mteb-bitext-mining-metric-f1
IN premise — summaries/2026/08/24/muennighoff-2022-mteb-s3-the-mteb-benchmark.md
Created 2026-08-25T02:58:20+00:00
MTEB bitext mining evaluation uses F1 as the primary metric, with closest pairs found via cosine similarity between embedded language sets.
Summary
When measuring how well text embeddings can find matching sentences across different languages, performance is scored on F1, which weights catching correct pairs and avoiding false matches equally. This sets the specific yardstick the system uses to judge whether an embedding model is good at cross-language text alignment, and cosine similarity between the embedded vectors is what determines which pairs are considered closest candidates.