<?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>Benchmark on Kent Yao</title>
    <link>https://yaooqinn.github.io/zh/tags/benchmark/</link>
    <description>Recent content in Benchmark on Kent Yao</description>
    <generator>Hugo -- 0.157.0</generator>
    <language>zh-cn</language>
    <lastBuildDate>Wed, 27 May 2026 00:00:00 +0000</lastBuildDate>
    <atom:link href="https://yaooqinn.github.io/zh/tags/benchmark/index.xml" rel="self" type="application/rss+xml" />
    <item>
      <title>-46% 还是 -2%?规则改写器只在自己家有效</title>
      <link>https://yaooqinn.github.io/zh/posts/query-engines/rule-rewrite-blindspot-dsb/</link>
      <pubDate>Wed, 27 May 2026 00:00:00 +0000</pubDate>
      <guid>https://yaooqinn.github.io/zh/posts/query-engines/rule-rewrite-blindspot-dsb/</guid>
      <description>在 TPC-H 10GB 上,一个学界 SOTA 的规则改写器把平均执行时间从 69.84s 砍到 37.57s,-46%。换到 DSB 10GB,同一个改写器把 32.62s 砍到 31.93s——只有 -2.1%。差距不在 query 难度,而在 benchmark 是不是它的训练集。&amp;ldquo;规则系统稳定可靠&amp;quot;很多时候是基准过拟合,不是工程事实。</description>
    </item>
  </channel>
</rss>
