Towards Self-Explainable Document Visual Question Answering with Chain-of-Explanation Predictions

TL;DR

This paper introduces CoExVQA, a self-explainable framework for Document Visual Question Answering that explicitly grounds reasoning steps, improving explainability and performance over existing models.

Contribution

It presents a novel chain-of-explanation approach that explicitly localizes evidence and answers, enhancing transparency and accuracy in DocVQA.

Findings

Achieves state-of-the-art explainable DocVQA performance on PFL-DocVQA.

Improves ANLS by 12% over current explainable baselines.

Enables direct inspection and verification of reasoning process.

Abstract

Document Visual Question Answering (DocVQA) requires vision-language models to reason not only about what information in a document is relevant to a question, but also where the answer is grounded on the page. Existing DocVQA models entangle question-relevant evidence and answer localization and operate largely as black boxes, offering limited means to verify how predictions depend on visual evidence. We propose CoExVQA, a self-explainable DocVQA framework with a grounded reasoning process through a chain-of-explanation design. CoExVQA first identifies question-relevant evidence, then explicitly localizes the answer region, and finally decodes the answer exclusively from the grounded region. Prediction via CoExVQA's chain-of-explanation enables direct inspection and verification of the reasoning process across modalities. Empirical results show that restricting decoding to grounded evidence achieves SotA explainable DocVQA performance on PFL-DocVQA, improving ANLS by 12% over the current explainable baselines while providing transparent and verifiable predictions.