Transfer Learning for Edge Classification on Dynamic Text-Attributed Graphs

2026-09-26

AI Summary

This study addresses the performance degradation of edge classification on dynamic textual attribute graphs under cross-domain distribution shifts by proposing a Spatio-Temporal Semantic Alignment (STSA) model. The method establishes a leave-one-domain-out evaluation protocol and integrates temporal difference features with node frequency characteristics. Furthermore, it leverages pretrained language model priors alongside contrastive semantic prediction objectives to achieve robust representation learning. Experimental results demonstrate that STSA significantly outperforms the Bag of Events baseline and existing state-of-the-art methods, achieving superior generalization performance for edge classification tasks on unseen domains.

Abstract

Learning transferable representations for dynamic text-attributed graphs (DyTAGs) requires models to capture underlying interaction dynamics that persist across domains. However, existing methods tend to overfit to domain-specific structural, temporal, and semantic patterns, limiting edge classification performance under distribution shifts. To expose and address this, we formally establish a leave-one-domain-out (LODO) transfer learning protocol for edge classification on DyTAGs. Under this protocol, we demonstrate that state-of-the-art self-supervised methods for dynamic graph learning perform poorly when transferred to unseen domains. Strikingly, existing methods underperform a structurally and temporally unaware Bag of Events (BoE) model we introduce, which inputs only unordered sequences of node and edge text features. Proceeding from the BoE, we propose Spatio-Temporal Semantic Alignment (STSA), which integrates a spatio-temporal encoder that fuses representations of time deltas and node occurrence frequencies into a unified manifold. STSA is trained with a Contrastive Semantic Forecasting objective, which anchors edge representations to a multi-domain textual latent space initialized by a pretrained language model, providing a robust prior that outperforms BoE and all existing methods we evaluate.

Problem

• Dynamic Text-Attributed Graphs
• Transfer Learning
• Edge Classification
• Distribution Shift
• Leave-One-Domain-Out

Innovation

• Dynamic Text-Attributed Graphs
• Transfer Learning
• Spatio-Temporal Semantic Alignment
• Contrastive Semantic Forecasting
• Bag of Events