Doc-750K: A Multimodal Document Understanding Dataset

OpenGVLab's Doc-750K dataset, referenced in the paper 'Docopilot: Improving Multimodal Models for Document-Level Understanding', is a collection of documents for training AI models. The dataset was last updated on July 22, 2025. It appears to contain a large number of document images, as suggested by unzipping instructions for image archives.

Use Cases

• Training multimodal AI models based on the document-level understanding task described.
• Benchmarking document parsing algorithms based on the multimodal nature of the data.
• Improving OCR and layout analysis models based on the likely inclusion of document images.
• Developing document question-answering systems based on the dataset's focus on understanding.

Strengths

• Dataset is associated with a published academic paper, suggesting a research-grade purpose.
• Last updated on 2025-07-22, indicating recent maintenance.

Limitations

• Description metadata is limited; actual data quality requires manual inspection after download.
• Column-level documentation is absent; field semantics must be inferred after download.
• Row count, file formats, and license are unknown, which may limit suitability assessment.

Provenance

Source

OpenGVLab

Freshness

Last updated 2025-07-22 10:24:06