Skip to content

Dataset · MMInstruction

Multimodal QA on arXiv figures and charts

MMInstruction/ArxivQA

Over 100k multiple-choice questions based on scientific images from arXiv articles.

What's inside

A large dataset containing between 100,000 and 1,000,000 English examples in JSONL format, complemented by image files from arXiv scientific publications. Each entry includes a link to an image, such as a scientific chart, and a multiple-choice question to test visual information understanding.

What you can do with it

  • Training and fine-tuning multimodal models (VLM) for scientific graph and diagram comprehension
  • Evaluating model capabilities in answering questions based on complex illustrations from articles
Tasks
Image → text
Languages
English
Size
100K–1M rows
Downloads this month
2.8K
Likes
38

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("MMInstruction/ArxivQA")