Skip to content

Dataset · U4R

arXiv scientific article benchmark for multimodal LLMs

U4R/DocGenome

An English-language dataset based on 500 thousand arXiv research articles for training and evaluating document processing.

What's inside

DocGenome was created by automatically annotating 500 thousand open-access scientific documents from 153 arXiv disciplines using the DocParser pipeline. The current English-language repository contains between 1 and 10 thousand records structuring multi-modal data from articles. The set is oriented toward question-answering tasks and text extraction from images.

What you can do with it

  • Training and testing multimodal LLMs (VLMs) on complex scientific document comprehension
  • Developing OCR systems and academic PDF layout parsing
  • Building Visual Question Answering (DocVQA) systems on charts and formulas
Tasks
Question answering, Image → text
Languages
English
Size
1K–10K rows
Downloads this month
23.3K
Likes
19

How to load it

from datasets import load_dataset

ds = load_dataset("U4R/DocGenome")