Get Started
Home
Topics
Search
Library
Research questionHow can LLMs improve rare-disease diagnosis while preserving ontology-based evidence trails?Ontology rankers connect candidate diseases to matched patient phenotypes, whereas LLM-generated differentials lack an equally traceable basis. The central difficulty is using both signals without making the resulting diagnosis harder to inspect.
AI
Evaluation & Benchmarks
Health
Machine Learning
Natural Language Processing
Latest papersRecent research connected to this question, newest first.Learning to Fuse LLMs with Ontology Rankers for Rare-Disease DiagnosisThe evidence concerns phenotype-based rare-disease diagnosis using ontology rankers and LLM-generated ranked differentials. The reported fusion model uses ranked-list behavior, system agreement, and ontology support; it was assessed on Phenopacket Store and RAMEDIS after removing a documented benchmark leakage pathway. Results include experiments with eight open LLMs and an API-based pairing with DeepSeek-V4-Flash without retraining; the source reports candidate-level ontology evidence for 90.8% of correct fused diagnoses.research paper · Sep 2, 2026
Related questions
How can language models reason iteratively to diagnose complex clinical cases safely and accurately?How can multimodal medical diagnosis identify informative evidence within each modality without sacrificing accuracy?How can LLM-derived knowledge bases disambiguate homonymous entities and preserve auditable fact provenance?How can LLMs reliably standardize incomplete biomedical metadata when valid schemas and terms require runtime access?