Metraipolis › archive › dc_ef0670ce653b

ARLoop: Fully-Open Pre-trained Looped Transformers with Attention Residuals — arXiv:2611.05234

Source: arXiv (en)
Original: https://arxiv.org/abs/2611.05234
License (as archived): CC0-1.0
Archived: 2026-11-04T18:38:48.540Z
SHA-256 of the archived text: 01bb9bf43fba4ac393a93b5a2b89776acfd2e75626a3c5a6c7c83b03e55b6da2
Stable URL of this page: https://metraipolis.com/documento/dc_ef0670ce653b

How to cite

ARLoop: Fully-Open Pre-trained Looped Transformers with Attention Residuals — arXiv:2611.05234 — arXiv (en), https://arxiv.org/abs/2611.05234 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_ef0670ce653b (Metraipolis archive dc_ef0670ce653b, sha256 01bb9bf43fba4ac393a93b5a2b89776acfd2e75626a3c5a6c7c83b03e55b6da2)

BibTeX

@misc{dcef0670ce653b,
  title = {ARLoop: Fully-Open Pre-trained Looped Transformers with Attention Residuals — arXiv:2611.05234},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2611.05234},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_ef0670ce653b; sha256 01bb9bf43fba4ac393a93b5a2b89776acfd2e75626a3c5a6c7c83b03e55b6da2},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2611.05234. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Looped transformers increase effective depth by repeatedly applying a shared set of transformer layers, offering a parameter-efficient way to scale computation. However, standard residual connections aggregate information across recurrent depth into a single stream, providing limited control over which intermediate representations from preceding layers should be reused. In this paper, we introduce ARLoop, a looped transformer architecture based on attention residuals, where the input of each layer is selectively aggregated through attention over model depth. Distinct from standard transformers, we find that looped transformers do not require access to the full residual history: sliding-window depth attention matches or outperforms full-history attention at larger recurrent depths. Across controlled experiments with multiple parameter scales (140M and 370M) and recurrent depths (R = 1 to 5), ARLoop consistently outperforms prior residual designs for looped transformers. More importantly, we further scale ARLoop to 0.6B and 1.2B parameters under a realistic LLM pretraining recipe, demonstrating that the gains transfer to realistic pretraining, with particularly strong improvements on math and coding tasks. Our 1.2B model, pretrained on only 0.5T tokens with three recurrences, outperforms Huginn-3.5B (the only fully-open looped LLM as of our release date) across all evaluated tasks and is competitive with similar-sized non-looped models such as Qwen3.5-2B despite being trained on substantially less data.

Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.CL, nuovi lavori del giorno).
Categorie arXiv: Computation and Language (cs.CL).
Pagina del lavoro: https://arxiv.org/abs/2611.05234 — 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 2026-11-04T18:32:52.173Z dalla pagina https://arxiv.org/list/cs.CL/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).