Paper 2601.15876
EvoCUA: Evolving Computer Use Agents via Learning from Scalable Synthetic Experience
- Published
- Jan 2026
- Research lab
- Independent
- Citations
- 31
- GitHub
- 336 stars
01 In brief
Summary
EvoCUA is a native computer-use agent that replaces static imitation with a self-sustaining evolutionary cycle integrating verifiable data synthesis, scalable interaction infrastructure, and iterative policy optimization.
A synthesis engine generates diverse tasks with executable validators, while a high-throughput sandbox platform orchestrates tens of thousands of asynchronous rollouts.
The training pipeline combines a cold start, rejection sampling fine-tuning, and step-level DPO to consolidate successes and correct failures.
On the OSWorld benchmark, EvoCUA-32B achieves a 56.7% success rate, surpassing the previous open-source SOTA OpenCUA-72B (45.0%) and the closed-weights UI-TARS-2 (53.1%).
Ablations show consistent gains across stages and foundation models, and scaling analysis indicates improvements with more compute and data.
The paper also proposes STEPO, a step-level policy optimization algorithm to address training-inference discrepancies in trajectory-level RL, with preliminary results showing superior performance over GRPO.
02 From the paper
Abstract
The development of native computer-use agents (CUA) represents a significant leap in multimodal AI. However, their potential is currently bottlenecked by the constraints of static data scaling. Existing paradigms relying primarily on passive imitation of static datasets struggle to capture the intricate causal dynamics inherent in long-horizon computer tasks. In this work, we introduce EvoCUA, a native computer use agentic model. Unlike static imitation, EvoCUA integrates data generation and policy optimization into a self-sustaining evolutionary cycle. To mitigate data scarcity, we develop a verifiable synthesis engine that autonomously generates diverse tasks coupled with executable validators. To enable large-scale experience acquisition, we design a scalable infrastructure orchestrating tens of thousands of asynchronous sandbox rollouts. Building on these massive trajectories, we propose an iterative evolving learning strategy to efficiently internalize this experience. This mechanism dynamically regulates policy updates by identifying capability boundaries -- reinforcing successful routines while transforming failure trajectories into rich supervision through error analysis and self-correction. Empirical evaluations on the OSWorld benchmark demonstrate that EvoCUA achieves a success rate of 56.7%, establishing a new open-source state-of-the-art. Notably, EvoCUA significantly outperforms the previous best open-source model, OpenCUA-72B (45.0%), and surpasses leading closed-weights models such as UI-TARS-2 (53.1%). Crucially, our results underscore the generalizability of this approach: the evolving paradigm driven by learning from experience yields consistent performance gains across foundation models of varying scales, establishing a robust and scalable path for advancing native agent capabilities.