Trustworthy Feature Importance Avoids Unrestricted Permutations

arXiv:2604.11253v1 Announce Type: cross Abstract: Feature importance methods using unrestricted permutations are flawed due to extrapolation errors; such errors appear in all non-trivial variable importance approaches. We propose three new approaches: conditional model reliance and Knockoffs with Gaussian transformation, and restricted ALE plot designs. Theoretical and numerical results show our strategies reduce/eliminate extrapolation.

Related

• DINO-QPM: Adapting Visual Foundation Models for Globally Interpretable Image Classification
• Stability of a Generalized Debiased Lasso with Applications to Resampling-Based Variable Selection
• Identifying Causal Effects Using a Single Proxy Variable
• Locket: Robust Feature-Locking Technique for Language Models

Source: arXiv cs.LG | 2026-04-14