Saved in:
Bibliographic Details
Main Authors: Vasu, Bhavan, Raffa, Giuseppe, Tadepalli, Prasad
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2601.13404
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866914265316720640
author Vasu, Bhavan
Raffa, Giuseppe
Tadepalli, Prasad
author_facet Vasu, Bhavan
Raffa, Giuseppe
Tadepalli, Prasad
contents While deep neural networks are extremely effective at classifying images, they remain opaque and hard to interpret. We introduce local and global explanation methods for black-box models that generate explanations in terms of human-recognizable primitive concepts. Both the local explanations for a single image and the global explanations for a set of images are cast as logical formulas in monotone disjunctive-normal-form (MDNF), whose satisfaction guarantees that the model yields a high score on a given class. We also present an algorithm for explaining the classification of examples into multiple classes in the form of a monotone explanation list over primitive concepts. Despite their simplicity and interpretability we show that the explanations maintain high fidelity and coverage with respect to the blackbox models they seek to explain in challenging vision datasets.
format Preprint
id arxiv_https___arxiv_org_abs_2601_13404
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Local-to-Global Logical Explanations for Deep Vision Models
Vasu, Bhavan
Raffa, Giuseppe
Tadepalli, Prasad
Computer Vision and Pattern Recognition
Artificial Intelligence
While deep neural networks are extremely effective at classifying images, they remain opaque and hard to interpret. We introduce local and global explanation methods for black-box models that generate explanations in terms of human-recognizable primitive concepts. Both the local explanations for a single image and the global explanations for a set of images are cast as logical formulas in monotone disjunctive-normal-form (MDNF), whose satisfaction guarantees that the model yields a high score on a given class. We also present an algorithm for explaining the classification of examples into multiple classes in the form of a monotone explanation list over primitive concepts. Despite their simplicity and interpretability we show that the explanations maintain high fidelity and coverage with respect to the blackbox models they seek to explain in challenging vision datasets.
title Local-to-Global Logical Explanations for Deep Vision Models
topic Computer Vision and Pattern Recognition
Artificial Intelligence
url https://arxiv.org/abs/2601.13404