# Oversmoothing in GNN

> Source: https://sukruyusufkaya.com/en/glossary/oversmoothing-in-gnn
> Updated: 2026-05-13T21:07:54.608Z
> Type: glossary
> Category: derin-ogrenme
**TLDR:** A problem in which node representations become too similar after excessive message passing, reducing discriminative power.

<p>Oversmoothing is one of the central issues that frequently appears as graph neural networks become deeper. When nodes repeatedly exchange information with neighbors, their representations may become increasingly homogeneous and lose discriminative power. This explains why simply stacking more GNN layers does not always improve performance. A significant portion of modern GNN research focuses on mitigating this problem.</p>