Dialogue Generation on Human Evaluation 100 dialogue contexts (test)
19Fluency (Win)DialogVED-Sampling
Evaluation results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| DialogVED-SamplingComparison Group=Group 2 (vs PLATO), Decoding=Sampling2022.04 | 19 | 13 | 22 | 24 | 31 | 14 | 24 | 17 | |
| DialogVEDComparison Group=Group 1 (vs PLATO), Decoding=Standard2022.04 | 16 | 11 | 38 | 16 | 13 | 11 | 26 | 14 |