<?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>latent-reasoning on Sparse Notes</title>
    <link>https://sparsenotes.com/tags/latent-reasoning/</link>
    <description>Recent content in latent-reasoning on Sparse Notes</description>
    <image>
      <url>https://sparsenotes.com/images/og-default.png</url>
      <link>https://sparsenotes.com/images/og-default.png</link>
    </image>
    <generator>Hugo -- gohugo.io</generator>
    <lastBuildDate>Sat, 30 May 2026 00:00:00 +0000</lastBuildDate><atom:link href="https://sparsenotes.com/tags/latent-reasoning/index.xml" rel="self" type="application/rss+xml" />
    <item>
      <title>Reasoning in Memory: Latent Reasoning Without Autoregressive Thoughts</title>
      <link>https://sparsenotes.com/posts/2026/05/papers/rim-working-memory/</link>
      <pubDate>Sat, 30 May 2026 00:00:00 +0000</pubDate>
      
      <guid>https://sparsenotes.com/posts/2026/05/papers/rim-working-memory/</guid>
      <description>Aichberger and Hochreiter propose RiM, a latent reasoning method that replaces autoregressively generated chain-of-thought tokens with fixed memory blocks processed in a single forward pass — matching or exceeding existing latent reasoning methods while decoupling internal computation from external communication.</description>
    </item>
    
  </channel>
</rss>
