LILT launches AURORA multilingual agent leaderboard

LILT launched AURORA, a multilingual AI leaderboard covering 4 enterprise agent benchmarks across non-English tasks. LILT’s announcement says the suite tests coding, customer support, long-context instruction following, and agentic reasoning with native-language experts; its early results name different leaders in Spanish, Japanese, and Serbian. Teams shipping customer-facing agents should add locale-specific acceptance tests before trusting an English leaderboard; the archive’s OSWorld baseline analysis shows why evaluation conditions matter as much as the headline score.