Attribution Upsampling should Redistribute, Not Interpolate

2026-03-16

AI Summary

Existing interpolation methods—such as bilinear and bicubic—often introduce aliasing, ringing, and boundary leakage when upsampling attribution maps, thereby distorting the true reasoning basis of models. This work reframes attribution upsampling as a semantic-boundary-aware quality redistribution problem and proposes Universal Semantic-aware Upsampling (USU). USU establishes the first axiomatic framework for attribution upsampling, formally proving that conventional interpolation violates essential fidelity criteria and deriving a unique ratio-based redistribution operator that satisfies all proposed axioms. Experiments on ImageNet, CIFAR-10, and CUB-200 demonstrate that USU rigorously preserves both attribution mass and importance ordering, yielding explanations that are semantically coherent and more trustworthy.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Reasoning under Uncertainty: Other Foundations of Reasoning under Uncertainty
• Machine Learning: Calibration & Uncertainty Quantification

Application Category

• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Responsible Web: Human-perceived consequences of algorithmic deployment on the web

Abstract

Attribution methods in explainable AI rely on upsampling techniques that were designed for natural images, not saliency maps. Standard bilinear and bicubic interpolation systematically corrupts attribution signals through aliasing, ringing, and boundary bleeding, producing spurious high-importance regions that misrepresent model reasoning. We identify that the core issue is treating attribution upsampling as an interpolation problem that operates in isolation from the model's reasoning, rather than a mass redistribution problem where model-derived semantic boundaries must govern how importance flows. We present Universal Semantic-Aware Upsampling (USU), a principled method that reformulates upsampling through ratio-form mass redistribution operators, provably preserving attribution mass and relative importance ordering. Extending the axiomatic tradition of feature attribution to upsampling, we formalize four desiderata for faithful upsampling and prove that interpolation structurally violates three of them. These same three force any redistribution operator into a ratio form; the fourth selects the unique potential within this family, yielding USU. Controlled experiments on models with known attribution priors verify USU's formal guarantees; evaluation across ImageNet, CIFAR-10, and CUB-200 confirms consistent faithfulness improvements and qualitatively superior, semantically coherent explanations.

Problem

• attribution upsampling
• saliency maps
• interpolation artifacts
• model reasoning
• faithful explanation

Innovation

• attribution upsampling
• mass redistribution
• semantic-aware
• faithful explanation
• axiomatic desiderata