Definition
Multilingual evaluation separates language-specific performance from global averages. It checks tone, formatting, market convention, policy wording, and task completion in each language.
AI Agent term
Multilingual Evaluation
Testing agents across the languages and markets where they will actually be used.
Why it matters
English performance does not guarantee Chinese, Japanese, Spanish, or other local-market readiness.
Example
An agent can rank high overall but lose in Japanese support because its tone is unnatural.