Testing embedding models on a hand-labelled journal

Automated excerpt

Every speed benchmark I had would stay green while search quietly returned the wrong entries. The right entry was usually somewhere in the top five, but the top hit was wrong in 8 of 15 questions. Only the model changed. 18 entries, 15 questions, top 5.

Selected automatically from source text; not independently written or fact-checked. Read the original for full context.