Visual Test-time Scaling for GUI Agent Grounding
This addresses the challenge of accurate action selection for Vision Language Model Agents in complex webpages, representing a strong specific gain in domain-specific interactive settings.
The paper tackles the problem of GUI agent grounding by introducing RegionFocus, a visual test-time scaling approach that dynamically zooms in on relevant regions to reduce clutter, resulting in performance gains of 28+% on Screenspot-pro and 24+% on WebVoyager benchmarks, achieving a new state-of-the-art grounding performance of 61.6% on ScreenSpot-Pro.
We introduce RegionFocus, a visual test-time scaling approach for Vision Language Model Agents. Understanding webpages is challenging due to the visual complexity of GUI images and the large number of interface elements, making accurate action selection difficult. Our approach dynamically zooms in on relevant regions, reducing background clutter and improving grounding accuracy. To support this process, we propose an image-as-map mechanism that visualizes key landmarks at each step, providing a transparent action record and enables the agent to effectively choose among action candidates. Even with a simple region selection strategy, we observe significant performance gains of 28+\% on Screenspot-pro and 24+\% on WebVoyager benchmarks on top of two state-of-the-art open vision language model agents, UI-TARS and Qwen2.5-VL, highlighting the effectiveness of visual test-time scaling in interactive settings. We achieve a new state-of-the-art grounding performance of 61.6\% on the ScreenSpot-Pro benchmark by applying RegionFocus to a Qwen2.5-VL-72B model. Our code will be released publicly at https://github.com/tiangeluo/RegionFocus.