model, implementation=re-implementation2018.11 | 0.796 | 0.719 | |
| DCNmodel_type=single model2018.11 | 0.759 | 0.662 | |
| DeFormerQuery dependency=Query-Agnostic2020.12 | 0.721 | — | |
| DilBERTQuery dependency=Query-Agnostic2020.12 | 0.72 | 0.63 |