Controllable Text Generation on AGNews (test)
16.94Output Perplexity (O-PPL)GPT2(raw)
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| GPT2(raw)Method Category=Zero-shot PLM2022.12 | 16.94 | 74.41 | 52.17 | 46.88 | 45.55 | |
| PFMethod Category=Lightweight method2022.12 | 20.27 | 32.35 | 68.44 | 59.17 | 32.73 | |
| Ctr-PFMethod Category=Lightweight method2022.12 | 20.41 | 33.9 | 83.21 | 60.34 | 31.2 | |
| GPT2Method Category=Finetuned PLM2022.12 | 22.22 | 23.46 | 83.08 | 54.93 | 39.93 | |
| Ground TruthMethod Category=Reference2022.12 | 31.04 | — | 94.89 | 67.24 | 23.31 | |
| PT(noise)+PLMethod Category=Self-Training with PLM, Backbone PLM=UniLM, Pseudo Labeler=BERT-base2022.12 | 32.36 | 16.64 | 89.7 | 53.79 | 47.95 | |
| PT(select)+PLMethod Category=Self-Training with PLM, Backbone PLM=UniLM, Pseudo Labeler=BERT-base2022.12 | 33.23 | 16.66 | 90.52 | 53.71 | 47.69 | |
| DuNSTMethod Category=Our method2022.12 | 34.73 | 33.58 | 93.59 | 59.42 | 37.02 | |
| T5Method Category=Finetuned PLM2022.12 | 48.33 | 32.12 | 88.43 | 58.06 | 37.01 | |
| UniLMMethod Category=Finetuned PLM2022.12 | 55.79 | 36.28 | 87.7 | 54.76 | 43.77 | |
| PTMethod Category=Self-Training with PLM, Backbone PLM=UniLM2022.12 | 57.4 | 40.95 | 86.36 | 52.35 | 46.41 | |
| PT(noise)Method Category=Self-Training with PLM, Backbone PLM=UniLM2022.12 | 58.59 | 45.32 | 85.27 | 53.35 | 46.57 |