Dialogue Generation on Anthropic-HH (test)
69.07Average Preference ScoreCal-DPO
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Cal-DPO2024.12 | 69.07 | 73.52 | 64.61 | |
| DDPOTemperature=0.72026.02 | 68.72 | — | — | |
| DPO*Temperature=0.72026.02 | 68.52 | — | — | |
| DPO*Temperature=0.12026.02 | 68.02 | — | — | |
| CPO2024.12 | 67.86 | 72.3 | 63.39 | |
| DDPOTemperature=0.12026.02 | 67.05 | — | — | |
| DPO2024.12 | 64.34 | 69.32 | 59.35 | |
| DPOP2024.12 | 63.34 | 68.77 | 57.91 | |
| SLiC2024.12 | 61.62 | 65.52 | 57.71 | |
| DPO+NLL2024.12 | 60.31 | 65.34 | 55.28 | |
| IPO2024.12 | 59.16 | 63.19 | 55.12 | |
| DPOTemperature=0.72026.02 | 58.99 | — | — | |
| f-DPO2024.12 | 56.3 | 60.21 | 52.38 | |
| DPOTemperature=0.12026.02 | 55.3 | — | — | |
| SFTTemperature=0.72026.02 | 49.48 | — | — | |
| SFTTemperature=0.12026.02 | 44 | — | — |