<?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>Data-Leakage on The Standard Error</title><link>https://jongha-jeon-dev.github.io/standarderror/tags/data-leakage/</link><description>Recent content in Data-Leakage on The Standard Error</description><generator>Hugo -- 0.123.7</generator><language>en-gb</language><lastBuildDate>Wed, 14 Oct 2026 00:00:00 +0000</lastBuildDate><atom:link href="https://jongha-jeon-dev.github.io/standarderror/tags/data-leakage/index.xml" rel="self" type="application/rss+xml"/><item><title>ML 3: Not Every Leak Leaks, and the Ones That Do Can Be Measured</title><link>https://jongha-jeon-dev.github.io/standarderror/lectures/ml-3-leakage/</link><pubDate>Thu, 08 Oct 2026 00:00:00 +0000</pubDate><guid>https://jongha-jeon-dev.github.io/standarderror/lectures/ml-3-leakage/</guid><description>&amp;#39;Fit every preprocessing step inside the folds&amp;#39; is right, and taught as if every violation were equally bad. Measured, seven of them range from nothing to a third of the accuracy scale.</description></item><item><title>ML 52: A Leaked Benchmark Shows in the Set, Not in the Item</title><link>https://jongha-jeon-dev.github.io/standarderror/lectures/ml-52-contamination/</link><pubDate>Wed, 14 Oct 2026 00:00:00 +0000</pubDate><guid>https://jongha-jeon-dev.github.io/standarderror/lectures/ml-52-contamination/</guid><description>A model trained on a benchmark&amp;#39;s answers scored higher on them and no higher on anything else. Its own likelihoods could not say which problems had leaked; a fresh set from the same source could say that some had.</description></item></channel></rss>