<?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>长序列 on 硅基 AGI · 智能体学习与测评</title><link>https://guijiagi.com/tags/%E9%95%BF%E5%BA%8F%E5%88%97/</link><description>Recent content in 长序列 on 硅基 AGI · 智能体学习与测评</description><generator>Hugo</generator><language>zh-cn</language><copyright>本站内容采用 CC BY-NC-SA 4.0 国际许可协议授权</copyright><lastBuildDate>Sun, 12 Jul 2026 17:50:00 +0800</lastBuildDate><atom:link href="https://guijiagi.com/tags/%E9%95%BF%E5%BA%8F%E5%88%97/index.xml" rel="self" type="application/rss+xml"/><item><title>深度解析Transformer-XL：长序列建模的突破</title><link>https://guijiagi.com/posts/article-06/</link><pubDate>Sun, 12 Jul 2026 17:50:00 +0800</pubDate><guid>https://guijiagi.com/posts/article-06/</guid><description>从原理到实践，全面解析Transformer-XL如何突破标准Transformer的序列长度限制</description></item><item><title>深度解析Transformer-XL：长序列建模的突破</title><link>https://guijiagi.com/posts/b2-5c0c8534/</link><pubDate>Sun, 12 Jul 2026 17:50:00 +0800</pubDate><guid>https://guijiagi.com/posts/b2-5c0c8534/</guid><description>从原理到实践，全面解析Transformer-XL如何突破标准Transformer的序列长度限制</description></item><item><title>序列并行：超长序列训练</title><link>https://guijiagi.com/posts/sequence-parallelism-explained/</link><pubDate>Thu, 02 Jul 2026 11:07:00 +0800</pubDate><guid>https://guijiagi.com/posts/sequence-parallelism-explained/</guid><description>解析序列并行的核心原理、DeepSpeed Ulysses与Megatron SP的区别与联系</description></item><item><title>线性注意力研究2026</title><link>https://guijiagi.com/posts/linear-attention-research-2026/</link><pubDate>Thu, 02 Jul 2026 10:58:00 +0800</pubDate><guid>https://guijiagi.com/posts/linear-attention-research-2026/</guid><description>从核函数近似到状态空间模型，梳理线性注意力在2026年的最新进展</description></item></channel></rss>