TabKD: Tabular Knowledge Distillation through Interaction Diversity of Learned Feature Bins

TL;DR

TabKD introduces a novel approach for data-free knowledge distillation in tabular data by focusing on interaction diversity, leading to improved model compression and higher agreement with teacher models.

Contribution

It proposes a new method that explicitly captures feature interaction coverage through adaptive feature bins and synthetic query generation, enhancing tabular model distillation.

Findings

Achieves highest student-teacher agreement in 14 out of 16 configurations.

Outperforms 5 state-of-the-art baselines on benchmark datasets.

Interaction coverage correlates strongly with distillation quality.

Abstract

Data-free knowledge distillation enables model compression without original training data, critical for privacy-sensitive tabular domains. However, existing methods does not perform well on tabular data because they do not explicitly address feature interactions, the fundamental way tabular models encode predictive knowledge. We identify interaction diversity, systematic coverage of feature combinations, as an essential requirement for effective tabular distillation. To operationalize this insight, we propose TabKD, which learns adaptive feature bins aligned with teacher decision boundaries, then generates synthetic queries that maximize pairwise interaction coverage. Across 4 benchmark datasets and 4 teacher architectures, TabKD achieves highest student-teacher agreement in 14 out of 16 configurations, outperforming 5 state-of-the-art baselines. We further show that interaction coverage strongly correlates with distillation quality, validating our core hypothesis. Our work establishes interaction-focused exploration as a principled framework for tabular model extraction.