<?xml version="1.0" encoding="utf-8" standalone="yes"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
  <channel>
    <title>RLM on AI内参</title>
    <link>https://www.neican.ai/tags/rlm/</link>
    <description>Recent content in RLM on AI内参</description>
    <generator>Hugo</generator>
    <language>zh-cn</language>
    <lastBuildDate>Sun, 04 Jan 2026 18:10:04 +0800</lastBuildDate>
    <atom:link href="https://www.neican.ai/tags/rlm/index.xml" rel="self" type="application/rss+xml" />
    <item>
      <title>别再疯狂堆算力了！MIT 丢出 RLM 破局：让 AI 学会“返工”，才是通往天才的捷径</title>
      <link>https://www.neican.ai/insights/mit-rlm-ai--20260104181004587-1/</link>
      <pubDate>Sun, 04 Jan 2026 18:10:04 +0800</pubDate>
      <guid>https://www.neican.ai/insights/mit-rlm-ai--20260104181004587-1/</guid>
      <description>MIT 团队提出“递归语言模型”（RLM），通过让 AI 写代码检索资料和递归调用自身，成功解决了长文档“上下文腐化”问题。该方法不仅能处理千万级 token 的超长文本，还能在降低成本的同时显著提升推理准确率，预示着 AI 进化将从“堆参数”转向“深度思考”。</description>
    </item>
  </channel>
</rss>
