ComputerRL: Scaling End-to-End Online Reinforcement Learning for Computer Use Agents
This work addresses the problem of inefficient and unstable training for desktop automation agents, enabling more skillful operation of digital workspaces, though it appears incremental in its method improvements.
The paper tackles the challenge of scaling end-to-end online reinforcement learning for autonomous desktop agents by introducing ComputerRL, a framework that unifies API and GUI interactions and employs distributed training and a novel training strategy, achieving a state-of-the-art accuracy of 48.9% on the OSWorld benchmark.
We introduce ComputerRL, a framework for autonomous desktop intelligence that enables agents to operate complex digital workspaces skillfully. ComputerRL features the API-GUI paradigm, which unifies programmatic API calls and direct GUI interaction to address the inherent mismatch between machine agents and human-centric desktop environments. Scaling end-to-end RL training is crucial for improvement and generalization across diverse desktop tasks; however, it remains challenging due to environmental inefficiency and instability during extended training. To support scalable and robust training, we develop a distributed RL infrastructure capable of orchestrating thousands of parallel virtual desktop environments to accelerate large-scale online RL. Furthermore, we propose Entropulse, a training strategy that alternates reinforcement learning with supervised fine-tuning, effectively mitigating entropy collapse during extended training runs. We employ ComputerRL on open models GLM-4-9B-0414 and GLM-4.1V-9B-Thinking, and evaluate them on the OSWorld benchmark. The AutoGLM-OS-9B achieves a new state-of-the-art accuracy of 48.9%, demonstrating significant improvements for general agents in desktop automation. Our code and the new OfficeWorld benchmark are available at https://github.com/thudm/ComputerRL. The algorithm and framework are adopted in building AutoGLM (Liu et al., 2024b).