Yufei Tong, Guanjie Cheng, Tianle Wu, Leilei Zhang, Hang Zheng, Peihan Wu, Jianwei Yin, Shuiguang Deng · 2026 · Preprints.org
DOI: 10.20944/preprints202610.2118.v1 · cited by 0 works (OpenAlex) · open-access PDF
Memahami dokumen visual sangat penting untuk memperoleh pengetahuan dari dokumen yang kaya visual dan menarik minat penelitian yang semakin meningkat. Survei yang ada menekankan arsitektur atau pendekatan teknis, dengan perhatian terbatas pada bagaimana kebocoran tugas membentuk pilihan teknis dan prioritas penelitian. Kami menyajikan survei yang berpusat pada tugas yang memecah pertanyaan visual dokumen panjang (VQA) menjadi empat komponen: pemindaian dokumen, organisasi bukti dan retrieval, organisasi generator dan pembelajaran, dan evaluasi. Untuk masing-masing, kami mengembangkan taksonomi yang halus dan motivasi pelacakan, tanggapan teknis, dan koneksi lintas komponen. Kami berfokus pada retrieval eksplisit sambil mencakup bacaan langsung. Analisis waktu kami melacak prioritas yang memperluas dari pemahaman konten multimodal ke pemahaman bukti yang cukup, tinggi, dan manajemen bukti yang berkelanj
Machine translation (Meta M2M100); the English original follows.
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.
Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese
Topics: Computer science, Task (project management), TRACE (psycholinguistics), Focus (optics), Verifiable secret sharing, Question answering
GET https://store.agentexchange.work/gw/papers/verify?title=...&year=... (0.003 USD), fetch this record as JSON with /gw/papers/paper?id=2610.31385, or list what it cites with /gw/papers/references?id=.... Pay per call in USDC (x402) or with a 19 USD card key for 1,000 calls. Catalogue · llms.txtSources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange