Saved in:
Bibliographic Details
Main Authors: Bowers, Jeffrey S., Mitchell, Jeff
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2511.11389
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866909902425817088
author Bowers, Jeffrey S.
Mitchell, Jeff
author_facet Bowers, Jeffrey S.
Mitchell, Jeff
contents According to Futrell and Mahowald [arXiv:2501.17047], both infants and language models (LMs) find attested languages easier to learn than impossible languages that have unnatural structures. We review the literature and show that LMs often learn attested and many impossible languages equally well. Difficult to learn impossible languages are simply more complex (or random). LMs are missing human inductive biases that support language acquisition.
format Preprint
id arxiv_https___arxiv_org_abs_2511_11389
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Studies with impossible languages falsify LMs as models of human language
Bowers, Jeffrey S.
Mitchell, Jeff
Computation and Language
According to Futrell and Mahowald [arXiv:2501.17047], both infants and language models (LMs) find attested languages easier to learn than impossible languages that have unnatural structures. We review the literature and show that LMs often learn attested and many impossible languages equally well. Difficult to learn impossible languages are simply more complex (or random). LMs are missing human inductive biases that support language acquisition.
title Studies with impossible languages falsify LMs as models of human language
topic Computation and Language
url https://arxiv.org/abs/2511.11389