DV-World: Benchmarking Data Visualization Agents in Real-World Scenarios

Real-world data visualization (DV) requires native environmental grounding, cross-platform evolution, and proactive intent alignment, yet existing benchmarks are often limited to code sandboxes, creation-only tasks, and fully specified intents. We introduce DV-World, a 260-task benchmark for evaluating DV agents across professional visualization lifecycles. DV-World covers three domains: DV-Sheet for native spreadsheet chart/dashboard creation and diagnostic repair; DV-Evolution for adapting reference visual artifacts to new data across programming paradigms; and DV-Interact for proactive intent alignment with a user simulator under ambiguous requirements. Its hybrid evaluation combines \textit{Table-value Alignment} for numerical precision with rubric-based \textit{MLLM-as-a-Judge} for semantic-visual assessment. Experiments show that state-of-the-art models achieve below 50\% overall performance, exposing major gaps in real-world DV capabilities. DV-World provides a realistic testbed for developing DV agents suited to enterprise workflows. Data and code are available at \url{dv-world-project.github.io}.

Charts and dashboards are widely used in real workplaces, but building and revising them often requires more than writing a short piece of code. A useful visualization agent must understand spreadsheet software, edit existing charts, adapt visualizations to new data, and ask clarifying questions when the user’s request is ambiguous. We introduce DV-World, a benchmark designed to test whether AI agents can handle these real-world visualization workflows. DV-World contains 260 tasks across three settings: working directly with spreadsheet charts and dashboards, adapting existing visualizations across different tools and programming styles, and interacting with users to clarify unclear requirements. To evaluate agents, DV-World checks both numerical correctness and visual-semantic quality using table-based alignment and rubric-guided multimodal judging. Our experiments show that even strong current models achieve below 50% overall performance, revealing major gaps between today’s agents and the needs of professional visualization work. DV-World provides a realistic testbed for developing more reliable and helpful visualization agents.