What can LLMs tell us about the mechanisms behind polarity illusions in humans? Experiments across model scales and training steps

TL;DR

This study investigates how large language models exhibit polarity illusions, revealing scale-dependent effects and suggesting that such illusions may arise from shallow processing rather than rational inference.

Contribution

It demonstrates how polarity illusions manifest differently across model scales and proposes a construction grammar-based synthesis for understanding these phenomena in LLMs.

Findings

NPI illusion weakens and disappears in larger models

Depth charge illusion strengthens with larger models

Implications for human sentence processing theories

Abstract

I use the Pythia scaling suite (Biderman et al. 2023) to investigate if and how two well-known polarity illusions, the NPI illusion and the depth charge illusion, arise in LLMs. The NPI illusion becomes weaker and ultimately disappears as model size increases, while the depth charge illusion becomes stronger in larger models. The results have implications for human sentence processing: it may not be necessary to assume "rational inference" mechanisms that convert ill-formed sentences into well-formed ones to explain polarity illusions, given that LLMs cannot plausibly engage in this kind of reasoning, especially at the implicit level of next-token prediction. On the other hand, shallow, "good enough" processing and/or partial grammaticalization of prescriptively ungrammatical structures may both occur in LLMs. I propose a synthesis of different theoretical accounts that is rooted in the basic tenets of construction grammar.