displayName
FLORES-200 devtest
benchmarkId
caseCount
1012
releasedAt
2022-07-06
composition
The FLORES-200 devtest split: 1,012 sentences professionally
translated from English into 200 languages (and pivots), used
as the canonical held-out evaluation set for many-to-many
machine-translation systems.
homepageUrl
https://github.com/facebookresearch/flores
description
The dev/devtest splits are the canonical held-out evaluation
sets for FLORES-200; vendors typically report spBLEU / chrF on
devtest.