**Large language models without grounding recover non-sensorimotor but not sensorimotor features of human concepts**

“_We found that learning constrained to the language domain captures human-level conceptual representation in non-sensorimotor dimensions such as valence and emotional arousal but yields impoverished representation of sensorimotor knowledge._”

Xu, Q., Peng, Y., Nastase, S.A. et al. Large language models without grounding recover non-sensorimotor but not sensorimotor features of human concepts. Nat Hum Behav (2025). https://doi.org/10.1038/s41562-025-02203-8.

#OpenAcces #OA #Article #DOI #Language #Linguistics #Science #Technology #Tech #AI #ArtificialIntelligence #LLMS #Psycholofy #Academia #Academics @ai @psychology