Proactive Dialogue Evaluation on DuRecDial ID 2.0 (test)
3.86ProactivityQwen 14B
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| Qwen 14Bmodel_size=14B, protocol=prompt-based2026.05 | 3.86 | 4.91 | 4.85 | 3.83 | |
| Qwen 32Bmodel_size=32B, protocol=prompt-based2026.05 | 3.86 | 4.91 | 4.85 | 4.01 | |
| LLaMA 8Bmodel_size=8B, protocol=prompt-based2026.05 | 3.74 | 4.82 | 4.73 | 4.17 | |
| Ours 0.3Bmodel_size=0.3B2026.05 | 2.45 | 4.41 | 4.3 | 2.16 | |
| Qwen 3Bmodel_size=3B, protocol=fine-tuning2026.05 | 2.21 | 4.3 | 4.31 | 2.05 | |
| LLaMA 1Bmodel_size=1B, protocol=fine-tuning2026.05 | 2.12 | 4.32 | 4.25 | 2.07 | |
| LLaMA 3Bmodel_size=3B, protocol=fine-tuning2026.05 | 2.08 | 4.23 | 4.18 | 1.98 |