Metraipolis › archive › dc_a0d3e72d697f
Temperon: Full-Time SAM Quality at a Third Less Wall-Clock — arXiv:2609.17575
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2609.17575
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-15T15:04:20.239Z
- SHA-256 of the archived text
37c839b8c975b2cb9ef3709c921c23405d0ca1ad1cf54108847b2c03e5a4104d- Stable URL of this page
https://metraipolis.com/documento/dc_a0d3e72d697f
How to cite
Temperon: Full-Time SAM Quality at a Third Less Wall-Clock — arXiv:2609.17575 — arXiv (en), https://arxiv.org/abs/2609.17575 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_a0d3e72d697f (Metraipolis archive dc_a0d3e72d697f, sha256 37c839b8c975b2cb9ef3709c921c23405d0ca1ad1cf54108847b2c03e5a4104d)
BibTeX
@misc{dca0d3e72d697f,
title = {Temperon: Full-Time SAM Quality at a Third Less Wall-Clock — arXiv:2609.17575},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2609.17575},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_a0d3e72d697f; sha256 37c839b8c975b2cb9ef3709c921c23405d0ca1ad1cf54108847b2c03e5a4104d},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.17575. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Sharpness-aware minimization (SAM) doubles the cost of every training step, yet its benefit concentrates where training ends. We study where an expensive training mode should be spent and propose Temperon: a plain-SGD explorer for the first 43% of the epoch budget, then one scheduled hand-off that gives the entire final cosine anneal to a SAM-wrapped Muon refiner. On CIFAR-10/100, SVHN and Tiny ImageNet (five seeds, times reported as epochs-to-target times an idle-GPU-calibrated epoch cost), Temperon matches the best full-time-SAM recipe on accuracy everywhere while reaching the hardest common target 35%, 34% and 32% sooner on three of the four, and sits a tier above the published SAM+SGD recipe at level cost. Ablations make the attribution exact: the Muon refiner is worth +0.85pp with everything else fixed; the explorer's shape and its restarts are worth nothing, and we withdraw them as contributions. Re-running the closest rival, late-phase SAM, at matched budget shows the frontier: it is fastest to every mid-level target, but the tier the Muon refiner buys (0.83 on CIFAR-100, 0.97 on CIFAR-10) is reached by no SGD-refined method in any seed, and on Tiny ImageNet, where Muon buys no tier, the rival simply wins -- the measured boundary of the method. The allocation law transfers to GPT-2 pretraining (full-SAM quality at -29% wall-clock) and GLUE fine-tuning (never worse than full-time SAM at a third of its SAM cost). Two constants organize the economics: skipping SAM early buys a fixed credit, and a Muon epoch costs 1.50x a SAM+SGD epoch on all four datasets. Finally, the hand-off cannot be timed from the trajectory: under cosine schedules the accuracy curve is plateau-then-surge, so the information lives in the schedule, making the scheduled switch principled rather than convenient. Code and a pip-installable implementation are released. Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-17 in cs.LG). Categorie arXiv: Machine Learning (cs.LG) ; Computer Vision and Pattern Recognition (cs.CV). Pagina del lavoro: https://arxiv.org/abs/2609.17575 — 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 2026-10-15T14:08:19.048Z dalla pagina https://arxiv.org/catchup/cs.LG/2026-09-17?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).