Abstract: Counterfactual subgraphs explain graph neural networks (GNNs) by answering the question: “How would the prediction change if a certain subgraph were absent in the input instance?” The ...