GFMate: Empowering Graph Foundation Models with Test-time Prompt Tuning

2026-05-14

AI Summary

Existing graph prompting methods are hindered by their reliance on source-domain binding and pretraining, limiting their generalization to new domains and diverse graph foundation models, while also overlooking the potential of unlabeled test data. To address these limitations, this work proposes GFMate—the first pretraining-agnostic test-time graph prompting framework. GFMate introduces centroid prompting and layer prompting, coupled with a complementary learning objective that jointly optimizes both labeled and unlabeled test samples. By decoupling from source-domain constraints, the method substantially enhances cross-domain generalization, achieving performance gains of up to 30.63% on average across twelve benchmark datasets, while maintaining high efficiency and broad applicability.

Abstract

Graph prompt tuning has shown great potential in graph learning by introducing trainable prompts to enhance the model performance in conventional single-domain scenarios. Recent research has extended graph prompts to improve Graph Foundation Models (GFMs) by few-shot tuning auxiliary prompts. Despite their progress, most existing methods embed source-domain information into prompts, which serve either as input to GFMs or encoded during model pre-training. Such prompt entanglement with specific source domains and GFM pre-training strategy restricts their generalisability to other domains and different GFMs. Furthermore, existing GFM prompts merely rely on few-shot tuning for adaptation, neglecting the rich information in unlabelled target domain test data. Motivated by these insights, this paper aims to empower GFMs with pre-training-agnostic test-time graph prompt tuning, named GFMate. GFMate introduces centroid and layer prompts applied after pre-training on target domains, avoiding entanglement with specific source domains and model pre-training. In addition, a test-time complementary learning objective is devised to exploit both labelled and unlabelled target domain data for effective test-time prompt tuning. Extensive experiments on 12 benchmark datasets demonstrate the superior performance and efficiency of GFMate, achieving improvements of up to 30.63%. Code is available at https://github.com/YanJiangJerry/GFMate.

Problem

• Graph Foundation Models
• prompt tuning
• domain generalization
• test-time adaptation
• unlabeled data

Innovation

• test-time prompt tuning
• Graph Foundation Models
• pre-training-agnostic
• centroid prompts
• complementary learning