DocAtlas: Long-Document Understanding as Mutable-State Interaction

• URL: http://arxiv.org/abs/2608.07527v1
• Date: Tue, 21 Jul 2026 09:45:34 GMT
• Title: DocAtlas: Long-Document Understanding as Mutable-State Interaction
• Abstract summary: We present DocAtlas, a system that treats long-document understanding as a mutable-state information-seeking process.<n>Given a document and a question, DocAtlas exposes search, reading, note-taking, and review tools, maintains a hierarchical tree and note store, and updates both as the agent records evidence.<n>With GPT-5.4, DocAtlas reaches 71.4% on MMLongBench-Doc, exceeding the human-expert reference of 65.8%.
• Score: 67.90400271544678
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Long-document understanding requires models to find and combine evidence across many pages, layouts, tables, figures, and charts. Existing retrieval-augmented systems usually select evidence from a static index before generation, while recent agentic systems add multi-turn tool use but often rely on frozen proprietary backbones whose behavior is set by prompts. We present DocAtlas, a system that treats long-document understanding as a mutable-state information-seeking process. We instantiate DocAtlas as a mutable document harness: an external environment that determines what document information is searched, read, stored, reviewed, and shown to the model at each step. Given a document and question, the harness exposes search, reading, note-taking, and review tools, maintains a hierarchical tree and note store, and updates both as the agent records evidence. DocAtlas combines self-improving retrieval, selective evidence access, and active working memory under a fixed context budget. The same harness supports inference-time use with large VLMs and end-to-end reinforcement learning for compact VLM agents. With GPT-5.4, DocAtlas reaches 71.4\% on MMLongBench-Doc, exceeding the human-expert reference of 65.8\%. A Qwen3.5-4B VLM trained with end-to-end RL in the DocAtlas environment reaches 63.7\%, compared with a 54.4\% direct-input baseline, showing that mutable document-harness design can improve compact document agents by a large margin.

Related papers

• DocNavRAG: Document-Structured Graph RAG with Stateful Evidence Construction for Complex Document Question Answering [22.38959921864722]
  We introduce DocNavRAG, which organizes document hierarchies and cross-region relations into a navigable graph.<n>Across four long- and multi-document QA benchmarks, DocNavRAG improves answer quality and context sufficiency over the strongest baseline by 7.8% and 17.7% on average.
  arXiv  Detail & Related papers  (2026-08-03T00:48:05Z)
• MAGE-RAG: Multigranular Adaptive Graph Evidence for Agentic Multimodal RAG in Long-Document QA [14.865891513918278]
  Long-document multimodal question answering requires a system to locate sparse evidence in long PDFs.<n>This paper proposes MAGE-RAG, a multigranular adaptive graph evidence framework for long-document multimodal QA.
  arXiv  Detail & Related papers  (2026-06-14T16:34:15Z)
• Trace Only What You Need: Structure-Aware On-Demand Hypergraph Memory for Long-Document Question Answering [11.17007249581283]
  Long-document question answering requires large language models to reason over evidence scattered across lengthy documents.<n>Existing structured RAG methods face three limitations: costly query-agnostic knowledge organization, insufficient use of original document structure, and no reuse of historical reasoning experience.<n>We propose DocTrace, a multi-agent RAG framework for long-document QA that supports query-triggered knowledge organization, document-structure-aware and experience-guided reasoning.
  arXiv  Detail & Related papers  (2026-06-09T14:29:06Z)
• Dr. DocBench: A Comprehensive Benchmark for Expert-Level and Difficult Document Parsing [53.41293908252118]
  We introduce Dr. DocBench, a difficulty-aware benchmark for expert-level document parsing.<n>Dr. DocBench spans 52 BISAC domains and selects documents through failure-based sampling.<n>It contains 4,514 annotated pages from long documents averaging around 100 pages, with 65k high-quality annotations for layout, reading order, hierarchical relations, and domain-specific visual contents.<n>Our analysis reveals substantial failures across subjects, content types, and structural attributes, highlighting Dr. DocBench as a comprehensive testbed for diagnosing and advancing document intelligence.
  arXiv  Detail & Related papers  (2026-05-31T18:35:30Z)
• DocQAC: Adaptive Trie-Guided Decoding for Effective In-Document Query Auto-Completion [8.602282135498108]
  DocQAC aims to enhance search productivity within long documents by helping users craft faster, more precise queries.<n>We propose a novel adaptive trie-guided decoding framework that uses user query prefixes to softly steer language models toward high-quality completions.<n>We evaluate our method on a newly introduced DocQAC benchmark derived from ORCAS, enriched with query-document pairs.
  arXiv  Detail & Related papers  (2026-04-20T13:30:45Z)
• Doc-V*:Coarse-to-Fine Interactive Visual Reasoning for Multi-Page Document VQA [71.42483000929614]
  Multi-page Document Visual Question Answering requires reasoning over semantics, layouts, and visual elements in long, visually dense documents.<n>We propose Doc-$V*$, an textbfOCR-free agentic framework that casts multi-page DocVQA as sequential evidence aggregation.
  arXiv  Detail & Related papers  (2026-04-15T11:12:27Z)
• BigDocs: An Open Dataset for Training Multimodal Models on Document and Code Tasks [57.589795399265945]
  We introduce BigDocs-7.5M, a high-quality, open-access dataset comprising 7.5 million multimodal documents across 30 tasks.<n>We also introduce BigDocs-Bench, a benchmark suite with 10 novel tasks.<n>Our experiments show that training with BigDocs-Bench improves average performance up to 25.8% over closed-source GPT-4o.
  arXiv  Detail & Related papers  (2024-12-05T21:41:20Z)
• M-Longdoc: A Benchmark For Multimodal Super-Long Document Understanding And A Retrieval-Aware Tuning Framework [75.95430061891828]
  We introduce M-LongDoc, a benchmark of 851 samples, and an automated framework to evaluate the performance of large multimodal models. We propose a retrieval-aware tuning approach for efficient and effective multimodal document reading.
  arXiv  Detail & Related papers  (2024-11-09T13:30:38Z)
• M3DocRAG: Multi-modal Retrieval is What You Need for Multi-page Multi-document Understanding [63.33447665725129]
  We introduce M3DocRAG, a novel multi-modal RAG framework that flexibly accommodates various document contexts. M3DocRAG can efficiently handle single or many documents while preserving visual information. We also present M3DocVQA, a new benchmark for evaluating open-domain DocVQA over 3,000+ PDF documents with 40,000+ pages.
  arXiv  Detail & Related papers  (2024-11-07T18:29:38Z)
• DAPR: A Benchmark on Document-Aware Passage Retrieval [57.45793782107218]
  We propose and name this task emphDocument-Aware Passage Retrieval (DAPR) While analyzing the errors of the State-of-The-Art (SoTA) passage retrievers, we find the major errors (53.5%) are due to missing document context. Our created benchmark enables future research on developing and comparing retrieval systems for the new task.
  arXiv  Detail & Related papers  (2023-05-23T10:39:57Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.