XCommonsense-BN: A Bangla Commonsense Reasoning Dataset

It is organized in a tabular format, where each entry includes a unique identifier, the reasoning category, the question in Bangla, four answer options labeled A through D, the correct option, and a short explanation in Bangla justifying the correct answer. The dataset comprises over 1,000 entries spanning five categories: causal, temporal, social, physical, and intentional reasoning, with each category containing at least 150–200 questions to ensure balanced coverage.

The dataset is provided in Excel formats encoded in UTF-8 to support the Bangla script. Sample entries illustrate typical questions, answer options, correct labels, and explanations, providing a representative view of the dataset’s structure and content. This dataset enables the development, evaluation, and benchmarking of machine learning models in Bangla commonsense reasoning tasks and contributes to research in low-resource language NLP.