Oversmoothing,"Oversquashing", Heterophily, Long-Range, and more: Demystifying Common Beliefs in Graph Machine Learning

2025-05-21

AI Summary

In graph machine learning, several widely adopted yet unverified core assumptions—such as oversmoothing, over-squashing, homophily-heterophily dichotomy, and long-range dependency confusion—lead to ill-defined problems and redundant, overlapping research directions. This paper systematically disentangles these long-conflated concepts for the first time, via rigorous theoretical analysis, formal conceptual modeling, and carefully constructed minimal counterexamples. Our contributions are threefold: (1) precise clarification of terminological semantics, dispelling prevalent empirical misconceptions; (2) advancement toward mathematically precise problem definitions and orthogonalization of research axes; and (3) establishment of a principled conceptual foundation and problem framework to guide interpretability analysis, generalization theory, and GNN architecture design. The work thus bridges critical gaps between intuition, formalism, and practice in graph representation learning.

Technology Category

• Machine Learning: Graph-based Machine Learning
• Data Mining & Knowledge Management: Graph Mining, Social Network Analysis & Community
• Reasoning under Uncertainty: Graphical Models

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Semantics and Knowledge: Methods, algorithms and applications for the development of semantic models, knowledge graphs and other forms of structured data models with machine-interpretable semantics
• Search and Retrieval-Augmented AI: Web query analysis, representation and understanding

Abstract

After a renaissance phase in which researchers revisited the message-passing paradigm through the lens of deep learning, the graph machine learning community shifted its attention towards a deeper and practical understanding of message-passing's benefits and limitations. In this position paper, we notice how the fast pace of progress around the topics of oversmoothing and oversquashing, the homophily-heterophily dichotomy, and long-range tasks, came with the consolidation of commonly accepted beliefs and assumptions that are not always true nor easy to distinguish from each other. We argue that this has led to ambiguities around the investigated problems, preventing researchers from focusing on and addressing precise research questions while causing a good amount of misunderstandings. Our contribution wants to make such common beliefs explicit and encourage critical thinking around these topics, supported by simple but noteworthy counterexamples. The hope is to clarify the distinction between the different issues and promote separate but intertwined research directions to address them.

Problem

• Clarifying misconceptions in graph machine learning beliefs
• Addressing oversmoothing and oversquashing issues in message-passing
• Distinguishing homophily-heterophily dichotomy and long-range tasks

Innovation

• Analyzing message-passing benefits and limitations
• Challenging common beliefs with counterexamples
• Promoting critical thinking in graph ML