Dialogue Response Generation on ESC (test)
88AccuracyMixed-Initiative Dialogue Prompting
Evaluation results
| Method | Links | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| Mixed-Initiative Dialogue PromptingStrategy=Prompt, Backbone=InstructGPT2023.05 | 88 | 3.72 | 3.8 | 3.81 | 90 | 91 | 3.19 | 0.52 | 64 | — | |
| Ground TruthStrategy=GT2023.05 | 85 | 3.57 | 3.6 | 3.61 | 90 | 90 | 3.03 | 0.56 | — | 36 | |
| Oracle-BlenderBotStrategy=Fine-tuning (FT)2023.05 | 81 | 3.57 | 3.63 | 3.55 | 89 | 87 | 3.25 | — | 44 | 48 |