Skip to content

Dataset · xlangai

OpenCUA desktop agent trajectories

xlangai/AgentNet

22.6K human-annotated tasks for computer-use agents on Windows, macOS, and Ubuntu.

What's inside

AgentNet is the first large-scale dataset of desktop computer-use agent trajectories: 22.6K human-annotated tasks for Windows, macOS, and Ubuntu. The data is designed for training and evaluating vision-language-action (VLA) computer-use models. Language is English, modality is image-text-to-text.

What you can do with it

  • Training VLA for computer use
  • Evaluating desktop agents
  • Fine-tuning OpenCUA
Tasks
Image text → text
Languages
English
Downloads this month
14.8K
Likes
99

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("xlangai/AgentNet")