Linear representations of grammaticality in neural language models
Computation and Language
Summary
The authors studied whether neural language models (NLMs) can tell if a sentence is grammatically correct, not just based on how likely they think the sentence is. Instead of using probabilities, they looked at the internal representations inside the models to see if grammatical and ungrammatical sentences are separated in the model’s 'thinking space.' They found that grammaticality is clearly encoded in these representations, beyond other factors like word frequency or meaning. This pattern holds across different grammar rules and even somewhat across languages, indicating that NLMs have a distinct way of representing grammar knowledge.
Authors
Jane Li, Najoung Kim
Abstract
Whether neural language models (NLMs) possess the ability to distinguish strings on the basis of their grammaticality remains a debated topic in the computational linguistics literature. Existing evidence has largely relied on probability-based measures, testing whether models assign higher probabilities to grammatical than ungrammatical strings. However, probability comparisons have been criticized as a measure for grammatical knowledge based on the assumption that grammaticality is inherently entangled with likelihood. Model-assigned probability is a function of many related sentence properties, such as lexical frequency, plausibility, and world knowledge. In this work, we move beyond probability-based evaluations and investigate whether grammaticality is encoded in the internal representations of NLMs. Using mass-mean probing, we test whether grammatical and ungrammatical sentences are systematically separated in representational space. We further examine the extent to which these representations are independent of sentence properties that are correlated with grammaticality, as well as their generalization across grammatical phenomena and languages. Our results provide evidence that grammaticality is robustly encoded in sentence representations of a wide range of pretrained NLMs, yielding clear representational separation on the dimension of grammaticality that cannot be fully explained by alternative sentence-level factors. Moreover, this encoding generalizes across a broad range of grammatical phenomena and to some degree, across languages, suggesting that grammaticality constitutes a coherent representational dimension in contemporary NLMs. These findings contribute new evidence to debates about the nature of syntactic knowledge in language models and offer a complementary framework for evaluating grammatical competence that is not dependent on string probabilities alone.