Search papers, labs, and topics across Lattice.
This paper tackles the challenge of generating executable scientific visualization pipelines from natural language descriptions using LLMs, specifically for web-based environments like vtk.js. They introduce a structure-aware retrieval-augmented generation workflow that leverages pipeline-aligned code examples to guide module selection, parameter configuration, and execution order. Experiments across various visualization tasks demonstrate that this structured, domain-specific context significantly improves pipeline executability and reduces the manual correction effort required.
Forget generic code generation – this work shows that structure-aware retrieval of domain-specific examples slashes the debugging needed to get LLMs to produce working scientific visualization pipelines.
Scientific visualization pipelines encode domain-specific procedural knowledge with strict execution dependencies, making their construction sensitive to missing stages, incorrect operator usage, or improper ordering. Thus, generating executable scientific visualization pipelines from natural-language descriptions remains challenging for large language models, particularly in web-based environments where visualization authoring relies on explicit code-level pipeline assembly. In this work, we investigate the reliability of LLM-based scientific visualization pipeline generation, focusing on vtk.js as a representative web-based visualization library. We propose a structure-aware retrieval-augmented generation workflow that provides pipeline-aligned vtk.js code examples as contextual guidance, supporting correct module selection, parameter configuration, and execution order. We evaluate the proposed workflow across multiple multi-stage scientific visualization tasks and LLMs, measuring reliability in terms of pipeline executability and human correction effort. To this end, we introduce correction cost as metric for the amount of manual intervention required to obtain a valid pipeline. Our results show that structured, domain-specific context substantially improves pipeline executability and reduces correction cost. We additionally provide an interactive analysis interface to support human-in-the-loop inspection and systematic evaluation of generated visualization pipelines.
