Same Graph Cross-Task Transfer in GNNs: Protocols and Predictors

2026-07-30

AI Summary

This work addresses the unreliability of existing graph neural network evaluations in cross-task transfer within the same graph—specifically between node classification (NC) and link prediction (LP)—which often suffer from data leakage and inconsistent data splits. The paper formally defines this transfer learning setting and introduces, for the first time, a leakage-free evaluation protocol that enforces fixed node/edge splits, shares a common message-passing graph (excluding test edges), and employs fixed negative sampling. Systematic experiments across backbone architectures such as GCN, GraphSAGE, and GPS reveal a strong directional bias in cross-task transfer: NC→LP consistently benefits homophilic graphs, whereas LP→NC is only advantageous in structure-dominated scenarios. Simple graph statistics like homophily effectively guide mechanism selection to avoid negative transfer, and the proposed CoTask Score reliably quantifies the multitask utility of shared encoders.

Abstract

Many real-world graphs support multiple predictive tasks over the same underlying structure, creating an opportunity to reuse supervision across node classification (NC) and link prediction (LP). However, existing evaluations often rely on incompatible splits, observed-graph assumptions, and negative sampling rules, making conclusions about same-graph cross-task transfer unreliable. We formalize same-graph NC-LP transfer and propose a leakage-free protocol that fixes node and edge splits, uses a shared message-passing graph that excludes evaluated edges, and employs fixed negatives for LP. Across three backbones (GCN, GraphSAGE, GPS), we find that transfer is strongly directional and predictable: NC $\to$ LP is consistently beneficial on homophilic graphs, while LP $\to$ NC is fragile and can even degrade accuracy under naive representation reuse. LP $\to$ NC becomes reliably positive mainly in a structure-dominant regime where LP is easy but NC is unsaturated, suggesting that LP acts as structural pretraining. Finally, we introduce the CoTask Score (CTS) to summarize joint NC+LP utility when a shared encoder must serve both tasks, and show that simple dataset statistics, especially homophily, can guide mechanism choice and help avoid negative transfer.

Problem

• cross-task transfer
• node classification
• link prediction
• graph neural networks
• homophily

Innovation

• cross-task transfer
• graph neural networks
• homophily
• leakage-free protocol
• CoTask Score