Computer-Using World Model

TL;DR

The paper introduces CUWM, a world model for desktop software that predicts UI state changes to improve agent decision-making in complex digital environments, enabling better planning and robustness.

Contribution

CUWM is a novel two-stage world model for desktop UI that predicts textual state changes and visualizes next states, trained on real user interactions and refined with reinforcement learning.

Findings

CUWM improves decision quality in Office tasks.

World-model-guided action search enhances robustness.

Offline training with reinforcement learning aligns predictions with environment structure.

Abstract

Agents operating in complex software environments benefit from reasoning about the consequences of their actions, as even a single incorrect user interface (UI) operation can derail long, artifact-preserving workflows. This challenge is particularly acute for computer-using scenarios, where real execution does not support counterfactual exploration, making large-scale trial-and-error learning and planning impractical despite the environment being fully digital and deterministic. We introduce the Computer-Using World Model (CUWM), a world model for desktop software that predicts the next UI state given the current state and a candidate action. CUWM adopts a two-stage factorization of UI dynamics: it first predicts a textual description of agent-relevant state changes, and then realizes these changes visually to synthesize the next screenshot. CUWM is trained on offline UI transitions collected from agents interacting with real Microsoft Office applications, and further refined with a lightweight reinforcement learning stage that aligns textual transition predictions with the structural requirements of computer-using environments. We evaluate CUWM via test-time action search, where a frozen agent uses the world model to simulate and compare candidate actions before execution. Across a range of Office tasks, world-model-guided test-time scaling improves decision quality and execution robustness.