Dataset · Hugging Face

Dharma-AI/DharmaOCR-Benchmark

Name: DharmaOCR-Benchmark
Creator: Dharma-AI
License: other
Keywords: task_categories:image-text-to-text, language:pt, license:other, size_categories:n<1K, format:parquet, modality:image, modality:text, library:datasets, library:dask, library:polars, library:mlcroissant, arxiv:2604.14314, region:us, ocr, benchmark

DharmaOCR-Benchmark Overview DharmaOCR-Benchmark is a 496-instance evaluation suite for OCR models focused on Brazilian Portuguese documents. It covers printed text, handwritten text, and legal/administrative documents — domains underrepresented in existing benchmarks like OCRBench and olmOCR-Bench. This benchmark evaluates not only transcription quality, but also text degeneration rate and unit inference cost as first-class metrics. Released alongside the… See the full description on the dataset page: https://huggingface.co/datasets/Dharma-AI/DharmaOCR-Benchmark.

View on Hugging FaceReport an issue

Author / organization	Dharma-AI
License (as declared)	other
Downloads	559 (as reported by Hugging Face)
Likes	6
Created	29 March 2026
Last updated at source	22 July 2026

ocrbenchmarkdocument-understandingbrazilian-portuguesetext-recognitionhandwriting-recognitionlegal-documents

Embed this entry

<iframe src="https://silytics.si/embed/item/a24bf52a-6573-4c50-8e78-cb9942b4d157" width="420" height="190" style="border:0" loading="lazy" title="Silytics entry"></iframe>

Free, no tracking. More about feeds and embeds.

Popularity, last 3 days

559 now (+10 since 2026-10-16). Counts as reported by the source, snapshotted daily by Silytics.

Connected in the knowledge graph

DharmaOCR: Specialized Small Language Models for Structured OCR that outperform Open-Source and Commercial BaselinesPaper · cites

Explore this entry in the graph

Related papers

DharmaOCR: Specialized Small Language Models for Structured OCR that outperform Open-Source and Commercial BaselinesPaper

Silytics shows public metadata from Hugging Face and links to the original. Information may be incomplete or out of date. Last refreshed 18 October 2026. Licenses and terms of the original apply to the work itself. Not financial, legal or professional advice.