A Comparative Study of XAI Methods for CNN and Transformer Models

This research paper, titled “A Comparative Study of XAI Methods for CNN and Transformer Models”, explores the effectiveness of various explainable artificial intelligence (XAI) methods in interpreting deep learning models, particularly Convolutional Neural Networks (CNNs) and Transformer architectures. We compare techniques such as LIME, SHAP, Grad-CAM, and attention visualization to assess their interpretability, trustworthiness, and performance impact. The study aims to bridge the gap between black-box AI models and human understanding, contributing toward more transparent and accountable AI systems. This version is made openly available for reference and citation via Zenodo.