In this technical report, we investigate the predictive performance differences of a rule-based approach and the GNN architectures NBFNet and A*Net with respect to knowledge graph completion. For the two most common benchmarks, we find that a substantial fraction of the performance difference can be explained by one unique negative pattern on each dataset that is hidden from the rule-based approach. Our findings add a unique perspective on the performance difference of different model classes for knowledge graph completion: Models can achieve a predictive performance advantage by penalizing scores of incorrect facts opposed to providing high scores for correct facts.
@article{arxiv.2412.05114,
title = {A*Net and NBFNet Learn Negative Patterns on Knowledge Graphs},
author = {Patrick Betz and Nathanael Stelzner and Christian Meilicke and Heiner Stuckenschmidt and Christian Bartelt},
journal= {arXiv preprint arXiv:2412.05114},
year = {2024}
}