<?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>PPA on 兜忆轩 | Cedric's Compiler Notes</title><link>https://douyixuan.github.io/tags/ppa/</link><description>Recent content in PPA 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>Mon, 05 Oct 2026 12:37:00 +0800</lastBuildDate><atom:link href="https://douyixuan.github.io/tags/ppa/index.xml" rel="self" type="application/rss+xml"/><item><title>020 · 在测试题上继续训练，还算测试吗？</title><link>https://douyixuan.github.io/posts/ai-silicon-papers/2026-10-05/</link><pubDate>Mon, 05 Oct 2026 12:37:00 +0800</pubDate><guid>https://douyixuan.github.io/posts/ai-silicon-papers/2026-10-05/</guid><description>Alpha-RTL 在每道 RTL 优化题上用 testbench 与 PPA 反馈在线更新 Qwen3-8B；结果显著优于冻结策略的 Agent，但也把评测对象从静态模型能力改成了带逐题训练预算的完整优化系统。</description></item></channel></rss>