Long-document visual question answering tasks and future research directions
A Task-Centric Survey of Long-Document Visual Question Answering: Taxonomy, Evolution, and Future Directions
Computer Vision and Pattern RecognitionComputation and LanguageInformation Retrieval
Summary
Understanding long documents that contain both images and text is tricky, especially when answering questions about them. The authors review how research breaks down the process into parts such as reading, finding evidence, generating answers, and checking accuracy. They show how research priorities have shifted towards managing lots of evidence and making answers reliable. The paper also suggests future work on improving the quality of evidence, verifying understanding, and handling limited resources.
What this means in practice
- •For document processing engineers: Design systems that better organize and retrieve information for answering questions about complex documents.
- •For enterprise data teams: Improve workflows that require accurate extraction and verification of knowledge from lengthy visual documents.
A survey. It maps existing work.
Authors
Yufei Tong, Guanjie Cheng, Tianle Wu, Leilei Zhang, Hang Zheng, Peihan Wu, Jianwei Yin, Shuiguang Deng
Abstract
Visual document understanding is essential for acquiring knowledge from visually rich documents and is attracting growing research interest. Existing surveys emphasize architectures or technical approaches, with limited attention to how task bottlenecks shape technical choices and research priorities. We present a task-centric survey that decomposes long-document visual question answering (VQA) into four components: document parsing, evidence organization and retrieval, generator organization and learning, and evaluation. For each, we develop a fine-grained taxonomy and trace motivations, technical responses, and cross-component connections. We focus on explicit retrieval while covering direct reading. Our temporal analysis traces expanding priorities from multimodal content understanding to sufficient, high-fidelity evidence acquisition and sustained evidence management. We outline future directions in evidence sufficiency, verifiable and correctable understanding, and budget-aware coordination, with an outlook on continual document understanding.