aiqbench analysis
This research advances Explainable AI (XAI) by integrating generative inpainting into perturbation-based methods like LIME, effectively solving the artifact problem inherent in traditional pixel-replacement techniques. By generating photorealistic out-of-distribution samples that align with the original data distribution, the study significantly improves the fidelity and reliability of visual model explanations. This methodological refinement directly addresses a core challenge in interpreting complex machine learning models for image data.
Inpainting Insights: Elevating Visual XAI with Photorealistic Perturbations
arXiv:2607.15482v2 Announce Type: replace Abstract: The increasing complexity of state-of-the-art machine learning models has made their behavior progressively harder to interpret, spurring rapid advancements in the field of eXplainable Artificial Intelligence (XAI). Among many methods proposed, perturbation-based approaches play a major role. By systematically altering (perturbing) input features, these approaches measure the impact on the model's predictions. For image data, traditional perturbation techniques, often involve replacing pixel values e.g., with a pre-defined color. However, such approaches, but also more refined deterministic techniques, generate unrealistic out-of-distribution samples and often leave visible artifacts, which can mislead the model and compromise explanation quality. In this work, we adjust LIME, a widely used perturbation-based method, to demonstrate how generative inpainting can improve perturbation-based explanations for images. We achieve photorealistic perturbed samples that align better with the original data distribution and enhance explanation quality.
Abstract:The increasing complexity of state-of-the-art machine learning models has made their behavior progressively harder to interpret, spurring rapid advancements in the field of eXplainable Artificial Intelligence (XAI). Among many methods proposed, perturbation-based approaches play a major role. By systematically altering (perturbing) input features, these approaches measure the impact on the model's predictions. For image data, traditional perturbation techniques, often involve replacing pixel values e.g., with a pre-defined color. However, such approaches, but also more refined deterministic techniques, generate unrealistic out-of-distribution samples and often leave visible artifacts, which can mislead the model and compromise explanation quality. In this work, we adjust LIME, a widely used perturbation-based method, to demonstrate how generative inpainting can improve perturbation-based explanations for images. We achieve photorealistic perturbed samples that align better with the original data distribution and enhance explanation quality.
Accepted preprint (v2, updated affiliations). 15 pages, 5 figures, 4 tables. This preprint has not undergone peer review or any post-submission improvements or corrections. The Version of Record of this contribution is published in ICPR 2026 International Workshops. ICPR 2026. Lecture Notes in Computer Science, vol 17114, and is available online at this https URL
Machine Learning (cs.LG)
arXiv:2607.15482 [cs.LG]
(or arXiv:2607.15482v2 [cs.LG] for this version)
Pattern Recognition. ICPR 2026 International Workshops. ICPR 2026. Lecture Notes in Computer Science, vol 17114. Springer, Cham, 2026, pp. 447-461
https://doi.org/10.1007/978-3-032-39388-3_32 DOI(s) linking to related resources
Full text under the publisher's open licence · 226 words
Related coverage
Inpainting Insights: Elevating Visual XAI with Photorealistic Perturbations
arXiv · Jul 20, 2026
NVIDIA Enables the Next Era Of Physical AI Research With Agent Skills For Autonomous Vehicles, Robotics And Vision AI
NVIDIA · Jun 3, 2026
How to Evaluate General-Purpose Robot Policies for Real-World Deployment
NVIDIA · Jul 12, 2026
TAPDreamer: Transferable Adversarial Patches for World Action Models
arXiv · Oct 6, 2026