Search papers, labs, and topics across Lattice.
The paper introduces Minimum Spanning Tree Compactness (MST-C), a graph-based metric, to quantify the structural compactness of attribution maps by evaluating their spread and cohesion. MST-C aggregates these geometric properties into a single score that favors attributions with salient points concentrated in small, cohesive clusters. Experiments demonstrate that MST-C effectively differentiates explanation methods, reveals structural differences between models, and serves as a diagnostic tool for attribution compactness, complementing existing complexity metrics.
Forget IoU, measuring the structural compactness of attribution maps with Minimum Spanning Trees reveals fundamental differences in how models explain themselves.
In the evaluation of attribution quality, the quantitative assessment of explanation legibility is particularly difficult, as it is influenced by varying shapes and internal organization of attributions not captured by simple statistics. To address this issue, we introduce Minimum Spanning Tree Compactness (MST-C), a graph-based structural metric that captures higher-order geometric properties of attributions, such as spread and cohesion. These components are combined into a single score that evaluates compactness, favoring attributions with salient points spread across a small area and spatially organized into few but cohesive clusters. We show that MST-C reliably distinguishes between explanation methods, exposes fundamental structural differences between models, and provides a robust, self-contained diagnostic for explanation compactness that complements existing notions of attribution complexity.
