This dataset is converted from CohereLabs/Global-MMLU-Lite into chat messages format for compatibility with causal LM evaluation pipelines.
Original dataset: CohereLabs/Global-MMLU-Lite
Paper: Global MMLU: Understanding and Addressing Cultural and Linguistic Biases in Multilingual Evaluation
messages: Chat format with user prompt (question + options A–D) and assistant answer (A/B/C/D)… See the full description on the dataset page:
https://huggingface.co/datasets/locailabs/gmmlu_lite.