<?xml version="1.0" encoding="utf-8" standalone="yes"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
  <channel>
    <title>Gcg on Tom Ganz</title>
    <link>https://thecout.com/tags/gcg/</link>
    <description>Recent content in Gcg on Tom Ganz</description>
    <generator>Hugo -- gohugo.io</generator>
    <language>en-us</language>
    <lastBuildDate>Sun, 04 Oct 2026 21:40:39 +0100</lastBuildDate>
    <atom:link href="https://thecout.com/tags/gcg/index.xml" rel="self" type="application/rss+xml" />
    <item>
      <title>Prompt injection classifiers are terrible security boundaries</title>
      <link>https://thecout.com/blog/injection/</link>
      <pubDate>Sun, 04 Oct 2026 10:00:00 +0100</pubDate>
      <guid>https://thecout.com/blog/injection/</guid>
      <description>Everyone&amp;rsquo;s advice for prompt injection is the same: stick a classifier in front of the model. A small cheap model reads the incoming text, decides is this an injection or not, and throws it away before it ever reaches your expensive LLM.&#xA;So using the recent trend of Laya/Jev models, a recent small non-autoregressive model that just scores an answer in one forward pass instead of generating tokens, I&amp;rsquo;ve built a pretty good prompt injection classifier using my smaller, better calibrated version of Laya called Smolaya.</description>
    </item>
  </channel>
</rss>
