600K examples for visual reasoning RL
A curated collection of 59 English datasets for training VLMs in logic, STEM, charts, and OCR.
- Image text → text
- 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
A curated collection of 59 English datasets for training VLMs in logic, STEM, charts, and OCR.
A dynamic English-language benchmark for assessing how well multimodal LLMs understand scientific journals and illustrations.
A set of under 1,000 complex multimodal questions with images for testing the limits of modern models.
Generated pages in Vietnamese and English with block annotations for training document parsing.
A large-scale multimodal benchmark for evaluating action understanding in video.
Multimodal dataset from Amazon Science to test VLM capability to find answers in bulky documents and images.
Mathematical reasoning benchmark over images: IQTest, FunctionQA, and PaperQA; languages en, zh, fa.
Over 20K document images across six languages for forgery detection and fake text analysis.
A multilingual collection of Wikimedia vector diagrams and maps with text coordinate annotations for evaluating image translation.
Multi-domain benchmark for geospatial intelligence and computer vision.
Millions of short daily first-person videos with detailed captions and Q&A from Reka AI.
Large-scale architecture photo base with metadata to train multimodal models in guessing the construction year.
English prompt dataset with image pairs for evaluating text-to-image preferences.
Multidisciplinary visual and textual questions for multimodal understanding evaluation.
A set of 4,000 video questions across 20 tasks to test time and action perception by multimodal models.
Vietnamese e-commerce data with product photos, descriptions, user reviews, and review images.
Large Conceptual Captions 12M dataset in WebDataset format for pre-training multimodal models.
A large open collection of English image-text pairs for high-quality text-to-image fine-tuning.
A large expert VQA dataset for analyzing the evolution, etymology, and structure of ancient Chinese characters.
Manga illustrations and English/Japanese texts for classification and VQA.
31k hours of image-prompted spontaneous speech across 105 Indian languages, with transcripts for ASR and multimodal tasks.
English pairs of images, questions, and answers gathered from arXiv research publications.
A curated selection of 8,000 English website screenshots for training models to understand web interface elements.
A large-scale corpus of screenshots from 87 applications with structured annotations for training computer-controlling agents.