VideoGUI: A Benchmark for GUI Automation from Instructional Videos

TL;DR

VideoGUI introduces a comprehensive benchmark from instructional videos to evaluate GUI automation systems across hierarchical levels, highlighting current models' struggles with complex, visual-centric GUI tasks.

Contribution

The paper presents VideoGUI, a multi-modal benchmark for assessing GUI assistants on complex, visual-centric tasks involving professional software, with detailed hierarchical evaluation metrics.

Findings

GPT-4o performs poorly on visual-centric GUI tasks

High-level planning remains a significant challenge for current models

Benchmark enables detailed analysis of GUI automation performance

Abstract

Graphical User Interface (GUI) automation holds significant promise for enhancing human productivity by assisting with computer tasks. Existing task formulations primarily focus on simple tasks that can be specified by a single, language-only instruction, such as "Insert a new slide." In this work, we introduce VideoGUI, a novel multi-modal benchmark designed to evaluate GUI assistants on visual-centric GUI tasks. Sourced from high-quality web instructional videos, our benchmark focuses on tasks involving professional and novel software (e.g., Adobe Photoshop or Stable Diffusion WebUI) and complex activities (e.g., video editing). VideoGUI evaluates GUI assistants through a hierarchical process, allowing for identification of the specific levels at which they may fail: (i) high-level planning: reconstruct procedural subtasks from visual conditions without language descriptions; (ii) middle-level planning: generate sequences of precise action narrations based on visual state (i.e., screenshot) and goals; (iii) atomic action execution: perform specific actions such as accurately clicking designated elements. For each level, we design evaluation metrics across individual dimensions to provide clear signals, such as individual performance in clicking, dragging, typing, and scrolling for atomic action execution. Our evaluation on VideoGUI reveals that even the SoTA large multimodal model GPT4o performs poorly on visual-centric GUI tasks, especially for high-level planning.