Skip to content

Dataset · lmms-lab-encoder

DocVQA for multimodal model evaluation

lmms-lab-encoder/DocVQA

Formatted DocVQA version for one-click evaluation of large multimodal models.

What's inside

This is a formatted version of the DocVQA dataset for the lmms-eval pipeline. Contains document images and text question-answer pairs; volume ranges from 10k to 100k examples. Designed for one-click evaluation of large multimodal models (LMMs) on the document VQA task.

What you can do with it

  • For LMM evaluation on DocVQA via lmms-eval
  • For testing VQA models on document images
  • For benchmarking the OCR/VQA pipeline
Size
10K–100K rows
Downloads this month
49.8K
Likes
90

How to load it

from datasets import load_dataset

ds = load_dataset("lmms-lab-encoder/DocVQA")