Explaining Model Robustness Through Saliency Maps and Natural Robustness Testing

Learn how to explain model robustness by combining saliency maps with natural robustness testing techniques in this 15-minute conference presentation from METACOG-25. Discover Brian Hu's research from KitWare on utilizing tools like NRTK (Natural Robustness Testing Kit) to better understand and visualize how machine learning models maintain performance under various conditions. Explore the intersection of explainable AI and model reliability through practical approaches that merge visual explanation methods with systematic robustness evaluation frameworks.