Official implementation of 'Vision Graph Prompting via Semantic Low-Rank Decomposition'. The paper has been accepted by ICML 2025. Vision GNN (ViG) demonstrates superior performance by representing ...