ShowUI-$\pi$: Flow-based Generative Models as GUI Dexterous Hands

TL;DR

ShowUI-$\pi$ introduces a flow-based generative model enabling continuous, dexterous GUI manipulation like dragging, combining discrete and continuous actions, and establishing a new benchmark for evaluating GUI drag capabilities.

Contribution

The paper presents the first flow-based generative model for GUI dexterous manipulation, integrating discrete and continuous actions, and provides a new dataset and benchmark for GUI drag tasks.

Findings

ShowUI-$\pi$ outperforms existing proprietary GUI agents on the ScreenDrag benchmark.

The model achieves 26.98 score with only 450M parameters, demonstrating efficiency.

The approach enables smooth, on-the-fly continuous control in GUI interactions.

Abstract

Building intelligent agents capable of dexterous manipulation is essential for achieving human-like automation in both robotics and digital environments. However, existing GUI agents rely on discrete click predictions (x,y), which prohibits free-form, closed-loop trajectories (e.g. dragging a progress bar) that require continuous, on-the-fly perception and adjustment. In this work, we develop ShowUI-$\pi$, the first flow-based generative model as GUI dexterous hand, featuring the following designs: (i) Unified Discrete-Continuous Actions, integrating discrete clicks and continuous drags within a shared model, enabling flexible adaptation across diverse interaction modes; (ii) Flow-based Action Generation for drag modeling, which predicts incremental cursor adjustments from continuous visual observations via a lightweight action expert, ensuring smooth and stable trajectories; (iii) Drag Training data and Benchmark, where we manually collect and synthesize 20K drag trajectories across five domains (e.g. PowerPoint, Adobe Premiere Pro), and introduce ScreenDrag, a benchmark with comprehensive online and offline evaluation protocols for assessing GUI agents' drag capabilities. Our experiments show that proprietary GUI agents still struggle on ScreenDrag (e.g. Operator scores 13.27, and the best Gemini-2.5-CUA reaches 22.18). In contrast, ShowUI-$\pi$ achieves 26.98 with only 450M parameters, underscoring both the difficulty of the task and the effectiveness of our approach. We hope this work advances GUI agents toward human-like dexterous control in digital world. The code is available at https://github.com/showlab/showui-pi.