Multi-hop Text QA on HotpotQA Full v1.1 (train)
82.2F1 ScoreHGN
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| HGNBackbone=RoBERTa-Large2021.09 | 82.2 | — | — | — | |
| ReasonBERT_RBackbone=RoBERTa-base2021.09 | 78.1 | 94 | 98 | 64.8 | |
| ReasonBERT_BBackbone=BERT-base2021.09 | 77.2 | 93.8 | 97.8 | 63.4 | |
| SplinterBackbone=Splinter-base2021.09 | 76.5 | 94.1 | 97.9 | 62.5 | |
| RoBERTaBackbone=RoBERTa-base2021.09 | 76.3 | 93.1 | 97.5 | 62.9 | |
| SpanBERTBackbone=SpanBERT-base2021.09 | 76.3 | 93.6 | 97.7 | 62.9 | |
| SSPTBackbone=SSPT-base2021.09 | 75.4 | 93.9 | 97.9 | 61.5 | |
| HGNBackbone=BERT-base2021.09 | 74.8 | — | — | — | |
| BERTBackbone=BERT-base2021.09 | 71.9 | 92.4 | 96.9 | 57.9 |