<?xml version="1.0" encoding="utf-8" standalone="yes"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
  <channel>
    <title>LLM效率 on AI内参</title>
    <link>https://www.neican.ai/tags/llm%E6%95%88%E7%8E%87/</link>
    <description>Recent content in LLM效率 on AI内参</description>
    <generator>Hugo</generator>
    <language>zh-cn</language>
    <lastBuildDate>Tue, 14 Oct 2025 21:10:13 +0800</lastBuildDate>
    <atom:link href="https://www.neican.ai/tags/llm%E6%95%88%E7%8E%87/index.xml" rel="self" type="application/rss+xml" />
    <item>
      <title>AI的元认知跃迁：Meta&#39;s“思维缓存”如何重塑大模型经济与智能边界</title>
      <link>https://www.neican.ai/insights/aimetas-20251014211013343-0/</link>
      <pubDate>Tue, 14 Oct 2025 21:10:13 +0800</pubDate>
      <guid>https://www.neican.ai/insights/aimetas-20251014211013343-0/</guid>
      <description>Meta AI通过“元认知复用”机制，使大模型能够学习并缓存重复推理模式，显著提升了推理效率，最高可将token消耗降低46%，同时保持准确率。这项创新不仅预示着AI运营成本的革命性降低和应用场景的拓展，更标志着大模型在实现类人经验式学习和元认知能力上的关键突破，深刻影响着AI的商业化前景与智能进化路径。</description>
    </item>
  </channel>
</rss>
