Revealing emergent human-like conceptual representations from language prediction
Published in Proceedings of the National Academy of Sciences (PNAS), 2025
Recommended citation: Ningyu Xu, Qi Zhang, Chenyang Du, Qinan Luo, Xipeng Qiu, Xuanjing Huang, Menghan Zhang. Revealing emergent human-like conceptual representations from language prediction. Proceedings of the National Academy of Sciences (PNAS), 2025. https://doi.org/10.1073/pnas.2512514122
People acquire concepts through rich physical and social experiences and use them to understand and navigate the world. In contrast, large language models (LLMs), trained solely through next-token prediction on text, exhibit strikingly human-like behaviors. Are these models developing concepts akin to those in humans? This study investigates whether LLMs develop human-like conceptual representations through language prediction alone. We compare conceptual representations in LLMs with those in humans using behavioral and neural data. Our findings reveal that LLMs spontaneously develop conceptual structures that closely resemble human conceptual organization, suggesting that language prediction may be a powerful mechanism for acquiring human-like conceptual knowledge.