Persona-grounded Dialogue on FoCus (test)
0.98Coh-UniEvalDPO
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| DPOBackbone=Qwen2 5B2026.02 | 0.98 | 0.06 | 0.21 | 0.29 | |
| PersoDPOBackbone=Qwen2 5B2026.02 | 0.7 | 0.18 | 0.3 | 0.44 | |
| Llama3.1 8BMode=Benchmark2026.02 | 0.55 | -0.22 | 0.17 | 0.33 | |
| Mistral 7BMode=Benchmark2026.02 | 0.53 | -0.23 | 0.21 | 0.3 | |
| Qwen2 7BMode=Benchmark2026.02 | 0.37 | -0.31 | 0.17 | 0.29 |