ExAct
Existing Counterfactual techniques may not support user preferences to keep certain features unchanged. Can we improve the performance of Counterfactual generation process while ensurinf that user preferences are reflected in the produced counterfactuals?