PulseAugur
实时 06:19:42
English(EN) Optimizer Memory Schedules for Outscaling the Overtraining Axis

新研究表明优化器性能随训练范围变化

一篇新研究论文探讨了在训练范围和参数数量增加时,不同优化器的性能表现。研究发现,像 Muon、SOAP、ADANAAdamW 这样的优化器的最佳超参数和相对性能会随着训练的延长而显著改变。具体来说,首选的学习率调度可能会逆转,权重衰减系数会随着超训练因子的平方根进行缩放。ADANA 在超训练水平较高时,尤其是在采用对数时间权重衰减和动量冷却的情况下,显示出比 AdamW 持续的缩放优势,并与 SOAP 竞争。在大多数测试范围内,Muon 和 SOAP 相对于 AdamW 提供了持续的代币效率优势。 AI

影响 研究结果表明,训练范围是选择和设计大规模模型优化器的关键因素。

排序理由 发表在 arXiv 上的研究论文,详细介绍了优化器性能。[lever_c_demoted from research: ic=1 ai=1.0]

在 arXiv cs.LG 阅读 →

AI 生成摘要 · Google Gemini · 来自 1 个来源。 我们如何撰写摘要 →

新研究表明优化器性能随训练范围变化

本文如何被排名

Signal score
32 / 100
Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket
Tool
发表在 arXiv 上的研究论文,详细介绍了优化器性能。[lever_c_demoted from research: ic=1 ai=1.0]
Source corroboration
Single-source cluster
Only one publisher covered this so far. Single-source stories can still rank when the publisher is high-authority, but they lack cross-source corroboration.
Topics
paper, model release
Editorial topic classification. Feeds into how the story surfaces on /topic/<slug> hub pages and into the per-entity coverage mix.
AI-industry relevance
High
Clearly on-topic for AI-industry coverage.
Story freshness
Breaking (< 6h)
Fresh story with cross-source coverage still developing. Ranking may shift as more sources report.

完整方法见我们的编辑标准

报道来源 [1]

  1. arXiv cs.LG TIER_1 English(EN) · Katie Everett, Shikai Qiu ·

    优化器内存调度以超规模化过度拟合轴

    arXiv:2609.04577v1 Announce Type: new Abstract: We investigate how optimizers scale across the overtraining axis and show that relative optimizer performance and optimal hyperparameters change substantially with training horizon. In particular, we study how matrix-preconditioned …