How Abstract Is Linguistic Generalization in Large Language Models? Experiments with Argument Structure
Top Cited Papers
Open Access
- 13 November 2023
- journal article
- research article
- Published by MIT Press in Transactions of the Association for Computational Linguistics
- Vol. 11, 1377-1395
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.1Keywords
All Related Versions
This publication has 29 references indexed in Scilit:
- What Does BERT Learn about the Structure of Language?2019
- Open Sesame: Getting inside BERT’s Linguistic Knowledge2019
- Targeted Syntactic Evaluation of Language Models2018
- What do RNN Language Models Learn about Filler–Gap Dependencies?2018
- Assessing the Ability of LSTMs to Learn Syntax-Sensitive DependenciesTransactions of the Association for Computational Linguistics, 2016
- Moving beyond Kučera and Francis: A critical evaluation of current word frequency norms and the introduction of a new and improved word frequency measure for American EnglishBehavior Research Methods, 2009
- Hard WordsLanguage Learning and Development, 2005
- Catastrophic forgetting in connectionist networksTrends in Cognitive Sciences, 1999
- The Structural Sources of Verb MeaningsLanguage Acquisition, 1990
- Catastrophic Interference in Connectionist Networks: The Sequential Learning Problem1989