We present SceneLoom, a VLM-powered system that enables context-aware coordination between data visualizations and real-world imagery to support expressive communication.

We propose a framework for integrating fine-tuned LLMs into domain-specific visualization systems. Tailor-Mind applies this framework in education to facilitate self-regulated learning.

TransforLearn is an interactive visual tutorial tool that helps beginners learn the Transformer model through architecture-driven and task-driven explorations.

We evaluate LLM-generated analogies in biology and physics with two-phrase studies, and presents a system to help teachers generate and refine analogies.