<?xml version="1.0" encoding="utf-8" standalone="yes"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
  <channel>
    <title>长上下文 on 知识铺的博客</title>
    <link>https://index.zshipu.com/ai001/tags/%E9%95%BF%E4%B8%8A%E4%B8%8B%E6%96%87/</link>
    <description>Recent content in 长上下文 on 知识铺的博客</description>
    <generator>Hugo -- gohugo.io</generator>
    <language>zh-CN</language>
    <lastBuildDate>Thu, 03 Sep 2026 00:00:00 +0000</lastBuildDate>
    <atom:link href="https://index.zshipu.com/ai001/tags/%E9%95%BF%E4%B8%8A%E4%B8%8B%E6%96%87/index.xml" rel="self" type="application/rss+xml" />
    <item>
      <title>Meta Muse Spark 1.3 用 21 倍差价直接买用户数据</title>
      <link>https://index.zshipu.com/ai001/post/20260903/Meta-Muse-Spark-1.3-%E7%94%A8-21-%E5%80%8D%E5%B7%AE%E4%BB%B7%E7%9B%B4%E6%8E%A5%E4%B9%B0%E7%94%A8%E6%88%B7%E6%95%B0%E6%8D%AE/</link>
      <pubDate>Thu, 03 Sep 2026 00:00:00 +0000</pubDate>
      <guid>https://index.zshipu.com/ai001/post/20260903/Meta-Muse-Spark-1.3-%E7%94%A8-21-%E5%80%8D%E5%B7%AE%E4%BB%B7%E7%9B%B4%E6%8E%A5%E4%B9%B0%E7%94%A8%E6%88%B7%E6%95%B0%E6%8D%AE/</guid>
      <description>Meta Muse Spark 1.3 把上下文窗口做到 100 万 token，同时推出两档定价，最高差价达到 21 倍。低价档要求用户提供数据，Meta 直接把用户生成的内容转化为模型训练资源。这一做法把长上下文模型的成本结构彻底暴露：不是单纯卖推理，而是用折扣买数据。 21 倍差价把用户数据直接标为训练燃料 Meta 为 Muse Spark 1.3 设置了两档定</description>
    </item>
  </channel>
</rss>
