LGAICVAug 14, 2025

On the Complexity-Faithfulness Trade-off of Gradient-Based Explanations

arXiv:2508.10490v11 citationsh-index: 4
Originality Incremental advance
AI Analysis

This work addresses the problem of noisy and difficult-to-interpret explanations in deep learning for researchers and practitioners, offering a principled analysis that is incremental to existing methods like GradCAM.

The paper tackles the trade-off between smoothness and faithfulness in gradient-based explanations for ReLU networks, introducing a spectral framework to analyze and quantify this trade-off, and validates it across various settings.

ReLU networks, while prevalent for visual data, have sharp transitions, sometimes relying on individual pixels for predictions, making vanilla gradient-based explanations noisy and difficult to interpret. Existing methods, such as GradCAM, smooth these explanations by producing surrogate models at the cost of faithfulness. We introduce a unifying spectral framework to systematically analyze and quantify smoothness, faithfulness, and their trade-off in explanations. Using this framework, we quantify and regularize the contribution of ReLU networks to high-frequency information, providing a principled approach to identifying this trade-off. Our analysis characterizes how surrogate-based smoothing distorts explanations, leading to an ``explanation gap'' that we formally define and measure for different post-hoc methods. Finally, we validate our theoretical findings across different design choices, datasets, and ablations.

Foundations

The foundational work for this paper's niche, ranked by how specifically the neighbourhood builds on it — not by global fame.

Your Notes