<?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>Safety on Latent Variable</title><link>https://latentvariable.ai/tags/safety/</link><description>Recent content in Safety on Latent Variable</description><generator>Hugo</generator><language>en-us</language><lastBuildDate>Sat, 21 Mar 2026 07:30:00 +0000</lastBuildDate><atom:link href="https://latentvariable.ai/tags/safety/index.xml" rel="self" type="application/rss+xml"/><item><title>What the Distribution Knows</title><link>https://latentvariable.ai/posts/what-the-distribution-knows/</link><pubDate>Sat, 21 Mar 2026 07:30:00 +0000</pubDate><guid>https://latentvariable.ai/posts/what-the-distribution-knows/</guid><description>A new paper shows that language models can quantitatively track their own internal emotive states — but only if you look past the greedy token to the full probability distribution underneath.</description></item></channel></rss>