<?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>weekly-papers-2026-07-11 on Sparse Notes</title>
    <link>https://sparsenotes.com/series/weekly-papers-2026-07-11/</link>
    <description>Recent content in weekly-papers-2026-07-11 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, 11 Jul 2026 00:00:00 +0000</lastBuildDate><atom:link href="https://sparsenotes.com/series/weekly-papers-2026-07-11/index.xml" rel="self" type="application/rss+xml" />
    <item>
      <title>Super Weights in LLMs and the Failure of Selective Training</title>
      <link>https://sparsenotes.com/posts/2026/07/papers/super-weights-2607.08733/</link>
      <pubDate>Sat, 11 Jul 2026 00:00:00 +0000</pubDate>
      
      <guid>https://sparsenotes.com/posts/2026/07/papers/super-weights-2607.08733/</guid>
      <description>Super Weights are outsized in importance but not in trainability: fine-tuning them in isolation collapses to random-guessing while vanilla LoRA succeeds with the same parameter budget.</description>
    </item>
    
    <item>
      <title>The Chubby Lock Service for Loosely-Coupled Distributed Systems</title>
      <link>https://sparsenotes.com/posts/2026/07/papers/chubby-lock-service-2006/</link>
      <pubDate>Sat, 11 Jul 2026 00:00:00 +0000</pubDate>
      
      <guid>https://sparsenotes.com/posts/2026/07/papers/chubby-lock-service-2006/</guid>
      <description>Mike Burrows&amp;#39; 2006 Chubby paper — how Google turned a Paxos-backed lock service into the coordination substrate underneath GFS, BigTable, and every service that came after.</description>
    </item>
    
    <item>
      <title>Who Needs DRAM? We Have Fiber</title>
      <link>https://sparsenotes.com/posts/2026/07/papers/fiber-memory-2607.08407/</link>
      <pubDate>Sat, 11 Jul 2026 00:00:00 +0000</pubDate>
      
      <guid>https://sparsenotes.com/posts/2026/07/papers/fiber-memory-2607.08407/</guid>
      <description>Fiber Memory: repurposing recirculating optical fiber as a shared, immutable delay-line store for LLM weights, cutting weight-delivery energy by 70% across 10,000 accelerators.</description>
    </item>
    
  </channel>
</rss>
