<?xml version="1.0" encoding="UTF-8"?><rss version="2.0"><channel><title>Noah Golmant</title><description>Personal site and blog</description><link>https://noahgolmant.com/</link><item><title>The local shape of LLM stable regions</title><link>https://noahgolmant.com/blog/stable-regions-residual-stream/</link><guid isPermaLink="true">https://noahgolmant.com/blog/stable-regions-residual-stream/</guid><description>Most of the time at inference, the Fisher pullback through the unembedding tells you exactly how far you can perturb the residual stream before the predictive distribution moves — and it&apos;s literally the second-order Taylor expansion of KL.</description><pubDate>Mon, 18 May 2026 00:00:00 GMT</pubDate></item><item><title>Revisiting pytorch-hessian-eigenthings, eight years later</title><link>https://noahgolmant.com/blog/hessian-eigenthings-v1/</link><guid isPermaLink="true">https://noahgolmant.com/blog/hessian-eigenthings-v1/</guid><description>A v1.0 rewrite of my old curvature analysis library: new operators, new algorithms, and a lot of numerical validation I should have written the first time.</description><pubDate>Thu, 14 May 2026 00:00:00 GMT</pubDate></item><item><title>Dynamical systems and an abstract view of autoregressive transformers</title><link>https://noahgolmant.com/blog/diving-into-attention/</link><guid isPermaLink="true">https://noahgolmant.com/blog/diving-into-attention/</guid><description>Understanding how attention can express logical propositions about correlations between tokens and other fun tips and tricks</description><pubDate>Tue, 08 Dec 2020 00:00:00 GMT</pubDate></item><item><title>Autoregressive transformers and lessons from enactivism</title><link>https://noahgolmant.com/blog/attention-please-intro/</link><guid isPermaLink="true">https://noahgolmant.com/blog/attention-please-intro/</guid><description>Understanding how attention can express logical propositions about correlations between tokens and other fun tips and tricks</description><pubDate>Sun, 06 Dec 2020 00:00:00 GMT</pubDate></item><item><title>Fine-tuned noise</title><link>https://noahgolmant.com/blog/sgd-noise/</link><guid isPermaLink="true">https://noahgolmant.com/blog/sgd-noise/</guid><description>Studying the covariance structure of mini-batch noise in stochastic gradient descent</description><pubDate>Wed, 18 Apr 2018 00:00:00 GMT</pubDate></item><item><title>Meta-learning and optimization</title><link>https://noahgolmant.com/blog/maml/</link><guid isPermaLink="true">https://noahgolmant.com/blog/maml/</guid><description>MAML Hunting</description><pubDate>Mon, 19 Feb 2018 00:00:00 GMT</pubDate></item><item><title>First-order methods almost always avoid saddle points</title><link>https://noahgolmant.com/blog/avoiding-saddle-points/</link><guid isPermaLink="true">https://noahgolmant.com/blog/avoiding-saddle-points/</guid><description>An application of dynamical systems theory to optimization</description><pubDate>Fri, 17 Nov 2017 00:00:00 GMT</pubDate></item></channel></rss>