<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:content="http://purl.org/rss/1.0/modules/content/"><channel><title>Scaling-Law on Xu'Blog</title><link>https://xuquant.com/tags/scaling-law/</link><description>Recent content in Scaling-Law on Xu'Blog</description><image><title>Xu'Blog</title><url>https://xuquant.com/og-default.png</url><link>https://xuquant.com/og-default.png</link></image><generator>Hugo -- 0.152.2</generator><language>zh</language><lastBuildDate>Sun, 07 Jun 2026 10:00:00 +0800</lastBuildDate><atom:link href="https://xuquant.com/tags/scaling-law/index.xml" rel="self" type="application/rss+xml"/><item><title>训练大模型的 Scaling Law：科学、工程与边界</title><link>https://xuquant.com/posts/foundation-models/chinchilla-and-modern-llm-training/</link><pubDate>Sun, 07 Jun 2026 10:00:00 +0800</pubDate><guid>https://xuquant.com/posts/foundation-models/chinchilla-and-modern-llm-training/</guid><description>Scaling law 是一套被研究透的科学，也是一个被行业系统性偏离的指南。从 Hestness 2017 的早期实证到 Rosenfeld 2020 的闭式解、Kaplan / Chinchilla 的拉锯、Besiroglu 2024 的复刻批判，再到训练栈的 4D 并行、FP8、Post-training RL，把 scaling law 的科学结论与工程演化串成一条主线。附 D3 拟合脆弱性 playground。</description></item></channel></rss>