Metraipolis › archive › dc_dd10dcf11392

LiFT: Loop Flow Transformers — arXiv:2610.05538

Source
arXiv
Original
https://arxiv.org/abs/2610.05538
License (as archived)
CC0-1.0
Archived
2026-10-06T05:26:58.022Z
SHA-256 of the archived text
dacf8ef99b9930c1bde99f7ca0b3cd7480bc877332b20d657f9b827da07cdfad
Stable URL of this page
https://metraipolis.com/documento/dc_dd10dcf11392

How to cite

LiFT: Loop Flow Transformers — arXiv:2610.05538 — arXiv, https://arxiv.org/abs/2610.05538 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_dd10dcf11392 (Metraipolis archive dc_dd10dcf11392, sha256 dacf8ef99b9930c1bde99f7ca0b3cd7480bc877332b20d657f9b827da07cdfad)
BibTeX
@misc{dcdd10dcf11392,
  title = {LiFT: Loop Flow Transformers — arXiv:2610.05538},
  howpublished = {arXiv},
  url = {https://arxiv.org/abs/2610.05538},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_dd10dcf11392; sha256 dacf8ef99b9930c1bde99f7ca0b3cd7480bc877332b20d657f9b827da07cdfad},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv, https://arxiv.org/abs/2610.05538. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

We introduce Loop Flow Transformers (LiFT), a family of looped generative models that scales computation by repeatedly applying a shared Diffusion Transformer (DiT) core, with only light changes to the standard architecture. Rather than asking every recurrent step for the final prediction, LiFT trains each step with a single regression target: a point on a straight path from the model's initial estimate to the flow-matching target. Because we index these targets by a continuous depth coordinate, a trained model can loop far beyond its training depth with no retraining, early exits, or other modifications. In our experiments, these longer rollouts improve generation, so inference computation can grow without adding parameters. On ImageNet at 256x256, LiFT-L/2 achieves an FID 3.34 points lower than our dense DiT-XL/2 baseline while using approximately 60% fewer parameters, 32% fewer training FLOPs, and 52% fewer inference FLOPs.

Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.AI, nuovi lavori del giorno).
Categorie arXiv: Machine Learning (cs.LG) ; Artificial Intelligence (cs.AI).
Pagina del lavoro: https://arxiv.org/abs/2610.05538 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-10-06T05:24:06.036Z dalla pagina https://arxiv.org/list/cs.AI/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).