<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Llm-Agents on Scholion</title><link>https://scholion.thluiz.com/tags/llm-agents/</link><description>Recent content in Llm-Agents on Scholion</description><generator>Hugo -- gohugo.io</generator><language>pt-BR</language><copyright>© 2026</copyright><lastBuildDate>Wed, 11 Nov 2026 20:43:33 +0000</lastBuildDate><atom:link href="https://scholion.thluiz.com/tags/llm-agents/index.xml" rel="self" type="application/rss+xml"/><item><title>You don’t need long context</title><link>https://scholion.thluiz.com/notes/you-dont-need-long-context/</link><pubDate>Wed, 11 Nov 2026 20:11:20 +0000</pubDate><guid>https://scholion.thluiz.com/notes/you-dont-need-long-context/</guid><description>The paper argues that frontier models can keep contexts short without losing much accuracy on long-horizon and long-input tasks, and that selective reading plus compaction can replace long contexts.</description></item></channel></rss>