Seq2Seq on QQP
66ROUGE-LLD4LG
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| LD4LGSampling=Oracle-5, Backbone=BART-base2022.12 | 66 | 68 | 46.6 | 87.2 | |
| DiffuSeqSampling=Oracle-52022.12 | 65.8 | 67.4 | 43.9 | 83.7 | |
| LD4LGSampling=Oracle-5, Backbone=FLAN-T5-Base2022.12 | 65.7 | 67.8 | 46 | 87.2 | |
| LD4LGSampling=MBR-5, Backbone=BART-base2022.12 | 61.1 | 63.3 | 40.3 | 86.2 | |
| LD4LGSampling=MBR-5, Backbone=FLAN-T5-Base2022.12 | 60.7 | 63 | 39.7 | 86.1 | |
| FLAN-T5-BaseSampling=Beam, Model Type=Fine-tuned LM2022.12 | 60.5 | 63 | 40.1 | 86.2 | |
| LD4LGSampling=Random, Backbone=BART-base2022.12 | 60.3 | 62.6 | 39 | 85.8 | |
| LD4LGSampling=Random, Backbone=FLAN-T5-Base2022.12 | 59.7 | 62.1 | 38.4 | 85.8 | |
| RDM-absorbingSampling=MBR-102022.12 | 59.5 | — | — | 84.7 | |
| BART-BaseSampling=Beam, Model Type=Fine-tuned LM2022.12 | 59.5 | 61.9 | 39 | 85.5 | |
| DiffusionBERTSampling=MBR-102022.12 | 58.9 | — | — | — | |
| DiffuSeqSampling=MBR-102022.12 | 58.8 | — | — | 83.7 | |
| RDM-multinomialSampling=MBR-102022.12 | 58.5 | — | — | 84.7 | |
| RDM-absorbingSampling=Random2022.12 | 57.9 | — | — | 83.7 | |
| RDM-multinomialSampling=Random2022.12 | 57.3 | — | — | 83.7 | |
| DiffuSeqSampling=Random2022.12 | 52.7 | 55.2 | 29.2 | 82.4 | |
| FLAN-T5-BaseSampling=Nucleus, Model Type=Fine-tuned LM2022.12 | 52.3 | 55 | 30.1 | 83.2 | |
| BART-BaseSampling=Nucleus, Model Type=Fine-tuned LM2022.12 | 48.3 | 51.5 | 28.1 | 79.9 |