Proactive Dialogue Generation on DuRecDial ID 2.0 (test)
1.28PPLQwen
Evaluation results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| QwenModel Size=32B, Evaluation Protocol=Prompt-based2026.05 | 1.28 | 26.47 | 19.2 | 9.9 | 2 | 11.3 | 37.19 | 3.8 | |
| QwenModel Size=14B, Evaluation Protocol=Prompt-based2026.05 | 1.33 | 26.51 | 19.1 | 9.7 | 2.1 | 12.5 | 42.66 | 2.28 | |
| LLaMAModel Size=8B, Evaluation Protocol=Prompt-based2026.05 | 1.51 | 26.19 | 19.9 | 10.9 | 1.9 | 11.4 | 35.99 | 3.42 | |
| LLaMAModel Size=3B, Evaluation Protocol=Fine-tuning2026.05 | 3.44 | 44.32 | 39.7 | 30 | 3.2 | 12.7 | 65.6 | 40.3 | |
| QwenModel Size=3B, Evaluation Protocol=Fine-tuning2026.05 | 3.47 | 43.07 | 38.4 | 28.3 | 3.3 | 13.9 | 63.77 | 33.65 | |
| LLaMAModel Size=1B, Evaluation Protocol=Fine-tuning2026.05 | 3.8 | 42.03 | 38.2 | 28.2 | 3.1 | 12.8 | 61.31 | 49.81 | |
| Our0.3B (soft)Inference Mode=Soft2026.05 | 3.93 | 44.87 | 41.6 | 31 | 2.2 | 8.7 | 61.17 | 19.77 | |
| Our0.3B (hard)Inference Mode=Hard2026.05 | 3.96 | 44.79 | 41.6 | 30.9 | 2.2 | 8.8 | 61.11 | 19.77 |