VizPilot: Automated Onboarding for SVG-based Composite Visualizations using Multimodal LLMs

2026-07-30

AI Summary

This work addresses the high cognitive load imposed on novice users by complex composite visualizations and the limitations of existing guidance approaches, which are often platform-dependent, labor-intensive to author, and lack generalizability. The paper proposes a novel automated method based on multimodal large language models (MLLMs) that, for the first time, reverse-engineers the structure of raw SVG-based visualizations and generates interactive guidance content. By integrating a two-stage visualization decomposition, structured explanation extraction, and precise mapping to SVG elements, the approach—implemented as a browser extension—supports both narrative-driven scrolling explanations and free-form exploration. Experimental results demonstrate that this solution significantly reduces authoring effort and user cognitive load, thereby substantially enhancing the usability and accessibility of composite visualizations.

Abstract

Composite visualizations integrate multiple visualizations to represent complex datasets effectively, but their intrinsic composite designs often impose a high initial cognitive load on novice users. Existing visualization onboarding approaches are typically platform-dependent, require substantial manual authoring effort, and struggle with the structural complexity of composite visualizations, limiting their general applicability. We present VizPilot, an automated visualization onboarding approach that reverse-engineers composite visualization structure to generate interactive onboarding experiences directly from raw visualization artifacts. VizPilot consists of two modules: a Composite Visualization Analyzer and an Onboarding Interface. Leveraging Multimodal Large Language Models (MLLMs), the Analyzer employs a two-stage pipeline that decomposes a visualization into visual components, extracts structured explanations, and maps them to precise SVG elements for reliable highlighting and interaction. Implemented as a browser extension, VizPilot requires only a brief visualization description and optional interaction source code from the visualization developer to automatically generate onboarding content. The Onboarding Interface supports both guided narrative scrollytelling and free exploration, enabling users to learn visualization components progressively or on demand. We evaluate VizPilot through a comparative analysis of different input modalities, a usage scenario demonstrating reduced authoring effort, and a user study assessing its impact on users' cognitive load. The results demonstrate that VizPilot effectively automates the authoring of onboarding experiences while improving the usability and accessibility of composite visualizations.

Problem

• composite visualizations
• visualization onboarding
• cognitive load
• SVG-based visualizations
• multimodal LLMs

Innovation

• Multimodal LLMs
• SVG-based visualization
• automated onboarding
• composite visualization
• visualization accessibility