On October 26, 2026, the research paper titled “Knowledge Graphs Quality Assessment in the Era of LLMs: A Conceptual Framework” was published as a new preprint on the arXiv platform (ID 2610.31684v1), focusing on the growing need for large language models (LLMs) to rely on knowledge graphs (KGs) to deliver accurate, comprehensive, and traceable answers, and systematically exploring the theoretical foundations and practical approaches for assessing KG quality.
This paper reviews the literature on quality assessment in the domains of data quality, information quality, linked data, knowledge graphs, and the intersection of KGs and LLMs, extending existing state-of-the-art frameworks into a taxonomy that encompasses 24 quality dimensions (QDs) and multiple quality metrics (QMs). It further proposes a general, domain- or task‑customizable, and practically applicable framework for evaluating KG quality. This framework seeks to address key questions: How is KG quality defined? In what scenarios can it be employed? And can it serve as a reliable knowledge‑grounding foundation for LLMs?
It should be noted that this work has so far been made publicly available only as an arXiv preprint and has not yet undergone peer review; the 24-dimensional quality framework and its specific indicator details presented in the paper are not elaborated in the abstract, and their validity and applicability remain to be verified by subsequent empirical studies. The paper was published at 04:00:00 UTC on October 26, 2026, which corresponds to October 26, 2026, in Singapore time, making it one of the latest academic developments of the day.