Metric: Rouge-1 (higher is better)
| # | Model↕ | Rouge-1▼ | Extra Data | Paper | Date↕ | Code |
|---|---|---|---|---|---|---|
| 1 | BART | 30.6 | No | BART: Denoising Sequence-to-Sequence Pre-trainin... | 2019-10-29 | Code |
| 2 | E-MCA | 30 | No | Using Local Knowledge Graph Construction to Scal... | 2019-10-18 | Code |
| 3 | Transformer Multitask + LayerDrop | 29.4 | No | Reducing Transformer Depth on Demand with Struct... | 2019-09-25 | Code |
| 4 | QG | 29.15 | No | Closed-book Question Generation via Contrastive ... | 2022-10-13 | Code |
| 5 | Multi-Inrerleave | 23.32 | No | Improving Conditioning in Context-Aware Sequence... | 2019-11-21 | - |