AI research
A research paper titled "BEACON--SP: Ontology--Grounded GraphRAG Framework for Clinical Suicide Risk Assessment" (arXiv:2610.09026), published on October 6, 2026, introduces an ontology--grounded Graph Retrieval--Augmented Generation (GraphRAG) architecture tailored for clinician--facing decision support in behavioral healthcare. Designed to resolve standard vector retrieval limitations in multi--hop, heterogeneous reasoning, the framework anchors dynamic patient knowledge graphs to a comprehensive domain ontology synthesizing the Three--Step Theory, the Integrated Motivational--Volitional Model, and the Suicide Social Determinants of Health Ontology. Using a two--pass retrieval mechanism that merges cosine--similarity seed identification with breadth--first graph traversal, BEACON--SP enables structured multi--hop reasoning across diagnoses, medications, temporal trajectories, and clinical risk factors, achieving a 76.4% clinician preference rate over traditional vector RAG baselines across a 1,500--query benchmark.