Challenges & Opportunities with LLM-Assisted Visualization Retargeting

2025-07-02

AI Summary

Existing visualization chart reuse faces high barriers: practitioners must simultaneously comprehend example code and novel data schemas, leading to inefficient and error-prone adaptation. This paper proposes two large language model (LLM)-assisted chart migration methods—direct code generation and structure-guided program synthesis—that jointly leverage visualization encoding grammar and data semantic analysis to achieve end-to-end, data-driven code transformation. Systematic evaluation across diverse chart types and complexity levels demonstrates that both approaches significantly reduce manual adaptation effort. Failure analysis identifies data preprocessing quality as the primary bottleneck, highlighting data alignment and structured prompt engineering as key directions for future improvement. To our knowledge, this is the first work to systematically introduce program synthesis paradigms into visualization reuse tasks, establishing a novel pathway toward low-code visualization development.

Technology Category

• Natural Language Processing: Code Generation / Program Synthesis from Natural Language
• Data Mining & Knowledge Management: Data Visualization & Summarization
• Computer Vision: Large Vision Models

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Web Mining and Content Analysis: Web data visualization

Abstract

Despite the ubiquity of visualization examples published on the web, retargeting existing custom chart implementations to new datasets remains difficult, time-intensive, and tedious. The adaptation process assumes author familiarity with both the implementation of the example as well as how the new dataset might need to be transformed to fit into the example code. With recent advances in Large Language Models (LLMs), automatic adaptation of code can be achieved from high-level user prompts, reducing the barrier for visualization retargeting. To better understand how LLMs can assist retargeting and its potential limitations, we characterize and evaluate the performance of LLM assistance across multiple datasets and charts of varying complexity, categorizing failures according to type and severity. In our evaluation, we compare two approaches: (1) directly instructing the LLM model to fully generate and adapt code by treating code as text inputs and (2) a more constrained program synthesis pipeline where the LLM guides the code construction process by providing structural information (e.g., visual encodings) based on properties of the example code and data. We find that both approaches struggle when new data has not been appropriately transformed, and discuss important design recommendations for future retargeting systems.

Problem

• LLM-assisted retargeting of custom visualizations to new datasets
• Evaluating LLM performance in adapting code for varying chart complexities
• Comparing direct code generation versus structured program synthesis approaches

Innovation

• LLM-assisted code adaptation for visualization retargeting
• Two approaches: direct code generation and guided synthesis
• Evaluation across datasets and chart complexities