INCLUDE-2.0: Region-Aware Multilingual LLM Benchmark

INCLUDE-2.0 is an extensive multilingual benchmark for region-aware factual evaluation across global languages and cultures.

• Publication: EMNLP 2026 Main Conference (“INCLUDE-2.0: Disentangling Knowledge Gaps in Multilingual LLM Evaluation”).
• Scale: Covers 91 languages from 86 countries (127 total language–country pairs) with over 220K curated samples drawn from native regional sources.
• Controlled Evaluation: Annotates regional knowledge types and pairs items with English translations, allowing researchers to decouple language fluency from knowledge of locally embedded facts.

• Resources: Hugging Face Dataset

  OpenReview