Datasets:

Eioss
/

CC-OCR-V2

license: mit
task_categories:
  - image-to-text
  - visual-question-answering
  - document-question-answering
language:
  - en
  - zh
tags:
  - ocr
  - document-understanding
  - multimodal
size_categories:
  - 1K<n<10K
viewer: false

CC-OCR V2: Benchmarking Large Multimodal Models for Literacy in Real-world Document Processing

Dataset Summary

CC-OCR V2 is a comprehensive and challenging OCR benchmark tailored to real-world document processing. It focuses on practical enterprise document processing tasks and incorporates hard and corner cases that are critical yet underrepresented in prior benchmarks.

The dataset comprises 7,093 high-difficulty samples covering 5 major OCR-centric tracks: Text Recognition, Document Parsing, Document Grounding, Key Information Extraction, and Document Question Answering.

Dataset Structure

The dataset is structured hierarchically by task and sub_task. Below is the statistical breakdown of the dataset:

Task	Sub-task	Samples
Extraction	business_transactions	340
	public_services	369
	regulated_records	300
Grounding	object_grounding	734
	text_grounding	734
Parsing	complex_table_parsing	300
	formula_parsing	100
	general_documents_parsing	300
	info_board_parsing	26
	molecular_parsing	100
QA	dashboard_qa	500
	financial_documents_qa	1000
	scientific_documents_qa	100
	user_interface_qa	400
Recognition	multi_lingual_recognition	640
	natural_scene_recognition	1150
Total		7093

Data Instances

Each sample in the dataset contains the following fields:

task (str): The primary track/category of the task (e.g., Extraction, QA, Parsing).
sub_task (str): The specific sub-category (e.g., business_transactions, financial_documents_qa).
scenario (str): The specific application scenario or document type.
question (str): The prompt or instruction given to the model.
images_list (str): A string containing the image file paths associated with the sample.
image (list of images): The actual images rendered by the viewer.
answer (str): The ground truth answer or expected output (often in JSON or structured text format).

Citation

@article{xu2026ccocr,
  title={CC-OCR V2: Benchmarking Large Multimodal Models for Literacy in Real-world Document Processing},
  author={Zhipeng Xu and Junhao Ji and Zulong Chen and Zhenghao Liu and Qing Liu and Chunyi Peng and Zubao Qin and Ze Xu and Jianqiang Wan and Jun Tang and Zhibo Yang and Shuai Bai and Dayiheng Liu},
  journal={arXiv preprint arXiv:2605.03903},
  year={2026}
}