CASE: Contrastive Activation for Saliency Estimation

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Williamson, Dane, Ji, Yangfeng, Dwyer, Matthew
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866915345684496384
author Williamson, Dane
Ji, Yangfeng
Dwyer, Matthew
author_facet Williamson, Dane
Ji, Yangfeng
Dwyer, Matthew
contents Saliency methods are widely used to visualize which input features are deemed relevant to a model's prediction. However, their visual plausibility can obscure critical limitations. In this work, we propose a diagnostic test for class sensitivity: a method's ability to distinguish between competing class labels on the same input. Through extensive experiments, we show that many widely used saliency methods produce nearly identical explanations regardless of the class label, calling into question their reliability. We find that class-insensitive behavior persists across architectures and datasets, suggesting the failure mode is structural rather than model-specific. Motivated by these findings, we introduce CASE, a contrastive explanation method that isolates features uniquely discriminative for the predicted class. We evaluate CASE using the proposed diagnostic and a perturbation-based fidelity test, and show that it produces faithful and more class-specific explanations than existing methods.
format Preprint
id arxiv_https___arxiv_org_abs_2506_07327
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle CASE: Contrastive Activation for Saliency Estimation
Williamson, Dane
Ji, Yangfeng
Dwyer, Matthew
Computer Vision and Pattern Recognition
Machine Learning
I.2.6; I.5.1; I.5.5; I.2.10
Saliency methods are widely used to visualize which input features are deemed relevant to a model's prediction. However, their visual plausibility can obscure critical limitations. In this work, we propose a diagnostic test for class sensitivity: a method's ability to distinguish between competing class labels on the same input. Through extensive experiments, we show that many widely used saliency methods produce nearly identical explanations regardless of the class label, calling into question their reliability. We find that class-insensitive behavior persists across architectures and datasets, suggesting the failure mode is structural rather than model-specific. Motivated by these findings, we introduce CASE, a contrastive explanation method that isolates features uniquely discriminative for the predicted class. We evaluate CASE using the proposed diagnostic and a perturbation-based fidelity test, and show that it produces faithful and more class-specific explanations than existing methods.
title CASE: Contrastive Activation for Saliency Estimation
topic Computer Vision and Pattern Recognition
Machine Learning
I.2.6; I.5.1; I.5.5; I.2.10
url https://arxiv.org/abs/2506.07327