TruthfulQA-multi is an automatically translated extension of the original TruthfulQA benchmark designed to evaluate truthfulness in Basque, Catalan, Galician, and Spanish. The dataset enables evaluating the ability of Large Language Models (LLMs) to maintain truthfulness across multiple languages.
