Skip to content

Dataset · Mike33

Multilingual chart QA benchmark

Mike33/PolyChartQA

Test set of 10K-100K chart examples with generation code and questions in 10 languages for VLM evaluation.

What's inside

The test portion of the multilingual PolyChartQA benchmark, sized between 10K and 100K records for visual question answering. It contains chart images, structured annotations, generation code, and question-answer pairs in 10 languages (including English, Arabic, Spanish, French, Hindi, and Bengali). The set tests model ability to read text in diagrams and perform numerical computations.

What you can do with it

  • Evaluating multilingual VLMs on chart and diagram comprehension
  • Testing model numerical reasoning capabilities on visual data
  • Testing OCR and structure extraction from diagrams across languages
Tasks
Visual question answering
Languages
Arabic, Bangla, English, Spanish, French, Hindi, Japanese, Russian and 2 more
Size
10K–100K rows
Downloads this month
34K
Likes
0

How to load it

from datasets import load_dataset

ds = load_dataset("Mike33/PolyChartQA")