Metraipolis › archive › dc_93a1cd31283e

A Tilted Bowl Is Not a Slippery Slope: Compressing Looped Models — arXiv:2609.39277

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.39277
License (as archived)
CC0-1.0
Archived
2026-10-07T13:14:24.466Z
SHA-256 of the archived text
53697ca0c513dfb77e0f8e4acc7ceccfd3857ce499f82ffc89b0f6648456cf3b
Stable URL of this page
https://metraipolis.com/documento/dc_93a1cd31283e

How to cite

A Tilted Bowl Is Not a Slippery Slope: Compressing Looped Models — arXiv:2609.39277 — arXiv (en), https://arxiv.org/abs/2609.39277 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_93a1cd31283e (Metraipolis archive dc_93a1cd31283e, sha256 53697ca0c513dfb77e0f8e4acc7ceccfd3857ce499f82ffc89b0f6648456cf3b)
BibTeX
@misc{dc93a1cd31283e,
  title = {A Tilted Bowl Is Not a Slippery Slope: Compressing Looped Models — arXiv:2609.39277},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.39277},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_93a1cd31283e; sha256 53697ca0c513dfb77e0f8e4acc7ceccfd3857ce499f82ffc89b0f6648456cf3b},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.39277. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Looped models reason by applying the same block of weights many times, so compressing that block saves memory traffic on every loop. Compressed looped models, however, often collapse, and the collapse is usually blamed on rounding error that accumulates from loop to loop. In this work we test that account on more than 30 models from five families and find, to our surprise, that it holds only for loops that never settle. When a loop settles, a fixed rounding error does not accumulate. It moves the point where the loop settles, much as tilting a bowl moves where a ball comes to rest, and the answer is lost only when the shift is larger than the readout tolerates. This picture lets us predict which models fail from a single label-free measurement, and it tells us why failed models recover: their loops still settle, so a few final loops with 8-bit weights bring the answer back. Motivated by these findings, we build a controller that stops when the model's halting head fires and then finishes with 8-bit loops. On Sudoku-Extreme and Maze-Hard it beats fixed-depth inference by up to 15 points under a third of the weight traffic.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-10-01 in cs.CL).
Categorie arXiv: Machine Learning (cs.LG) ; Computation and Language (cs.CL).
Pagina del lavoro: https://arxiv.org/abs/2609.39277 — 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 2026-10-07T12:44:51.959Z dalla pagina https://arxiv.org/catchup/cs.CL/2026-10-01?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).