AutoGUIWorld: Image Generators as Visual World Models for GUI Agent

2026-10-01

AI Summary

This study addresses the high cost and limited diversity inherent in training GUI agents on real-world environments by pioneering the use of an image generator as a visual world model to synthesize high-quality interaction trajectories without executing actual software. The proposed approach integrates the visual priors of generative models with the task knowledge of planners, achieving data synthesis through structured scene sampling, atomic action planning, iterative screenshot editing, and spatial annotation filtering. In total, 79,000 synthetic samples are generated. Following fine-tuning on this dataset, the resulting model achieves an average score of 40.8% on OSWorld and a success rate of 32.2% on ScienceBoard, demonstrating the efficacy of synthesizing scalable and diverse training data for GUI agent development.

Abstract

GUI agents require high-quality interaction trajectories to learn how software environments respond to actions, maintain state, and support multi-step workflows. However, the diversity of available trajectories is constrained by the applications, interface states, and workflows accessible in the underlying environments. Expanding this coverage requires deploying increasingly diverse and complex software, with specialized applications imposing additional installation, configuration, and runtime costs. We introduce AutoGUIWorld, a data generation framework that combines the visual priors of image generators with the task knowledge of a planner to synthesize GUI interaction trajectories without deploying or running the corresponding software environments. AutoGUIWorld samples initial GUI scenes from structured specifications of operating-system context, visual appearance, and interface state, and generates tasks conditioned on those scenes. A planner then specifies atomic actions and their intended visual consequences, while an image generator iteratively edits the current screenshot to produce subsequent observations. Action grounding and transition-level quality filtering yield 79,266 spatially annotated step-level training samples across Ubuntu, Windows, macOS, and Chrome. Fine-tuning Qwen3.5-35B-A3B on AutoGUIWorld trajectories improves the mean task score on OSWorld from 33.0% to 40.8% and the task success rate on ScienceBoard from 14.0% to 32.2%. These results show that generated trajectories improve GUI-agent performance on real desktop and scientific tasks.

Problem

• GUI agent
• interaction trajectories
• data diversity
• visual world model

Innovation

• GUI Agent
• Visual World Models
• Data Generation Framework
• Image Generators
• Trajectory Synthesis