himalaya-ai/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:… See the full description on the dataset page: https://huggingface.co/datasets/himalaya-ai/ocr-document-processing-eval.
1723
1---2configs:3- config_name: default4 data_files:5 - split: test6 path: ocr_document_processing_eval.viewer.ocr.parquet7task_categories:8- image-to-text9language:10- ne11- hi12- mr13tags:14- ocr15- devanagari16- glm-ocr17pretty_name: Ocr Document Processing Eval18---19 20# ocr_document_processing_eval21 22Document-processing OCR proxy for digitization, KYC-like numeric fields, and RAG extraction checks.23 24- Repo: `himalaya-ai/ocr-document-processing-eval`25- Task: `document_processing_ocr`26- Main raw file: `*.ocr.jsonl` with `image`, `ocr`, `source_repo`, and language/provenance columns.27- Optional fine-tuning/eval file: `*.sharegpt.json` with `messages` and `images`.28 29## Core Columns30 31- `id`: unique sample identifier32- `image`: relative path to the image file33- `ocr`: ground-truth text label34 35## Source Mix36 37- `indic_vision_bench_deva_ocr` `ocr/test`: 40%38- `nayana_bench_deva_documents` `default/hi`: 20%39- `nayana_bench_deva_documents` `default/mr`: 15%40- `devanagari_digits_mixed` `default/train`: 15%41- `hindi_handwritten_word_ocr` `default/test`: 10%42 43## Notes44 45Generated by `scripts/sample_ocr_eval_sets.py` from the GLM fine-tuning workspace.46If a source only exposes a train split, keep the deterministic held-out row ids out of SFT/training runs.47