Dialogue commonsense evaluation on DECO (test)
0.3Pearson CorrelationACCENT
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| ACCENTbackbone=T5-base, training_data=DECO training set2023.05 | 0.3 | 0.3 | |
| Cross-encoderbackbone=COMET, training_data=DECO training set2023.05 | 0.17 | 0.17 | |
| Cross-encodertraining_data=DECO training set2023.05 | 0.15 | 0.15 | |
| MLP Regressorfeatures=DialoGPT + ConceptNet, training_data=DECO training set2023.05 | 0.11 | 0.01 | |
| FED-understandabletype=off-the-shelf2023.05 | -0.12 | -0.07 | |
| FED-appropriatetype=off-the-shelf2023.05 | -0.16 | -0.1 |