Abstract: Graph neural networks, with their message-passing mechanism, can model complex structural dependencies in graphs and have demonstrated excellent performance in knowledge graph completion ...