Datasets:

himalaya-ai
/

ocr-document-processing-eval

configs:
  - config_name: default
    data_files:
      - split: test
        path: ocr_document_processing_eval.viewer.ocr.parquet
task_categories:
  - image-to-text
language:
  - ne
  - hi
  - mr
tags:
  - ocr
  - devanagari
  - glm-ocr
pretty_name: Ocr Document Processing Eval

ocr_document_processing_eval

Document-processing OCR proxy for digitization, KYC-like numeric fields, and RAG extraction checks.

Repo: himalaya-ai/ocr-document-processing-eval
Task: document_processing_ocr
Main raw file: *.ocr.jsonl with image, ocr, source_repo, and language/provenance columns.
Optional fine-tuning/eval file: *.sharegpt.json with messages and images.

Core Columns

id: unique sample identifier
image: relative path to the image file
ocr: ground-truth text label

Source Mix

indic_vision_bench_deva_ocr ocr/test: 40%
nayana_bench_deva_documents default/hi: 20%
nayana_bench_deva_documents default/mr: 15%
devanagari_digits_mixed default/train: 15%
hindi_handwritten_word_ocr default/test: 10%

Notes

Generated by scripts/sample_ocr_eval_sets.py from the GLM fine-tuning workspace. If a source only exposes a train split, keep the deterministic held-out row ids out of SFT/training runs.