Low-Frequency Shortcuts in Texture-Driven Visual Learning

2026-06-02

AI Summary

This study addresses the tendency of neural networks to rely on low-frequency shortcut features in texture-driven tasks, which undermines both in-distribution and out-of-distribution generalization. The work reveals, for the first time, that models disproportionately depend on low-frequency components despite discriminative cues being primarily encoded in high-frequency details. To mitigate this bias, the authors propose a spectral pruning strategy that explicitly suppresses low-frequency information during both training and inference, thereby encouraging balanced utilization of the full frequency spectrum. Experimental results demonstrate that this approach improves in-distribution accuracy by up to 8% and enhances robustness against low-frequency out-of-distribution perturbations by as much as 40%, while also uncovering a performance trade-off under high-frequency perturbations.

Abstract

Neural networks suffer from shortcut learning, where learned features generalize well to the training set but not to in-distribution (ID) or out-of-distribution (OOD) test sets. Existing studies are all based on a few standard benchmarks, which are shape-driven. Numerous application domains, however, are texture-driven. In this work, we present shortcut learning analysis for texture-driven domains, and compare it with that of a standard benchmark. We show that texture-driven domains suffer from low-frequency shortcuts. They make the majority of their decisions based on a few low-frequency components (LFCs) with a skewed spectral behavior, despite that their classification information is in higher-frequency, fine-grained details. Pruning LFCs from training and test sets eliminates the shortcut and provides a more balanced spectral behavior, improving the ID accuracy by up to 8%. We show that low-frequency shortcuts make the models highly vulnerable to OOD corruptions, leading up to 70% accuracy drop compared to the ID accuracy. Pruning LFCs significantly improves robustness to low-frequency corruptions, by up to 40%, and introduces a trade-off for high-frequency corruptions; the balanced spectral behavior provides a better generalization performance, whereas the increased dependence on high-frequency features reduces it. OOD accuracy depends on the interaction between these two factors.

Problem

• shortcut learning
• texture-driven
• low-frequency components
• out-of-distribution robustness
• spectral bias

Innovation

• shortcut learning
• texture-driven
• low-frequency components
• spectral behavior
• out-of-distribution robustness