Saved in:
Bibliographic Details
Main Authors: Zhou, Xiangyu, Xiao, Chenhan, Weng, Yang
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2602.17107
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866917281692385280
author Zhou, Xiangyu
Xiao, Chenhan
Weng, Yang
author_facet Zhou, Xiangyu
Xiao, Chenhan
Weng, Yang
contents Shapley value-based methods have become foundational in explainable artificial intelligence (XAI), offering theoretically grounded feature attributions through cooperative game theory. However, in practice, particularly in vision tasks, the assumption of feature independence breaks down, as features (i.e., pixels) often exhibit strong spatial and semantic dependencies. To address this, modern SHAP implementations now include the Owen value, a hierarchical generalization of the Shapley value that supports group attributions. While the Owen value preserves the foundations of Shapley values, its effectiveness critically depends on how feature groups are defined. We show that commonly used segmentations (e.g., axis-aligned or SLIC) violate key consistency properties, and propose a new segmentation approach that satisfies the $T$-property to ensure semantic alignment across hierarchy levels. This hierarchy enables computational pruning while improving attribution accuracy and interpretability. Experiments on image and tabular datasets demonstrate that O-Shap outperforms baseline SHAP variants in attribution precision, semantic coherence, and runtime efficiency, especially when structure matters.
format Preprint
id arxiv_https___arxiv_org_abs_2602_17107
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Owen-based Semantics and Hierarchy-Aware Explanation (O-Shap)
Zhou, Xiangyu
Xiao, Chenhan
Weng, Yang
Artificial Intelligence
Shapley value-based methods have become foundational in explainable artificial intelligence (XAI), offering theoretically grounded feature attributions through cooperative game theory. However, in practice, particularly in vision tasks, the assumption of feature independence breaks down, as features (i.e., pixels) often exhibit strong spatial and semantic dependencies. To address this, modern SHAP implementations now include the Owen value, a hierarchical generalization of the Shapley value that supports group attributions. While the Owen value preserves the foundations of Shapley values, its effectiveness critically depends on how feature groups are defined. We show that commonly used segmentations (e.g., axis-aligned or SLIC) violate key consistency properties, and propose a new segmentation approach that satisfies the $T$-property to ensure semantic alignment across hierarchy levels. This hierarchy enables computational pruning while improving attribution accuracy and interpretability. Experiments on image and tabular datasets demonstrate that O-Shap outperforms baseline SHAP variants in attribution precision, semantic coherence, and runtime efficiency, especially when structure matters.
title Owen-based Semantics and Hierarchy-Aware Explanation (O-Shap)
topic Artificial Intelligence
url https://arxiv.org/abs/2602.17107