U-CECE: A Universal Multi-Resolution Framework for Conceptual Counterfactual Explanations

TMLR Paper9003 Authors

Abstract: As AI models grow more complex, explainability is essential for building trust, yet concept-based counterfactual methods still face a trade-off between expressivity and efficiency. Representing underlying concepts as atomic sets is fast but misses relational context, whereas full graph representations are more faithful but require solving the NP-hard Graph Edit Distance (GED) problem. We propose U-CECE, a unified, model-agnostic multi-resolution framework for conceptual counterfactual explanations that adapts to data regime and compute budget. U-CECE spans three levels of expressivity: atomic concepts for broad explanations, relational sets-of-sets for simple interactions, and structural graphs for full semantic structure. At the structural level, both a precision-oriented transductive mode based on supervised Graph Neural Networks (GNNs) and a scalable inductive mode based on unsupervised graph autoencoders (GAEs) are supported. Experiments on the structurally divergent CUB and Visual Genome datasets characterize the efficiency-expressivity trade-off across levels, while human surveys and LVLM-based evaluation show that the retrieved structural counterfactuals are semantically equivalent to, and often preferred over, exact GED-based ground-truth explanations.

Submission Type: Long submission (more than 12 pages of main content)

Previous TMLR Submission Url: https://openreview.net/forum?id=6YBMR1u1N5

Changes Since Last Submission: This is a resubmission of our previously withdrawn manuscript. We proactively withdrew the earlier version after discovering a data leakage issue (test set contamination in the training set) that affected the inductive experiments in Section 5.4.2. For this resubmission, we corrected the data split and reran all affected experiments. Consequently, Table 3 and its associated analysis in Section 5.4.2 have been updated with the correct results. We also updated the accompanying code to reflect this fix. Finally, we made minor presentation improvements throughout the paper and updated a few references.

Assigned Action Editor: ~Vicenç_Gómez1

Submission Number: 9003