Elamur Research⌕
☰
home / papers / 2026 / DocLens: A Tool-Augmented Multi-Agent Framework for Long Visual Document Understanding

DocLens: A Tool-Augmented Multi-Agent Framework for Long Visual Document Understanding

Paper · 2026 · Dawei Zhu, Rui Meng, Jiefeng Chen et al.

Summary

Long visual documents, such as financial reports, papers and manuals, spread their information over many pages of text, tables, charts and figures. Current Vision-Language Models (VLMs) handle them poorly, mainly because they cannot find the right evidence. They miss relevant pages, they miss small details inside visual elements, and they often make up answers when the document has none. DocLens is a training-free multi-agent framework built around document-parsing tools to address this. Its Lens Module has two agents. A Page Navigator gives the VLM OCR text alongside page images to pick evidence pages, and an Element Localizer uses layout detection to crop tables, charts and figures for close inspection. A Reasoning Module then samples several candidate answers and has an Adjudicator choose the most consistent one. With Gemini-2.5-Pro as the backbone, DocLens sets new state-of-the-art results on MMLongBench-Doc and FinRAGBench-V and beats the reported human-expert score. It is particularly strong on vision-heavy and unanswerable questions, which shows that precise evidence localization is key to reliable document QA.

Key points

  • Motivating failures: On MMLongBench-Doc, Gemini-2.5-Pro alone recalls only 68% of evidence pages. Vector retrievers (ColBERT, ColPali) reach just 55.3% Recall@10. Models invent answers for more than half of the unanswerable questions.
  • Page recall: The Page Navigator reaches 97.3% evidence-page recall. It runs repeated sampling at temperature above zero, takes the union of the selected pages, and splits long documents into chunks when they exceed the context window.
  • Evaluation setup:
    • Benchmarks: MMLongBench-Doc (documents average 47.5 pages; human experts score 65.8) and FinRAGBench-V (dense, newspaper-like layouts, with block-level visual citation).
    • Backbones: Gemini-2.5-Pro, Gemini-2.5-Flash and Claude-4-Sonnet.
  • Baselines and gains: DocLens is compared with screenshot-only and screenshot+OCR settings and with the agentic systems MACT, M3DocRAG, MDocAgent and SimpleDoc. Gains are largest for the weaker backbones.
  • Stated limitations: Some visual elements remain hard even after zooming in. The framework is also not tailored to particular professional fields, for instance documents from law, medicine or finance.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

 Institutions

🎓Peking UniversityBeijing · China315 works · 43,267 citations

 Fields, topics and concepts

📚 Shelves 5The field of science this work belongs to (OpenAlex fields).Computer Science · 18,614 worksArtificial Intelligence · 17,965 worksArts and Humanities · 17,450 worksLanguage and Linguistics · 17,296 worksComputer Vision and Pattern Recognition · 3,486 works

Cite

Cite as Dawei Zhu, Rui Meng, Jiefeng Chen, Sujian Li, Tomas Pfister and Jinsung Yoon. 2026. DocLens: A Tool-Augmented Multi-Agent Framework for Long Visual Document Understanding. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers).

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2026.acl-long.1234

@inproceedings{zhu2026doclens,
  title = {{DocLens: A Tool-Augmented Multi-Agent Framework for Long Visual Document Understanding}},
  author = {Dawei Zhu, Rui Meng, Jiefeng Chen, Sujian Li, Tomas Pfister and Jinsung Yoon},
  booktitle = {Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)},
  year = {2026},
  doi = {10.18653/v1/2026.acl-long.1234},
  url = {https://doi.org/10.18653/v1/2026.acl-long.1234}
}

References

36 references, 21 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Jianlv Chen, Shitao Xiao, Peitian Zhang, Kun Luo, Defu Lian, and Zheng Liu. 2024. Bge m3-embedding: Multi-lingual, multi-functionality, multi-granularity text embeddings through self-knowledge distillation. Preprint, arXiv:2402.03216.
  2. Jaemin Cho, Debanjan Mahata, Ozan Irsoy, Yujie He, and Mohit Bansal. 2024. M3docrag: Multimodal retrieval is what you need for multi-page multi-document understanding. arXiv preprint arXiv:2411.04952. find
  3. ClaudeTeam. 2025. Claude sonnet 4: Hybrid reasoning model with superior intelligence for high-volume use cases, and 200k context window. https://www.anthropic.com/claude/sonnet. ↗
  4. Gheorghe Comanici, Eric Bieber, Mike Schaekermann, Ice Pasupat, Noveen Sachdeva, Inderjit Dhillon, Marcel Blistein, Ori Ram, Dan Zhang, Evan Rosen, et al. 2025. Gemini 2.5: Pushing the frontier with advanced reasoning, multimodality, long context, and next generation agentic capabilities. arXiv prep…
  5. Chao Deng, Jiale Yuan, Pi Bu, Peijie Wang, Zhong-Zhi Li, Jian Xu, Xiao-Hui Li, Yuan Gao, Jun Song, Bo Zheng, et al. 2024. Longdocurl: a comprehensive multimodal long document benchmark integrating understanding, reasoning, and locating. arXiv preprint arXiv:2412.18424.
  6. Kuicai Dong, Yujing Chang, Xin Deik Goh, Dexun Li, Ruiming Tang, and Yong Liu. 2025a. Benchmarking retrieval-augmented multimodal generation for document question answering. Preprint, arXiv:2501.04008. — [arXiv:2501.04008](https://arxiv.org/abs/2501.04008) find
  7. Kuicai Dong, Yujing Chang, Xin Deik Goh, Dexun Li, Ruiming Tang, and Yong Liu. 2025b. Mmdocir: Benchmarking multi-modal retrieval for long documents. arXiv preprint arXiv:2501.08828. — [arXiv:2501.08828](https://arxiv.org/abs/2501.08828) find
  8. Manuel Faysse, Hugues Sibille, Tony Wu, Bilel Omrani, Gautier Viaud, Céline Hudelot, and Pierre Colombo. 2024. Colpali: Efficient document retrieval with vision language models. arXiv preprint arXiv:2407.01449.
  9. Dong Guo, Faming Wu, Feida Zhu, Fuxing Leng, Guang Shi, Haobin Chen, Haoqi Fan, Jian Wang, Jianyu Jiang, Jiawei Wang, Jingji Chen, Jingjia Huang, Kang Lei, Liping Yuan, Lishu Luo, Pengfei Liu, Qinghao Ye, Rui Qian, Shen Yan, Shixiong Zhao, Shuai Peng, Shuangye Li, Sihang Yuan, Sijin Wu, Tianheng Che… — Seed1.5-VL Technical Report
  10. Siwei Han, Peng Xia, Ruiyi Zhang, Tong Sun, Yun Li, Hongtu Zhu, and Huaxiu Yao. 2025. Mdocagent: A multi-modal multi-agent framework for document understanding. arXiv preprint arXiv:2503.13964. — [arXiv:2503.13964](https://arxiv.org/abs/2503.13964) find
  11. SU Hongjin, Howard Yen, Mengzhou Xia, Weijia Shi, Niklas Muennighoff, Han-yu Wang, Liu Haisu, Quan Shi, Zachary S Siegel, Michael Tang, et al. 2025. Bright: A realistic and challenging benchmark for reasoning-intensive retrieval. In The Thirteenth International Conference on Learning Representations…
  12. Chelsi Jain, Yiran Wu, Yifan Zeng, Jiale Liu, Zhenwen Shao, Qingyun Wu, Huazheng Wang, et al. 2025. Simpledoc: Multi-modal document understanding with dual-cue page retrieval and iterative refinement. arXiv preprint arXiv:2506.14035.
  13. Jiaheng Liu, Dawei Zhu, Zhiqi Bai, Yancheng He, Huanxuan Liao, Haoran Que, Zekun Wang, Chenchen Zhang, Ge Zhang, Jiebin Zhang, et al. 2025. A comprehensive survey on long context language modeling. arXiv preprint arXiv:2503.17407.
  14. Xueguang Ma, Shengyao Zhuang, Bevan Koopman, Guido Zuccon, Wenhu Chen, and Jimmy Lin. 2024a. Visa: Retrieval augmented generation with visual source attribution. arXiv preprint arXiv:2412.14457.
  15. Yubo Ma, Yuhang Zang, Liangyu Chen, Meiqi Chen, Yizhu Jiao, Xinze Li, Xinyuan Lu, Ziyu Liu, Yan Ma, Xiaoyi Dong, et al. 2024b. Mmlongbench-doc: Benchmarking long-context document understanding with visualizations. Advances in Neural Information Processing Systems, 37:95963–96010.
  16. Ahmed Masry, Do Xuan Long, Jia Qing Tan, Shafiq Joty, and Enamul Hoque. 2022. Chartqa: A benchmark for question answering about charts with visual and logical reasoning. arXiv preprint arXiv:2203.10244.
  17. Minesh Mathew, Dimosthenis Karatzas, and CV Jawahar. 2021. Docvqa: A dataset for vqa on document images. In Proceedings of the IEEE/CVF winter conference on applications of computer vision, pages 2200–2209.
  18. Junbo Niu, Zheng Liu, Zhuangcheng Gu, Bin Wang, Linke Ouyang, Zhiyuan Zhao, Tao Chu, Tianyao He, Fan Wu, Qintong Zhang, Zhenjiang Jin, Guang Liang, Rui Zhang, Wenzheng Zhang, Yuan Qu, Zhifei Ren, Yuefeng Sun, Yuanhong Zheng, Dongsheng Ma, Zirui Tang, Boyu Niu, Ziyang Miao, Hejun Dong, Siyi Qian, Jun… — MinerU2.5: A Decoupled Vision-Language Model for Efficient High-Resolution Document Parsing
  19. OpenAITeam. 2025. Introducing gpt-5. https://openai.com/index/introducing-gpt-5/. ↗
  20. QwenTeam. 2025. Qwen3-vl: Sharper vision, deeper thought, broader action. https://qwen.ai/blog?id= 99f0335c4ad9ff6153e517418d48535ab6d8afef& from=research.latest-advancements-list. ↗
  21. Keshav Santhanam, Omar Khattab, Jon Saad-Falcon, Christopher Potts, and Matei Zaharia. 2021. Colbertv2: Effective and efficient retrieval via lightweight late interaction. arXiv preprint arXiv:2112.01488.
  22. Ryota Tanaka, Taichi Iki, Taku Hasegawa, Kyosuke Nishida, Kuniko Saito, and Jun Suzuki. 2025. Vdocrag: Retrieval-augmented generation over visually-rich documents. In Proceedings of the Computer Vision and Pattern Recognition Conference, pages 24827–24837. find
  23. Ryota Tanaka, Kyosuke Nishida, Kosuke Nishida, Taku Hasegawa, Itsumi Saito, and Kuniko Saito. 2023. Slidevqa: A dataset for document visual question answering on multiple images. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 37, pages 13636–13645.
  24. Core Team, Zihao Yue, Zhenru Lin, Yifan Song, Weikun Wang, Shuhuai Ren, Shuhao Gu, Shicheng Li, Peidian Li, Liang Zhao, Lei Li, Kainan Bao, Hao Tian, Hailin Zhang, Gang Wang, Dawei Zhu, Cici, Chenhong He, Bowen Ye, Bowen Shen, Zihan Zhang, Zihan Jiang, Zhixian Zheng, Zhichao Song, Zhenbo Luo, Yue Yu… — MiMo-VL Technical Report
  25. Liang Wang, Nan Yang, Xiaolong Huang, Binxing Jiao, Linjun Yang, Daxin Jiang, Rangan Majumder, and Furu Wei. 2022. Text embeddings by weakly-supervised contrastive pre-training. arXiv preprint arXiv:2212.03533.
  26. Qiuchen Wang, Ruixue Ding, Zehui Chen, Weiqi Wu, Shihang Wang, Pengjun Xie, and Feng Zhao. 2025a. Vidorag: Visual document retrieval-augmented generation via dynamic iterative reasoning agents. arXiv preprint arXiv:2502.18017.
  27. Qiuchen Wang, Ruixue Ding, Yu Zeng, Zehui Chen, Lin Chen, Shihang Wang, Pengjun Xie, Fei Huang, and Feng Zhao. 2025b. Vrag-rl: Empower vision-perception-based rag for visually rich information understanding via iterative reasoning with reinforcement learning. arXiv preprint arXiv:2505.22019.
  28. Extending embedding models for long context retrieval. In Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing, pages 802–816. find
  29. Anni Zou, Wenhao Yu, Hongming Zhang, Kaixin Ma, Deng Cai, Zhuosheng Zhang, Hai Zhao, and Dong Yu. 2024. Docbench: A benchmark for evaluating llm-based document reading systems. Preprint, arXiv:2407.10701. find
  30. Xuezhi Wang, Jason Wei, Dale Schuurmans, Quoc V Le, Ed H Chi, Sharan Narang, Aakanksha Chowdhery, and Denny Zhou. 2023. Self-consistency improves chain of thought reasoning in language models. In The Eleventh International Conference on Learning Representations.
  31. An Yang, Anfeng Li, Baosong Yang, Beichen Zhang, Binyuan Hui, Bo Zheng, Bowen Yu, Chang Gao, Chengen Huang, Chenxu Lv, Chujie Zheng, Dayiheng Liu, Fan Zhou, Fei Huang, Feng Hu, Hao Ge, Haoran Wei, Huan Lin, Jialong Tang, Jian Yang, Jianhong Tu, Jianwei Zhang, Jianxin Yang, Jiaxi Yang, Jing Zhou, Jin… — Qwen3 Technical Report
  32. Shi Yu, Chaoyue Tang, Bokai Xu, Junbo Cui, Junhao Ran, Yukun Yan, Zhenghao Liu, Shuo Wang, Xu Han, Zhiyuan Liu, et al. 2024. Visrag: Vision-based retrieval-augmented generation on multi-modality documents. arXiv preprint arXiv:2410.10594. — [arXiv:2410.10594](https://arxiv.org/abs/2410.10594) find
  33. Xinlei Yu, Zhangquan Chen, Yudong Zhang, Shilin Lu, Ruolin Shen, Jiangning Zhang, Xiaobin Hu, Yanwei Fu, and Shuicheng Yan. 2025. Visual document understanding and question answering: A multi-agent collaboration framework with test-time scaling. arXiv preprint arXiv:2508.03404. — [arXiv:2508.03404](https://arxiv.org/abs/2508.03404) find
  34. Wenjin Zhang, Yunhao Wang, Guoxuan Li, Zhaoyang Yu, Jing Ao, Jingping Liu, Yidan Fan, Chang Zong, Huorong Li, Jianchen Xiao, Biao Qin, and Tong Ruan. 2024. Uda: A benchmark suite for retrieval augmented generation in real-world document analysis. Preprint, arXiv:2406.09153. find
  35. Suifeng Zhao, Zhuoran Jin, Sujian Li, and Jun Gao. 2025. Finragbench-v: A benchmark for multimodal rag with visual citation in the financial domain. arXiv preprint arXiv:2505.17471. find
  36. Dawei Zhu, Liang Wang, Nan Yang, Yifan Song, Wenhao Wu, Furu Wei, and Sujian Li. 2024. Longembed: find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/zhu-2026-doclens-tool-augmented-multi-agent/