Lung-R1: A Knowledge Graph-Guided LLM for Pulmonary Diagnostic Reasoning

Lung-R1 is a novel large language model (LLM) designed for pulmonary diagnostic reasoning, leveraging the LungKG knowledge graph, which consists of 59,038 nodes and 164,308 edges across various entity and relation types. The model, with a size of 14 billion parameters, utilizes KG-constrained reasoning-chain construction and KG-guided reinforcement learning, achieving state-of-the-art results in a 20-system evaluation, particularly in EMR Diagnosis with a score of 4.3583, outperforming existing baselines. This advancement is significant for practitioners as it enhances the integration of structured medical knowledge with patient-specific data for improved diagnostic accuracy in pulmonary diseases.