JMLR-13B For MedQA

Large Language Models (LLMs) have demonstrated a remarkable potential in medical knowledge acquisition and question-answering. However, LLMs can potentially hallucinate and yield factually incorrect outcomes, even with domain-specific pretraining. Previously, retrieval augmented generation (RAG) has limited success in addressing hallucinations. Unlike previous methods in RAG where the retrieval model was trained separately from the LLM, we introduce JMLR (for Jointly trains LLM and information Retrieval (IR)) during the fine-tuning phase. The synchronized training mechanism enhances JMLR's ability to retrieve clinical guidelines and leverage medical knowledge to reason and answer questions and reduces the demand for computational resources. We evaluated JMLR on the important medical question answering application. Our experimental results demonstrate that JMLR-13B (70.5%) outperforms a previous state-of-the-art open-source model using conventional pre-training and fine-tuning Meditron-70B (68.9%) and Llama2-13B with RAG (54.9%) on a medical question-answering dataset. JMLR-13B (148 GPU hours) also trains much faster than Meditron-70B (42630 GPU hours). Through this work, we provide a new and efficient knowledge enhancement tool for healthcare, demonstrating the potential of integrating IR and LLM training for medical question-answering systems. The code, along with selected retrieval data that can be made public, is included in the supplementary material and will be made publicly accessible with CC-BY 4.0 license upon the paper's acceptance.

Model	Parameter	Open Access	MedQA	Amboss	MMLU	MedMCQA	Average
GPT-4	?	No	74.7	82.1	88.4	69.5	78.6
ChatGPT	?	No	50.2	49.1	69.4	51.0	54.9
Meditron	70B	Yes	60.7	76.4	73.6	65.1	68.9
RAG	13B	Yes	59.9	76.9	69.9	64.2	67.7
JLMR	13B	Yes	62.5	81.2	72.8	65.5	70.5