End-to-end dialog modeling on MultiWOZ 2.0
94.3Inform ScoreESAinsTOD
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| ESAinsTODBase Model=Llama 2 7B2026.03 | 94.3 | 87.1 | 21.48 | 112.18 | |
| Llama 2 7B (FT)fine-tuned exclusively=true2026.03 | 92.8 | 84.9 | 20.95 | 109.8 | |
| SPACEfine-tuned individually=true2026.03 | 91.5 | 84.7 | 19.3 | 107.4 | |
| MTTOD2026.03 | 90.99 | 82.58 | 20.25 | 107.04 | |
| PPTODfine-tuned individually=true2026.03 | 89.2 | 79.4 | 18.62 | 102.92 | |
| ESAinsTOD (GT)Base Model=Llama 2 7B, ground-truth system actions=true2026.03 | 88.2 | 81.6 | 21.87 | 106.77 | |
| TOP+NOD2026.03 | 86.9 | 76.2 | 20.58 | 102.13 | |
| SOLOIST2026.03 | 85.5 | 72.9 | 16.54 | 95.74 | |
| UBARoracle dialogue states=true2026.03 | 85.1 | 71.02 | 16.21 | 94.27 | |
| MinTLoracle dialogue states=true2026.03 | 84.88 | 74.91 | 17.89 | 97.78 | |
| SimpleTOD2026.03 | 84.4 | 70.1 | 15.01 | 92.26 |