DeepYardDeepYard
C

CUA-Gym

Reinforcement learning benchmark suite for training GUI automation agents with verifiable rewards

Open SourceFree

About

CUA-Gym is a research framework providing a scalable training environment and standardized task suite for computer-use agents. Designed for RLVR (Reinforcement Learning with Verifiable Rewards) training, it offers deterministic evaluation of GUI automation agents across benchmark tasks. Enables reproducible research on autonomous desktop interaction and computer control agents with ground-truth reward signals.

Details

Type
Integrations
Language

Tags

autonomousevaluationopen-sourceframeworkpython