Fact Extraction and Verification on FEVER (test)
75.96Label Accuracy (LA)KGAT (CorefRoBERTa_LARGE)
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| KGAT (CorefRoBERTa_LARGE)Backbone=CorefRoBERTa_LARGE2020.04 | 75.96 | — | — | 72.3 | |
| KGAT (CorefBERT_LARGE)Backbone=CorefBERT_LARGE2020.04 | 74.37 | — | — | 70.86 | |
| KGATBackbone=RoBERTa_LARGE2020.04 | 74.07 | — | — | 70.38 | |
| KGATBackbone=BERT_LARGE2020.04 | 73.61 | — | — | 70.24 | |
| CICD2021.05 | 73.1 | 49.7 | — | — | |
| KGAT (CorefBERT_BASE)Backbone=CorefBERT_BASE2020.04 | 72.88 | — | — | 69.82 | |
| KGATBackbone=BERT_BASE2020.04 | 72.81 | — | — | 69.4 | |
| Full system (single)Mode=single2019.09 | 72.56 | 74.62 | 67.26 | — | |
| SR-MRSBackbone=BERT2020.04 | 72.56 | — | — | 67.26 | |
| GEARBackbone=BERT2020.04 | 71.6 | — | — | 67.1 | |
| KGAT2021.05 | 71.6 | 48.5 | — | — | |
| BERT ConcatBackbone=BERT2020.04 | 71.01 | — | — | 65.64 | |
| GEAR2021.05 | 70.8 | 47.4 | — | — | |
| Nie (2019)2019.09 | 68.16 | 52.81 | 64.23 | — | |
| Yoneda (2018)2019.09 | 67.44 | 35.21 | 62.34 | — | |
| HAN2021.05 | 66.9 | 44.6 | — | — | |
| Hanselowski (2018)2019.09 | 65.22 | 37.33 | 61.32 | — | |
| NSMN2021.05 | 62.1 | 39.8 | — | — |