Like human brains, large language models reason about diverse data in a general way

A new study shows LLMs represent different data types based on their underlying meaning and reason about data in their dominant language.