Text Generation On Emnlp2017 Wmt
평가 지표
BLEU-2
BLEU-3
BLEU-4
BLEU-5
평가 결과
이 벤치마크에서 각 모델의 성능 결과
모델 이름 | BLEU-2 | BLEU-3 | BLEU-4 | BLEU-5 | Paper Title | Repository |
|---|---|---|---|---|---|---|
| SeqGAN | 0.859 | 0.6015 | 0.4541 | 0.4498 | SeqGAN: Sequence Generative Adversarial Nets with Policy Gradient | |
| PPOGAN | 0.905 | 0.692 | 0.47 | 0.322 | Improving GAN Training with Probability Ratio Clipping and Sample Reweighting | |
| RankGAN | 0.778 | 0.478 | 0.411 | 0.463 | Adversarial Ranking for Language Generation | |
| LeakGAN | 0.956 | 0.819 | 0.627 | 0.498 | Long Text Generation via Adversarial Training with Leaked Information | |
| RelGAN | 0.881 | 0.705 | 0.501 | 0.319 | RelGAN: Relational Generative Adversarial Networks for Text Generation |
0 of 5 row(s) selected.