<?xml version="1.0" encoding="utf-8" standalone="yes"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
  <channel>
    <title>ML Evaluation on Programmer.ie: Modern AI programming</title>
    <link>http://programmer.ie/tags/ml-evaluation/</link>
    <description>Recent content in ML Evaluation on Programmer.ie: Modern AI programming</description>
    <generator>Hugo</generator>
    <language>en-us</language>
    <lastBuildDate>Wed, 05 Aug 2026 23:24:45 +0100</lastBuildDate>
    <atom:link href="http://programmer.ie/tags/ml-evaluation/index.xml" rel="self" type="application/rss+xml" />
    <item>
      <title>RELATE: Searching Embeddings by Relation, Not Just Similarity</title>
      <link>http://programmer.ie/post/relate/</link>
      <pubDate>Wed, 05 Aug 2026 23:24:45 +0100</pubDate>
      <guid>http://programmer.ie/post/relate/</guid>
      <description>&lt;p&gt;Embeddings are everywhere in modern AI.&lt;/p&gt;&#xA;&lt;p&gt;They power semantic search, retrieval-augmented generation, recommendations, clustering, duplicate detection, code search, memory systems, and many of the mechanisms through which an AI system decides what information is relevant.&lt;/p&gt;&#xA;&lt;p&gt;Yet most systems interrogate embeddings in essentially the same way:&lt;/p&gt;&#xA;&lt;blockquote&gt;&#xA;&lt;p&gt;Take two vectors and calculate cosine similarity.&lt;/p&gt;&lt;/blockquote&gt;&#xA;&lt;p&gt;That is useful. But it also makes a strong assumption.&lt;/p&gt;&#xA;&lt;p&gt;It assumes that the information we care about is expressed directly through the default geometry of the embedding space.&lt;/p&gt;</description>
    </item>
  </channel>
</rss>
