http://programmer.ie/tags/semantic-checksum/ Recent content in Semantic Checksum on Programmer.ie: Modern AI programming Hugo en-us Mon, 07 Sep 2026 15:00:00 +0000 http://programmer.ie/books/embeddings-from-first-principles/22-chapter/ Mon, 07 Sep 2026 15:00:00 +0000 http://programmer.ie/books/embeddings-from-first-principles/22-chapter/ <p><em>Part VII — What Survives Transformation</em></p> <h2 id="two-vectors-for-one-document">Two vectors for one document</h2> <div class="highlight"><pre tabindex="0" style="color:#f8f8f2;background-color:#272822;-moz-tab-size:4;-o-tab-size:4;tab-size:4;"><code class="language-text" data-lang="text"><span style="display:flex;"><span>full_doc → E(full_doc) one vector </span></span><span style="display:flex;"><span>summary(doc) → E(summary(doc)) one vector </span></span></code></pre></div><p>Both are single vectors of the same dimension. One was produced from 4,000 words; the other from 120. The question:</p> <blockquote> <p><strong>How much of the document&rsquo;s representational geometry survives when the document is compressed — and is the surviving-or-not a signal we can use?</strong></p></blockquote> <p>If <code>E(summary)</code> lands where <code>E(full_doc)</code> lands <em>and</em> behaves like it under retrieval and clustering, the summary is a faithful compression in the sense the embedding cares about. If it drifts, something was dropped — and <em>which</em> direction it drifted may tell us what.</p>