Large Language Models as Neurolinguistic Subjects: Discrepancy between Performance and Competence

TL;DR

This paper explores the discrepancy between LLMs' linguistic performance and underlying competence by introducing a neurolinguistic assessment approach, revealing that models excel more in form than in meaning, with implications for evaluating language understanding.

Contribution

It introduces a neurolinguistic evaluation method combining minimal pair and diagnostic probing, and provides new multilingual datasets to better assess LLMs' linguistic competence.

Findings

LLMs show higher competence in form than in meaning.

Psycholinguistic and neurolinguistic assessments reveal performance-competence discrepancy.

Instruction tuning improves performance but not underlying competence.

Abstract

This study investigates the linguistic understanding of Large Language Models (LLMs) regarding signifier (form) and signified (meaning) by distinguishing two LLM assessment paradigms: psycholinguistic and neurolinguistic. Traditional psycholinguistic evaluations often reflect statistical rules that may not accurately represent LLMs' true linguistic competence. We introduce a neurolinguistic approach, utilizing a novel method that combines minimal pair and diagnostic probing to analyze activation patterns across model layers. This method allows for a detailed examination of how LLMs represent form and meaning, and whether these representations are consistent across languages. We found: (1) Psycholinguistic and neurolinguistic methods reveal that language performance and competence are distinct; (2) Direct probability measurement may not accurately assess linguistic competence; (3) Instruction tuning won't change much competence but improve performance; (4) LLMs exhibit higher competence and performance in form compared to meaning. Additionally, we introduce new conceptual minimal pair datasets for Chinese (COMPS-ZH) and German (COMPS-DE), complementing existing English datasets.