AncientDoc: A Benchmark for Chinese Ancient Document Understanding with 2,973 Pages

2,973 pages of Chinese ancient documents form a benchmark for multimodal large model evaluation. The dataset, created by ByteDance, is designed for tasks ranging from optical character recognition to knowledge reasoning. It was last updated on the platform in September 2025.

Use Cases

• Train OCR models for historical Chinese scripts based on the image data.
• Benchmark document layout analysis and understanding for ancient documents.
• Evaluate multimodal large language models on knowledge reasoning tasks with historical content.

Strengths

• Specifically designed as a benchmark for Chinese ancient document understanding.
• Covers 2,973 pages, providing a defined scale for evaluation.
• Supports multi-task evaluation from recognition to reasoning.

Limitations

• Column-level documentation and sample data are unavailable, requiring field semantics to be inferred after download.
• The specific license, file formats, and geographic origin of the documents are not provided in the input.

Provenance

Source

ByteDance

Freshness

Last updated 2025-09-08 09:09:46.