Understanding how sentences are internally represented in the human brain, as
well as in large language models (LLMs) such as ChatGPT, is a major challenge
for cognitive science. Classic linguistic theories propose that the brain
represents a sentence by parsing it into hierarchically organized constituents.
In contrast, LLMs do not explicitly parse linguistic constituents and their
latent representations remains poorly explained. Here, we demonstrate that
humans and LLMs construct similar latent representations of hierarchical
linguistic constituents by analyzing their behaviors during a novel one-shot
learning task, in which they infer which words should be deleted from a
sentence. Both humans and LLMs tend to delete a constituent, instead of a
nonconstituent word string. In contrast, a naive sequence processing model that
has access to word properties and ordinal positions does not show this
property. Based on the word deletion behaviors, we can reconstruct the latent
constituency tree representation of a sentence for both humans and LLMs. These
results demonstrate that a latent tree-structured constituency representation
can emerge in both the human brain and LLMs.

人脑与大型语言模型在句子内部表征方面具有相似的层级结构，通过删除单词来推断句子的潜在分层结构，并证明了这种结构在人脑和语言模型中都能出现。

人类和大型语言模型中潜在成分表示的主动运用

Active Use of Latent Constituency Representation in both Humans and  Large Language Models

Corpus-based grammar induction generally relies on hand-parsed training data
to learn the structure of the language. Unfortunately, the cost of building
large annotated corpora is prohibitively expensive. This work aims to improve
the induction strategy when there are few labels in the training data. We show
that the most informative linguistic constituents are the higher nodes in the
parse trees, typically denoting complex noun phrases and sentential clauses.
They account for only 20% of all constituents. For inducing grammars from
sparsely labeled training data (e.g., only higher-level constituent labels), we
propose an adaptation strategy, which produces grammars that parse almost as
well as grammars induced from fully labeled corpora. Our results suggest that
for a partial parser to replace human annotators, it must be able to
automatically extract higher-level constituents rather than base noun phrases.

本研究旨在提高在训练数据中标注少量标签的情况下诱导策略。我们发现最具信息量的语言成分是树状结构中更高层的节点，通常代表复杂的名词短语和句子从句，这些仅占所有成分的 20％。针对标注数据稀疏（例如仅有高层成分标签）的语法诱导，提出了一种适应策略，可以产生与从完全标注的语料库诱导的语法几乎一样好的语法。我们的结果表明，为了取代人工注释员，部分解析器必须能够自动提取更高层次的成分，而不是基本名词短语。