<?xml version="1.0" encoding="utf-8"?><feed xmlns="http://www.w3.org/2005/Atom" ><generator uri="https://jekyllrb.com/" version="3.10.0">Jekyll</generator><link href="https://fvmassoli.github.io/feed.xml" rel="self" type="application/atom+xml" /><link href="https://fvmassoli.github.io/" rel="alternate" type="text/html" /><updated>2026-11-07T17:53:21+00:00</updated><id>https://fvmassoli.github.io/feed.xml</id><title type="html">Fabio Valerio Massoli</title><subtitle>Research scientist leading work on efficient reasoning for LLMs at Qualcomm AI Research: RL post-training, looped language models, and information theory.</subtitle><author><name>Fabio Valerio Massoli</name><email>fabiovaleriomassoli@gmail.com</email></author><entry><title type="html">Why looped language models?</title><link href="https://fvmassoli.github.io/blog/2026/11/hello-world/" rel="alternate" type="text/html" title="Why looped language models?" /><published>2026-11-01T00:00:00+00:00</published><updated>2026-11-01T00:00:00+00:00</updated><id>https://fvmassoli.github.io/blog/2026/11/hello-world</id><content type="html" xml:base="https://fvmassoli.github.io/blog/2026/11/hello-world/"><![CDATA[<p>This is an example post. Write in Markdown; math works when <code class="language-plaintext highlighter-rouge">math: true</code> is set in the front matter:</p>

\[I(Y; P \mid X) = \mathbb{E}_{x,p}\left[ D_{\mathrm{KL}}\big(p(Y \mid x, p)\,\|\,p(Y \mid x)\big) \right]\]

<p>Delete this file once you publish your first real post.</p>]]></content><author><name>Fabio Valerio Massoli</name><email>fabiovaleriomassoli@gmail.com</email></author><category term="looped LMs" /><summary type="html"><![CDATA[A short note on recurrent depth and test-time compute.]]></summary></entry></feed>