Metraipolis › archive › dc_e43af8ea76ee

There Is No Free Loop: When Does Looping a Trained Model Pay for Its Inference Compute? — arXiv:2610.42030

Source: arXiv (en)
Original: https://arxiv.org/abs/2610.42030
License (as archived): CC0-1.0
Archived: 2026-11-03T07:13:30.017Z
SHA-256 of the archived text: ccc076130baa11873b2f1fb8367ac2c10ae1a8c217229a07a375a873f53eb2a8
Stable URL of this page: https://metraipolis.com/documento/dc_e43af8ea76ee

How to cite

There Is No Free Loop: When Does Looping a Trained Model Pay for Its Inference Compute? — arXiv:2610.42030 — arXiv (en), https://arxiv.org/abs/2610.42030 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_e43af8ea76ee (Metraipolis archive dc_e43af8ea76ee, sha256 ccc076130baa11873b2f1fb8367ac2c10ae1a8c217229a07a375a873f53eb2a8)

BibTeX

@misc{dce43af8ea76ee,
  title = {There Is No Free Loop: When Does Looping a Trained Model Pay for Its Inference Compute? — arXiv:2610.42030},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2610.42030},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_e43af8ea76ee; sha256 ccc076130baa11873b2f1fb8367ac2c10ae1a8c217229a07a375a873f53eb2a8},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.42030. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Looping re-applies a block of layers so that each token passes through the block more than once. Adding a loop to a trained model is attractive because it leaves the parameter count, and hence the disk size, the memory footprint and the training cost unchanged. Unfortunately, every generated token pays for the loop. Re-applying m of N layers makes each token about (N+m)/N times as expensive. Loops that a model learns during large-scale pretraining justify this price, but whether loops added after pretraining justify this cost remains unclear. We compare a looped model with two simple baselines. We first compare it with the same model trained identically without the loop (equal parameters). Secondly, we compare it with the best unlooped use of the same inference floating-point operations (equal FLOPs). Across Qwen3 and Qwen3.5 models from 0.8B to 9B parameters, only the looped 0.8B model beats both baselines. At 4B and 9B, the looped model does not beat the first baseline, i.e. the same model trained without the loop. This is not because the benchmarks are too easy, since on the same benchmarks the unlooped 9B model still scores up to 13 points higher than the 4B model, showing that there is ample room for improvement. Introducing loops in an already pretrained model therefore only pays for small models with a fixed parameter budget.

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/2610.42030 — 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 2026-11-03T07:09:31.890Z dalla pagina https://arxiv.org/list/cs.AI/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).