Domain-Specific Knowledge Graphs in RAG-Enhanced Healthcare LLMs
This work addresses the challenge of reliable healthcare information retrieval for medical professionals and researchers, but it is incremental as it builds on existing RAG methods with domain-specific adaptations.
The study tackled the problem of improving trustworthy domain-specific reasoning in healthcare LLMs by evaluating whether domain knowledge graphs enhance Retrieval-Augmented Generation, finding that precise, scope-matched retrieval yields consistent gains while indiscriminate graph unions reduce accuracy, with smaller models benefiting more from well-scoped retrieval.
Large Language Models (LLMs) generate fluent answers but can struggle with trustworthy, domain-specific reasoning. We evaluate whether domain knowledge graphs (KGs) improve Retrieval-Augmented Generation (RAG) for healthcare by constructing three PubMed-derived graphs: $\mathbb{G}_1$ (T2DM), $\mathbb{G}_2$ (Alzheimer's disease), and $\mathbb{G}_3$ (AD+T2DM). We design two probes: Probe 1 targets merged AD T2DM knowledge, while Probe 2 targets the intersection of $\mathbb{G}_1$ and $\mathbb{G}_2$. Seven instruction-tuned LLMs are tested across retrieval sources {No-RAG, $\mathbb{G}_1$, $\mathbb{G}_2$, $\mathbb{G}_1$ + $\mathbb{G}_2$, $\mathbb{G}_3$, $\mathbb{G}_1$+$\mathbb{G}_2$ + $\mathbb{G}_3$} and three decoding temperatures. Results show that scope alignment between probe and KG is decisive: precise, scope-matched retrieval (notably $\mathbb{G}_2$) yields the most consistent gains, whereas indiscriminate graph unions often introduce distractors that reduce accuracy. Larger models frequently match or exceed KG-RAG with a No-RAG baseline on Probe 1, indicating strong parametric priors, whereas smaller/mid-sized models benefit more from well-scoped retrieval. Temperature plays a secondary role; higher values rarely help. We conclude that precision-first, scope-matched KG-RAG is preferable to breadth-first unions, and we outline practical guidelines for graph selection, model sizing, and retrieval/reranking. Code and Data available here - https://github.com/sydneyanuyah/RAGComparison