MedG-KRP: Medical Graph Knowledge Representation Probing

TL;DR

This paper introduces a knowledge graph-based method to evaluate and visualize the biomedical reasoning abilities of large language models, aiming to improve their reliability for clinical applications.

Contribution

It presents a novel approach using medical knowledge graphs to assess and interpret LLMs' reasoning in biomedical contexts, addressing limitations of traditional benchmarks.

Findings

GPT-4 performs best in human review but worst in ground truth comparison.

PalmyraMed-70b shows the opposite pattern.

The method enables visualization of LLMs' reasoning pathways in medicine.

Abstract

Large language models (LLMs) have recently emerged as powerful tools, finding many medical applications. LLMs' ability to coalesce vast amounts of information from many sources to generate a response-a process similar to that of a human expert-has led many to see potential in deploying LLMs for clinical use. However, medicine is a setting where accurate reasoning is paramount. Many researchers are questioning the effectiveness of multiple choice question answering (MCQA) benchmarks, frequently used to test LLMs. Researchers and clinicians alike must have complete confidence in LLMs' abilities for them to be deployed in a medical setting. To address this need for understanding, we introduce a knowledge graph (KG)-based method to evaluate the biomedical reasoning abilities of LLMs. Essentially, we map how LLMs link medical concepts in order to better understand how they reason. We test GPT-4, Llama3-70b, and PalmyraMed-70b, a specialized medical model. We enlist a panel of medical students to review a total of 60 LLM-generated graphs and compare these graphs to BIOS, a large biomedical KG. We observe GPT-4 to perform best in our human review but worst in our ground truth comparison; vice-versa with PalmyraMed, the medical model. Our work provides a means of visualizing the medical reasoning pathways of LLMs so they can be implemented in clinical settings safely and effectively.