Metric: Rouge-L (higher is better)
| # | Model↕ | Rouge-L▼ | Extra Data | Paper | Date↕ | Code |
|---|---|---|---|---|---|---|
| 1 | Masque (NarrativeQA + MS MARCO) | 59.87 | Yes | Multi-style Generative Reading Comprehension | 2019-01-08 | - |
| 2 | BERT-QA with Hard EM objective | 58.8 | No | A Discrete Hard EM Approach for Weakly Supervise... | 2019-09-11 | Code |
| 3 | Masque (NarrativeQA only) | 54.74 | No | Multi-style Generative Reading Comprehension | 2019-01-08 | - |
| 4 | ConZNet | 46.67 | No | - | - | - |
| 5 | DecaProp | 44.69 | No | Densely Connected Attention Propagation for Read... | 2018-11-10 | Code |
| 6 | MHPGM + NOIC | 44.16 | No | Commonsense for Generative Multi-Hop Question An... | 2018-09-17 | Code |
| 7 | BiAttention + DCU-LSTM | 41.44 | No | - | - | - |
| 8 | BiDAF | 36.74 | No | Bidirectional Attention Flow for Machine Compreh... | 2016-11-05 | Code |
| 9 | FiD+Distil | 32 | No | Distilling Knowledge from Reader to Retriever fo... | 2020-12-08 | Code |