Attribute-guided Dynamic Prompt Learning for Graph Neural Networks

Graph Neural Networks (GNNs) have achieved remarkable success in analyzing graph-structured data, with their performance dependent on the graph structure. However, the graph structure at test time often differs from that used during training. This phenomenon, known as graph structure shifts, leads to significant performance degradation in GNNs. Existing methods tackle this problem by improving the robustness of GNNs, but they often overlook representation deviation caused by structure shifts. To address this limitation, we propose an attribute-guided dynamic prompt learning model that generates prompt vectors to approximate the intrinsic information of nodes. With these prompt vectors, the trained GNNs are expected to maintain their performance under graph structure shifts. Unlike previous prompt-based methods that learn unified prompt vectors for all nodes, we obtain node-level prompts by encoding node attributes that provide the unique information. Given the diversity of shifted graph structures, we introduce a structure-aware adaptation mechanism that adjusts the prompt vectors based on the input graph. Furthermore, we apply gradient-based attacks to generate perturbed graphs, encouraging the model to generalize to unseen structures. Extensive experiments on multiple benchmark datasets demonstrate the effectiveness and robustness of our model.