DocVQA: A Dataset for VQA on Document Images

DocVQA is a Visual Question Answering benchmark that consists of 50,000 questions covering 12,000+ document images. This implementation solves and scores the “validation” split.

Contributed By

@evanmiller-anthropic

Code

src/inspect_evals/docvqa

Paper

https://arxiv.org/abs/2007.00398

Overview

DocVQA is a Visual Question Answering benchmark that consists of 50,000 questions covering 12,000+ document images. This implementation solves and scores the “validation” split.

Usage

Installation

Install with pip install inspect-evals, or uv sync from a checkout of this repository.

Running evaluations

uv run inspect eval inspect_evals/docvqa --model openai/gpt-5-nano

You can also import tasks as normal Python objects and run them from python:

from inspect_ai import eval
from inspect_evals.docvqa import docvqa
eval(docvqa)

Drop uv run if you manage dependencies yourself. Log viewing (inspect view) and default-model setup are documented in the Inspect Evals README.

Options

You can control a variety of options from the command line. For example:

uv run inspect eval inspect_evals/docvqa --limit 10 --sample-shuffle
uv run inspect eval inspect_evals/docvqa --max-connections 10
uv run inspect eval inspect_evals/docvqa --temperature 0.5

See uv run inspect eval --help for all available options.

Parameters

docvqa

  • shuffle (bool): (default: True)

Dataset

The DocVQA dataset contains a “validation” split and a “test” split. To prevent leakage into training data, the authors of DocVQA have chosen to hold back the answers to the “test” split. Scoring on the “test” split requires coordinating with the DocVQA authors.

Each split contains several questions about each image. Here is an example image:

Diabetes in Australia

And associated example questions:

  • How many females are affected by diabetes?
  • What percentage of cases can be prevented?
  • What could lead to blindness or stroke diabetes?

The model is tasked to answer each question by referring to the image. The prompts are based on OpenAI’s simple-evals.

Scoring

DocVQA computes the Average Normalized Levenstein Similarity:

Average Normalized Levenstein Similarity definition

Changelog

[2-A] - 2026-02-16

  • Migrate version to new scheme. See #907.

[1.0.1] - 2025-12-18

  • Adds backoff policy for functions that connect to huggingface servers.