Graph Prompt

Graph prompting is a novel approach in graph representation learning that adapts pre-trained graph neural networks (GNNs) to new downstream tasks by adding "prompts"—small pieces of information—to the input graph, rather than retraining the entire model. Current research focuses on optimizing prompt design and placement, exploring reinforcement learning for efficient prompt selection, and addressing vulnerabilities to backdoor attacks. This paradigm offers improved data efficiency and adaptability compared to traditional fine-tuning, with applications spanning recommendation systems, biological network analysis, and other areas requiring efficient knowledge transfer in graph-structured data.