Dataset · microsoft
Business Document Question-Answering Benchmark
microsoft/SynthDocQA
Synthetic business PDFs and nearly 10,000 questions on tables, charts, forms, and text for evaluating Document QA systems.
What's inside
Contains 100 synthetic PDF documents from various business domains and 9,798 English questions about them. Each question is tied to a specific page element (table, chart, form, note, or text block) and includes structured rules to verify the expected answer. Published by Microsoft as a benchmark for document understanding systems.
What you can do with it
- Evaluate multimodal LLMs and RAG systems in retrieval tasks on complex PDFs
- Test models for data extraction from tables, forms, and charts
- Tasks
- Question answering, Document question answering
- Languages
- English
- Size
- 1K–10K rows
- Downloads this month
- 2.2K
- Likes
- 8
How to load it
from datasets import load_dataset
ds = load_dataset("microsoft/SynthDocQA")