Evaluating sensorimotor knowledge in large language models
eScholarship (California Digital Library) · 18.06.2025
DOI: 10.48448/qr5p-gb82 ↗Kurz gefasst
Fachartikel in der Zeitschrift eScholarship (California Digital Library), erschienen am 18.06.2025. Verfasst von Cognitive Science Society 2025 und 2 weiteren Autor:innen.
Abstract (Originalsprache)
Large Language Models (LLMs) process language but lack direct sensorimotor experience. This study assesses their ability to estimate human perceptual ratings using the Lancaster Sensorimotor Norms. We prompt LLMs with the original rating task instructions and analyze their correlations with human norms. Results suggest LLMs struggle with embodied cognition, highlighting limitations in computational models of sensory meaning. Future research should explore fine-tuning LLMs on sensorimotor data to enhance embodied representations.
Schlagwörter
Zitieren
APA
2025, C. S. S., Hollander, & John (2025). Evaluating sensorimotor knowledge in large language models. eScholarship (California Digital Library). https://doi.org/10.48448/qr5p-gb82
Vancouver
2025 CSS, Hollander, John. Evaluating sensorimotor knowledge in large language models. eScholarship (California Digital Library). 2025. doi:10.48448/qr5p-gb82
BibTeX
@article{2025,
author = {2025, Cognitive Science Society and Hollander and John},
title = {Evaluating sensorimotor knowledge in large language models},
journal = {eScholarship (California Digital Library)},
year = {2025},
doi = {10.48448/qr5p-gb82}
}
EUVIMED