How Abstract Is Linguistic Generalization in Large Language Models? Experiments with Argument Structure

Michael Wilson, Jackson Petty, Robert Frank


Abstract
Language models are typically evaluated on their success at predicting the distribution of specific words in specific contexts. Yet linguistic knowledge also encodes relationships between contexts, allowing inferences between word distributions. We investigate the degree to which pre-trained transformer-based large language models (LLMs) represent such relationships, focusing on the domain of argument structure. We find that LLMs perform well in generalizing the distribution of a novel noun argument between related contexts that were seen during pre-training (e.g., the active object and passive subject of the verb spray), succeeding by making use of the semantically organized structure of the embedding space for word embeddings. However, LLMs fail at generalizations between related contexts that have not been observed during pre-training, but which instantiate more abstract, but well-attested structural generalizations (e.g., between the active object and passive subject of an arbitrary verb). Instead, in this case, LLMs show a bias to generalize based on linear order. This finding points to a limitation with current models and points to a reason for which their training is data-intensive.1
Anthology ID:
2023.tacl-1.78
Volume:
Transactions of the Association for Computational Linguistics, Volume 11
Month:
Year:
2023
Address:
Cambridge, MA
Venue:
TACL
SIG:
Publisher:
MIT Press
Note:
Pages:
1377–1395
Language:
URL:
https://aclanthology.org/2023.tacl-1.78
DOI:
10.1162/tacl_a_00608
Bibkey:
Cite (ACL):
Michael Wilson, Jackson Petty, and Robert Frank. 2023. How Abstract Is Linguistic Generalization in Large Language Models? Experiments with Argument Structure. Transactions of the Association for Computational Linguistics, 11:1377–1395.
Cite (Informal):
How Abstract Is Linguistic Generalization in Large Language Models? Experiments with Argument Structure (Wilson et al., TACL 2023)
Copy Citation:
PDF:
https://aclanthology.org/2023.tacl-1.78.pdf
Video:
 https://aclanthology.org/2023.tacl-1.78.mp4