<?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>Knowledge-Transfer on 兜忆轩 | Cedric's Compiler Notes</title><link>https://douyixuan.github.io/tags/knowledge-transfer/</link><description>Recent content in Knowledge-Transfer on 兜忆轩 | Cedric's Compiler Notes</description><image><title>兜忆轩 | Cedric's Compiler Notes</title><url>https://douyixuan.github.io/images/papermod-cover.png</url><link>https://douyixuan.github.io/images/papermod-cover.png</link></image><generator>Hugo -- 0.146.0</generator><language>en-us</language><lastBuildDate>Tue, 20 Oct 2026 12:24:00 +0800</lastBuildDate><atom:link href="https://douyixuan.github.io/tags/knowledge-transfer/index.xml" rel="self" type="application/rss+xml"/><item><title>035 · 成功经验，为什么反而让 kernel 变慢？</title><link>https://douyixuan.github.io/posts/ai-silicon-papers/2026-10-20/</link><pubDate>Tue, 20 Oct 2026 12:24:00 +0800</pubDate><guid>https://douyixuan.github.io/posts/ai-silicon-papers/2026-10-20/</guid><description>MetaKernelBench 只传递文字经验，让同一模型在另一种 DSL 中重新解同一道 kernel 题。它提醒我们：源实现成功，不等于经验对接收者有益；经验也可能把一次搜索的局限带到下一次。</description></item></channel></rss>