A Additive feature attribution methods unify existing explainers for GNNs In this section, we analyze the vanilla gradient-based explainers and GNNExplainer [ 24