The proposed model fuses the multiple modalities of a function including text, graph, and image representations to detect function-level vulnerability - GuoXinrong/MVulD ...