A*Net and NBFNet Learn Negative Patterns on Knowledge Graphs
This provides a unique perspective on model performance differences for knowledge graph completion, though it is incremental as it explains existing results rather than introducing new methods.
The study investigated performance differences between rule-based and GNN-based models (A*Net and NBFNet) for knowledge graph completion, finding that a substantial fraction of the difference is explained by unique negative patterns hidden from rule-based approaches.
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.