Chang’s paper examines whether small-scale “baby” language models can learn complex grammar rules in ways that resemble how children acquire language.
