Position: Graph Matching Systems Deserve Better Benchmarks
Indradyumna Roy, Saswat Meher, Eeshaan Jain, Soumen Chakrabarti, Abir De
Abstract
Data sets used in recent work on graph similarity scoring and matching tasks suffer from significant limitations. Using Graph Edit Distance (GED) as a showcase, we highlight pervasive issues such as train-test leakage and poor generalization, which have misguided the community's understanding and assessment of the capabilities of a method or model. These limitations arise, in part, because preparing labeled data is computationally expensive for combinatorial graph problems. We establish some key properties of GED that enable scalable data augmentation for training, and adversarial test set generation. Together, our analysis, experiments and insights establish new, sound guidelines for designing and evaluating future neural networks, and suggest open challenges for future research.
BibTeX
@inproceedings{
roy2025position,
title={Position: Graph Matching Systems Deserve Better Benchmarks},
author={Indradyumna Roy and Saswat Meher and Eeshaan Jain and Soumen Chakrabarti and Abir De},
booktitle={Forty-second International Conference on Machine Learning Position Paper Track},
year={2025},
url={https://openreview.net/forum?id=GYZLed4d3M}
}