Our paper "Context-aware multimodal AI navigates hidden pathways in five centuries of art evolution" has been published in PNAS!
posted on July 24, 2026

Our paper “Context-aware multimodal AI navigates hidden pathways in five centuries of art evolution” has been published in Proceedings of the National Academy of Sciences (PNAS)!

Summary

This study presents a computational framework leveraging context-aware multimodal generative AI (such as Stable Diffusion and CLIP) to analyze five centuries of Western art history. By projecting artworks into high-dimensional semantic spaces, the research disentangles formal visual properties (such as color and composition) from contextual semantic features (such as subject matter and historical narrative). The findings reveal that contextual AI representations form continuous topological pathways aligning with historical artistic periods, styles, and individual artists, allowing for quantitative tracking of semantic evolution—such as the gradual shift from religious motifs to secular themes across centuries.