<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Latent-Communication on skulldc</title><link>https://skulldc.net/tags/latent-communication/</link><description>Recent content in Latent-Communication on skulldc</description><generator>Hugo</generator><language>en-us</language><lastBuildDate>Sun, 08 Nov 2026 10:25:10 +0000</lastBuildDate><atom:link href="https://skulldc.net/tags/latent-communication/index.xml" rel="self" type="application/rss+xml"/><item><title>PLatour tests whether a small model really uses a larger model's prompt representations</title><link>https://skulldc.net/2026/11/platour-tests-whether-a-small-model-really-uses-a-larger-model-s-promp/</link><pubDate>Sun, 08 Nov 2026 10:25:10 +0000</pubDate><guid>https://skulldc.net/2026/11/platour-tests-whether-a-small-model-really-uses-a-larger-model-s-promp/</guid><description>A paper on prefill-to-autoregressive cross-model transfer releases an open framework and finds that some reported gains appear even when the sender&amp;rsquo;s signal carries no information.</description></item></channel></rss>