Task-oriented dialogue on MultiWOZ 2.0
91.8Inform RateLAVA
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| LAVA2021.02 | 91.8 | 81.8 | 12 | — | |
| Human2021.02 | 91 | 82.7 | — | — | |
| MD-SequicityVariant=Multi-domain setting2021.02 | 86.6 | 71.6 | 16.8 | — | |
| SOLOIST2021.02 | 85.5 | 72.9 | 16.5 | — | |
| SimpleTOD2021.02 | 84.4 | 70.1 | 15.1 | — | |
| AuGPTBackbone=GPT-2-small2021.02 | 83.1 | 70.1 | 17.2 | — | |
| DAMD2021.02 | 76.3 | 60.4 | 16.6 | — | |
| DarwinTODBackbone=Qwen3-8B, Setting=Zero-shot2026.01 | 75.25 | 70.96 | 0.1917 | 92.28 | |
| DarwinTODBackbone=Qwen2.5-7B, Setting=Zero-shot2026.01 | 74.72 | 69.12 | 0.19 | 90.92 | |
| DarwinTODBackbone=Llama3-8B, Setting=Zero-shot2026.01 | 74.18 | 68.26 | 0.1925 | 90.47 | |
| AgentTODBackbone=Llama3-8B, Setting=Zero-shot2026.01 | 67.5 | 45.98 | 0.1424 | 70.98 | |
| TOATODSetting=Zero-shot2026.01 | 10.59 | 5.21 | 0.0383 | 11.73 | |
| GALAXYSetting=Zero-shot2026.01 | 9.1 | 4.76 | 0.0361 | 10.54 |