---
title: "A Multistage Extraction Pipeline for Long Scanned Financial"
description: "The authors, from OCBC, build a modular pipeline with four stages: pages are cleaned (cropping, rotation and… 31.9 points. 36 references."
authors: ["Han Yuxuan", "Yuanxing Zhang", "Yushuo Wang", "Yichao Jin"]
year: 2026
venue: "Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 6: Industry Track)"
source: "https://aclanthology.org/2026.acl-industry.99/"
doi: "10.18653/v1/2026.acl-industry.99"
licence: "CC BY 4.0 (ACL Anthology)"
licence_url: "https://creativecommons.org/licenses/by/4.0/"
summary_licence: "Elamur Cosmic License"
canonical: "https://research.elamur.ai/free/en/papers/yuxuan-2026-multistage-extraction-pipeline-long-scanned/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/yuxuan-2026-multistage-extraction-pipeline-long-scanned/)

[home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [2026](https://md.research.elamur.ai/free/en/papers/2026/page/20/) / A Multistage Extraction Pipeline for Long Scanned Financial Documents: An Empirical Study in Industrial KYC Workflows

# A Multistage Extraction Pipeline for Long Scanned Financial Documents: An Empirical Study in Industrial KYC Workflows

**Authors:** Han Yuxuan, [Yuanxing Zhang](https://md.research.elamur.ai/free/en/authors/yuanxing-zhang/), Yushuo Wang and Yichao Jin · **Year:** 2026 · **Venue:** [Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 6: Industry Track)](https://md.research.elamur.ai/free/en/venues/acl/) ([ACL 2026](https://md.research.elamur.ai/free/en/venues/acl/2026/)) · **Publisher:** [Association for Computational Linguistics](https://md.research.elamur.ai/free/en/publishers/association-for-computational-linguistics/) · **Licence:** [CC BY 4.0 (ACL Anthology)](https://creativecommons.org/licenses/by/4.0/)

**Links:** [Publisher's page](https://aclanthology.org/2026.acl-industry.99/) · [PDF](https://aclanthology.org/2026.acl-industry.99.pdf) · [DOI 10.18653/v1/2026.acl-industry.99](https://doi.org/10.18653/v1/2026.acl-industry.99) · [OpenAlex](https://openalex.org/W7166815242)

**Citations:** 0 citations (OpenAlex, 2026-10-14)

## Summary
Banks running KYC and compliance checks must pull specific financial fields (revenue, net profit, dividends, etc.) out of long, scanned, often multilingual customer submissions. In these documents, the relevant content sits on only a few of dozens of noisy pages. Feeding whole PDFs straight into a vision–language model is costly and gives unreliable results.

The authors, from OCBC, build a modular pipeline with four stages:
- pages are cleaned (cropping, rotation and skew fixes, rescaling, contrast normalization);
- each page is transcribed with multilingual OCR;
- a hybrid BM25 plus sentence-embedding retriever ranks pages using field-specific queries;
- only the top-ranked pages go to a compact VLM with structured, language-adapted prompts.

They tested the system on 120 internal production KYC documents (roughly 3,000 pages). It outperforms direct PDF-to-VLM extraction for every OCR–VLM pairing, by up to 31.9 percentage points, without increasing latency. The study gives practical evidence that careful system design, especially locating the right pages first, lets small VLMs handle realistic document extraction in regulated settings.

## Key points
- Evaluation covers 2 OCR engines (PaddleOCRv3, EasyOCR) × 3 VLMs (MiniCPM-o-2.6, Gemma-3-27b-it, Qwen3-VL-8B-Instruct). The best pairing, PaddleOCR with MiniCPM-o-2.6, reaches 87.27% field-level accuracy, and gains over the baseline range from 20.7 to 31.9 points.
- Retrieval cuts the pages sent to the VLM by about 70%, so token usage drops while per-page inference time stays similar to the baseline.
- Ablations rank the modules clearly:
  - dropping retrieval costs 16.8–24.0 points;
  - dropping image preprocessing costs 6.2–16.3 points;
  - structured prompting adds little on average but helps with corner cases.
- Payslips score far higher than financial statements (96.92% vs. 83.95% in the best setup). However, the pipeline's gain is much larger for statements (over 40 points vs. about 8).
- Main error sources are inconsistent financial terminology, OCR failures on poor scans or handwriting, and ambiguous currency units (e.g., IDR'000, juta). Queries and prompts are hand-crafted per field, and the corpus is internal and not released.

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

📚 Shelves (5): [🖥️ Computer Science](https://md.research.elamur.ai/free/en/shelves/computer-science/) · 18,614 works · [🤖 Artificial Intelligence](https://md.research.elamur.ai/free/en/shelves/computer-science/artificial-intelligence/) · 17,965 works · [🎭 Arts and Humanities](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/) · 17,450 works · [🗣️ Language and Linguistics](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/language-and-linguistics/) · 17,296 works · [👁️ Computer Vision and Pattern Recognition](https://md.research.elamur.ai/free/en/shelves/computer-science/computer-vision-and-pattern-recognition/) · 3,486 works

## Cite

Cite as: Han Yuxuan, Yuanxing Zhang, Yushuo Wang and Yichao Jin. 2026. A Multistage Extraction Pipeline for Long Scanned Financial Documents: An Empirical Study in Industrial KYC Workflows. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 6: Industry Track).

The paper itself, not this page. BibTeX from the library's record of the work.

```bibtex
@inproceedings{yuxuan2026multistage,
  title = {{A Multistage Extraction Pipeline for Long Scanned Financial Documents: An Empirical Study in Industrial KYC Workflows}},
  author = {Han Yuxuan, Yuanxing Zhang, Yushuo Wang and Yichao Jin},
  booktitle = {Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 6: Industry Track)},
  year = {2026},
  doi = {10.18653/v1/2026.acl-industry.99},
  url = {https://doi.org/10.18653/v1/2026.acl-industry.99}
}
```

- [← previous A multimodal system to support disaster risk management](https://md.research.elamur.ai/free/en/papers/lucena-2026-sistema-multimodal-de-apoio-ao/)
- [next → A Multitask Transformer for Offensive Language Detection and Target Identification in HateBR](https://md.research.elamur.ai/free/en/papers/silva-2026-multitask-transformer-offensive-language-detection/)
- [↑ back to papers of 2026](https://md.research.elamur.ai/free/en/papers/2026/page/20/)

## References

36 references, 13 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. Riaz Ahmad, Saeeda Naz, and Imran Razzak. 2021. Efficient skew detection and correction in scanned document images through clustering of probabilistic hough transforms. Pattern recognition letters, 152:93–99. [find](https://scholar.google.com/scholar?q=Riaz+Ahmad%2C+Saeeda+Naz%2C+and+Imran+Razzak%2E+2021%2E+Efficient+skew+detection+and+correction+in+scanned+document+images+through+clustering+of+probabilistic+hough+transforms%2E+Pattern+recognition+letters%2C+15)
2. Hayato Aida, Kosuke Takahashi, and Takahiro Omi. 2025. Enhancing large vision-language models with layout modality for table question answering on japanese annual securities reports. arXiv preprint arXiv:2505.17625. — [arXiv:2505.17625](https://arxiv.org/abs/2505.17625) [find](https://scholar.google.com/scholar?q=Hayato+Aida%2C+Kosuke+Takahashi%2C+and+Takahiro+Omi%2E+2025%2E+Enhancing+large+vision%2Dlanguage+models+with+layout+modality+for+table+question+answering+on+japanese+annual+securities+reports%2E+arXiv+preprint+ar)
3. Shaheera Saba Mohd Naseem Akhter and Priti P Rege. 2020. Improving skew detection and correction in different document images using a deep learning approach. In 2020 11th International Conference on Computing, Communication and Networking Technologies (ICCCNT), pages 1–6, Online. IEEE, IEEE. [find](https://scholar.google.com/scholar?q=Shaheera+Saba+Mohd+Naseem+Akhter+and+Priti+P+Rege%2E+2020%2E+Improving+skew+detection+and+correction+in+different+document+images+using+a+deep+learning+approach%2E+In+2020+11th+International+Conference+on+C)
4. Zahra Anvari and Vassilis Athitsos. 2021. A survey on deep learning based document image enhancement. arXiv preprint arXiv:2112.02719. — [arXiv:2112.02719](https://arxiv.org/abs/2112.02719) [find](https://scholar.google.com/scholar?q=Zahra+Anvari+and+Vassilis+Athitsos%2E+2021%2E+A+survey+on+deep+learning+based+document+image+enhancement%2E+arXiv+preprint+arXiv%3A2112%2E02719%2E+%E2%80%94+%5BarXiv%3A2112%2E02719%5D%28https%3A%2F%2Farxiv%2Eorg%2Fabs%2F2112%2E02719%29)
5. Shuai Bai, Yuxuan Cai, Ruizhe Chen, Keqin Chen, Xionghui Chen, Zesen Cheng, Lianghao Deng, Wei Ding, Chang Gao, Chunjiang Ge, and 1 others. 2025. [Qwen3-vl technical report](https://md.research.elamur.ai/free/en/papers/bai-2025-qwen3-vl-technical-report/). arXiv preprint arXiv:2511.21631.
6. Ethan Bradley, Muhammad Roman, Karen Rafferty, and Barry Devereux. 2026. Synfintabs: A dataset of synthetic financial tables for information and table extraction. In Document Analysis and Recognition – ICDAR 2025 Workshops, pages 85–100, Cham, Switzerland. Springer Nature. [find](https://scholar.google.com/scholar?q=Ethan+Bradley%2C+Muhammad+Roman%2C+Karen+Rafferty%2C+and+Barry+Devereux%2E+2026%2E+Synfintabs%3A+A+dataset+of+synthetic+financial+tables+for+information+and+table+extraction%2E+In+Document+Analysis+and+Recognition+)
7. Kai Chen, Mathias Seuret, Jean Hennebert, and Rolf Ingold. 2017. Convolutional neural networks for page segmentation of historical document images. In 2017 14th IAPR International Conference on Document Analysis and Recognition (ICDAR), volume 1, pages 965–970. IEEE. [find](https://scholar.google.com/scholar?q=Kai+Chen%2C+Mathias+Seuret%2C+Jean+Hennebert%2C+and+Rolf+Ingold%2E+2017%2E+Convolutional+neural+networks+for+page+segmentation+of+historical+document+images%2E+In+2017+14th+IAPR+International+Conference+on+Docume)
8. Zhiyu Chen, Wenhu Chen, Charese Smiley, Sameena Shah, Iana Borova, Dylan Langdon, Reema Moussa, Matt Beane, Ting-Hao Huang, Bryan Routledge, and William Yang Wang. 2021. [Finqa: A dataset of numerical reasoning over financial data](https://md.research.elamur.ai/free/en/papers/chen-2021-finqa-dataset-numerical-reasoning-over/). In Proceedings of the 2021 Conference on Empirical Methods in Natural…
9. Cheng Cui, Tingquan Gao, Shengyu Wei, Yuning Du, Ruoyu Guo, Shuilong Dong, Bin Lu, Ying Zhou, Xueying Lv, Qiwen Liu, and 1 others. 2021. Pp-lcnet: A lightweight cpu convolutional neural network. arXiv preprint arXiv:2109.15099. — [arXiv:2109.15099](https://arxiv.org/abs/2109.15099) [find](https://scholar.google.com/scholar?q=Cheng+Cui%2C+Tingquan+Gao%2C+Shengyu+Wei%2C+Yuning+Du%2C+Ruoyu+Guo%2C+Shuilong+Dong%2C+Bin+Lu%2C+Ying+Zhou%2C+Xueying+Lv%2C+Qiwen+Liu%2C+and+1+others%2E+2021%2E+Pp%2Dlcnet%3A+A+lightweight+cpu+convolutional+neural+network%2E+arXiv)
10. Cheng Cui, Ting Sun, Manhui Lin, Tingquan Gao, Yubo Zhang, Jiaxuan Liu, Xueqing Wang, Zelun Zhang, Changda Zhou, Hongen Liu, Yue Zhang, Wenyu Lv, Kui Huang, Yichao Zhang, Jing Zhang, Jun Zhang, Yi Liu, Dianhai Yu, and Yanjun Ma. 2025. Paddleocr 3.0 technical report. arXiv preprint arXiv:2507.05595. — [arXiv:2507.05595](https://arxiv.org/abs/2507.05595) [find](https://scholar.google.com/scholar?q=Cheng+Cui%2C+Ting+Sun%2C+Manhui+Lin%2C+Tingquan+Gao%2C+Yubo+Zhang%2C+Jiaxuan+Liu%2C+Xueqing+Wang%2C+Zelun+Zhang%2C+Changda+Zhou%2C+Hongen+Liu%2C+Yue+Zhang%2C+Wenyu+Lv%2C+Kui+Huang%2C+Yichao+Zhang%2C+Jing+Zhang%2C+Jun+Zhang%2C+Yi+Liu)
11. Jake Poznanski, Aman Rangapur, Jon Borchardt, Jason Dunkelberger, Christopher Wilhelm, Kyle Lo, and Luca Soldaini. 2025. [olmocr: Unlocking trillions of tokens in pdfs with vision language models](https://md.research.elamur.ai/free/en/papers/poznanski-2025-olmocr-unlocking-trillions-tokens-pdfs/). In Proceedings of the CODEML Workshop at ICML 2025.
12. Manuel Faysse, Hugues Sibille, Tony Wu, Bilel Omrani, Gautier Viaud, CELINE HUDELOT, and Pierre Colombo. 2025. [Colpali: Efficient document retrieval with vision language models](https://md.research.elamur.ai/free/en/papers/faysse-2024-colpali-efficient-document-retrieval-vision/). In The Thirteenth International Conference on Learning Representations (ICLR) 2025.
13. Ravi K. Rajendran, Biplob Debnath, Murugan Sankaradass, and Srimat Chakradhar. 2025. Ecodoc: A cost-efficient multimodal document processing system for enterprises using llms. In Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 6: Industry Track). Assoc… [find](https://scholar.google.com/scholar?q=Ravi+K%2E+Rajendran%2C+Biplob+Debnath%2C+Murugan+Sankaradass%2C+and+Srimat+Chakradhar%2E+2025%2E+Ecodoc%3A+A+cost%2Defficient+multimodal+document+processing+system+for+enterprises+using+llms%2E+In+Proceedings+of+the+63)
14. Dianyuan Han. 2013. Comparison of commonly used image interpolation methods. In Conference of the 2nd International Conference on Computer Science and Electronics Engineering (ICCSEE 2013), pages 1556–1559. Atlantis Press. [find](https://scholar.google.com/scholar?q=Dianyuan+Han%2E+2013%2E+Comparison+of+commonly+used+image+interpolation+methods%2E+In+Conference+of+the+2nd+International+Conference+on+Computer+Science+and+Electronics+Engineering+%28ICCSEE+2013%29%2C+pages+1556)
15. Nils Reimers and Iryna Gurevych. 2019. [Sentence-bert: Sentence embeddings using siamese bert-networks](https://md.research.elamur.ai/free/en/papers/reimers-2019-sentence-bert-sentence-embeddings-siamese/). In Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing (EMNLP-IJCNLP), pages 3982–3992.
16. Anwen Hu, Haiyang Xu, Jiabo Ye, Ming Yan, Liang Zhang, Bo Zhang, Ji Zhang, Qin Jin, Fei Huang, and Jingren Zhou. 2024. [mplug-docowl 1.5: Unified structure learning for ocr-free document understanding](https://md.research.elamur.ai/free/en/papers/hu-2024-mplug-docowl-1-5-unified/). In Findings of the Association for Computational Linguistics: EMNLP 2024, pages 3096–3120, Miami, F…
17. Ali M Reza. 2004. Realization of the contrast limited adaptive histogram equalization (clahe) for real-time image enhancement. Journal of VLSI signal processing systems for signal, image and video technology, 38(1):35–44. [find](https://scholar.google.com/scholar?q=Ali+M+Reza%2E+2004%2E+Realization+of+the+contrast+limited+adaptive+histogram+equalization+%28clahe%29+for+real%2Dtime+image+enhancement%2E+Journal+of+VLSI+signal+processing+systems+for+signal%2C+image+and+video+tec)
18. Anwen Hu, Haiyang Xu, Liang Zhang, Jiabo Ye, Ming Yan, Ji Zhang, Qin Jin, Fei Huang, and Jingren Zhou. 2025. [mplug-docowl2: High-resolution compressing for OCR-free multi-page document understanding](https://md.research.elamur.ai/free/en/papers/hu-2025-mplug-docowl2-high-resolution-compressing/). In Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1…
19. Li Im Tan, Wai San Phang, Kim On Chin, and Anthony Patricia. 2015. Rule-based sentiment analysis for financial news. In 2015 IEEE International Conference on Systems, Man, and Cybernetics, pages 1601–1606. IEEE. [find](https://scholar.google.com/scholar?q=Li+Im+Tan%2C+Wai+San+Phang%2C+Kim+On+Chin%2C+and+Anthony+Patricia%2E+2015%2E+Rule%2Dbased+sentiment+analysis+for+financial+news%2E+In+2015+IEEE+International+Conference+on+Systems%2C+Man%2C+and+Cybernetics%2C+pages+1601%E2%80%93)
20. JaidedAI. 2020. Easyocr. [find](https://scholar.google.com/scholar?q=JaidedAI%2E+2020%2E+Easyocr%2E)
21. Wang Jiapeng, Jin Lianwen, and Ding Kai. 2022. Lilt: A simple yet effective language-independent layout transformer for structured document understanding. In Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 7747–7757, Dublin, Irel… [find](https://scholar.google.com/scholar?q=Wang+Jiapeng%2C+Jin+Lianwen%2C+and+Ding+Kai%2E+2022%2E+Lilt%3A+A+simple+yet+effective+language%2Dindependent+layout+transformer+for+structured+document+understanding%2E+In+Proceedings+of+the+60th+Annual+Meeting+of+)
22. Geewook Kim, Teakgyu Hong, Moonbin Yim, JeongYeon Nam, Jinyoung Park, Jinyeong Yim, Wonseok Hwang, Sangdoo Yun, Dongyoon Han, and Seunghyun Park. 2022. [Ocr-free document understanding transformer](https://md.research.elamur.ai/free/en/papers/kim-2021-ocr-free-document-understanding-transformer/). In Computer Vision – ECCV 2022: 17th European Conference, Tel Aviv, Israel, October 23–27, 2022, Procee…
23. Sheikh Mahmudul and Conlon Sumali. 2012. A rule-based system to extract financial information. Journal of Computer Information Systems, 52(4):10–19. [find](https://scholar.google.com/scholar?q=Sheikh+Mahmudul+and+Conlon+Sumali%2E+2012%2E+A+rule%2Dbased+system+to+extract+financial+information%2E+Journal+of+Computer+Information+Systems%2C+52%284%29%3A10%E2%80%9319%2E)
24. Stephen Robertson and Hugo Zaragoza. 2009. The probabilistic relevance framework: Bm25 and beyond. Foundations and Trends in Information Retrieval, 3(4):333–389. [find](https://scholar.google.com/scholar?q=Stephen+Robertson+and+Hugo+Zaragoza%2E+2009%2E+The+probabilistic+relevance+framework%3A+Bm25+and+beyond%2E+Foundations+and+Trends+in+Information+Retrieval%2C+3%284%29%3A333%E2%80%93389%2E)
25. Kuldeep Singh, Simerjot Kaur, and Charese Smiley. 2024. Finqapt: Empowering financial decisions with end-to-end llm-driven question answering pipeline. In Proceedings of the 5th ACM International Conference on AI in Finance, pages 266–273. [find](https://scholar.google.com/scholar?q=Kuldeep+Singh%2C+Simerjot+Kaur%2C+and+Charese+Smiley%2E+2024%2E+Finqapt%3A+Empowering+financial+decisions+with+end%2Dto%2Dend+llm%2Ddriven+question+answering+pipeline%2E+In+Proceedings+of+the+5th+ACM+International+Conf)
26. Ray Smith. 2007. An overview of the tesseract ocr engine. In Ninth international conference on document analysis and recognition (ICDAR 2007), volume 2, pages 629–633. IEEE. [find](https://scholar.google.com/scholar?q=Ray+Smith%2E+2007%2E+An+overview+of+the+tesseract+ocr+engine%2E+In+Ninth+international+conference+on+document+analysis+and+recognition+%28ICDAR+2007%29%2C+volume+2%2C+pages+629%E2%80%93633%2E+IEEE%2E)
27. Archita Srivastava, Abhas Kumar, Rajesh Kumar, and Prabhakar Srinivasan. 2025. Enhancing financial vqa in vision language models using intermediate structured representations. arXiv preprint arXiv:2501.04675. — [arXiv:2501.04675](https://arxiv.org/abs/2501.04675) [find](https://scholar.google.com/scholar?q=Archita+Srivastava%2C+Abhas+Kumar%2C+Rajesh+Kumar%2C+and+Prabhakar+Srinivasan%2E+2025%2E+Enhancing+financial+vqa+in+vision+language+models+using+intermediate+structured+representations%2E+arXiv+preprint+arXiv%3A250)
28. Jin Khye Tan, En Jun Choong, Ethan Jeremiah Chitty, Yan Pheng Choo, John Hsin Yang Wong, and Chern Eu Cheah. 2025. Fine-tuning vision-language models for markdown conversion of financial tables in malaysian audited financial reports. arXiv preprint arXiv:2508.05669. Accepted at ICDE 2025 research pa… — [arXiv:2508.05669](https://arxiv.org/abs/2508.05669) [find](https://scholar.google.com/scholar?q=Jin+Khye+Tan%2C+En+Jun+Choong%2C+Ethan+Jeremiah+Chitty%2C+Yan+Pheng+Choo%2C+John+Hsin+Yang+Wong%2C+and+Chern+Eu+Cheah%2E+2025%2E+Fine%2Dtuning+vision%2Dlanguage+models+for+markdown+conversion+of+financial+tables+in+mal)
29. Guobo Xie and Wen Lu. 2013. Image edge detection based on opencv. International Journal of Electronics and Electrical Engineering, 1(2):104–106. [find](https://scholar.google.com/scholar?q=Guobo+Xie+and+Wen+Lu%2E+2013%2E+Image+edge+detection+based+on+opencv%2E+International+Journal+of+Electronics+and+Electrical+Engineering%2C+1%282%29%3A104%E2%80%93106%2E)
30. Xu Yang, Xu Yiheng, Lv Tengchao, Cui Lei, Wei Furu, Wang Guoxin, Lu Yijuan, Florencio Dinei, Zhang Cha, Che Wanxiang, Zhang Min, and Zhou Lidong. 2021. [Layoutlmv2: Multi-modal pre-training for visually-rich document understanding](https://md.research.elamur.ai/free/en/papers/xu-2021-layoutlmv2-multi-modal-pre-training/). In Proceedings of the 59th Annual Meeting of the Association for Comp…
31. Yuan Yao, Tianyu Yu, Ao Zhang, Chongyi Wang, Junbo Cui, Hongji Zhu, Tianchi Cai, Haoyu Li, Weilin Zhao, Zhihui He, and 1 others. 2024. [Minicpm-v: A gpt-4v level mllm on your phone](https://md.research.elamur.ai/free/en/papers/yao-2024-minicpm-v-gpt-4v/). arXiv preprint arXiv:2408.01800.
32. Jiabo Ye, Anwen Hu, Haiyang Xu, Qinghao Ye, Ming Yan, Guohai Xu, Chenliang Li, Junfeng Tian, Qi Qian, Ji Zhang, Qin Jin, Liang He, Xin Lin, and Fei Huang. 2023. [Ureader: Universal ocr-free visually-situated language understanding with multimodal large language model](https://md.research.elamur.ai/free/en/papers/ye-2023-ureader-universal-ocr-free-visually/). In Findings of the Association f…
33. Xu Yiheng, Li Minghao, Cui Lei, Huang Shaohan, Wei Furu, and Zhou Ming. 2020. [Layoutlm: Pre-training of text and layout for document image understanding](https://md.research.elamur.ai/free/en/papers/xu-2019-layoutlm-pre-training-text-layout/). In Proceedings of the 26th ACM SIGKDD international conference on knowledge discovery & data mining, pages 1192–1200.
34. Shi Yu, Chaoyue Tang, Bokai Xu, Junbo Cui, Junhao Ran, Yukun Yan, Zhenghao Liu, Shuo Wang, Xu Han, Zhiyuan Liu, and Maosong Sun. 2025. Visrag: Vision-based retrieval-augmented generation on multi-modality documents. In The Thirteenth International Conference on Learning Representations (ICLR) 2025. [find](https://scholar.google.com/scholar?q=Shi+Yu%2C+Chaoyue+Tang%2C+Bokai+Xu%2C+Junbo+Cui%2C+Junhao+Ran%2C+Yukun+Yan%2C+Zhenghao+Liu%2C+Shuo+Wang%2C+Xu+Han%2C+Zhiyuan+Liu%2C+and+Maosong+Sun%2E+2025%2E+Visrag%3A+Vision%2Dbased+retrieval%2Daugmented+generation+on+multi%2Dmoda)
35. Huang Yupan, Lv Tengchao, Cui Lei, Lu Yutong, and Wei Furu. 2022. [Layoutlmv3: Pre-training for document ai with unified text and image masking](https://md.research.elamur.ai/free/en/papers/huang-2022-layoutlmv3-pre-training-document-ai/). In Proceedings of the 30th ACM international conference on multimedia, pages 4083–4091.
36. Li Zihan, Huang Wenhui, and Cui Lei. 2024. Docllm: Document representation learning with large language models. In Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 8529–8548. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Li+Zihan%2C+Huang+Wenhui%2C+and+Cui+Lei%2E+2024%2E+Docllm%3A+Document+representation+learning+with+large+language+models%2E+In+Proceedings+of+the+62nd+Annual+Meeting+of+the+Association+for+Computational+Linguisti)

*Updated 2 November 2026*