# Knowledge-Graph Grounding Helps LLMs Only for Out-of-Training Knowledge: A Controlled Study on Clinical Question Answering

arXiv: https://arxiv.org/abs/2606.22419 · HTML: https://mlsanity.com/paper/2606.22419/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Placed in Medical and biomedical RAG, one of 15 directions in Retrieval-augmented generation.
- Builds on 3 of its 9 references that are in the map.
- First posted on arXiv 21 Jun 2026.

## Placed in

- Retrieval-augmented generation / [Medical and biomedical RAG](https://mlsanity.com/rag/medical-and-biomedical-rag/)

## Builds on (most cited first, 3 of 3)

- [What Disease does this Patient Have? A Large-scale Open Domain Question Answering Dataset from Medical Exams](https://mlsanity.com/paper/2009.13081/) (2020)
- [ClashEval: Quantifying the tug-of-war between an LLM's internal prior and external evidence](https://mlsanity.com/paper/2404.10198/) (2024)
- [Knowledge Graphs as the Missing Data Layer for LLM-Based Industrial Asset Operations](https://mlsanity.com/paper/2605.26874/) (2026)

## Built on by (most cited first, 1 of 1)

- [The Answer Path and the Grounding Instruction in LLM Question Answering over Knowledge Graphs](https://mlsanity.com/paper/2609.10237/) (2026)

## Abstract (arXiv)

A recent Nature Medicine study reports that general-purpose frontier LLMs outperform specialized retrieval-augmented clinical tools on medical benchmarks, and that retrieval can hurt strong models. We ask the natural follow-up: does structured knowledge-graph (KG) grounding change this, and when does grounding help at all? We contribute two results. First, a reproduction: the study's headline HealthBench score (~88) is the Consensus variant, not full HealthBench, where frontier models and ideal completions both score ~46-47 under a physician-calibrated grader (agreement 82.5%); we reproduce GPT-5.2 Consensus =90.9 and flag a score-deflating grader bug. Second, a knowledge-boundary result. Using a graph+vector engine (samyama-graph) over the public biomedical KG PrimeKG, neither naive triple retrieval nor an agentic natural-language-to-Cypher loop (82% successful queries) improves MedQA across a weak-to-strong model ladder (all |Delta| <= 3.4). On a synthetic counterfactual KG, and on a hybrid benchmark mixing known and novel facts, the identical pipeline lifts out-of-training accuracy from chance to ~100% (+68 to +79) while adding nothing on known facts (a no-LLM arm answers both). Across three regimes (no-knowledge, graph-aided, hybrid), grounding helps only insofar as the decisive fact lies outside the model's training -- public-KG facts are redundant, private and novel data are where it pays -- matching the study's institutional-data caveat.