Aguvis: Unified Pure Vision Agents for Autonomous GUI Interaction
This addresses the problem of platform-specific and text-reliant GUI automation for users and developers, representing a novel advancement rather than an incremental improvement.
The paper tackles the challenge of automating GUI tasks by introducing Aguvis, a unified vision-based framework that operates directly on screen images and incorporates structured reasoning, achieving state-of-the-art performance across offline and real-world online benchmarks.
Automating GUI tasks remains challenging due to reliance on textual representations, platform-specific action spaces, and limited reasoning capabilities. We introduce Aguvis, a unified vision-based framework for autonomous GUI agents that directly operates on screen images, standardizes cross-platform interactions and incorporates structured reasoning via inner monologue. To enable this, we construct Aguvis Data Collection, a large-scale dataset with multimodal grounding and reasoning annotations, and develop a two-stage training pipeline that separates GUI grounding from planning and reasoning. Experiments show that Aguvis achieves state-of-the-art performance across offline and real-world online benchmarks, marking the first fully autonomous vision-based GUI agent that operates without closed-source models. We open-source all datasets, models, and training recipes at https://aguvis-project.github.io to advance future research.