Proactive Dialogue Evaluation on DuRecDial OOD 2.0 (test)
4.07ProactivityQwen 32B
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| Qwen 32Bmodel_size=32B, protocol=prompt-based2026.05 | 4.07 | 4.91 | 4.86 | 4.13 | |
| Qwen 14Bmodel_size=14B, protocol=prompt-based2026.05 | 4.02 | 4.92 | 4.83 | 3.83 | |
| LLaMA 8Bmodel_size=8B, protocol=prompt-based2026.05 | 3.78 | 4.82 | 4.69 | 4.03 | |
| Ours 0.3Bmodel_size=0.3B2026.05 | 2.71 | 4.33 | 3.98 | 1.98 | |
| Qwen 3Bmodel_size=3B, protocol=fine-tuning2026.05 | 2.35 | 4.29 | 4.09 | 1.88 | |
| LLaMA 3Bmodel_size=3B, protocol=fine-tuning2026.05 | 2.32 | 4.29 | 4.07 | 1.89 | |
| LLaMA 1Bmodel_size=1B, protocol=fine-tuning2026.05 | 2.31 | 4.33 | 3.97 | 1.89 |