Metraipolis › archive › dc_fa1b0306a925

Jina-OCR-v1: Efficient Document Parsing with Speculative Decoding and Dense Verifiable Rewards — arXiv:2609.03181

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.03181
License (as archived)
CC0-1.0
Archived
2026-10-17T01:40:00.925Z
SHA-256 of the archived text
2af834de97e3ce8e7832702e9da42728f10b9d6a8d79e7e42105399a9638c730
Stable URL of this page
https://metraipolis.com/documento/dc_fa1b0306a925

How to cite

Jina-OCR-v1: Efficient Document Parsing with Speculative Decoding and Dense Verifiable Rewards — arXiv:2609.03181 — arXiv (en), https://arxiv.org/abs/2609.03181 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_fa1b0306a925 (Metraipolis archive dc_fa1b0306a925, sha256 2af834de97e3ce8e7832702e9da42728f10b9d6a8d79e7e42105399a9638c730)
BibTeX
@misc{dcfa1b0306a925,
  title = {Jina-OCR-v1: Efficient Document Parsing with Speculative Decoding and Dense Verifiable Rewards — arXiv:2609.03181},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.03181},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_fa1b0306a925; sha256 2af834de97e3ce8e7832702e9da42728f10b9d6a8d79e7e42105399a9638c730},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.03181. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

We present Jina-OCR-v1, an end-to-end document parsing model built to serve on low-budget GPUs. It combines the compressed-vision encoder and the 3B mixture-of-experts decoder of DeepSeek-OCR, which activates about 570M parameters per token, with a FastMTP speculative decoding head that shares a single draft block recursively across K=3 prediction steps. Greedy verification makes decoding lossless. Post-training combines instruction alignment, robustness fine-tuning on difficult documents, and GRPO under dense verifiable rewards: deterministic formula, table, and structural checks that award partial credit. The training data mixes cleaned public corpora with targeted synthetic pages. At the default dynamic-resolution setting, Jina-OCR-v1 scores 91.14 on OmniDocBench v1.6 and 83.4 on olmOCR-Bench, and reaches the highest page throughput in our comparison at 2.57 pages per second. On a low-budget GPU such as the NVIDIA L4, FastMTP doubles decoding speed over greedy autoregressive decoding. The model is publicly available at this https URL .

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-04 in cs.CL).
Categorie arXiv: Computation and Language (cs.CL) ; Computer Vision and Pattern Recognition (cs.CV).
Pagina del lavoro: https://arxiv.org/abs/2609.03181 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-10-17T01:11:43.502Z dalla pagina https://arxiv.org/catchup/cs.CL/2026-09-04?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).