Research map / Large language models / Retrieval and knowledge
Assessing Automated Fact-Checking for Medical LLM Responses with Knowledge Graphs
Shasha Zhou, Mingyu Huang, Jack Cole, Charles Britton, Ming Yin, Jan Wolber, Ke Li
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use KG-based FAITH scoring instead of NLP and LLM-based metrics | automated factuality metric · MS-AKT, clinician judgments (16 questions, 20 clinicians) | Pearson correlation with clinicians: BLEU-4 0.081 -> FAITH 0.696 | Improved |
| use FAITH to trigger Reject-to-Answer instead of model uncertainty | GPT-4o · MedQA | accuracy | Improved |
| add KG-retrieval RAG triggered by FAITH | GPT-4o · MedQA | factuality of justification | Improved |
| add multi-round conversation to claim extraction prompt (ablation) | FAITH claim extractor · BioRED and DDI13 | recall | Improved |
| add critical analysis prompt to claim extraction (ablation) | FAITH claim extractor · BioRED and DDI13 | precision | Improved |
| replace GPT-4o extractor with Llama 3 or Llama 3.1 (ablation) | FAITH claim extractor · BioRED and DDI13 | extraction performance | Hurt |
| replace GPT-4o extractor with GPT-4o-mini (ablation) | FAITH claim extractor · BioRED and DDI13 | extraction performance | No clear effect |
| delete 20% of nodes or insert 20% noisy edges in UMLS (ablation) | FAITH · UMLS perturbation | performance | Hurt |
| delete 20% of random edges in UMLS (ablation) | FAITH · UMLS perturbation | performance | No clear effect |
| swap UMLS for PrimeKG or OGBL-biokg (ablation) | FAITH · LLM ranking | relative ranking of LLMs | No clear effect |
Closest papers by meaning
- Reliable and Diverse Evaluation of LLM Medical Knowledge Mastery (ICLR 2025 · similarity 0.88)
- MedCalc-Bench: Evaluating Large Language Models for Medical Calculations (NeurIPS 2024 · similarity 0.88)
- ER-Reason: A Benchmark Dataset for LLM Clinical Reasoning in the Emergency Room (NeurIPS 2026 · similarity 0.88)
- Reasoning-Enhanced Healthcare Predictions with Knowledge Graph Community Retrieval (ICLR 2025 · similarity 0.88)
- MedAlign: A Clinician-Generated Dataset for Instruction Following with Electronic Medical Records (AAAI 2024 · similarity 0.88)
- Evaluating LLMs Across Multi-Cognitive Levels: From Medical Knowledge Mastery to Scenario-Based Problem Solving (ICML 2025 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
