1 paper
Yu Wang, Sharon Li
The paper investigates whether unified multimodal models share a common semantic space by introducing cross-branch semantic steering, showing that semantic directions from the unde…