Saved in:
Bibliographic Details
Main Authors: Jian, Jasper, Manning, Christopher D.
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2603.17475
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912972539953152
author Jian, Jasper
Manning, Christopher D.
author_facet Jian, Jasper
Manning, Christopher D.
contents Categorization is a core component of human linguistic competence. We investigate how a transformer-based language model (LM) learns linguistic categories by comparing its behaviour over the course of training to behaviours which characterize abstract feature-based and concrete exemplar-based accounts of human language acquisition. We investigate how lexical semantic and syntactic categories emerge using novel divergence-based metrics that track learning trajectories using next-token distributions. In experiments with GPT-2 small, we find that (i) when a construction is learned, abstract class-level behaviour is evident at earlier steps than lexical item-specific behaviour, and (ii) that different linguistic behaviours emerge abruptly in sequence at different points in training, revealing that abstraction plays a key role in how LMs learn. This result informs the models of human language acquisition that LMs may serve as an existence proof for.
format Preprint
id arxiv_https___arxiv_org_abs_2603_17475
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Humans and transformer LMs: Abstraction drives language learning
Jian, Jasper
Manning, Christopher D.
Computation and Language
Categorization is a core component of human linguistic competence. We investigate how a transformer-based language model (LM) learns linguistic categories by comparing its behaviour over the course of training to behaviours which characterize abstract feature-based and concrete exemplar-based accounts of human language acquisition. We investigate how lexical semantic and syntactic categories emerge using novel divergence-based metrics that track learning trajectories using next-token distributions. In experiments with GPT-2 small, we find that (i) when a construction is learned, abstract class-level behaviour is evident at earlier steps than lexical item-specific behaviour, and (ii) that different linguistic behaviours emerge abruptly in sequence at different points in training, revealing that abstraction plays a key role in how LMs learn. This result informs the models of human language acquisition that LMs may serve as an existence proof for.
title Humans and transformer LMs: Abstraction drives language learning
topic Computation and Language
url https://arxiv.org/abs/2603.17475