---
title: "DeepSeek-OCR: Contexts Optical Compression (2025)"
description: "This design keeps memory low on high-resolution pages and produces few tokens. 42 references, cited by 48 papers here."
authors: ["Haoran Wei", "Yaofeng Sun", "Yukun Li"]
year: 2025
venue: "arXiv:2510.18234"
source: "https://arxiv.org/abs/2510.18234"
doi: "10.48550/arxiv.2510.18234"
arxiv: "2510.18234"
licence: "CC BY 4.0 (arXiv)"
licence_url: "http://creativecommons.org/licenses/by/4.0/"
summary_licence: "Elamur Cosmic License"
canonical: "https://research.elamur.ai/free/en/papers/wei-2025-deepseek-ocr-contexts-optical/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/wei-2025-deepseek-ocr-contexts-optical/)

[home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [2025](https://md.research.elamur.ai/free/en/papers/2025/page/2/) / DeepSeek-OCR: Contexts Optical Compression

# DeepSeek-OCR: Contexts Optical Compression

**Authors:** Haoran Wei, [Yaofeng Sun](https://md.research.elamur.ai/free/en/authors/yaofeng-sun/) and [Yukun Li](https://md.research.elamur.ai/free/en/authors/yukun-li/) · **Year:** 2025 · **Venue:** arXiv:2510.18234 · **Licence:** [CC BY 4.0 (arXiv)](http://creativecommons.org/licenses/by/4.0/)

**Links:** [Publisher's page](https://arxiv.org/abs/2510.18234) · [PDF](https://arxiv.org/pdf/2510.18234) · [DOI 10.48550/arxiv.2510.18234](https://doi.org/10.48550/arxiv.2510.18234) · [arXiv 2510.18234](https://arxiv.org/abs/2510.18234) · [OpenAlex](https://openalex.org/W4416055645)

**Citations:** 4 citations (OpenAlex, 2026-10-14) · [48](https://md.research.elamur.ai/free/en/papers/wei-2025-deepseek-ocr-contexts-optical/cited-by/) in our library, of the 141,840 papers whose bibliographies we read

## Summary
Because attention cost grows quadratically with sequence length, long contexts are expensive for LLMs. The authors ask whether rendering text as an image could carry the same content in far fewer tokens. OCR serves as a measurable test: the model must recover the original text from a small set of vision tokens. DeepSeek-OCR pairs a new encoder, DeepEncoder (~380M parameters), with a DeepSeek3B-MoE-A570M decoder. DeepEncoder chains three parts in sequence: an 80M SAM-base stage using window attention, a 16× convolutional token compressor, and a 300M CLIP-large stage with global attention. This design keeps memory low on high-resolution pages and produces few tokens. On the Fox benchmark, decoding is about 97% precise when text tokens outnumber vision tokens by less than ten to one, and it drops to roughly 60% at 20×. On OmniDocBench, it leads end-to-end models while using the fewest vision tokens. The work is a proof-of-concept for compressing long histories and for modeling gradual "forgetting" in LLMs. It is also a practical tool for producing pretraining data, and the code and weights are open.

## Key points
- On Fox, precision is 96%+ at 9–10× compression, about 90% at 10–12×, and about 60% at 20× (text tokens per vision token).
- On OmniDocBench, it outperforms GOT-OCR2.0 (256 tokens/page) with only 100 vision tokens, and it beats MinerU2.0 (6000+ tokens/page on average) with fewer than 800.
- Throughput is 200k+ pages/day on a single A100-40G, or about 33M pages/day on 20 nodes with 8 A100-40G each.
- Beyond OCR, the model handles charts, chemical formulas, simple geometry, and natural images. It has no SFT stage, so it is not a chatbot and needs completion-style prompts.
- Code and weights are available at github.com/deepseek-ai/DeepSeek-OCR. Planned future checks include interleaved digital-optical pretraining and needle-in-a-haystack tests.

*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 (2): [🖥️ Computer Science](https://md.research.elamur.ai/free/en/shelves/computer-science/) · 9,085 works · [👁️ Computer Vision and Pattern Recognition](https://md.research.elamur.ai/free/en/shelves/computer-science/computer-vision-and-pattern-recognition/) · 1,282 works

🧭 Topics (1): [🪶 Efficient models & compression](https://md.research.elamur.ai/free/en/topics/efficient-models/) · 77 works

## Cite

Cite as: Haoran Wei, Yaofeng Sun and Yukun Li. 2025. DeepSeek-OCR: Contexts Optical Compression. arXiv:2510.18234.

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

```bibtex
@misc{wei2025deepseek,
  title = {{DeepSeek-OCR: Contexts Optical Compression}},
  author = {Haoran Wei, Yaofeng Sun and Yukun Li},
  year = {2025},
  doi = {10.48550/arxiv.2510.18234},
  eprint = {2510.18234},
  archivePrefix = {arXiv},
  url = {https://doi.org/10.48550/arxiv.2510.18234}
}
```

- [← previous Kimi-Audio Technical Report](https://md.research.elamur.ai/free/en/papers/kimiteam-2025-kimi-audio-technical-report/)
- [next → Pragmatics in the Era of Large Language Models: A Survey on Datasets, Evaluation, Opportunities and Challenges](https://md.research.elamur.ai/free/en/papers/ma-2025-pragmatics-era-large-language-models/)
- [↑ back to papers of 2025](https://md.research.elamur.ai/free/en/papers/2025/page/2/)

## References

42 references, 14 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. Marker. URL [https://github.com/datalab-to/marker](https://github.com/datalab-to/marker) . [↗](https://github.com/datalab-to/marker)
2. Mathpix. URL [https://mathpix.com/](https://mathpix.com/) . [↗](https://mathpix.com/)
3. Ocrflux, 2025. URL [https://github.com/chatdoc-com/OCRFlux](https://github.com/chatdoc-com/OCRFlux) . [↗](https://github.com/chatdoc-com/OCRFlux)
4. G. AI. Gemini 2.5-pro, 2025. URL [https://gemini.google.com/](https://gemini.google.com/) . [↗](https://gemini.google.com/)
5. S. Bai, K. Chen, X. Liu, J. Wang, W. Ge, S. Song, K. Dang, P. Wang, S. Wang, J. Tang, H. Zhong, Y. Zhu, M. Yang, Z. Li, J. Wan, P. Wang, W. Ding, Z. Fu, Y. Xu, J. Ye, X. Zhang, T. Xie, Z. Cheng, H. Zhang, Z. Yang, H. Xu, and J. Lin. [Qwen2.5-vl technical report](https://md.research.elamur.ai/free/en/papers/bai-2025-qwen2-5-vl-technical/). arXiv preprint arXiv:2502.13923, 2025.
6. L. Blecher, G. Cucurull, T. Scialom, and R. Stojnic. Nougat: Neural optical understanding for academic documents. arXiv preprint arXiv:2308.13418, 2023. [find](https://scholar.google.com/scholar?q=L%2E+Blecher%2C+G%2E+Cucurull%2C+T%2E+Scialom%2C+and+R%2E+Stojnic%2E+Nougat%3A+Neural+optical+understanding+for+academic+documents%2E+arXiv+preprint+arXiv%3A2308%2E13418%2C+2023%2E)
7. J. Chen, L. Kong, H. Wei, C. Liu, Z. Ge, L. Zhao, J. Sun, C. Han, and X. Zhang. Onechart: Purify the chart structural extraction via one auxiliary token. In Proceedings of the 32nd ACM International Conference on Multimedia, pages 147–155, 2024. [find](https://scholar.google.com/scholar?q=J%2E+Chen%2C+L%2E+Kong%2C+H%2E+Wei%2C+C%2E+Liu%2C+Z%2E+Ge%2C+L%2E+Zhao%2C+J%2E+Sun%2C+C%2E+Han%2C+and+X%2E+Zhang%2E+Onechart%3A+Purify+the+chart+structural+extraction+via+one+auxiliary+token%2E+In+Proceedings+of+the+32nd+ACM+International+C)
8. Z. Chen, W. Wang, H. Tian, S. Ye, Z. Gao, E. Cui, W. Tong, K. Hu, J. Luo, Z. Ma, et al. [How far are we to gpt-4v? closing the gap to commercial multimodal models with open-source suites](https://md.research.elamur.ai/free/en/papers/chen-2024-far-gpt-4v-closing/). arXiv preprint arXiv:2404.16821, 2024.
9. C. Cui, T. Sun, M. Lin, T. Gao, Y. Zhang, J. Liu, X. Wang, Z. Zhang, C. Zhou, H. Liu, et al. Paddleocr 3.0 technical report. arXiv preprint arXiv:2507.05595, 2025. — [arXiv:2507.05595](https://arxiv.org/abs/2507.05595) [find](https://scholar.google.com/scholar?q=C%2E+Cui%2C+T%2E+Sun%2C+M%2E+Lin%2C+T%2E+Gao%2C+Y%2E+Zhang%2C+J%2E+Liu%2C+X%2E+Wang%2C+Z%2E+Zhang%2C+C%2E+Zhou%2C+H%2E+Liu%2C+et+al%2E+Paddleocr+3%2E0+technical+report%2E+arXiv+preprint+arXiv%3A2507%2E05595%2C+2025%2E+%E2%80%94+%5BarXiv%3A2507%2E05595%5D%28https%3A%2F%2Farxiv%2Eo)
10. M. Dehghani, J. Djolonga, B. Mustafa, P. Padlewski, J. Heek, J. Gilmer, A. Steiner, M. Caron, R. Geirhos, I. Alabdulmohsin, et al. Patch n’ pack: Navit, a vision transformer for any aspect ratio and resolution. Advances in Neural Information Processing Systems, 36:3632–3656, 2023. [find](https://scholar.google.com/scholar?q=M%2E+Dehghani%2C+J%2E+Djolonga%2C+B%2E+Mustafa%2C+P%2E+Padlewski%2C+J%2E+Heek%2C+J%2E+Gilmer%2C+A%2E+Steiner%2C+M%2E+Caron%2C+R%2E+Geirhos%2C+I%2E+Alabdulmohsin%2C+et+al%2E+Patch+n%E2%80%99+pack%3A+Navit%2C+a+vision+transformer+for+any+aspect+ratio+and+r)
11. H. Feng, S. Wei, X. Fei, W. Shi, Y. Han, L. Liao, J. Lu, B. Wu, Q. Liu, C. Lin, et al. Dolphin: Document image parsing via heterogeneous anchor prompting. arXiv preprint arXiv:2505.14059, 2025. [find](https://scholar.google.com/scholar?q=H%2E+Feng%2C+S%2E+Wei%2C+X%2E+Fei%2C+W%2E+Shi%2C+Y%2E+Han%2C+L%2E+Liao%2C+J%2E+Lu%2C+B%2E+Wu%2C+Q%2E+Liu%2C+C%2E+Lin%2C+et+al%2E+Dolphin%3A+Document+image+parsing+via+heterogeneous+anchor+prompting%2E+arXiv+preprint+arXiv%3A2505%2E14059%2C+2025%2E)
12. Y. Goyal, T. Khot, D. Summers-Stay, D. Batra, and D. Parikh. [Making the v in vqa matter: Elevating the role of image understanding in visual question answering](https://md.research.elamur.ai/free/en/papers/goyal-2016-making-v-vqa-matter-elevating/). In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 6904–6913, 2017.
13. J. Gu, X. Meng, G. Lu, L. Hou, N. Minzhe, X. Liang, L. Yao, R. Huang, W. Zhang, X. Jiang, et al. Wukong: A 100 million large-scale chinese cross-modal pre-training benchmark. Advances in Neural Information Processing Systems, 35:26418–26431, 2022. [find](https://scholar.google.com/scholar?q=J%2E+Gu%2C+X%2E+Meng%2C+G%2E+Lu%2C+L%2E+Hou%2C+N%2E+Minzhe%2C+X%2E+Liang%2C+L%2E+Yao%2C+R%2E+Huang%2C+W%2E+Zhang%2C+X%2E+Jiang%2C+et+al%2E+Wukong%3A+A+100+million+large%2Dscale+chinese+cross%2Dmodal+pre%2Dtraining+benchmark%2E+Advances+in+Neural+Inform)
14. High-flyer. HAI-LLM: Efficient and lightweight training tool for large models, 2023. URL [https://www.high-flyer.cn/en/blog/hai-llm](https://www.high-flyer.cn/en/blog/hai-llm) . [↗](https://www.high-flyer.cn/en/blog/hai-llm)
15. S. Iyer, X. V. Lin, R. Pasunuru, T. Mihaylov, D. Simig, P. Yu, K. Shuster, T. Wang, Q. Liu, P. S. Koura, et al. [Opt-iml: Scaling language model instruction meta learning through the lens of generalization](https://md.research.elamur.ai/free/en/papers/iyer-2022-opt-iml-scaling-language-model/). arXiv preprint arXiv:2212.12017, 2022.
16. S. Kazemzadeh, V. Ordonez, M. Matten, and T. Berg. [Referitgame: Referring to objects in photographs of natural scenes](https://md.research.elamur.ai/free/en/papers/kazemzadeh-2014-referitgame-referring-objects-photographs-natural/). In Proceedings of the 2014 conference on empirical methods in natural language processing (EMNLP), pages 787–798, 2014.
17. A. Kirillov, E. Mintun, N. Ravi, H. Mao, C. Rolland, L. Gustafson, T. Xiao, S. Whitehead, A. C. Berg, W.-Y. Lo, et al. [Segment anything](https://md.research.elamur.ai/free/en/papers/kirillov-2023-segment-anything/). arXiv preprint arXiv:2304.02643, 2023.
18. Z. Li, Y. Liu, Q. Liu, Z. Ma, Z. Zhang, S. Zhang, Z. Guo, J. Zhang, X. Wang, and X. Bai. Monkeyocr: Document parsing with a structure-recognition-relation triplet paradigm. arXiv preprint arXiv:2506.05218, 2025. [find](https://scholar.google.com/scholar?q=Z%2E+Li%2C+Y%2E+Liu%2C+Q%2E+Liu%2C+Z%2E+Ma%2C+Z%2E+Zhang%2C+S%2E+Zhang%2C+Z%2E+Guo%2C+J%2E+Zhang%2C+X%2E+Wang%2C+and+X%2E+Bai%2E+Monkeyocr%3A+Document+parsing+with+a+structure%2Drecognition%2Drelation+triplet+paradigm%2E+arXiv+preprint+arXiv%3A2506%2E0)
19. A. Liu, B. Feng, B. Wang, B. Wang, B. Liu, C. Zhao, C. Dengr, C. Ruan, D. Dai, D. Guo, et al. Deepseek-v2: A strong, economical, and efficient mixture-of-experts language model. arXiv preprint arXiv:2405.04434, 2024. [find](https://scholar.google.com/scholar?q=A%2E+Liu%2C+B%2E+Feng%2C+B%2E+Wang%2C+B%2E+Wang%2C+B%2E+Liu%2C+C%2E+Zhao%2C+C%2E+Dengr%2C+C%2E+Ruan%2C+D%2E+Dai%2C+D%2E+Guo%2C+et+al%2E+Deepseek%2Dv2%3A+A+strong%2C+economical%2C+and+efficient+mixture%2Dof%2Dexperts+language+model%2E+arXiv+preprint+arXiv%3A2)
20. A. Liu, B. Feng, B. Xue, B. Wang, B. Wu, C. Lu, C. Zhao, C. Deng, C. Zhang, C. Ruan, et al. Deepseek-v3 technical report. arXiv preprint arXiv:2412.19437, 2024. [find](https://scholar.google.com/scholar?q=A%2E+Liu%2C+B%2E+Feng%2C+B%2E+Xue%2C+B%2E+Wang%2C+B%2E+Wu%2C+C%2E+Lu%2C+C%2E+Zhao%2C+C%2E+Deng%2C+C%2E+Zhang%2C+C%2E+Ruan%2C+et+al%2E+Deepseek%2Dv3+technical+report%2E+arXiv+preprint+arXiv%3A2412%2E19437%2C+2024%2E)
21. C. Liu, H. Wei, J. Chen, L. Kong, Z. Ge, Z. Zhu, L. Zhao, J. Sun, C. Han, and X. Zhang. Focus anywhere for fine-grained multi-page document understanding. arXiv preprint arXiv:2405.14295, 2024. — [arXiv:2405.14295](https://arxiv.org/abs/2405.14295) [find](https://scholar.google.com/scholar?q=C%2E+Liu%2C+H%2E+Wei%2C+J%2E+Chen%2C+L%2E+Kong%2C+Z%2E+Ge%2C+Z%2E+Zhu%2C+L%2E+Zhao%2C+J%2E+Sun%2C+C%2E+Han%2C+and+X%2E+Zhang%2E+Focus+anywhere+for+fine%2Dgrained+multi%2Dpage+document+understanding%2E+arXiv+preprint+arXiv%3A2405%2E14295%2C+2024%2E+%E2%80%94+%5BarX)
22. I. Loshchilov and F. Hutter. Sgdr: Stochastic gradient descent with warm restarts. arXiv preprint arXiv:1608.03983, 2016. [find](https://scholar.google.com/scholar?q=I%2E+Loshchilov+and+F%2E+Hutter%2E+Sgdr%3A+Stochastic+gradient+descent+with+warm+restarts%2E+arXiv+preprint+arXiv%3A1608%2E03983%2C+2016%2E)
23. I. Loshchilov and F. Hutter. Decoupled weight decay regularization. In ICLR, 2019. [find](https://scholar.google.com/scholar?q=I%2E+Loshchilov+and+F%2E+Hutter%2E+Decoupled+weight+decay+regularization%2E+In+ICLR%2C+2019%2E)
24. A. Masry, D. X. Long, J. Q. Tan, S. Joty, and E. Hoque. [Chartqa: A benchmark for question answering about charts with visual and logical reasoning](https://md.research.elamur.ai/free/en/papers/masry-2022-chartqa-benchmark-question-answering-about/). arXiv preprint arXiv:2203.10244, 2022.
25. A. Nassar, A. Marafioti, M. Omenetti, M. Lysak, N. Livathinos, C. Auer, L. Morin, R. T. de Lima, Y. Kim, A. S. Gurbuz, et al. Smoldocling: An ultra-compact vision-language model for end-to-end multi-modal document conversion. arXiv preprint arXiv:2503.11576, 2025. — [arXiv:2503.11576](https://arxiv.org/abs/2503.11576) [find](https://scholar.google.com/scholar?q=A%2E+Nassar%2C+A%2E+Marafioti%2C+M%2E+Omenetti%2C+M%2E+Lysak%2C+N%2E+Livathinos%2C+C%2E+Auer%2C+L%2E+Morin%2C+R%2E+T%2E+de+Lima%2C+Y%2E+Kim%2C+A%2E+S%2E+Gurbuz%2C+et+al%2E+Smoldocling%3A+An+ultra%2Dcompact+vision%2Dlanguage+model+for+end%2Dto%2Dend+multi%2Dm)
26. OpenAI. [Gpt-4 technical report](https://md.research.elamur.ai/free/en/papers/openai-2023-gpt-4-technical-report/), 2023.
27. L. Ouyang, Y. Qu, H. Zhou, J. Zhu, R. Zhang, Q. Lin, B. Wang, Z. Zhao, M. Jiang, X. Zhao, et al. Omnidocbench: Benchmarking diverse pdf document parsing with comprehensive annotations. In Proceedings of the Computer Vision and Pattern Recognition Conference, pages 24838–24848, 2025. [find](https://scholar.google.com/scholar?q=L%2E+Ouyang%2C+Y%2E+Qu%2C+H%2E+Zhou%2C+J%2E+Zhu%2C+R%2E+Zhang%2C+Q%2E+Lin%2C+B%2E+Wang%2C+Z%2E+Zhao%2C+M%2E+Jiang%2C+X%2E+Zhao%2C+et+al%2E+Omnidocbench%3A+Benchmarking+diverse+pdf+document+parsing+with+comprehensive+annotations%2E+In+Proceedings+)
28. J. Poznanski, A. Rangapur, J. Borchardt, J. Dunkelberger, R. Huff, D. Lin, C. Wilhelm, K. Lo, and L. Soldaini. olmocr: Unlocking trillions of tokens in pdfs with vision language models. arXiv preprint arXiv:2502.18443, 2025. [find](https://scholar.google.com/scholar?q=J%2E+Poznanski%2C+A%2E+Rangapur%2C+J%2E+Borchardt%2C+J%2E+Dunkelberger%2C+R%2E+Huff%2C+D%2E+Lin%2C+C%2E+Wilhelm%2C+K%2E+Lo%2C+and+L%2E+Soldaini%2E+olmocr%3A+Unlocking+trillions+of+tokens+in+pdfs+with+vision+language+models%2E+arXiv+preprint)
29. A. Radford, J. W. Kim, C. Hallacy, A. Ramesh, G. Goh, S. Agarwal, G. Sastry, A. Askell, P. Mishkin, J. Clark, et al. [Learning transferable visual models from natural language supervision](https://md.research.elamur.ai/free/en/papers/radford-2021-learning-transferable-visual-models-natural/). In International conference on machine learning, pages 8748–8763. PMLR, 2021.
30. Rednote. dots.ocr, 2025. URL [https://github.com/rednote-hilab/dots.ocr](https://github.com/rednote-hilab/dots.ocr) . [↗](https://github.com/rednote-hilab/dots.ocr)
31. C. Schuhmann, R. Vencu, R. Beaumont, R. Kaczmarczyk, C. Mullis, A. Katta, T. Coombes, J. Jitsev, and A. Komatsuzaki. [Laion-400m: Open dataset of clip-filtered 400 million image-text pairs](https://md.research.elamur.ai/free/en/papers/schuhmann-2021-laion-400m-open-dataset-clip/). arXiv preprint arXiv:2111.02114, 2021.
32. A. Singh, V. Natarajan, M. Shah, Y. Jiang, X. Chen, D. Batra, D. Parikh, and M. Rohrbach. Towards vqa models that can read. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 8317–8326, 2019. [find](https://scholar.google.com/scholar?q=A%2E+Singh%2C+V%2E+Natarajan%2C+M%2E+Shah%2C+Y%2E+Jiang%2C+X%2E+Chen%2C+D%2E+Batra%2C+D%2E+Parikh%2C+and+M%2E+Rohrbach%2E+Towards+vqa+models+that+can+read%2E+In+Proceedings+of+the+IEEE%2FCVF+conference+on+computer+vision+and+pattern+rec)
33. T. Sun, C. Cui, Y. Du, and Y. Liu. Pp-doclayout: A unified document layout detection model to accelerate large-scale data construction. arXiv preprint arXiv:2503.17213, 2025. [find](https://scholar.google.com/scholar?q=T%2E+Sun%2C+C%2E+Cui%2C+Y%2E+Du%2C+and+Y%2E+Liu%2E+Pp%2Ddoclayout%3A+A+unified+document+layout+detection+model+to+accelerate+large%2Dscale+data+construction%2E+arXiv+preprint+arXiv%3A2503%2E17213%2C+2025%2E)
34. B. Wang, C. Xu, X. Zhao, L. Ouyang, F. Wu, Z. Zhao, R. Xu, K. Liu, Y. Qu, F. Shang, et al. Mineru: An open-source solution for precise document content extraction. arXiv preprint arXiv:2409.18839, 2024. [find](https://scholar.google.com/scholar?q=B%2E+Wang%2C+C%2E+Xu%2C+X%2E+Zhao%2C+L%2E+Ouyang%2C+F%2E+Wu%2C+Z%2E+Zhao%2C+R%2E+Xu%2C+K%2E+Liu%2C+Y%2E+Qu%2C+F%2E+Shang%2C+et+al%2E+Mineru%3A+An+open%2Dsource+solution+for+precise+document+content+extraction%2E+arXiv+preprint+arXiv%3A2409%2E18839%2C+202)
35. P. Wang, S. Bai, S. Tan, S. Wang, Z. Fan, J. Bai, K. Chen, X. Liu, J. Wang, W. Ge, et al. Qwen2-vl: Enhancing vision-language model’s perception of the world at any resolution. arXiv preprint arXiv:2409.[12191, 2024](https://md.research.elamur.ai/free/en/papers/wang-2024-qwen2-vl-enhancing-vision/).
36. H. Wei, L. Kong, J. Chen, L. Zhao, Z. Ge, J. Yang, J. Sun, C. Han, and X. Zhang. Vary: Scaling up the vision vocabulary for large vision-language model. In European Conference on Computer Vision, pages 408–424. Springer, 2024. [find](https://scholar.google.com/scholar?q=H%2E+Wei%2C+L%2E+Kong%2C+J%2E+Chen%2C+L%2E+Zhao%2C+Z%2E+Ge%2C+J%2E+Yang%2C+J%2E+Sun%2C+C%2E+Han%2C+and+X%2E+Zhang%2E+Vary%3A+Scaling+up+the+vision+vocabulary+for+large+vision%2Dlanguage+model%2E+In+European+Conference+on+Computer+Vision%2C+page)
37. H. Wei, L. Kong, J. Chen, L. Zhao, Z. Ge, E. Yu, J. Sun, C. Han, and X. Zhang. Small language model meets with reinforced vision vocabulary. arXiv preprint arXiv:2401.12503, 2024. [find](https://scholar.google.com/scholar?q=H%2E+Wei%2C+L%2E+Kong%2C+J%2E+Chen%2C+L%2E+Zhao%2C+Z%2E+Ge%2C+E%2E+Yu%2C+J%2E+Sun%2C+C%2E+Han%2C+and+X%2E+Zhang%2E+Small+language+model+meets+with+reinforced+vision+vocabulary%2E+arXiv+preprint+arXiv%3A2401%2E12503%2C+2024%2E)
38. H. Wei, C. Liu, J. Chen, J. Wang, L. Kong, Y. Xu, Z. Ge, L. Zhao, J. Sun, Y. Peng, et al. General ocr theory: Towards ocr-2.0 via a unified end-to-end model. arXiv preprint arXiv:2409.01704, 2024. — [arXiv:2409.01704](https://arxiv.org/abs/2409.01704) [find](https://scholar.google.com/scholar?q=H%2E+Wei%2C+C%2E+Liu%2C+J%2E+Chen%2C+J%2E+Wang%2C+L%2E+Kong%2C+Y%2E+Xu%2C+Z%2E+Ge%2C+L%2E+Zhao%2C+J%2E+Sun%2C+Y%2E+Peng%2C+et+al%2E+General+ocr+theory%3A+Towards+ocr%2D2%2E0+via+a+unified+end%2Dto%2Dend+model%2E+arXiv+preprint+arXiv%3A2409%2E01704%2C+2024%2E+%E2%80%94+%5B)
39. H. Wei, Y. Yin, Y. Li, J. Wang, L. Zhao, J. Sun, Z. Ge, X. Zhang, and D. Jiang. Slow perception: Let’s perceive geometric figures step-by-step. arXiv preprint arXiv:2412.20631, 2024. — [arXiv:2412.20631](https://arxiv.org/abs/2412.20631) [find](https://scholar.google.com/scholar?q=H%2E+Wei%2C+Y%2E+Yin%2C+Y%2E+Li%2C+J%2E+Wang%2C+L%2E+Zhao%2C+J%2E+Sun%2C+Z%2E+Ge%2C+X%2E+Zhang%2C+and+D%2E+Jiang%2E+Slow+perception%3A+Let%E2%80%99s+perceive+geometric+figures+step%2Dby%2Dstep%2E+arXiv+preprint+arXiv%3A2412%2E20631%2C+2024%2E+%E2%80%94+%5BarXiv%3A2412%2E206)
40. Z. Wu, X. Chen, Z. Pan, X. Liu, W. Liu, D. Dai, H. Gao, Y. Ma, C. Wu, B. Wang, et al. [Deepseek-vl2: Mixture-of-experts vision-language models for advanced multimodal understanding](https://md.research.elamur.ai/free/en/papers/wu-2024-deepseek-vl2-mixture-experts/). arXiv preprint arXiv:2412.10302, 2024.
41. W. Yu, Z. Yang, L. Li, J. Wang, K. Lin, Z. Liu, X. Wang, and L. Wang. [Mm-vet: Evaluating large multimodal models for integrated capabilities](https://md.research.elamur.ai/free/en/papers/yu-2023-mm-vet-evaluating-large-multimodal/). arXiv preprint arXiv:2308.02490, 2023.
42. J. Zhu, W. Wang, Z. Chen, Z. Liu, S. Ye, L. Gu, H. Tian, Y. Duan, W. Su, J. Shao, et al. [Internvl3: Exploring advanced training and test-time recipes for open-source multimodal models](https://md.research.elamur.ai/free/en/papers/zhu-2025-internvl3-exploring-advanced-training/). arXiv preprint arXiv:2504.10479, 2025.

*Updated 24 October 2026*