Question Generation on SQuAD1.1
Metric: BLEU-4 (higher is better)
Metric: BLEU-4 (higher is better)
| # | Model↕ | BLEU-4▼ | Extra Data | Paper | Date↕ | Code |
|---|---|---|---|---|---|---|
| 1 | ERNIE-GENLARGE (beam size=5) | 25.41 | Yes | ERNIE-GEN: An Enhanced Multi-Flow Pre-training a... | 2020-01-26 | Code |
| 2 | BART (TextBox 2.0) | 25.08 | No | TextBox 2.0: A Text Generation Library with Pre-... | 2022-12-26 | Code |
| 3 | ProphetNet + ASGen | 24.44 | Yes | - | - | - |
| 4 | UniLMv2 | 24.43 | Yes | UniLMv2: Pseudo-Masked Language Models for Unifi... | 2020-02-28 | Code |
| 5 | ProphetNet + syn. mask + localness | 24.37 | No | Enhancing Pre-trained Models with Text Structure... | 2022-09-09 | - |
| 6 | ProphetNet | 23.91 | Yes | ProphetNet: Predicting Future N-gram for Sequenc... | 2020-01-13 | Code |
| 7 | UniLM + ASGen | 23.7 | Yes | - | - | - |
| 8 | UniLM | 22.78 | Yes | Unified Language Model Pre-training for Natural ... | 2019-05-08 | Code |
| 9 | BERTSQG | 22.17 | Yes | - | - | Code |
| 10 | Selector & NQG++ | 15.874 | No | Mixture Content Selection for Diverse Sequence G... | 2019-09-04 | Code |
| 11 | MPQG | 13.91 | No | - | - | Code |
| 12 | RNN +attn +copy | 13.5 | No | Evaluating Rewards for Question Generation Models | 2019-02-28 | Code |
| 13 | NQG++ | 13.27 | No | Neural Question Generation from Text: A Prelimin... | 2017-04-06 | Code |