Dialogue Planning on P4G (test)
3.12Average Turn (AT)UP-NRPA
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| UP-NRPABackbone=Qwen2.5 14B, level=22026.04 | 3.12 | 98.49 | |
| DialogXpertBackbone=Qwen2.5 14B2026.04 | 3.34 | 91.29 | |
| UP-NRPABackbone=Qwen2.5 14B, level=12026.04 | 3.4 | 91.84 | |
| DialogXpertBackbone=Qwen1-1.8B2026.04 | 3.97 | 87.93 | |
| UP-NRPABackbone=GPT-4o mini, level=22026.04 | 4.36 | 100 | |
| UP-NRPABackbone=GPT-4o mini, level=12026.04 | 4.91 | 97.47 | |
| DialogXpertBackbone=Vicuna 13B2026.04 | 5.07 | 81.32 | |
| LDPPBackbone=Qwen1-1.8B2026.04 | 5.57 | 79.5 | |
| LDPP w/o 2nd StageBackbone=Qwen1-1.8B2026.04 | 6.14 | 76 | |
| LDPP w/o 3rd StageBackbone=Qwen1-1.8B2026.04 | 6.84 | 57 | |
| UDPBackbone=GPT-4o mini2026.04 | 7.705 | 59.8 | |
| ProCoT2026.04 | 7.98 | 54.3 | |
| UDP w/o RLBackbone=GPT-4o mini2026.04 | 8 | 53.3 | |
| UDP w/o PTBackbone=GPT-4o mini2026.04 | 8.017 | 51.3 | |
| ICL-AIF2026.04 | 8.06 | 46.5 | |
| PPDPPBackbone=Vicuna 13B2026.04 | 8.185 | 46.3 | |
| TRIPBackbone=GPT3.52026.04 | 8.2 | 49.5 | |
| Standard2026.04 | 8.32 | 46.8 | |
| GDP-Zero2026.04 | 9.12 | 32.8 |