A tokenization algorithm that iteratively merges the most frequent adjacent tokens to build a vocabulary.