InduOCRBench: An OCR Benchmark for Industrial RAG Systems

InduOCRBench is an OCR benchmark for industrial Retrieval-Augmented Generation systems, created by qihoo360 and released in 2026. It covers 11 challenging document types observed in real-world enterprise workflows. The dataset was last updated on 2026-05-08.

Use Cases

• Benchmarking OCR robustness for downstream RAG utility based on the described industrial document types.
• Evaluating document understanding systems on real-world enterprise workflows mentioned in the description.
• Training or fine-tuning OCR models for challenging industrial document formats.
• Researching the gap between character-level OCR metrics and actual RAG performance.

Strengths

• Focuses on 11 challenging industrial document types from real-world workflows.
• Addresses a specific gap between traditional OCR metrics and downstream RAG utility.
• Dataset is associated with a paper accepted to ACL 2026 Industry Track.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count, file formats, and sample data are unknown, which may limit suitability assessment.
• License information is unavailable, which could restrict usage.

Provenance

Source

qihoo360

Collection Method

Likely curated from real-world enterprise workflows to cover challenging document types.

Time Range

Release and update dates are in 2026; temporal coverage of the underlying data is unknown.

Freshness

Last updated 2026-05-08 08:25:33; freshness should be verified.

Geography

Spatial coverage is not specified.