Daily AI news, sourcedWhere it stands
As of Oct 12, SenseTime, Tsinghua and NTU researchers propose Looped Diffusion Transformer (Looped-DiT), which reuses shared Transformer blocks several times inside each denoising step.
- Official
Looped-DiT paper shows a 260M looped text-to-image model beating a 6.5× larger baseline with less inference compute
Posted on arXiv on Oct 12, 2026 (source: arxiv.org).
SenseTime, Tsinghua and NTU researchers propose Looped Diffusion Transformer (Looped-DiT), which reuses shared Transformer blocks several times inside each denoising step. With deep supervision and self-modulating attention, a 260M-parameter looped model surpasses a non-looped model about 6.5× larger across text-to-image benchmarks while using about 4.9× less inference compute; loop depth can also be changed at sampling time. Paper: arXiv 2609.40305.
Source: arXiv
- Official
- An institution announces something that has already happened, through its own channel or a founder or executive account.
- Statement by …
- A named person's views, predictions, plans or comments, including remarks reported by media when the original is unavailable.
- Reported by …
- Only media or another third party is saying it. We name the outlet.
- Unconfirmed
- Circulating online (posts, comments, leaks). No company, named person or outlet has confirmed it.
Items older than 72 hours never go in Top stories and show their original date.