Preprint
Aug 2026
Domain-Grounded Candidate Selection for Agentic Image Editing: A Shadow Removal Case
The results suggest that commercial vision-language models do not replace classic low-level vision priors; instead, such priors remain useful for constraining and steering physically underconstrained generation.
Shi-Lin Hu, Jing-Yi Xu, Dimitris Samaras et al.
· 1 citation