Toward AI VIS Co-Scientists: A General and End-to-End Agent Harness for Solving Complex Data Visualization Tasks

2026-05-20

AI Summary

This work addresses the challenge that creating complex scientific visualizations typically demands interdisciplinary expertise and hinders researchers from efficiently producing them autonomously. To overcome this, we propose the first end-to-end multi-agent collaborative framework capable of automatically generating fully functional, view-coordinated, and highly customized single-page visualization applications from only raw data and high-level task descriptions. The framework integrates task planning, environment configuration, interface implementation, and validation mechanisms, supporting iterative refinement through intermediate artifacts and a documented instruction stream. Evaluated on multiple real-world scientific scenarios from the IEEE SciVis Contest, our system successfully constructed complete visualizations entirely autonomously, effectively meeting domain experts’ requirements and representing a significant step toward a general-purpose AI research partner.

Abstract

The ability to inspect, interpret, and communicate complex data is crucial for virtually any scientific endeavor, but often requires significant expertise outside the core domain ranging from data management and analysis to visualization design and implementation. We present an end-to-end agentic harness that, based on only the data and a high level description of the tasks, independently designs custom visual analysis applications (VIS apps). This represents an important step towards a general AI co-scientist envisioned by many as an autonomous system that can autonomously execute long horizon tasks based on high-level directions. Our proposed VIS co-scientist is an essential component of this broader AI co-scientist vision: a harness that can autonomously analyze data and design visualization solutions using a collection of agents and specialized skills that coordinate exploratory analysis, plan, configure the environment, implement, validate the interface, and most importantly evaluate the overall task completion. Each stage produces document and instruction artifacts that guide downstream work and enable iterative refinement. We validate this approach on IEEE SciVis Contests spanning multiple science and engineering fields. These contests serve as ideal proving grounds because they encode real-world complexity: ambiguous requirements, diverse data modalities, design trade-offs, and task-driven validation. Given only the data and target tasks, our system autonomously produces functional single-page VIS Apps with verified linked-view behavior, highly customized to domain experts' specified tasks and needs.

Problem

• Data Visualization
• AI Co-Scientist
• Complex Data Analysis
• Autonomous Visualization
• End-to-End Agent

Innovation

• AI co-scientist
• end-to-end agent harness
• automated visualization design
• multi-agent coordination
• task-driven VIS apps