Machine Reading Comprehension on SQuAD (dev)
91.2F1 Score (F1)+SRL (BERT_LARGE)
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| +SRL (BERT_LARGE)Backbone=BERT-Large, Additional features=SRL, Implementation=Our implementation2018.09 | 91.2 | 84.5 | 3.3 | |
| BERT_LARGEBackbone=BERT-Large, Status=Previous SOTA, Source=Published2018.09 | 90.9 | 84.1 | — | |
| Baseline (BERT_LARGE)Backbone=BERT-Large, Implementation=Our implementation2018.09 | 90.9 | 84.2 | — | |
| +SRL (BERT_BASE)Backbone=BERT-Base, Additional features=SRL, Implementation=Our implementation2018.09 | 88.8 | 81.7 | 2.6 | |
| BERT_BASEBackbone=BERT-Base, Source=Published2018.09 | 88.5 | 80.8 | — | |
| Baseline (BERT_BASE)Backbone=BERT-Base, Implementation=Our implementation2018.09 | 88.5 | 81.3 | — | |
| R.M. ReaderSource=Published2018.09 | 86.3 | 78.9 | — | |
| +SRL (ELMo)Backbone=ELMo, Additional features=SRL, Implementation=Our implementation2018.09 | 86 | 78.5 | 5.4 | |
| BiDAF+ELMoBackbone=BiDAF, Additional features=ELMo, Source=Published2018.09 | 85.6 | — | — | |
| Baseline (ELMo)Backbone=ELMo, Implementation=Our implementation2018.09 | 85.2 | 77.5 | — | |
| MQAN single-taskTask=single-task, Source=Published2018.09 | 75.5 | — | — | |
| MQAN multi-taskTask=multi-task, Source=Published2018.09 | 74.3 | — | — |