Dialogue Generation on Judicial Dialogue Human Evaluation (test)
3.99CSFull Model
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| Full Model2026.05 | 3.99 | 4.53 | 4.32 | 4.63 | 4.37 | |
| Full ModelAblation=w/o Appraisal Agent2026.05 | 3.83 | 3.89 | 4.42 | 3.89 | 4.01 | |
| Full ModelAblation=w/o Goal Relevance2026.05 | 3.77 | 4.26 | 4.21 | 4.42 | 4.17 | |
| Full ModelAblation=w/o Novelty Reward2026.05 | 3.74 | 4.32 | 4.26 | 4.37 | 4.17 | |
| SaulLM-7B2026.05 | 3.73 | 3.21 | 4.05 | 3 | 3.5 | |
| Full ModelAblation=w/o Succinct Reward2026.05 | 3.45 | 4.05 | 4.11 | 4.05 | 3.92 |