MME-Unify benchmark for multimodal models
Evaluation set of images and English text across 10 tasks and 30 subtasks for testing U-MLLMs.
- Multiple choice
- Question answering
- Visual question answering
- 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
Evaluation set of images and English text across 10 tasks and 30 subtasks for testing U-MLLMs.
Dataset for benchmarking counterfactual vision-language models (VLMs).
A multimodal English science question dataset with multiple-choice options and images.
A VQA dataset with geometric diagram images and English text question-answer pairs, containing about 1 to 10k examples.
About 25.5 thousand pairs of scanned book pages and human-verified Telugu text from Wikisource.
100K to 1M question-answer pairs with reasoning over text-dense images across 5 European languages.
A collection of images and English questions to test VLM ability to find subtle differences in similar pictures.
182k alternating text and image examples for step-by-step visual reasoning.
Apple's extension of the DataComp-1B dataset with synthetic captions, embeddings, and metadata for image-text pairs.
64,878 English QA pairs for 40,263 spine images to evaluate multimodal models.
A set of 1,505 visual questions to test multimodal model image reasoning capabilities.
Image and Chinese text dataset for evaluating MLLMs in multi-image attribute extraction.
49K English and Chinese questions about cats and dogs with images, audio, video, and text.
Medical image dataset with questions, answers, and semantic annotations in English and Chinese.
Massive English dataset for Document VQA used in Idefics3 fine-tuning.
Curated English visual-text tasks for fair evaluation of large multimodal models.
Part of a massive 1 trillion token multimodal dataset for pre-training models in document and page understanding.
An English dataset of 10K–100K examples featuring scene descriptions, trajectories, and VQA tasks for embodied AI.
A dataset of annotated CCTV clips for evaluating VLMs in industrial safety.
Large-scale multimodal data corpus for training a universal reasoning model.
English chart-question pairs to evaluate whether VLMs actually analyze images instead of guessing.
Large DataComp-1B image-text corpus with improved captions from LLaVA-1.5-LLaMA3-8B.
Large-scale benchmark for comparative summarization of chart pairs by language models.
Instructional dataset for educational tasks and multimodal learning.