Abstract: Pretrained Language Models (PLMs) and Graph Neural Networks (GNNs) have emerged as promising approaches for software vulnerability detection. However, existing methods still face limitations ...