CLAIMar 10, 2025

Language Models Fail to Introspect About Their Knowledge of Language

arXiv:2503.07513v320 citationsh-index: 5Has Code
Originality Incremental advance
AI Analysis

This addresses the problem of interpretability and validation of linguistic methods in AI research, showing that prompted responses should not be conflated with models' knowledge, which is incremental as it builds on prior arguments.

The study investigated whether large language models (LLMs) can introspect about their internal linguistic knowledge, such as grammaticality, by comparing their prompted responses to direct string probability measurements across 21 models, and found no evidence of privileged self-access, indicating they cannot introspect.

There has been recent interest in whether large language models (LLMs) can introspect about their own internal states. Such abilities would make LLMs more interpretable, and also validate the use of standard introspective methods in linguistics to evaluate grammatical knowledge in models (e.g., asking "Is this sentence grammatical?"). We systematically investigate emergent introspection across 21 open-source LLMs, in two domains where introspection is of theoretical interest: grammatical knowledge and word prediction. Crucially, in both domains, a model's internal linguistic knowledge can be theoretically grounded in direct measurements of string probability. We then evaluate whether models' responses to metalinguistic prompts faithfully reflect their internal knowledge. We propose a new measure of introspection: the degree to which a model's prompted responses predict its own string probabilities, beyond what would be predicted by another model with nearly identical internal knowledge. While both metalinguistic prompting and probability comparisons lead to high task accuracy, we do not find evidence that LLMs have privileged "self-access". By using general tasks, controlling for model similarity, and evaluating a wide range of open-source models, we show that LLMs cannot introspect, and add new evidence to the argument that prompted responses should not be conflated with models' linguistic generalizations.

Code Implementations1 repo
Foundations

The foundational work for this paper's niche, ranked by how specifically the neighbourhood builds on it — not by global fame.

Your Notes