Single-Turn User Simulation on PEARL (test)
0.9373BertScoreUserSim-Qwen 8B
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| UserSim-Qwen 8BMode=Fine-tuned, Parameters=8B2026.03 | 0.9373 | 0.3093 | 30.7637 | |
| UserSim-Llama 8BMode=Fine-tuned, Parameters=8B2026.03 | 0.9364 | 0.3023 | 31.2219 | |
| Qwen3 32BMode=Zero-shot, Parameters=32B2026.03 | 0.8921 | 0.3409 | 31.6797 | |
| Llama3.1 70BMode=Zero-shot, Parameters=70B2026.03 | 0.887 | 0.2786 | 43.4067 |