Comparative Assessment on Topical-Chat
60.89Coherencehard BT-σ
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| hard BT-σaggregation=reliability-aware hard2026.02 | 60.89 | 53.45 | 67.36 | 58.9 | |
| soft BT2026.02 | 60.05 | 53.87 | 61.87 | 58.2 | |
| hard BT2026.02 | 59.31 | 50.25 | 62.57 | 56.9 | |
| BT-σaggregation=reliability-aware soft2026.02 | 59.02 | 56.3 | 63.49 | 60.56 | |
| BT-σ-asptype=aspect-dependent variant2026.02 | 58.94 | 54.92 | 65.58 | 60.71 | |
| Temp-BTtype=supervised baseline, calibration=temperature scaling2026.02 | 56.88 | 52.21 | 63.86 | 60.65 | |
| Avg-Prob2026.02 | 56.01 | 49.39 | 61.62 | 51.53 |