Dialogue on Dialogue (test)
8.84FluencySFT
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| SFTBackbone=LLaMA2-13B2024.06 | 8.84 | 7.77 | 8.49 | 7.43 | 7.31 | |
| HBATBackbone=LLaMA2-13B, Optimization=PPO2024.06 | 8.8 | 7.7 | 8.48 | 7.39 | 7.89 | |
| HBATBackbone=LLaMA2-13B, Optimization=DPO2024.06 | 8.79 | 7.8 | 8.45 | 7.51 | 7.96 | |
| DPOBackbone=LLaMA2-13B, Optimization=DPO2024.06 | 8.67 | 7.13 | 8.54 | 7.63 | 7.84 | |
| RLHFBackbone=LLaMA2-13B, Optimization=PPO2024.06 | 8.39 | 7.62 | 8.87 | 7.47 | 7.72 |