Skip to content

Dataset · sal4ahm

Scientific chart questions and answers

sal4ahm/RealCQA

28 thousand chart images and 2 million question-answer pairs for complex logical tasks.

What's inside

It contains 28,266 images of real scientific charts and about 2 million corresponding text question-answer pairs. The dataset was prepared for a scientific paper at the ICDAR 2023 conference to test models in logical reasoning and visual data understanding tasks.

What you can do with it

  • Training VLM models to answer complex questions on charts and diagrams
  • Evaluating neural network capabilities for logical analysis of visual information
Size
10K–100K rows
Downloads this month
25.5K
Likes
5

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("sal4ahm/RealCQA")