Base model: Llama1-7B, trained on MedMCQA + PubMedQA

Full training

Acc: 0.44