MAGPrompt: Message-Adaptive Graph Prompt Tuning for Graph Neural Networks

2026-02-05

AI Summary

This work addresses the performance limitations of pre-trained graph neural networks (GNNs) when transferred to downstream tasks, which often arise from a mismatch between pre-training objectives and task-specific requirements. Existing graph prompting methods struggle to effectively modulate neighbor interactions during message passing. To overcome this, we propose Message-adaptive Graph Prompting (MAGP), a novel approach that, for the first time, directly embeds learnable prompts into the GNN message-passing mechanism. With the backbone parameters frozen, MAGP employs task-specific prompt vectors to reweight incoming neighbor messages, enabling fine-grained and task-adaptive information aggregation. The method is compatible with mainstream GNN architectures and pre-training strategies, significantly outperforming existing graph prompting techniques in few-shot settings and matching the performance of full-parameter fine-tuning under full-data conditions.

Technology Category

• Natural Language Processing: Prompt Engineering / Prompting
• Machine Learning: Graph-based Machine Learning
• Data Mining & Knowledge Management: Graph Mining, Social Network Analysis & Community

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Pre-trained graph neural networks (GNNs) transfer well, but adapting them to downstream tasks remains challenging due to mismatches between pre-training objectives and task requirements. Graph prompt tuning offers a parameter-efficient alternative to fine-tuning, yet most methods only modify inputs or representations and leave message passing unchanged, limiting their ability to adapt neighborhood interactions. We propose message-adaptive graph prompt tuning, which injects learnable prompts into the message passing step to reweight incoming neighbor messages and add task-specific prompt vectors during message aggregation, while keeping the backbone GNN frozen. The approach is compatible with common GNN backbones and pre-training strategies, and applicable across downstream settings. Experiments on diverse node- and graph-level datasets show consistent gains over prior graph prompting methods in few-shot settings, while achieving performance competitive with fine-tuning in full-shot regimes.

Problem

• graph neural networks
• prompt tuning
• message passing
• downstream task adaptation
• parameter-efficient learning

Innovation

• message-adaptive prompting
• graph prompt tuning
• parameter-efficient adaptation
• message passing
• pre-trained GNNs