<?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>AI/ML on 3rd layer</title><link>https://3rdlayer.uk/ja/categories/ai/ml/</link><description>Recent content in AI/ML on 3rd layer</description><generator>Hugo -- 0.154.5</generator><language>ja-JP</language><lastBuildDate>Sun, 16 Feb 2025 00:00:00 +0000</lastBuildDate><atom:link href="https://3rdlayer.uk/ja/categories/ai/ml/index.xml" rel="self" type="application/rss+xml"/><item><title>Sparse AutoencoderでGPT-2の感情を操る</title><link>https://3rdlayer.uk/ja/posts/sae-steering-gpt2/</link><pubDate>Sun, 16 Feb 2025 00:00:00 +0000</pubDate><guid>https://3rdlayer.uk/ja/posts/sae-steering-gpt2/</guid><description>OpenAIのPretrained SAEでGPT-2内部の感情Featureを発見し、SAEをゼロから学習するまで。Feature Patchingで&amp;rsquo;good person&amp;rsquo;を&amp;rsquo;shit&amp;rsquo;に変える実験。</description></item></channel></rss>