1 paper
Minh-Quan Le, Armand Comas, Alexandros Lattas +7
The paper proposes a self‑correcting framework called SC‑CMJP that couples image understanding and generation via cross‑modal attention in masked diffusion models, and introduces a…