A new paper introduces HEG-TKG, a system designed to address the "Provenance Gap" in clinical AI, where large language models often fabricate citations. The HEG-TKG system grounds clinical claims in temporal knowledge graphs derived from PubMed records and curated sources, ensuring 100% evidence verifiability. Evaluations showed that HEG-TKG matches baseline clinical feature coverage while providing verifiable citations, and it demonstrated significant resistance to injected clinical errors. The system is designed for on-premise deployment using open-source models to maintain patient data privacy. AI
Summary written by None from 1 source. How we write summaries →
IMPACT Enhances trust in clinical AI by ensuring verifiable citations, potentially accelerating adoption in healthcare settings.
RANK_REASON This is a research paper detailing a novel system for improving the verifiability of AI-generated clinical information. [lever_c_demoted from research: ic=1 ai=1.0]