Tag
This paper investigates syntactic convergence in instruction-tuned large language models, finding that they reuse human syntax more than humans themselves do in dialogue contexts, with instruction-tuning increasing reliance on syntactic patterns.