OCR Document Processing Eval: A Benchmark for Digitization and Information Extraction

An evaluation dataset for optical character recognition and document processing tasks. The dataset is hosted on Hugging Face by the author himalaya-ai and was last updated on June 2, 2026. It appears to be a proxy for testing digitization, KYC-like numeric field extraction, and RAG extraction checks.

Use Cases

• Benchmarking OCR accuracy for document digitization based on the described proxy task.
• Evaluating model performance on extracting numeric fields for KYC-like applications as mentioned in the description.
• Testing Retrieval-Augmented Generation (RAG) systems on extracted document text for information retrieval checks.
• Fine-tuning or evaluating multimodal models that process images and text, based on the optional fine-tuning file format.

Strengths

• Includes a main raw file with image, OCR text, source repository, and language/provenance columns.
• Provides an optional fine-tuning/eval file structured for conversational AI with messages and images.
• Last update timestamp is precisely recorded as 2026-06-02 16:02:33.

Limitations

• Row count, file size, and specific column definitions are unknown, which may limit suitability assessment.
• Column-level documentation is absent; field semantics must be inferred after download.
• License information is unknown, which could restrict commercial or research use.

Provenance

Source

huggingface (himalaya-ai/ocr-document-processing-eval)

Collection Method

Aggregated from multiple sources including indic_vision_bench_deva_ocr and nayana_bench_deva_documents.

Freshness

Last updated 2026-06-02 16:02:33; freshness should be verified.