An Instance-Level Profiling Framework for Graph-Structured Data

Tianqi Zhao, Russa Biswas, Megha Khosla

Track: Type E (Late-Breaking Abstracts)

Keywords: Data-centric graph machine learning, data visualization, graph machine learning evaluation

Abstract: Graph machine learning models often achieve similar overall performance yet behave differently at the node level—failing on different subsets of nodes with varying reliability. Standard evaluation metrics such as accuracy obscure these fine-grained differences, making it difficult to diagnose when and where models fail. We introduce NodePro, a node profiling framework that combines data-centric signals capturing feature dissimilarity, label uncertainty, and structural ambiguity with model-centric measures of prediction confidence and consistency to provide fine-grained insights into node-level behavior and failure modes.

Submission Number: 97