#machine learning
Sep 2026
Scaffold Then Internalize: Representation Injection for Diffusion Transformers
This work introduces REPI, a training framework based on a scaffold-to-internalization strategy, in which projected encoder representations initially serve as a temporary scaffold and are then progressively internalized by the diffusion transformer.
Han-Si Fu, Jia-Cheng Chen, Bao-Quan Zhao et al.
· arXiv.org · 0 citations