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

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 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 LP acts as structural pretraining. Finally, we introduce 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.

When studying networks like social platforms or academic citation databases, researchers often want to answer multiple questions at once: who are these people, and who will connect with each other? Typically, AI models for each question are trained separately, even when they share the same underlying network. We asked whether training a model on one question could automatically make it better at the other. Surprisingly, we found the answer is asymmetric: models trained to classify nodes consistently become better at predicting connections, but the reverse rarely works and can even make things worse. We also discovered that a simple property of the network, namely whether connected nodes tend to be similar to each other, can predict in advance which direction of knowledge sharing will help. Along the way, we propose a cleaner framework that prevents information leaking between training and testing.