IAAILooped-DiT
Looped Diffusion Transformer
Scales text-to-image computation by repeating shared Transformer blocks within each denoising step. Deep supervision across intermediate loops and self-modulating attention fix the weak supervision and eroding local detail that make naive looping fail.
Inside the method
Each denoising step: [ Shared DiT blocks ] × r (deep supervision)
Simplified conceptual schematic. Consult the paper for the complete architecture.
- Recurrence family
- Core recurrence
- Depth control
- Fixed loops per denoising step
- KV / state strategy
- Not applicable (diffusion)
Reading note
Beats non-looped baselines at matched parameters and compute; a 260M looped model surpasses a 6.5× larger model on several text-to-image benchmarks with 4.9× less inference compute. More loops help more than more denoising steps at a fixed budget.
Sources checked 2026-10-28. This catalog does not imply independent reproduction.
Related works
Cite this work
@misc{loopeddit2026,
title = {Looped Diffusion Transformer},
author = {Yong Xien Chng and Tianyi Chen and Wenwen Tong and Haiwen Diao and Zhongang Cai and Lei Yang and Ziwei Liu and Lewei Lu and Dahua Lin and Gao Huang},
year = {2026},
eprint = {2609.40305},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2609.40305}
}