UltraCUA: A Foundation Model for Computer Use Agents with Hybrid Action
This addresses the problem of cascading failures and inefficiency in computer-use agents for users relying on automated tasks, representing a novel integration rather than an incremental improvement.
The paper tackles the problem of multimodal agents for computer use being limited to primitive GUI actions, which leads to performance bottlenecks, by introducing UltraCUA, a foundation model that integrates GUI primitives with high-level programmatic tool calls. The result shows substantial improvements, such as a 22% relative improvement on OSWorld and a 21.7% success rate on WindowsAgentArena, while being 11% faster in execution steps.
Multimodal agents for computer use rely exclusively on primitive actions (click, type, scroll) that require accurate visual grounding and lengthy execution chains, leading to cascading failures and performance bottlenecks. While other agents leverage rich programmatic interfaces (APIs, MCP servers, tools), computer-use agents (CUAs) remain isolated from these capabilities. We present UltraCUA, a foundation model that bridges this gap through hybrid action -- seamlessly integrating GUI primitives with high-level programmatic tool calls. To achieve this, our approach comprises four key components: (1) an automated pipeline that scales programmatic tools from software documentation, open-source repositories, and code generation; (2) a synthetic data engine producing over 17,000 verifiable tasks spanning real-world computer-use scenarios; (3) a large-scale high-quality hybrid action trajectory collection with both low-level GUI actions and high-level programmatic tool calls; and (4) a two-stage training pipeline combining supervised fine-tuning with online reinforcement learning, enabling strategic alternation between low-level and high-level actions. Experiments with our 7B and 32B models demonstrate substantial improvements over state-of-the-art agents. On OSWorld, UltraCUA models achieve an average 22% relative improvement over base models, while being 11% faster in terms of steps. Out-of-domain evaluation on WindowsAgentArena shows our model reaches 21.7% success rate, outperforming baselines trained on Windows data. The hybrid action mechanism proves critical, reducing error propagation while maintaining execution efficiency.