Developing a purpose-built 13B Llama2-based LLM for medical conversations outperforms GPT-4 in accuracy and completeness.
