Multi-hop Text QA on HotpotQA 1% v1.1 (train)
63.1F1ReasonBERT_R
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| ReasonBERT_RBackbone=RoBERTa-base2021.09 | 63.1 | 50.2 | |
| ReasonBERT_BBackbone=BERT-base2021.09 | 57.6 | 45.3 | |
| SplinterBackbone=Splinter-base2021.09 | 57 | 44.2 | |
| SpanBERTBackbone=SpanBERT-base2021.09 | 56.5 | 44.1 | |
| RoBERTaBackbone=RoBERTa-base2021.09 | 56 | 43.1 | |
| SSPTBackbone=SSPT-base2021.09 | 54.7 | 41.8 | |
| BERTBackbone=BERT-base2021.09 | 39.8 | 28.6 |