<?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>Long Context on AI Tech Blog</title>
    <link>https://jesamkim.github.io/ai-tech-blog/tags/long-context/</link>
    <description>Recent content in Long Context on AI Tech Blog</description>
    <generator>Hugo -- 0.147.6</generator>
    <language>ko</language>
    <lastBuildDate>Sat, 25 Jul 2026 10:00:00 +0900</lastBuildDate>
    <atom:link href="https://jesamkim.github.io/ai-tech-blog/tags/long-context/index.xml" rel="self" type="application/rss+xml" />
    <item>
      <title>Inkling 해부: 975B MoE 오픈웨이트가 노리는 &#39;기업 커스터마이징&#39;</title>
      <link>https://jesamkim.github.io/ai-tech-blog/posts/2026-07-25-inkling-975b-moe-open-weights-anatomy/</link>
      <pubDate>Sat, 25 Jul 2026 10:00:00 +0900</pubDate>
      <guid>https://jesamkim.github.io/ai-tech-blog/posts/2026-07-25-inkling-975b-moe-open-weights-anatomy/</guid>
      <description>Thinking Machines Lab의 첫 오픈웨이트 모델 Inkling(975B MoE)의 아키텍처 선택을 분해합니다. sliding-window 55/66, learned relative-position bias, kernel-4 convolution이 왜 &amp;#39;벤치마크 1등&amp;#39;보다 &amp;#39;파인튜닝 베이스&amp;#39;를 겨냥하는지 살펴봅니다.</description>
    </item>
  </channel>
</rss>
