Skip to content

Dataset · earakely-scale

AI agent benchmark for desktop multitasking with humans

earakely-scale/CuaDriverBench-AgentEnv

Environment and execution logs for evaluating AI agents operating a computer alongside a human user.

What's inside

The dataset includes 100,000 to 1,000,000 records containing video, text, and tabular data in English for reinforcement learning and image-text-to-text tasks. In the Cua Driver Bench on AgentEnv scenarios, the agent performs interface tasks (e.g., clicking buttons beneath pop-up notifications) while the user types in a parallel window.

What you can do with it

  • Testing computer use AI agents (CUA) under human background activity
  • Training multimodal navigation models in cluttered interfaces with overlaps
Tasks
Reinforcement learning, Image text → text
Languages
English
Size
100K–1M rows
Downloads this month
4.8K
Likes
2

How to load it

from datasets import load_dataset

ds = load_dataset("earakely-scale/CuaDriverBench-AgentEnv")