datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
omnidocbench-render-compare
OmniDocBench Render-and-Compare
This dataset contains the rendered HTML reconstructions and comparison images produced
by a render-and-compare pipeline — a reference-free visual similarity evaluation
framework for OCR systems.
Overview
The pipeline processes each page of OmniDocBench through
a Qwen3.5-122B-A10B OCR model, renders the structured output back to a PNG via HTML
(reconstructed.png), and compares it against the original page scan (masked_original.png)
using… See the full description on the dataset page: https://huggingface.co/datasets/gt-free-ocr-metrics/omnidocbench-render-compare.persian-ocr-bench
Persian OCR Evaluation Dataset
This benchmark contains paired Persian document images and UTF-8 transcription
targets for OCR evaluation. Each JSONL row references one image under
bench_data/images/ and contains its transcription in the text field.
Schema
image: image path relative to the repository
id: stable image identifier
page: page number, currently 1
type: evaluation item type, currently transcription
text: reference transcription
language: fa
checked:… See the full description on the dataset page: https://huggingface.co/datasets/mshojaei77/persian-ocr-bench.SDS-KoPub-OCR
SDS-KoPub OCR Results & Embeddings
OCR layout parsing results and VL embeddings for the
SDS-KoPub-VDR-Benchmark
corpus (40,781 Korean public document pages).
Contents
File
Description
Size
ocr_results.jsonl
GLM-OCR structured layout results (regions, markdown, bbox, labels)
40,781 records
parsed_texts.jsonl
Extracted text per page (embedding input)
40,781 records
embeddings/corpus_regions.npy
Region multimodal embeddings (image+caption)
(21052, 2048)… See the full description on the dataset page: https://huggingface.co/datasets/Forturne/SDS-KoPub-OCR.chandra-27b-ocr-v1
Chandra OCR + Qwen3.6-27B-AWQ v1.0 Dataset
Overview
OCR metadata dataset extracted from 422,692 Korean government gazette pages (1994-2026) using the Chandra OCR + Qwen3.6-27B-AWQ pipeline on AMD MI300X 192GB ROCm.
Dataset Files
File
Records
Description
pety_ocr_results.jsonl
1,237
OCR from scanned PETY PDFs (1994-2010)
document_texts_sample.jsonl
10,000
Sampled digital-born searchThema PDFs (2001-2026)
extracted_persons.jsonl
22… See the full description on the dataset page: https://huggingface.co/datasets/yakdoli/chandra-27b-ocr-v1.OCR-Reasoning
OCR-Reasoning Benchmark: Unveiling the True Capabilities of MLLMs in Complex Text-Rich Image Reasoning
Project Page | Paper | GitHub
OCR-Reasoning is a comprehensive benchmark designed to systematically assess Multimodal Large Language Models (MLLMs) on text-rich image reasoning tasks. The benchmark comprises 1,069 human-annotated examples spanning 6 core reasoning abilities: spatial reasoning, numerical analysis, mathematical reasoning, enumerative reasoning, logical reasoning… See the full description on the dataset page: https://huggingface.co/datasets/mx262/OCR-Reasoning.persian-ocr-benchmark
Persian OCR Evaluation Dataset
This benchmark contains paired Persian document images and UTF-8 transcription
targets for OCR evaluation. Each JSONL row references one image under
bench_data/images/ and contains its transcription in the text field.
Schema
image: image path relative to the repository
id: stable image identifier
page: page number, currently 1
type: evaluation item type, currently transcription
text: reference transcription
language: fa
checked:… See the full description on the dataset page: https://huggingface.co/datasets/PersianML/persian-ocr-benchmark.wzf-jinwen-ocr
吴镇烽《金文通鉴》OCR 数据集 (Wzf Jinwen Tongjian OCR Dataset)
⚠️ PRIVATE / RESEARCH USE ONLY — 源自吴镇烽老师《金文通鉴》商业出版数据库。
仅供个人学术研究和比赛使用，严禁公开分发。详见底部"数据来源与版权"。
概要
来自吴镇烽《金文通鉴》（青铜器铭文数据库）的高质量金文 OCR 训练数据。
40,095 张图像（拓本 25,049 + 器物 15,046）
19,720 件独立器物
10,275 种字符（含 PUA 金文专字 + CJK 扩展平面）
释文 ground truth 由吴镇烽老师学术级整理
目录
images/
├── rubbing/ # 拓本图（OCR 训练主力，质量高）
└── vessel/ # 器物全照（含部分铭文，可作辅助/预训练）
train.jsonl # 31,997 行
val.jsonl # 3,910 行
test.jsonl # 4… See the full description on the dataset page: https://huggingface.co/datasets/Raphael2099/wzf-jinwen-ocr.arxiv-chandra-ocr-2-include-images-first50-20260415
arXiv OCR with Chandra OCR 2
This output bundle stores OCR results for arXiv PDFs using datalab-to/chandra-ocr-2.
Summary
Output dataset: nielsr/arxiv-chandra-ocr-2-include-images-first50-20260415
Output bucket: hf://buckets/nielsr/arxiv-chandra-ocr-2-include-images-first50-20260415
Source paper IDs in input list: 27,584
Processed IDs recorded in state/processed_ids.txt: 50
Successes: 50
Partial successes: 0
Errors: 0
Next shard index: 10
Updated at:… See the full description on the dataset page: https://huggingface.co/datasets/nielsr/arxiv-chandra-ocr-2-include-images-first50-20260415.odia-handwritten-ocr
Odia Handwritten OCR Dataset
Dataset Description
This dataset contains 182,152 handwritten Odia character images prepared for training OCR models. The dataset covers all 47 OHCS (Odia Handwritten Character Set) characters with balanced class distribution.
Dataset Summary
Language: Odia (ଓଡ଼ିଆ)
Task: Optical Character Recognition (OCR)
Total Images: 182,152
Character Classes: 47
Image Format: Grayscale JPG (32x32 pixels)
Splits: Train (145,717), Validation (18… See the full description on the dataset page: https://huggingface.co/datasets/tell2jyoti/odia-handwritten-ocr.omnidocbench-qwen-ocr-logprobs
OmniDocBench Qwen OCR Log-Probabilities
This dataset provides token-level and bounding-box-level OCR log-probabilities produced by
running Qwen3.5-122B-A10B (via vLLM) on the original page scans of the
OmniDocBench benchmark.
It is a reference-free auxiliary signal — no ground-truth text is used.
Dataset Structure
ocr_logprobs/
<page_id>/
ocr_logprobs.json # full per-token logprobs + top-5 alternatives
ocr_html.html # raw HTML output from the OCR model… See the full description on the dataset page: https://huggingface.co/datasets/gt-free-ocr-metrics/omnidocbench-qwen-ocr-logprobs.ocr-correction
OCR (Optical Character Recognition) Correction Dataset
This dataset comprises OCR-corrected text samples from English books and newspapers sourced from the Internet Archive. It provides pairs of raw OCR text and their AI-corrected versions, designed for OCR correction tasks.
Dataset Structure
Data Instances
Each instance contains:
input: Raw OCR text with errors
output: Corrected text
Example:
{
"input": "\n\n(ii) The income of Tarai and Bhabar… See the full description on the dataset page: https://huggingface.co/datasets/agentlans/ocr-correction.DeepSeek-OCR
DeepSeek-OCR-Dataset
A high-quality multimodal dataset derived from the official paper “DeepSeek-OCR: Contexts Optical Compression for Ultra-Long Document Understanding”. This dataset contains 2,200 structured image–text pairs covering diverse document types—including academic papers, financial reports, textbooks, slides, newspapers, charts, chemical formulas, and geometric figures—extracted and reformatted to support end-to-end OCR, layout parsing, and vision-language pretraining… See the full description on the dataset page: https://huggingface.co/datasets/amishor/DeepSeek-OCR.Legal_domain_ocr_extracted_Nepali_sft_dataset
Nepali Legal SFT Dataset — Software Development & Operation Committee Order, 2083
Dataset Summary
This dataset contains 31 single-turn instruction/response pairs in Nepali (Devanagari script), derived from a single Government of Nepal legal instrument:
सफ्टवेयर विकास तथा सञ्चालन समिति (गठन) आदेश, २०८३
(Software Development and Operation Committee (Formation) Order, 2083)
The order was issued by the Government of Nepal under Section 3 of the विकास समिति ऐन, २०१३… See the full description on the dataset page: https://huggingface.co/datasets/Somtharu181coder/Legal_domain_ocr_extracted_Nepali_sft_dataset.arabic-ocr-printed-500k
Arabic Printed OCR Lines — Synthetic, 500k
A general-purpose printed Arabic text-line recognition corpus: 500,000 train + 2,000 val
line images with labels, built to fine-tune line-recognition models (PaddleOCR PP-OCR rec
CTC/MultiHead, TrOCR, etc.). Real line-crop printed-Arabic data does not exist at this scale
on the Hub, so this corpus is rendered synthetically with diverse fonts + real Arabic text and
a documented label/decoding contract.
Why this exists… See the full description on the dataset page: https://huggingface.co/datasets/medyas/arabic-ocr-printed-500k.LLAVA-graph-OCRCleanedpleias-post-ocr-correction-chonkie-aligned-en
PleIAs Post-OCR Correction — Chonkie-Aligned Semantic Chunks
This dataset is a semantically chunked and span-aligned derivative of PleIAs/Post-OCR-Correction.
Each record contains:
an OCR hypothesis chunk from the original text field;
a corresponding post-OCR correction output chunk from the corrected_text field;
metadata inherited from the PleIAs dataset;
character spans linking each chunk back to the original source document;
alignment diagnostics produced during filtering.… See the full description on the dataset page: https://huggingface.co/datasets/emanuelaboros/pleias-post-ocr-correction-chonkie-aligned-en.ACL-23-Paper-OCR-Markdown
ACL 2023 Paper in Markdown after OCR
This dataset contains 2150 papers from Association for Computational Linguistics (ACL) 2023:
Long Papers (912 papers)
Short Papers (185 papers)
System Demonstrations (59 paper)
Student Research Workshop (35 papers)
Industry Track (77 papers)
Tutorial Abstracts (7 papers)
Findings (902 papers)
This dataset is processed and compiled by @hu_yifei as part of open-source effort from the Open Research Assistant Project.
OCR process
The… See the full description on the dataset page: https://huggingface.co/datasets/yifeihu/ACL-23-Paper-OCR-Markdown.jfk-pdfs-ocr
JFK OCRed files
Use colab: https://colab.research.google.com/drive/1xsXHYIaEU50tS0eTwuI5QdkGrbvjHQpW?usp=sharing to get more records.
OCRFlux-pubtabnet-cross
OCRFlux-pubtabnet-cross
PDF documents are typically paginated, which often results in tables or paragraphs being split across consecutive pages. Accurately detecting and merging such cross-page structures is crucial to avoid generating incomplete or fragmented content.
The merging of two table fragments is especially challenging. For example, the table spanning multiple pages will repeat the header of the first page on the second page. Another difficult scenario is that the table… See the full description on the dataset page: https://huggingface.co/datasets/ChatDOC/OCRFlux-pubtabnet-cross.Latin-OCR-Artifacts
Latin sentences sourced from The Latin Library, converted to images, were subsequently degraded via OCRODEG.
OCR (via Kraken and Tesseract) transcriptions were generated. The dataset was then augmented with several synthetic noise patterns in order to emulate the more severe corruption found in many older digitizations.
If you use this in your work, please cite:
@misc{mccarthy2025LACOROCR,
author = {McCarthy, A. M.},
title = {{Latin OCR Artifacts}},
year = {2025}… See the full description on the dataset page: https://huggingface.co/datasets/aimgo/Latin-OCR-Artifacts.OCRFlux-bench-cross
OCRFlux-bench-cross
PDF documents are typically paginated, which often results in tables or paragraphs being split across consecutive pages. Accurately detecting and merging such cross-page structures is crucial to avoid generating incomplete or fragmented content.
The detection task can be formulated as follows: given the Markdowns of two consecutive pages—each structured as a list of Markdown elements (e.g., paragraphs and tables)—the goal is to identify the indexes of elements… See the full description on the dataset page: https://huggingface.co/datasets/ChatDOC/OCRFlux-bench-cross.traditional-chinese-historical-ocr-lo-chia-luen
Traditional Chinese Historical OCR Dataset
(Lo Chia-Lun Manuscripts)
This dataset consists of manually annotated OCR text crops derived from the Lo Chia-Lun Manuscript Collection (羅家倫文稿), hosted by the National Chengchi University Library.
The dataset is designed to support research on Traditional Chinese OCR, particularly for historical documents characterized by vertical layouts, handwritten or semi-printed glyphs, and long-form text lines.
Due to archival and… See the full description on the dataset page: https://huggingface.co/datasets/ZihCiLin/traditional-chinese-historical-ocr-lo-chia-luen.omnidocbench-render-compare-sample
OmniDocBench Render-and-Compare — Sample
This is a 60-page stratified sample of
gt-free-ocr-metrics/omnidocbench-render-compare
(the full dataset is ~10 GB).
It is provided to help reviewers explore the data without downloading the full dataset,
as recommended by the NeurIPS 2025 Datasets & Benchmarks Track guidelines.
Sampling Methodology
Pages were selected by stratified random sampling from the full dataset:
Each page in ocr_all (1 355 pages) was assigned to one… See the full description on the dataset page: https://huggingface.co/datasets/gt-free-ocr-metrics/omnidocbench-render-compare-sample.OCRFlux-bench-single
OCRFlux-bench-single
OCRFlux-bench-single is a benchmark of 2000 PDF pages and their ground-truth Markdowns sampled from our private document datasets, which are labeled manually with multi-round checking.
This dataset can be used to measure the performance of OCR systems in single-page parsing.
Quick links:
🤗 Model
🛠️ Code
Data Mix
Table 1: Pages breakdown by language
Language
Pages
English
1000
Chinese
1000
Total
2000… See the full description on the dataset page: https://huggingface.co/datasets/ChatDOC/OCRFlux-bench-single.OCRFlux-pubtabnet-single
OCRFlux-pubtabnet-single
OCRFlux-pubtabnet-single is a benchmark of 9064 table images and their corresponding ground-truth HTML, which are derived from the public PubTabNet benchmark with some format transformations.
This dataset can be used to measure the performance of OCR systems in single-page table parsing.
Quick links:
🤗 Model
🛠️ Code
Data Mix
Table 1: Tables breakdown by complexity (whether they contain rowspan or colspan cells)… See the full description on the dataset page: https://huggingface.co/datasets/ChatDOC/OCRFlux-pubtabnet-single.arxiv-chandra-ocr-2-include-images-demo-2604-08626-20260416
arXiv OCR with Chandra OCR 2
This output bundle stores OCR results for arXiv PDFs using datalab-to/chandra-ocr-2.
Summary
Output dataset: nielsr/arxiv-chandra-ocr-2-include-images-demo-2604-08626-20260416
Output bucket: hf://buckets/nielsr/arxiv-chandra-ocr-2-include-images-demo-2604-08626-20260416
Source paper IDs in input list: 1
Processed IDs recorded in state/processed_ids.txt: 1
Successes: 1
Partial successes: 0
Errors: 0
Next shard index: 1
Updated at:… See the full description on the dataset page: https://huggingface.co/datasets/nielsr/arxiv-chandra-ocr-2-include-images-demo-2604-08626-20260416.pleias-post-ocr-correction-chonkie-aligned-fr
PleIAs Post-OCR Correction — Chonkie-Aligned Semantic Chunks
This dataset is a semantically chunked and span-aligned derivative of PleIAs/Post-OCR-Correction.
Each record contains:
an OCR hypothesis chunk from the original text field;
a corresponding post-OCR correction output chunk from the corrected_text field;
metadata inherited from the PleIAs dataset;
character spans linking each chunk back to the original source document;
alignment diagnostics produced during filtering.… See the full description on the dataset page: https://huggingface.co/datasets/emanuelaboros/pleias-post-ocr-correction-chonkie-aligned-fr.arxiv-chandra-ocr-2-include-images-demo-2604-14148-20260416
arXiv OCR with Chandra OCR 2
This output bundle stores OCR results for arXiv PDFs using datalab-to/chandra-ocr-2.
Summary
Output dataset: nielsr/arxiv-chandra-ocr-2-include-images-demo-2604-14148-20260416
Output bucket: hf://buckets/nielsr/arxiv-chandra-ocr-2-include-images-demo-2604-14148-20260416
Source paper IDs in input list: 1
Processed IDs recorded in state/processed_ids.txt: 1
Successes: 1
Partial successes: 0
Errors: 0
Next shard index: 1
Updated at:… See the full description on the dataset page: https://huggingface.co/datasets/nielsr/arxiv-chandra-ocr-2-include-images-demo-2604-14148-20260416.Bourse_Casablanca_Rapports_Financiers_Bruts_OCR
Bourse Casablanca : Rapports Financiers Bruts (OCR)
Présentation
Ce dataset contient l'extraction OCR intégrale des Rapports Annuels et Financiers de 58 sociétés cotées à la Bourse de Casablanca (BVC).
L'archive couvre l'intégralité de l'historique disponible sur le site officiel de la Bourse de Casablanca pour chaque émetteur, transformant des documents PDF complexes en données HTML exploitables.
Chiffres clés
Périmètre : 58 sociétés de la cote marocaine.… See the full description on the dataset page: https://huggingface.co/datasets/OumarDicko/Bourse_Casablanca_Rapports_Financiers_Bruts_OCR.ocr_data_question_28k_Qwen3-8B