Predictions into Human-Interpretable Logical Rules

Neural Information Processing Systems 

These explainers, however, only explain an instance (e.g., a graph) and fail to uncover the combinatorial reasoning learned by a G