Question Answering on Natural Questions (long)
Metric: F1 (higher is better)
Metric: F1 (higher is better)
| # | Model↕ | F1▼ | Extra Data | Paper | Date↕ | Code |
|---|---|---|---|---|---|---|
| 1 | DensePhrases | 79.6 | No | Learning Dense Representations of Phrases at Scale | 2020-12-23 | Code |
| 2 | Cluster-Former (#C=512) | 76.5 | No | Cluster-Former: Clustering-based Sparse Transfor... | 2020-09-13 | - |
| 3 | Locality-Sensitive Hashing | 75.5 | No | Reformer: The Efficient Transformer | 2020-01-13 | Code |
| 4 | Sparse Attention | 74.5 | No | Generating Long Sequences with Sparse Transformers | 2019-04-23 | Code |
| 5 | BERTwwm + SQuAD 2 | 68.2 | No | Frustratingly Easy Natural Question Answering | 2019-09-11 | - |
| 6 | BERTjoint | 64.7 | No | A BERT Baseline for the Natural Questions | 2019-01-24 | Code |
| 7 | DecAtt + DocReader | 54.8 | No | - | - | Code |
| 8 | DrQA | 46.1 | No | Reading Wikipedia to Answer Open-Domain Questions | 2017-03-31 | Code |