AI research

A research paper titled "Ontology--guided multi--agent reasoning for explainable differential diagnosis" (arXiv:2610.36325), published on October 26, 2026, by researchers at the Technical University of Cluj--Napoca, presents an ontology--grounded multi--agent architecture designed to overcome black--box pattern matching and hallucination in LLM--based clinical reasoning. While large language models demonstrate proficiency in medical question answering, they frequently fail at formal clinical deduction and generate plausible yet unverified conclusions. To enforce logical rigor, the proposed framework grounds diagnostic inference in an expert--curated medical ontology and decomposes clinical inquiry into auditable steps executed by specialized agents—including symptom extraction, active follow--up interviewing to target discriminative symptoms, and iterative diagnosis refinement. By constraining generative models to evidence--backed ontological associations, the system delivers structured, traceable, and explainable differential diagnostic reasoning.