CFGNP Framework for Direct Counterfactual Prediction via GNNs

The Counterfactual Graph Neural Process (CFGNP) enables end-to-end counterfactual inference directly from observational data. This method bypasses the need to estimate an intermediate causal graph, preventing error propagation common in traditional two-step causal modeling approaches.