Natural Language Generation on MultiWOZ 2.2 (test)
90InformTOATOD_base
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| TOATOD_baseBackbone=T5-base, Trainable Parameters=36M2023.05 | 90 | 79.8 | 17.04 | 101.94 | |
| MarsBackbone=T5-small, Trainable Parameters=60M2023.05 | 88.9 | 78 | 19.9 | 103.4 | |
| MTTODBackbone=T5-base, Trainable Parameters=360.9M2023.05 | 85.9 | 76.5 | 19 | 100.2 | |
| TOATOD_smallBackbone=T5-small, Trainable Parameters=7.9M2023.05 | 85.8 | 74 | 18 | 97.9 | |
| BORTBackbone=T5-small, Trainable Parameters=60M2023.05 | 85.5 | 77.4 | 17.9 | 99.4 | |
| GALAXYBackbone=UniLM, Trainable Parameters=340M2023.05 | 85.4 | 75.7 | 19.64 | 100.2 | |
| RSTODBackbone=T5-small, Trainable Parameters=105.5M2023.05 | 83.5 | 75 | 18 | 97.3 | |
| UBARBackbone=GPT2, Trainable Parameters=1.5B2023.05 | 83.4 | 70.3 | 17.6 | 94.4 | |
| SOLOISTBackbone=GPT2, Trainable Parameters=1.5B2023.05 | 82.3 | 72.4 | 13.6 | 90.9 | |
| MinTLBackbone=BART-large, Trainable Parameters=440M2023.05 | 73.7 | 65.4 | 19.4 | 89 |