Abstract: Graph Neural Networks (GNNs) frequently face class imbalance issues, especially in heterogeneous graphs. Existing GNNs often assume balanced class sizes, which isn’t true in many cases.