analiza aiqbench
This research investigates how large language models handle linguistic productivity by testing whether they utilize human-like strategies of coercion and preemption. The study finds that while larger models successfully adapt to novel words through semantic adjustment (coercion), they fail to employ negative evidence to avoid plausible but unseen structures (preemption). This asymmetry highlights a fundamental difference between model behavior and human grammar acquisition, specifically regarding the absence of negative evidence in current architectures.
Linguistic Productivity in Large Language Models: Models Coerce, but do not Preempt
arXiv:2606.02953v2 Announce Type: replace Abstract: Usage-based theories of grammar explain language users' ability to interpret and produce novel combinations of words through two complementary inferential strategies. In cases of coercion, a language user, when encountering a word that is mismatched to its grammatical context, reinterprets that word in a way that resolves the conflict. In cases of preemption, the user infers that an otherwise plausible structure is unavailable or dispreferred because it fails to occur in contexts where it would be expected. Large Language Models are also usage-based, in the sense that the structures of language are learned through exposure to vast amounts of text. Here, we test whether or not the opposing forces of coercion and preemption also encourage and constrain linguistic productivity in LLMs. We demonstrate across models of various sizes and architectures that larger models recognize constructional productivity in cases of coercion, even when these involve novel words. However, we also show that even the largest models do not replicate the conservative language strategy of preemption for novel words: in particular, they do not appear to draw on negative evidence when making choices among competing grammatical options. We conclude that coercion is detectable because it is both surprising and semantically motivated: when a word is mismatched to its syntactic context, the word's meaning can be adjusted to fit the context. Preemption has no such signature. There is no record of the lexeme-construction pairings that were avoided--only of those that were produced. This asymmetry helps to explain why models coerce (create mismatched sign combinations) but do not preempt (avoid unconventional yet semantically defensible ones).
Abstract:Usage-based theories of grammar explain language users' ability to interpret and produce novel combinations of words through two complementary inferential strategies. In cases of coercion, a language user, when encountering a word that is mismatched to its grammatical context, reinterprets that word in a way that resolves the conflict. In cases of preemption, the user infers that an otherwise plausible structure is unavailable or dispreferred because it fails to occur in contexts where it would be expected. Large Language Models are also usage-based, in the sense that the structures of language are learned through exposure to vast amounts of text. Here, we test whether or not the opposing forces of coercion and preemption also encourage and constrain linguistic productivity in LLMs. We demonstrate across models of various sizes and architectures that larger models recognize constructional productivity in cases of coercion, even when these involve novel words. However, we also show that even the largest models do not replicate the conservative language strategy of preemption for novel words: in particular, they do not appear to draw on negative evidence when making choices among competing grammatical options. We conclude that coercion is detectable because it is both surprising and semantically motivated: when a word is mismatched to its syntactic context, the word's meaning can be adjusted to fit the context. Preemption has no such signature. There is no record of the lexeme-construction pairings that were avoided--only of those that were produced. This asymmetry helps to explain why models coerce (create mismatched sign combinations) but do not preempt (avoid unconventional yet semantically defensible ones).
Computation and Language (cs.CL)
arXiv:2606.02953 [cs.CL]
(or arXiv:2606.02953v2 [cs.CL] for this version)
https://doi.org/10.48550/arXiv.2606.02953 Focus to learn more arXiv-issued DOI via DataCite
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Text integral sub licența deschisă a publicației · 354 cuvinte
Subiecte conexe
Examining form-based linguistic productivity in people and large language models
arXiv · 27 oct. 2026
Linguistic Productivity in Large Language Models: Models Coerce, but do not Preempt
arXiv · 3 iun. 2026
How Human-Like Are Large Language Models? A Register-Aware Linguistic Evaluation Framework
arXiv · 1 sept. 2026
DecompSR: A dataset for decomposed analyses of compositional multihop spatial reasoning
arXiv · 12 iun. 2026