Skip to content

Dataset · Oitaaa

Complex operational PDF document analysis benchmark

Oitaaa/GDP.pdf

100 real operational PDFs and related questions from 10 industries to evaluate multimodal document analysis.

What's inside

Contains 100 real operational PDF files (reports, contracts, applications, records) and prompts for them. Data is drawn from actual workflows across 10 fields, including finance, medicine, law, science, and construction. The benchmark is referenced in Anthropic Fable 5 and Mythos 5 model cards.

What you can do with it

  • Evaluating multimodal LLMs on tasks involving dense PDFs with complex formatting
  • Testing RAG systems and parsers for enterprise document workflows
Tasks
Document question answering
Size
up to 1K rows
Downloads this month
31.6K
Likes
1

How to load it

from datasets import load_dataset

ds = load_dataset("Oitaaa/GDP.pdf")