Search papers, labs, and topics across Lattice.
The paper introduces CounterFlowNet, a generative method using conditional Generative Flow Networks (GFlowNets) to produce counterfactual explanations by sequentially modifying input features. CounterFlowNet addresses limitations of existing methods by generating multiple high-quality explanations that are sparse, applicable to heterogeneous tabular data, and consistent with user-defined constraints. Experiments on eight datasets demonstrate that CounterFlowNet achieves better trade-offs between validity, sparsity, plausibility, and diversity while fully satisfying user-defined constraints.
CounterFlowNet generates more diverse and plausible counterfactual explanations than existing methods by framing the search as a sequential decision process learned by a GFlowNet.
Counterfactual explanations (CFs) provide human-interpretable insights into model's predictions by identifying minimal changes to input features that would alter the model's output. However, existing methods struggle to generate multiple high-quality explanations that (1) affect only a small portion of the features, (2) can be applied to tabular data with heterogeneous features, and (3) are consistent with the user-defined constraints. We propose CounterFlowNet, a generative approach that formulates CF generation as sequential feature modification using conditional Generative Flow Networks (GFlowNet). CounterFlowNet is trained to sample CFs proportionally to a user-specified reward function that can encode key CF desiderata: validity, sparsity, proximity and plausibility, encouraging high-quality explanations. The sequential formulation yields highly sparse edits, while a unified action space seamlessly supports continuous and categorical features. Moreover, actionability constraints, such as immutability and monotonicity of features, can be enforced at inference time via action masking, without retraining. Experiments on eight datasets under two evaluation protocols demonstrate that CounterFlowNet achieves superior trade-offs between validity, sparsity, plausibility, and diversity with full satisfaction of the given constraints.
