Quantitative Attribution with Counterfactuals - Datasets

We include a set of datasets for classification tasks. Each of these datasets includes a pre-trained black-box classifier, which successfully solves the classification task. These datasets were used to show how Quantitative Attributions with Counterfactuals (QuAC) can help explain how the classifiers are solving these tasks. 

Several of these datasets are synthetic, made especially to show how certain features are difficult to explain with conventional attribution methods. Other datasets show the real-world applications of QuAC to solving difficult biological problems.