Out-of-Distribution Generalization in Graph Foundation Models

TL;DR

The paper tackles out-of-distribution generalization (OOD) in Graph Foundation Models (GFMs) by formalizing distribution shifts along structure, domain, modality, and task, and by introducing a unified problem formulation with $G=(V,E,X,M)$, $p_{ ext{src}}$, $p_{ ext{tgt}}$, and latent factors $\Phi=(\Phi_{\text{struct}},\Phi_{\text{dom}},\Phi_{\text{mod}},\Phi_{\text{task}})$. It categorizes GFMs into homogeneous-task and heterogeneous-task families, reviewing representative approaches (e.g., multi-graph pretraining, domain alignment, invariant learning, and instruction-based adaptation) and their evaluation practices. The survey synthesizes common evaluation protocols across topology/feature shifts, cross-domain transfer, modality variation, and task changes, emphasizing that a multi-faceted evaluation is essential to capture OOD behavior. It also outlines open directions such as universal graph vocabularies, scaling laws, theory for graph distribution shifts, and standardized benchmarking to accelerate progress in robust, transferable GFMs.

Abstract

Graphs are a fundamental data structure for representing relational information in domains such as social networks, molecular systems, and knowledge graphs. However, graph learning models often suffer from limited generalization when applied beyond their training distributions. In practice, distribution shifts may arise from changes in graph structure, domain semantics, available modalities, or task formulations. To address these challenges, graph foundation models (GFMs) have recently emerged, aiming to learn general-purpose representations through large-scale pretraining across diverse graphs and tasks. In this survey, we review recent progress on GFMs from the perspective of out-of-distribution (OOD) generalization. We first discuss the main challenges posed by distribution shifts in graph learning and outline a unified problem setting. We then organize existing approaches based on whether they are designed to operate under a fixed task specification or to support generalization across heterogeneous task formulations, and summarize the corresponding OOD handling strategies and pretraining objectives. Finally, we review common evaluation protocols and discuss open directions for future research. To the best of our knowledge, this paper is the first survey for OOD generalization in GFMs.

Theorems & Definitions (1)

• Definition 1: OOD Generalization in Graph Foundation Models