The Structural Sources of Verb Meaning Revisited: Large Language Models Display Syntactic Bootstrapping
Xiaomeng Zhu R. Thomas McCoy Robert Frank · 2025
Read The Structural Sources of Verb Meaning Revisited: Large Language Models Display Syntactic Bootstrapping by Xiaomeng Zhu R. Thomas McCoy Robert Frank (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2508-12482.txt.
Description
Xiaomeng Zhu, Thomas McCoy, and Robert Frank ask in a 2025 paper whether language models learn verb meanings the way children may. One long held idea says children use the sentence frames around a verb to infer what the verb means. The authors test whether models lean on word order and grammar in the same way.
They trained two model types on altered data that removed either grammar cues or word pairing cues. Verb knowledge fell more when grammar cues were removed, and verbs for mental states suffered most. Noun knowledge, by contrast, fell more when word pairings were scrambled. The results back the grammar based account and show how model training can test ideas about child learning.
Catalog record
| CREATOR | Xiaomeng Zhu R. Thomas McCoy Robert Frank |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 39 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2508-12482