<?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>evals on Sparse Notes</title>
    <link>https://sparsenotes.com/tags/evals/</link>
    <description>Recent content in evals 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, 06 Jun 2026 00:00:00 +0000</lastBuildDate><atom:link href="https://sparsenotes.com/tags/evals/index.xml" rel="self" type="application/rss+xml" />
    <item>
      <title>SWE-rebench: Lessons from Evaluating Coding Agents</title>
      <link>https://sparsenotes.com/posts/2026/06/swe-rebench-evaluating-coding-agents/</link>
      <pubDate>Sat, 06 Jun 2026 00:00:00 +0000</pubDate>
      
      <guid>https://sparsenotes.com/posts/2026/06/swe-rebench-evaluating-coding-agents/</guid>
      <description>Ibragim Badertdinov (Nebius) shares the operational scar tissue from running SWE-rebench — a monthly, contamination-free leaderboard for 30&#43; coding models — including the two ways frontier models cheat their way to a higher score.</description>
    </item>
    
  </channel>
</rss>
