counterfactual explanations
Counterfactual explanations provide insights into model behavior by illustrating how slight changes in input features would have led to different outcomes. These are particularly useful for making black-box models interpretable and understanding the decision-making process.
- Collective Counterfactual Explanations: Balancing Individual Goals and Collective Dynamics
- DiCoFlex: Model-Agnostic Diverse Counterfactuals with Flexible Control
- ElliCE: Efficient and Provably Robust Algorithmic Recourse via the Rashomon Sets
- Few-Shot Knowledge Distillation of LLMs With Counterfactual Explanations
- From Counterfactuals to Trees: Competitive Analysis of Model Extraction Attacks
- LeapFactual: Reliable Visual Counterfactual Explanation Using Conditional Flow Matching
- Performative Validity of Recourse Explanations