Data-centric Prompt Tuning for Dynamic Graphs
Yufei Peng, Cheng Yang, Zhengjie Fan, Chuan Shi
TL;DR
This work tackles the challenge of adapting pre-trained dynamic graph embeddings to diverse downstream tasks under few-shot conditions. It proposes DDGPrompt, a data-centric prompting framework that refines the input via a unified node expression feature matrix and three prompt matrices (temporal bias, edge weight, feature mask) before feeding into a frozen backbone. A self-supervised contrastive pretraining stage complements the prompt design, and downstream tuning trains only the prompts and a classifier, blending task-agnostic representation with task-specific adaptation. Experimental results on four dynamic graph datasets show robust improvements over strong baselines and prior prompt methods, with favorable time/memory profiles and strong performance in sparse-data regimes. The approach offers a model-agnostic, efficient path to cross-task transfer and practical deployment in evolving graph settings.
Abstract
Dynamic graphs have attracted increasing attention due to their ability to model complex and evolving relationships in real-world scenarios. Traditional approaches typically pre-train models using dynamic link prediction and directly apply the resulting node temporal embeddings to specific downstream tasks. However, the significant differences among downstream tasks often lead to performance degradation, especially under few-shot settings. Prompt tuning has emerged as an effective solution to this problem. Existing prompting methods are often strongly coupled with specific model architectures or pretraining tasks, which makes it difficult to adapt to recent or future model designs. Moreover, their exclusive focus on modifying node or temporal features while neglecting spatial structural information leads to limited expressiveness and degraded performance. To address these limitations, we propose DDGPrompt, a data-centric prompting framework designed to effectively refine pre-trained node embeddings at the input data level, enabling better adaptability to diverse downstream tasks. We first define a unified node expression feature matrix that aggregates all relevant temporal and structural information of each node, ensuring compatibility with a wide range of dynamic graph models. Then, we introduce three prompt matrices (temporal bias, edge weight, and feature mask) to adjust the feature matrix completely, achieving task-specific adaptation of node embeddings. We evaluate DDGPrompt under a strict few-shot setting on four public dynamic graph datasets. Experimental results demonstrate that our method significantly outperforms traditional methods and prompting approaches in scenarios with limited labels and cold-start conditions.