Preference Evaluation on HH-Harmless
60Win RateDLMA-13B
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| DLMA-13BComparison Baseline=Llama2-13B2024.02 | 60 | 15 | 25 | |
| DLMA-7BComparison Baseline=RLAIF-7B2024.02 | 59 | 21 | 20 | |
| DLMA-7BComparison Baseline=Llama2-7B2024.02 | 58 | 19 | 23 | |
| DLMA-13BComparison Baseline=CD-13B2024.02 | 55 | 16 | 29 | |
| DLMA-13BComparison Baseline=RLAIF-13B2024.02 | 52 | 14 | 34 | |
| DLMA-7BComparison Baseline=CD-7B2024.02 | 51 | 22 | 27 | |
| DLMA-13BComparison Baseline=RLCD-13B2024.02 | 49 | 20 | 31 | |
| DLMA-7BComparison Baseline=RLCD-7B2024.02 | 41 | 27 | 32 |