Adapting Visualizations inContext:
A Natural Language Grammar Supported by Large Language Models

We designed a declarative visualization grammar compatible with Large Language Models, and an LLM-based controller to adapt the AR visualization according to designer's requirements and contextual information in real-time. The main challenge is how to divide the responsibilities of humans and Al in their collaboration, as LLM-based Al shows strong capabilities in some decision-making work.

* This work is still in progress.