Table 6 Final training parameter settings
From: Evaluating clinical AI summaries with large language models as judges
Model | Strategy | Batch Size | Learning Rate | Beta |
|---|---|---|---|---|
Mixtral 8 × 22B | SFT | 16 | 1e-7 | – |
Mixtral 8 × 22B | DPO | 1 | 1e-7 | 0.7 |
Llama 3.1 8B | SFT | 32 | 1e-4 | – |
Llama 3.1 8B | DPO | 1 | 1e-7 | 0.5 |
