dorsal/arxiv
View SchemaReading or Reasoning? Format Decoupled Reinforcement Learning for Document OCR
| Authors | Yufeng Zhong, Lei Chen, Zhixiong Zeng, Xuanle Zhao, Deyang Jiang, Liming Zheng, Jing Huang, Haibo Qiu, Peng Shi, Siqi Yang, Lin Ma |
|---|---|
| Categories | |
| ArXiv ID | 2601.08834vv1 |
| URL | https://arxiv.org/abs/2601.08834 |
| License | http://creativecommons.org/licenses/by/4.0/ |
Abstract
Reading text from images or scanned documents via OCR models has been a longstanding focus of researchers. Intuitively, text reading is perceived as a straightforward perceptual task, and existing work primarily focuses on constructing enriched data engineering to enhance SFT capabilities. In this work, we observe that even advanced OCR models exhibit significantly higher entropy in formatted text (\emph{e.g.}, formula, table, etc.) compared to plain text, often by an order of magnitude. These statistical patterns reveal that advanced OCR models struggle with high output uncertainty when dealing with format sensitive document, suggesting that reasoning over diverse reading pathways may improve OCR performance. To address this, we propose format decoupled reinforcement learning (FD-RL), which leverages high-entropy patterns for targeted optimization. Our approach employs entropy-based data filtration strategy to identify format-intensive instances, and adopt format decoupled rewards tailored to different format types, enabling format-level validation rather than token-level memorization. FD-RL achieves an average score of 90.41 on OmniDocBench, setting a new record for end-to-end models on this highly popular benchmark. More importantly, we conduct comprehensive ablation studies over data, training, filtering, and rewarding strategies, thoroughly validating their effectiveness.
{
"annotation_id": "c935b93b-040b-4b93-8f75-8d9787d88b64",
"date_created": "2026-02-17T05:53:19.292000Z",
"date_modified": "2026-02-17T05:53:19.292000Z",
"file_hash": "81030da1fb5793d75abc20fb24717913b8882300c61152346a4752a0ce1e787b",
"private": false,
"record": {
"abstract": "Reading text from images or scanned documents via OCR models has been a longstanding focus of researchers. Intuitively, text reading is perceived as a straightforward perceptual task, and existing work primarily focuses on constructing enriched data engineering to enhance SFT capabilities. In this work, we observe that even advanced OCR models exhibit significantly higher entropy in formatted text (\\emph{e.g.}, formula, table, etc.) compared to plain text, often by an order of magnitude. These statistical patterns reveal that advanced OCR models struggle with high output uncertainty when dealing with format sensitive document, suggesting that reasoning over diverse reading pathways may improve OCR performance. To address this, we propose format decoupled reinforcement learning (FD-RL), which leverages high-entropy patterns for targeted optimization. Our approach employs entropy-based data filtration strategy to identify format-intensive instances, and adopt format decoupled rewards tailored to different format types, enabling format-level validation rather than token-level memorization. FD-RL achieves an average score of 90.41 on OmniDocBench, setting a new record for end-to-end models on this highly popular benchmark. More importantly, we conduct comprehensive ablation studies over data, training, filtering, and rewarding strategies, thoroughly validating their effectiveness.",
"arxiv_id": "2601.08834",
"authors": [
"Yufeng Zhong",
"Lei Chen",
"Zhixiong Zeng",
"Xuanle Zhao",
"Deyang Jiang",
"Liming Zheng",
"Jing Huang",
"Haibo Qiu",
"Peng Shi",
"Siqi Yang",
"Lin Ma"
],
"categories": [
"cs.CV",
"cs.AI",
"cs.CL"
],
"license": "http://creativecommons.org/licenses/by/4.0/",
"title": "Reading or Reasoning? Format Decoupled Reinforcement Learning for Document OCR",
"url": "https://arxiv.org/abs/2601.08834",
"version": "v1"
},
"schema_id": "dorsal/arxiv",
"source": {
"execution_id": "d09564e9-57ee-42e3-8acf-df2a663542cc",
"id": "arXiv Dataset",
"type": "Model",
"variant": "snapshot-2026-01-17",
"version": "0.1.0"
},
"user_id": 1000002
}