A simplified one-stage domain adaptation protocol can effectively train a specialized Chinese medical language model, HuatuoGPT-II, that outperforms general language models and rivals proprietary models in the medical domain.
