Search papers, labs, and topics across Lattice.
The authors introduce DV-World, a new benchmark designed to evaluate data visualization (DV) agents in realistic, professional settings, addressing limitations of existing benchmarks like code-sandbox confinement and perfect intent assumptions. DV-World includes tasks spanning spreadsheet manipulation, adapting visualizations across programming paradigms, and aligning with ambiguous user intents. Experiments using a hybrid evaluation framework combining numerical precision and MLLM-as-a-Judge reveal that state-of-the-art models achieve less than 50% overall performance, highlighting significant gaps in real-world DV capabilities.
Current AI models are surprisingly inept at real-world data visualization tasks, failing more than half the time on a new benchmark designed to mimic enterprise workflows.
Real-world data visualization (DV) requires native environmental grounding, cross-platform evolution, and proactive intent alignment. Yet, existing benchmarks often suffer from code-sandbox confinement, single-language creation-only tasks, and assumption of perfect intent. To bridge these gaps, we introduce DV-World, a benchmark of 260 tasks designed to evaluate DV agents across real-world professional lifecycles. DV-World spans three domains: DV-Sheet for native spreadsheet manipulation including chart and dashboard creation as well as diagnostic repair; DV-Evolution for adapting and restructuring reference visual artifacts to fit new data across diverse programming paradigms and DV-Interact for proactive intent alignment with a user simulator that mimics real-world ambiguous requirements. Our hybrid evaluation framework integrates Table-value Alignment for numerical precision and MLLM-as-a-Judge with rubrics for semantic-visual assessment. Experiments reveal that state-of-the-art models achieve less than 50% overall performance, exposing critical deficits in handling the complex challenges of real-world data visualization. DV-World provides a realistic testbed to steer development toward the versatile expertise required in enterprise workflows. Our data and code are available at \href{https://github.com/DA-Open/DV-World}{this project page}.
