Where OCR models are actually scored. Benchmark datasets with results tables first, then per-language leaderboard Spaces. Ordered by likes.
Daniel van Strien PRO
davanstrien
AI & ML interests
Machine Learning Librarian
Recent Activity
upvoted an article about 4 hours ago
Building a Fast Multilingual OCR Model with Synthetic Data updated a bucket about 7 hours ago
davanstrien/trackio-bucket upvoted an article about 7 hours ago
Meet North Micro Vision: A 2.4B Native-Resolution Vision-Language ModelOrganizations
Text-line and region recognisers, plus detection and recognition pipelines for documents, manga and text in photographs.
-
small-models-for-glam/kraken-ppocrv6-medium
Image-to-Text β’ Updated β’ 15 -
PaddlePaddle/PP-OCRv6_medium_rec
Image-to-Text β’ Updated β’ 64.2k β’ 29 -
kha-white/manga-ocr-base
Image-to-Text β’ Updated β’ 1.11M β’ 180 -
JustANormalTinkerer/hayai-ocr-v2
Image-to-Text β’ 0.2B β’ Updated β’ 4.78k β’ 8
OCR for particular languages and writing systems, from whole-page readers to dedicated text-line recognisers.
-
typhoon-ai/typhoon-ocr1.5-2b
Image-Text-to-Text β’ 2B β’ Updated β’ 362k β’ β’ 30 -
sbintuitions/sarashina2.2-ocr
Image-to-Text β’ 4B β’ Updated β’ 1.86k β’ 33 -
5CD-AI/Vintern-1B-v3_5
Image-Text-to-Text β’ 0.9B β’ Updated β’ 5.35k β’ 125 -
NAMAA-Space/Qari-OCR-0.4.0-VL-4B-Instruct
Image-to-Text β’ Updated β’ 908 β’ 9
Recent Hub datasets of video game gameplay (frames/video + actions, replays, trajectories) for imitation learning and game-playing agents.
-
davanstrien/reasoning-required
Viewer β’ Updated β’ 5k β’ 111 β’ 20 -
davanstrien/ModernBERT-based-Reasoning-Required
Text Classification β’ 0.1B β’ Updated β’ 21 β’ 11 -
davanstrien/fineweb-with-reasoning-scores-and-topics
Viewer β’ Updated β’ 10k β’ 38 β’ 2 -
davanstrien/fine-reasoning-questions
Viewer β’ Updated β’ 244 β’ 69 β’ 19
Maths reasoning datasets found using https://huggingface.co/spaces/librarian-bots/huggingface-datasets-semantic-search
Example of using distilabel to generate synthetic triplets data for fine-tuning a Sentence Transformer model
-
bigcode/self-oss-instruct-sc2-exec-filter-50k
Viewer β’ Updated β’ 50.7k β’ 38.1k β’ 107 -
davanstrien/similarity-dataset-sc2-8b
Viewer β’ Updated β’ 2.32k β’ 54 β’ 6 -
davanstrien/code-prompt-similarity-model
Sentence Similarity β’ 0.1B β’ Updated β’ 80 β’ 6 -
davanstrien/abstract-wiki
Viewer β’ Updated β’ 5k β’ 18 β’ 2
πΈ This is a collection of synthetic datasets built to help improve the ability of open language models to better write haikus through the use of DPO
A collection of datasets that probably support DPO
-
HuggingFaceH4/ultrafeedback_binarized
Viewer β’ Updated β’ 187k β’ 24.5k β’ 348 -
mlabonne/orpo-dpo-mix-40k
Viewer β’ Updated β’ 44.2k β’ 1.56k β’ 310 -
argilla/OpenHermesPreferences
Viewer β’ Updated β’ 989k β’ 988 β’ 214 -
argilla/distilabel-capybara-dpo-7k-binarized
Viewer β’ Updated β’ 7.56k β’ 21.6k β’ 184
Curated OCR models for documents, languages, handwriting and text in images. Browse five collections with short practical notes.
Handwriting, historical print and manuscript recognition. Notes on languages, text-line segmentation and transcription conventions.
-
Riksarkivet/trocr-base-handwritten-hist-swe-2
Image-to-Text β’ 0.4B β’ Updated β’ 98.3k β’ 18 -
BDRC/tibetan-ocr
Image-Text-to-Text β’ 0.8B β’ Updated β’ 9.33k β’ 4 -
isaacmg/qwen3-vl-8b-hebrew-v19a-ckpt
Image-Text-to-Text β’ Updated β’ 54 -
magistermilitum/tridis_v2_HTR_historical_manuscripts
0.6B β’ Updated β’ 380 β’ 7
Roughly ordered by recent releases, useful updates and current usage. Practical OCR and document parsing models. Reviewed September 2026.
-
ATH-MaaS/OvisOCR2
Image-Text-to-Text β’ 0.9B β’ Updated β’ 96.5k β’ β’ 463 -
StarDoc-AI/TeleOCR
Image-Text-to-Text β’ 1B β’ Updated β’ 29.7k β’ 49 -
nvidia/NVIDIA-Nemotron-Parse-2.0
Image-Text-to-Text β’ 0.9B β’ Updated β’ 108k β’ 114 -
tencent/HunyuanOCR
Image-Text-to-Text β’ 1B β’ Updated β’ 652k β’ 819
The reasoning datasets that defined 2025. Part 1 of Datasets Wrapped 2025. #DatasetsWrapped2025
Creating a smol model for tl;dr-ing the hub
-
davanstrien/Smol-Hub-tldr
Text Generation β’ 0.4B β’ Updated β’ 83 β’ 11 -
Semantic Hugging Face Hub Search
π93Search Hugging Face datasets and models by meaning
-
davanstrien/hub-tldr-dataset-summaries-llama
Viewer β’ Updated β’ 5k β’ 28 β’ 1 -
davanstrien/hub-tldr-model-summaries-llama
Viewer β’ Updated β’ 5k β’ 44 β’ 1
A collection of demos for various approaches to synthetic data generation
Papers about synthetic dataset generation
-
Better Synthetic Data by Retrieving and Transforming Existing Datasets
Paper β’ 2404.14361 β’ Published β’ 2 -
Generative AI for Synthetic Data Generation: Methods, Challenges and the Future
Paper β’ 2403.04190 β’ Published β’ 1 -
Best Practices and Lessons Learned on Synthetic Data for Language Models
Paper β’ 2404.07503 β’ Published β’ 32 -
A Multi-Faceted Evaluation Framework for Assessing Synthetic Data Generated by Large Language Models
Paper β’ 2404.14445 β’ Published
This collection contains models, datasets and spaces related to historic language models i.e. language models trained on historic data
-
dbmdz/bert-base-finnish-europeana-cased
Fill-Mask β’ 0.1B β’ Updated β’ 31 -
dbmdz/bert-base-historic-english-cased
Fill-Mask β’ 0.1B β’ Updated β’ 109 β’ 1 -
Livingwithmachines/erwt-year
Fill-Mask β’ Updated β’ 20 -
dbmdz/bert-base-historic-dutch-cased
Fill-Mask β’ 0.1B β’ Updated β’ 70 β’ 2
Datasets suitable for Image Preference Optimization based on their colum names
Where OCR models are actually scored. Benchmark datasets with results tables first, then per-language leaderboard Spaces. Ordered by likes.
Curated OCR models for documents, languages, handwriting and text in images. Browse five collections with short practical notes.
Text-line and region recognisers, plus detection and recognition pipelines for documents, manga and text in photographs.
-
small-models-for-glam/kraken-ppocrv6-medium
Image-to-Text β’ Updated β’ 15 -
PaddlePaddle/PP-OCRv6_medium_rec
Image-to-Text β’ Updated β’ 64.2k β’ 29 -
kha-white/manga-ocr-base
Image-to-Text β’ Updated β’ 1.11M β’ 180 -
JustANormalTinkerer/hayai-ocr-v2
Image-to-Text β’ 0.2B β’ Updated β’ 4.78k β’ 8
Handwriting, historical print and manuscript recognition. Notes on languages, text-line segmentation and transcription conventions.
-
Riksarkivet/trocr-base-handwritten-hist-swe-2
Image-to-Text β’ 0.4B β’ Updated β’ 98.3k β’ 18 -
BDRC/tibetan-ocr
Image-Text-to-Text β’ 0.8B β’ Updated β’ 9.33k β’ 4 -
isaacmg/qwen3-vl-8b-hebrew-v19a-ckpt
Image-Text-to-Text β’ Updated β’ 54 -
magistermilitum/tridis_v2_HTR_historical_manuscripts
0.6B β’ Updated β’ 380 β’ 7
OCR for particular languages and writing systems, from whole-page readers to dedicated text-line recognisers.
-
typhoon-ai/typhoon-ocr1.5-2b
Image-Text-to-Text β’ 2B β’ Updated β’ 362k β’ β’ 30 -
sbintuitions/sarashina2.2-ocr
Image-to-Text β’ 4B β’ Updated β’ 1.86k β’ 33 -
5CD-AI/Vintern-1B-v3_5
Image-Text-to-Text β’ 0.9B β’ Updated β’ 5.35k β’ 125 -
NAMAA-Space/Qari-OCR-0.4.0-VL-4B-Instruct
Image-to-Text β’ Updated β’ 908 β’ 9
Roughly ordered by recent releases, useful updates and current usage. Practical OCR and document parsing models. Reviewed September 2026.
-
ATH-MaaS/OvisOCR2
Image-Text-to-Text β’ 0.9B β’ Updated β’ 96.5k β’ β’ 463 -
StarDoc-AI/TeleOCR
Image-Text-to-Text β’ 1B β’ Updated β’ 29.7k β’ 49 -
nvidia/NVIDIA-Nemotron-Parse-2.0
Image-Text-to-Text β’ 0.9B β’ Updated β’ 108k β’ 114 -
tencent/HunyuanOCR
Image-Text-to-Text β’ 1B β’ Updated β’ 652k β’ 819
Recent Hub datasets of video game gameplay (frames/video + actions, replays, trajectories) for imitation learning and game-playing agents.
The reasoning datasets that defined 2025. Part 1 of Datasets Wrapped 2025. #DatasetsWrapped2025
-
davanstrien/reasoning-required
Viewer β’ Updated β’ 5k β’ 111 β’ 20 -
davanstrien/ModernBERT-based-Reasoning-Required
Text Classification β’ 0.1B β’ Updated β’ 21 β’ 11 -
davanstrien/fineweb-with-reasoning-scores-and-topics
Viewer β’ Updated β’ 10k β’ 38 β’ 2 -
davanstrien/fine-reasoning-questions
Viewer β’ Updated β’ 244 β’ 69 β’ 19
Creating a smol model for tl;dr-ing the hub
-
davanstrien/Smol-Hub-tldr
Text Generation β’ 0.4B β’ Updated β’ 83 β’ 11 -
Semantic Hugging Face Hub Search
π93Search Hugging Face datasets and models by meaning
-
davanstrien/hub-tldr-dataset-summaries-llama
Viewer β’ Updated β’ 5k β’ 28 β’ 1 -
davanstrien/hub-tldr-model-summaries-llama
Viewer β’ Updated β’ 5k β’ 44 β’ 1
Maths reasoning datasets found using https://huggingface.co/spaces/librarian-bots/huggingface-datasets-semantic-search
A collection of demos for various approaches to synthetic data generation
Example of using distilabel to generate synthetic triplets data for fine-tuning a Sentence Transformer model
-
bigcode/self-oss-instruct-sc2-exec-filter-50k
Viewer β’ Updated β’ 50.7k β’ 38.1k β’ 107 -
davanstrien/similarity-dataset-sc2-8b
Viewer β’ Updated β’ 2.32k β’ 54 β’ 6 -
davanstrien/code-prompt-similarity-model
Sentence Similarity β’ 0.1B β’ Updated β’ 80 β’ 6 -
davanstrien/abstract-wiki
Viewer β’ Updated β’ 5k β’ 18 β’ 2
Papers about synthetic dataset generation
-
Better Synthetic Data by Retrieving and Transforming Existing Datasets
Paper β’ 2404.14361 β’ Published β’ 2 -
Generative AI for Synthetic Data Generation: Methods, Challenges and the Future
Paper β’ 2403.04190 β’ Published β’ 1 -
Best Practices and Lessons Learned on Synthetic Data for Language Models
Paper β’ 2404.07503 β’ Published β’ 32 -
A Multi-Faceted Evaluation Framework for Assessing Synthetic Data Generated by Large Language Models
Paper β’ 2404.14445 β’ Published
πΈ This is a collection of synthetic datasets built to help improve the ability of open language models to better write haikus through the use of DPO
This collection contains models, datasets and spaces related to historic language models i.e. language models trained on historic data
-
dbmdz/bert-base-finnish-europeana-cased
Fill-Mask β’ 0.1B β’ Updated β’ 31 -
dbmdz/bert-base-historic-english-cased
Fill-Mask β’ 0.1B β’ Updated β’ 109 β’ 1 -
Livingwithmachines/erwt-year
Fill-Mask β’ Updated β’ 20 -
dbmdz/bert-base-historic-dutch-cased
Fill-Mask β’ 0.1B β’ Updated β’ 70 β’ 2
A collection of datasets that probably support DPO
-
HuggingFaceH4/ultrafeedback_binarized
Viewer β’ Updated β’ 187k β’ 24.5k β’ 348 -
mlabonne/orpo-dpo-mix-40k
Viewer β’ Updated β’ 44.2k β’ 1.56k β’ 310 -
argilla/OpenHermesPreferences
Viewer β’ Updated β’ 989k β’ 988 β’ 214 -
argilla/distilabel-capybara-dpo-7k-binarized
Viewer β’ Updated β’ 7.56k β’ 21.6k β’ 184
Datasets suitable for Image Preference Optimization based on their colum names