/

← All benchmarks

Global-MMLU

multilingual · knowledge↗ Quelle

49Evidence

not yet measured

What it measures

MMLU in 42 Sprachen, kulturell geprueft statt nur maschinell uebersetzt.

What it does not measure

Generieren in der Zielsprache.

How the score is composed

Currency50

Discrimination50

Headroom50

Contamination safety33

Gaming resistance55

Methodological rigor80

Reproducibility60

Details

Benchmaxxing risk: medium
Difficulty: medium
Public test set: yes
Human-verified: yes
LLM-judged: no
Skala: 0–100 % correct (höher besser)

Best measured models

No results ingested yet.