Skip to content

Dataset · VLR-CVC

DocVQA 2026 document benchmark

VLR-CVC/DocVQA-2026

A small document QA benchmark across eight domains for multimodal reasoning.

What's inside

This is the evaluation dataset DocVQA 2026 for the ICDAR 2026 competition on multimodal document reasoning. Questions require reasoning over documents from eight domains: business reports, scientific papers, slides, posters, maps, comics, infographics, and engineering drawings. The data is in English, featuring image and text modalities, with a size of less than 1k examples.

What you can do with it

  • Evaluating VLMs on document question answering
  • Multimodal reasoning benchmark
  • Evaluating OCR+LLM pipelines
Tasks
Visual question answering, Document question answering, Image text → text, Question answering
Languages
English
Size
up to 1K rows
Downloads this month
1K
Likes
74

How to load it

from datasets import load_dataset

ds = load_dataset("VLR-CVC/DocVQA-2026")