Emotional Support Dialogue on ESConv (test)
-12.1EngagingnessBBMHR: davinci
Evaluation Results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| BBMHR: davinciExpert Backbone=davinci2023.05 | -12.1 | -1.96 | 1.26 | -8.6 | -7.09 | 0.91 | -10.93 | |
| BBMHR: GPT1Expert Backbone=GPT12023.05 | -23.17 | -9.89 | -12.51 | -18.48 | -20.07 | -10.43 | -26.2 | |
| BBMHR: adaExpert Backbone=ada2023.05 | -24.02 | -7.04 | -7.16 | -11.52 | -15.59 | -2.48 | -19.41 | |
| BBMHR: GPT2Expert Backbone=GPT22023.05 | -24.82 | -8.15 | -3.64 | -14.02 | -19.65 | -9.21 | -22.33 | |
| BBMHMethod=BlenderBot with Mixed Humans2023.05 | -26.07 | -21.6 | -11.95 | -10.53 | -22.9 | -12.47 | -30.19 | |
| in-context davinciBackbone=davinci, Prompting=in-context2023.05 | -35.87 | -28.89 | -24.29 | -14.33 | -29.65 | -24.29 | -47.3 | |
| BBMethod=BlenderBot2023.05 | -36.78 | -22.92 | -15.67 | -28.91 | -30.15 | -17.64 | -42.68 |