Softwaregym creates diverse apps to train smarter computer-use agents
SoftwareGym: Scaling Computer-Use Agent Training with Diverse Synthesized Software Environments
Multiagent SystemsSoftware Engineering
Summary
It’s hard to teach computer programs how to use software because there aren’t enough different and reliable software examples to learn from. The authors introduce SoftwareGym, a system that automatically builds many different software applications with working buttons and menus. These synthesized apps help train computer-use agents better without depending on real software. Models trained with SoftwareGym’s apps perform better on new, unseen tasks, showing that this approach can help make smarter agents who can use different software tools.
What this means in practice
- •For software developers: Generate diverse test applications automatically to train and evaluate software automation agents without extensive manual setup.
- •For enterprise ai teams: Use synthesized software environments to improve AI agents’ ability to perform complex workflows across multiple application types securely and reliably.
Authors
Tian Jin, Wenhao Wang, Jing Kang, Haoting Shi, Rui Ye, Yijun Lu, Siheng Chen
Abstract
Computer-use agents (CUAs) are rapidly improving, yet synthesizing diverse and reliable software environments at scale remains challenging. Native application environments require substantial construction and integration effort, which limits application coverage and increases the risk of application-specific overfitting. Browser-hosted environments provide limited interaction depth and workflow complexity. These limitations motivate SoftwareGym, an automated pipeline for synthesizing GUI software environments with diverse interfaces, executable functionality, and runtime-verified behavior. It generates structurally diverse designs, realizes them as runnable applications, and uses GUI evaluation with iterative repair to ensure functional reliability. Using SoftwareGym, we synthesize 182 applications across twelve categories and achieve a mean GUI task pass rate of 91.5%. We construct SoftwareGym-Bench, a 120-task evaluation suite spanning the same twelve categories, and use trajectories collected from these environments for CUA training. The resulting model improves Pass@3 from 41.39% to 56.97% on the strictly out-of-domain OSWorld benchmark and achieves 40.0% on SoftwareGym-Bench. These results demonstrate the potential of synthesized software environments as scalable and transferable supervision for CUA training.