<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Latent Diffusion on Yiwen Cai</title><link>https://yiwen-cai.github.io/tags/latent-diffusion/</link><description>Recent content in Latent Diffusion on Yiwen Cai</description><generator>Hugo -- gohugo.io</generator><language>zh-cn</language><managingEditor>caiyiwen.cs@foxmail.com (Yiwen Cai)</managingEditor><webMaster>caiyiwen.cs@foxmail.com (Yiwen Cai)</webMaster><copyright>© 2026 Yiwen Cai</copyright><lastBuildDate>Mon, 19 Oct 2026 10:09:31 +0800</lastBuildDate><atom:link href="https://yiwen-cai.github.io/tags/latent-diffusion/index.xml" rel="self" type="application/rss+xml"/><item><title>MIT 6.S184: Flow Matching and Diffusion Models — 第 6 章：构建大规模图像与视频生成器</title><link>https://yiwen-cai.github.io/notes/diffusion/large-scale-generators/</link><pubDate>Sun, 18 Oct 2026 00:00:00 +0000</pubDate><author>caiyiwen.cs@foxmail.com (Yiwen Cai)</author><guid>https://yiwen-cai.github.io/notes/diffusion/large-scale-generators/</guid><description>从「网络到底长什么样」出发，讲清时间、类别与文本如何嵌入成向量，Diffusion Transformer 的 patchify、AdaLN 与交叉注意力，U-Net 的编解码结构；再推导 VAE 的重建项、先验项与重参数化，说明为什么现代生成模型都在潜空间里训练；最后用 Stable Diffusion 3 与 Movie Gen Video 把前五章的方法串起来。</description></item></channel></rss>