Constrained Question Generation on interrogative question generation (test)
43.7ROUGENEUROLOGIC*
Evaluation results
| Method | Links | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| NEUROLOGIC*Backbone=GPT-2, Search Strategy=greedy2021.12 | 43.7 | 14.7 | 28 | 20.9 | 47.7 | 100 | 2.83 | 2.77 | 2.74 | 2.76 | |
| NEUROLOGIC*Backbone=GPT-2, Search Strategy=sample2021.12 | 43.5 | 14.6 | 28.2 | 20.8 | 47.8 | 100 | 2.83 | 2.75 | 2.76 | 2.73 | |
| NEUROLOGIC*Backbone=GPT-2, Search Strategy=beam2021.12 | 42.9 | 14.4 | 27.8 | 20.3 | 46.9 | 100 | 2.81 | 2.86 | 2.76 | 2.75 | |
| TSMHBackbone=GPT-22021.12 | 42 | 4.3 | 25.9 | 10.4 | 37.7 | 92.7 | 2.35 | 2.28 | 2.37 | 2.22 | |
| NEUROLOGICBackbone=GPT-22021.12 | 38.8 | 11.2 | 24.5 | 18 | 41.7 | 90.6 | 2.78 | 2.71 | 2.49 | 2.51 | |
| CGMHBackbone=GPT-22021.12 | 28.8 | 2 | 18 | 5.5 | 21.5 | 18.3 | 2.28 | 2.34 | 2.11 | 2.02 |