Don’t Be Afraid Of Over-Smoothing And Over-Squashing

Over-smoothing and over-squashing have been extensively studied in the literature on Graph Neural Networks (GNNs) over the past years. In this paper, we challenge this prevailing focus in GNN research, arguing that these phenomena are less critical for practical applications than assumed. We find that over-smoothing rarely limits model performance in real-world datasets. We draw this conclusion from extensive experiments on standard community benchmark datasets, demonstrating that accuracy and over-smoothing are mostly uncorrelated and that optimal model depths remain small even with mitigation techniques, thus highlighting the negligible role of over-smoothing. Similarly, we challenge the idea that over-squashing is always detrimental in practical applications and question the relevance of information exchange between structural communities along bottleneck edges. The results of our experiments show that architectural interventions designed to mitigate over-squashing fail to yield significant performance gains. We call for a paradigm shift in theoretical research, urging a diligent analysis of learning tasks and datasets using statistics that measure the underlying distribution of label-relevant information to better understand their localisation and factorisation.