From Model to Data (M2D): Shifting Complexity from GNNs to Graphs for Transparent Graph Learning

2026-05-07

AI Summary

Although graph neural networks (GNNs) achieve strong performance, their black-box nature hinders understanding of the mechanisms underlying their success. To address this, this work proposes the Model-to-Data (M2D) distillation framework, which explicitly materializes the complexity of a teacher GNN into the data space for the first time. By leveraging knowledge distillation to guide graph structure enhancement and node feature augmentation, M2D constructs an enriched graph capable of encapsulating the teacher’s behavior, enabling a simple student model to replicate its performance. This approach preserves original accuracy while substantially improving model transparency, allowing direct observation and analysis of internal GNN mechanisms—such as attention-based aggregation and fairness objectives—and thereby establishing a new paradigm for interpretable graph learning.

Abstract

Graph Neural Networks (GNNs) achieve high performance but can be opaque to humans, making it difficult to understand and compare the many proposed architectures. While existing explainability methods attribute individual predictions to nodes, edges, or features, they do not provide architectural transparency or explain the fundamental performance gap between simple and more complex models. To address this limitation, we introduce Model-to-Data (M2D) distillation, a new framework that increases transparency by transferring model complexity into the data space. M2D distills the teacher model into an augmented graph with enriched features and structure, enabling a simple student to match the teacher's performance. By materializing model behavior in the data, our approach allows humans to inspect architectural advantages directly. We show that M2D reveals underlying mechanisms such as fairness objectives and attention-based aggregation in an interpretable way, enhancing GNN transparency while preserving performance.

Problem

• Graph Neural Networks
• Model Transparency
• Explainability
• Architectural Comparison
• Interpretability

Innovation

• Model-to-Data distillation
• Graph Neural Networks
• Interpretability
• Knowledge Distillation
• Transparent Graph Learning