Skip to content

Dataset · OpenGVLab

Benchmark for multimodal GUI agents across six platforms

OpenGVLab/MMBench-GUI

A test suite for evaluating multimodal agents working with graphical user interfaces across six platforms.

What's inside

A hierarchical benchmark by OpenGVLab consisting of screen images and text tasks to evaluate GUI automation agents. It covers Windows, macOS, Linux, iOS, Android, and web interfaces, assessing content comprehension, element retrieval, and task execution.

What you can do with it

  • Evaluating multimodal models as AI agents for computer and smartphone control
  • Testing accuracy in recognizing and locating clickable elements in screenshots
Tasks
Image text → text
Downloads this month
748
Likes
38

How to load it

from datasets import load_dataset

ds = load_dataset("OpenGVLab/MMBench-GUI")