DEXTER: Diffusion-Guided EXplanations with TExtual Reasoning for Vision Models

Mubarak Shah (University of Central Florida) · Simone Carnemolla (University of Catania) · Matteo Pennisi (University of Catania) · Sarinda Samarasinghe (University of Central Florida) · Giovanni Bellitto (University of Catania) · Simone Palazzo (University of Catania) · Daniela Giordano (University of Catania) · Concetto Spampinato (University of Catania)

activation maximizationbias explanationclass-conditional imagesclass-level bias reportingdebiasingdexterdiffusion modelsinterpretable outputsmachine learning modelsnatural language reportsquantitative evaluationsslice discoverytextual explanationstransparent aivisual classifiers

Understanding and explaining the behavior of machine learning models is essential for building transparent and trustworthy AI systems. We introduce DEXTER, a data-free framework that employs diffusion models and large language models to generate global, textual explanations of visual classifiers. DEXTER operates by optimizing text prompts to synthesize class-conditional images that strongly activate a target classifier. These synthetic samples are then used to elicit detailed natural language reports that describe class-specific decision patterns and biases. Unlike prior work, DEXTER enables natural language explanation about a classifier's decision process without access to training data or ground-truth labels. We demonstrate DEXTER's flexibility across three tasks—activation maximization, slice discovery and debiasing, and bias explanation—each illustrating its ability to uncover the internal mechanisms of visual classifiers. Quantitative and qualitative evaluations, including a user study, show that DEXTER produces accurate, interpretable outputs. Experiments on ImageNet, Waterbirds, CelebA, and FairFaces confirm that DEXTER outperforms existing approaches in global model explanation and class-level bias reporting. Code is available at https://github.com/perceivelab/dexter.