Language Models Generalize to Human-like Word Order Preferences

Summary

A central question in language acquisition is whether linguistic biases can emerge from general learning mechanisms operating over underdetermined input.