LLMs coerce but do not preempt, study finds

Arxiv CS.CL (NLP/language models)

Paper argues LLMs exhibit coercive productivity but lack preemption, a key mechanism in usage-based grammar. The study distinguishes frequency-driven entrenchment from preemption via statistical inference.

llm

Arxiv CS.CL (NLP/language models)·Claire BonialClaire Benet PostLaura MichaelisHarish Tayyar Madabushi

8 days ago