Graph-Enhanced LLM Architecture for Reliable Data Management: Design Principles and Proof of Concept
- 21 May 2026
- conference paper
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-6
Abstract
Large Language Models (LLMs) demonstrate significant capabilities in natural language understanding, generation, and reasoning, yet their deployment in structured data contexts is hindered by hallucinations, limited explainability, and unreliable outputs. This paper presents a Graph-Enhanced LLM Architecture that integrates structured knowledge graphs with LLMs to provide traceable, auditable, and explainable responses for AI-assisted data management. The proposed architecture combines natural language interpretation, graph-grounded retrieval, schema-aware validation, and retrieval-augmented generation (RAG) in a modular pipeline. Beyond the conceptual architecture, the paper includes a proof-of-concept implementation and a small feasibility demonstration illustrating how graph grounding improves factual support and traceability for representative data management queries. The primary contribution lies in the architectural design principles and integration strategy for graph-enhanced LLM systems. A comprehensive large-scale empirical benchmark evaluation is left for future work.Keywords
This publication has 13 references indexed in Scilit:
- Graph Machine Learning in the Era of Large Language Models (LLMs)ACM Transactions on Intelligent Systems and Technology, 2025
- A Survey on Hallucination in Large Language Models: Principles, Taxonomy, Challenges, and Open QuestionsACM Transactions on Information Systems, 2025
- Survey of Hallucination in Natural Language GenerationACM Computing Surveys, 2023
- Knowledge-graph-based explainable AI: A systematic reviewJournal of Information Science, 2022
- Knowledge GraphsACM Computing Surveys, 2021
- On the Dangers of Stochastic Parrots2021
- Foundations of Modern Query Languages for Graph DatabasesACM Computing Surveys, 2017
- HoloCleanProceedings of the VLDB Endowment, 2017
- Detecting data errorsProceedings of the VLDB Endowment, 2016
- Big Data Integration2015