DocXEval Framework for Multimodal Document Explainability
October 28, 2026
DocXEval adapts XAI methods and metrics to the specific context of text, layout, and image integration in document understanding. The framework evaluates 13 explainability methods across 4 models and 8 datasets to address black-box risks in high-stakes document processing.
HOW THIS AFFECTS YOU
●
builderYou can use this to better understand and audit the decisions of your multimodal document models.
●
researcherThis provides a standardized way to measure explainability specifically for multimodal document tasks.