Uncertainty-Aware Subset Selection for Robust Visual Explainability under Distribution Shifts

2025-12-09

AI Summary

Existing visual explanation methods perform well under in-distribution (ID) conditions but yield redundant, unstable, and uncertainty-sensitive interpretations under out-of-distribution (OOD) settings. To address this, we propose a training-free robust subset selection framework that— for the first time—integrates hierarchical gradient-based uncertainty estimation with submodular optimization. Our approach employs adaptive weight perturbation to guide the selection of diverse, information-rich salient regions. Crucially, it imposes no additional computational or architectural burden on the underlying model. Extensive experiments across multiple ID and OOD benchmark datasets demonstrate that our method simultaneously enhances explanation fidelity and stability in both ID and OOD regimes, effectively mitigating redundancy and instability. This work establishes a novel paradigm for trustworthy visual explanation under distributional shift.

Technology Category

• Computer Vision: Interpretability, Explainability, and Transparency
• Reasoning under Uncertainty: Other Foundations of Reasoning under Uncertainty
• Machine Learning: Calibration & Uncertainty Quantification

Application Category

• User Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalization
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs

Abstract

Subset selection-based methods are widely used to explain deep vision models: they attribute predictions by highlighting the most influential image regions and support object-level explanations. While these methods perform well in in-distribution (ID) settings, their behavior under out-of-distribution (OOD) conditions remains poorly understood. Through extensive experiments across multiple ID-OOD sets, we find that reliability of the existing subset based methods degrades markedly, yielding redundant, unstable, and uncertainty-sensitive explanations. To address these shortcomings, we introduce a framework that combines submodular subset selection with layer-wise, gradient-based uncertainty estimation to improve robustness and fidelity without requiring additional training or auxiliary models. Our approach estimates uncertainty via adaptive weight perturbations and uses these estimates to guide submodular optimization, ensuring diverse and informative subset selection. Empirical evaluations show that, beyond mitigating the weaknesses of existing methods under OOD scenarios, our framework also yields improvements in ID settings. These findings highlight limitations of current subset-based approaches and demonstrate how uncertainty-driven optimization can enhance attribution and object-level interpretability, paving the way for more transparent and trustworthy AI in real-world vision applications.

Problem

• Improves robustness of visual explanations under distribution shifts
• Addresses redundancy and instability in subset selection methods
• Enhances attribution fidelity without additional training or models

Innovation

• Combines submodular selection with gradient-based uncertainty estimation
• Uses adaptive weight perturbations to estimate uncertainty
• Guides submodular optimization with uncertainty for robust explanations