Can Graph Foundation Models Generalize Over Architecture?

TL;DR

This paper argues that graph foundation models need to adapt their architecture at inference time to truly generalize across diverse graph tasks, and introduces a framework for such adaptive architecture discovery.

Contribution

The paper proposes a novel framework for adaptive GNN architecture at inference time, enabling zero-shot generalization across heterogeneous graph tasks without retraining.

Findings

Adaptive architecture improves zero-shot generalization.

Fixed-backbone GFMs underperform on tasks with different architectural needs.

The approach enhances robustness on synthetic and real-world benchmarks.

Abstract

Graph foundation models (GFMs) have recently attracted interest due to the promise of graph neural network (GNN) architectures that generalize zero-shot across graphs of arbitrary scales, feature dimensions, and domains. While existing work has demonstrated this ability empirically across diverse real-world benchmarks, these tasks share a crucial hidden limitation: they admit a narrow set of effective GNN architectures. In particular, current domain-agnostic GFMs rely on fixed architectural backbones, implicitly assuming that a single message-passing regime suffices across tasks. In this paper, we argue that architecture adaptivity is a necessary requirement for true GFMs. We show that existing approaches are non-robust to task-dependent architectural attributes and, as a case study, use range as a minimal and measurable axis along which this limitation becomes explicit. With theoretical analysis and controlled synthetic experiments, we demonstrate that fixed-backbone GFMs provably under-reach on tasks whose architectural requirements differ from those seen at training time. To address this issue, we introduce a framework that adapts effective GNN architecture at inference time by discovering and mixing task-specific linear graph operators, enabling zero-shot generalization across tasks with heterogeneous architectural requirements, without retraining. We validate our approach on arbitrary-range synthetic tasks and a suite of real-world benchmarks, demonstrating improved performance and robustness over existing domain-agnostic GFMs.