<?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>Long-Context on danilchenko.dev</title><link>https://danilchenko.dev/tags/long-context/</link><description>Recent content in Long-Context on danilchenko.dev</description><generator>Hugo</generator><language>en-us</language><lastBuildDate>Sat, 11 Apr 2026 06:00:00 +0000</lastBuildDate><atom:link href="https://danilchenko.dev/tags/long-context/index.xml" rel="self" type="application/rss+xml"/><item><title>TriAttention Compresses KV Cache 10.7x — How Trigonometry Fixed Long-Context Reasoning</title><link>https://danilchenko.dev/posts/2026-04-11-triattention-kv-cache-compression-long-reasoning/</link><pubDate>Sat, 11 Apr 2026 06:00:00 +0000</pubDate><guid>https://danilchenko.dev/posts/2026-04-11-triattention-kv-cache-compression-long-reasoning/</guid><description>TriAttention uses pre-RoPE vector concentration and trigonometric scoring to compress KV cache 10.7x while matching full attention accuracy on reasoning tasks.</description></item></channel></rss>