<ul><li>Effective explanation methods are crucial for ensuring transparency in high-stakes machine learning models.</li><li>A new framework based on spectral analysis of explanation outcomes has been proposed.</li><li>The framework uncovers two factors of explanation quality: stability and target sensitivity.</li><li>Experiments on MNIST and ImageNet demonstrate the trade-offs between these factors in popular evaluation techniques.</li></ul>

Uncovering the Structure of Explanation Quality with Spectral Analysis

Discover more