Task-Oriented Dialogue on MultiWOZ low-resource 2.1
29.52Average Precision (AP)VLK-RL
Evaluation results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| VLK-RLBackbone=Qwen-14B, Training Environment=Low-resource from scratch2026.04 | 29.52 | 23.51 | 20.58 | 11.05 | 12.17 | 19.48 | 24.21 | |
| VLK-RLBackbone=Qwen-7B, Training Environment=Low-resource from scratch2026.04 | 27.89 | 21.76 | 19.41 | 9.83 | 10.34 | 20.62 | 24.71 | |
| VLK-RLBackbone=GPT-4o-mini, Training Environment=Low-resource from scratch2026.04 | 26.51 | 20.53 | 18.27 | 9.12 | 8.9 | 21.35 | 25.1 | |
| CAPIDTraining Environment=Low-resource from scratch2026.04 | 21.46 | 16.59 | 14.73 | 6.18 | 5.24 | 23.1 | 26.4 | |
| GDP-ZeroTraining Environment=Low-resource from scratch2026.04 | 20.59 | 15.87 | 14.32 | 5.62 | 4.48 | 23.74 | 26.81 | |
| TransferTODTraining Environment=Low-resource from scratch2026.04 | 19.84 | 14.92 | 13.79 | 5.41 | 4.12 | 24.25 | 26.94 | |
| GALAXYTraining Environment=Low-resource from scratch2026.04 | 18.75 | 13.16 | 11.64 | 5.17 | 3.26 | 24.83 | 27.35 | |
| ACGOSTraining Environment=Low-resource from scratch2026.04 | 16.28 | 11.54 | 9.79 | 4.21 | 1.93 | 26.1 | 27.92 | |
| PPOTraining Environment=Low-resource from scratch2026.04 | 14.56 | 9.35 | 7.37 | 3.12 | 0 | — | 28.45 |