Generalizing GNNs with Tokenized Mixture of Experts

TL;DR

This paper introduces STEM-GNN, a novel framework that enhances the robustness and generalization of graph neural networks under distribution shifts and perturbations by combining mixture-of-experts encoding, token stabilization, and Lipschitz regularization.

Contribution

The paper proposes STEM-GNN, a pretrain-then-finetune approach with innovative components to improve GNN robustness and generalization across diverse graph tasks and shifts.

Findings

STEM-GNN outperforms existing methods on nine benchmarks.

It achieves better robustness to degree, homophily shifts, and corruptions.

Maintains competitive performance on clean graphs.

Abstract

Deployed graph neural networks (GNNs) are frozen at deployment yet must fit clean data, generalize under distribution shifts, and remain stable to perturbations. We show that static inference induces a fundamental tradeoff: improving stability requires reducing reliance on shift-sensitive features, leaving an irreducible worst-case generalization floor. Instance-conditional routing can break this ceiling, but is fragile because shifts can mislead routing and perturbations can make routing fluctuate. We capture these effects via two decompositions separating coverage vs selection, and base sensitivity vs fluctuation amplification. Based on these insights, we propose STEM-GNN, a pretrain-then-finetune framework with a mixture-of-experts encoder for diverse computation paths, a vector-quantized token interface to stabilize encoder-to-head signals, and a Lipschitz-regularized head to bound output amplification. Across nine node, link, and graph benchmarks, STEM-GNN achieves a stronger three-way balance, improving robustness to degree/homophily shifts and to feature/edge corruptions while remaining competitive on clean graphs.