2.3m reasoning examples from Qwen3-VL-235B
Raw corpus of 2.3 million multimodal tasks and reasoning chains (8.8 billion tokens) before quality filtering.
- Visual question answering
- Question answering
- Text generation
- English
Hugging Face datasets with plain-language descriptions: what's inside and what models to train on them. 4,788 in the catalog so far, with new ones added every hour.
Answered by Gemini 4 · the chat is kept only in this tab
Raw corpus of 2.3 million multimodal tasks and reasoning chains (8.8 billion tokens) before quality filtering.
Approximately 900K question-answer pairs (40% image-based, 60% text-based) for boosting multimodal model reasoning logic.
Over 1 million illustrations extracted from digitized historical books from the British Library (1510–1900).
A collection of images and short videos with questions from Scale AI to evaluate models' ability to understand hidden context.
Over 1,200 instructions and screenshots from iOS, Android, macOS, Windows, and the web for evaluating multimodal models.
From 1 to 10k document pages for comparing OCR and vision-language models across three tracks.
Over a million AI-generated images with detailed English captions for text-to-image and image-to-text tasks.
A large multimodal dataset with real website designs and their code for training frontend models.
A large multimodal dataset for understanding scientific images: images and text for image-to-text tasks.
An augmented version of ImageNet-1K with over a million records containing text descriptions, bounding boxes, and label error notes.
Chart images and English texts to evaluate how multimodal models write code.
Images of medieval manuscripts and their text transcriptions in six European languages.
A large multimodal set of 1 million to 10 million image-text pairs in English.
80K image-instruction pairs with responses from 12 vision-language models evaluated by GPT-4V.
From 1 to 10 million document image-text pairs in English for training VLM models.
37.6 million image-text pairs from Wikipedia across 108 languages for multimodal retrieval and training.
A large open dataset of 218M images with expanded descriptions enriched with web facts.
English images and text for training and validating trustworthy multimodal models via RLAIF.
1.8M multimodal examples with detailed reasoning distilled from Qwen3-VL-235B-Thinking.
6k+ top-voted NSFW images from CivitAI with prompt and NSFW-level metadata.
Printed sheet music images with region annotations for testing optical music recognition systems.
400,476 videos with previews, view counts, explicit flags, and captions for video-text models.
Image and caption pairs with identical words in different orders to test vision-language models.
Over a million images of complex mathematical expressions for training OCR and image-to-text systems.