<?xml version="1.0" encoding="UTF-8"?><rss version="2.0"><channel><title>Harishchandra Pal · notes tagged “llms”</title><description>Notes tagged llms.</description><link>https://harishchandra.net/</link><language>en</language><item><title>Why token count for Hindi and English differ</title><link>https://harishchandra.net/writing/why-hindi-costs-more-tokens-than-english/</link><guid isPermaLink="true">https://harishchandra.net/writing/why-hindi-costs-more-tokens-than-english/</guid><description>Hindi used to cost six times more tokens than English for the same sentence. Then it stopped. Both the problem and the fix come out of how BPE builds its vocabulary.</description><pubDate>Sat, 05 Sep 2026 00:00:00 GMT</pubDate><category>llms</category><category>tokenization</category><category>hindi</category></item></channel></rss>