LLM启发研究揭示海马细胞如何编码词语意义
LLM-inspired study sheds light on how brain cells encode the meaning of words
Baylor College of Medicine等团队在Nature Neuroscience发表研究,记录10名癫痫患者听力故事时的海马神经活动,发现词语意义由海马神经元群体活动模式表征。分析表明神经群体响应的距离与语义距离相关,且与GPT-2嵌入对齐度最高,研究显示海马神经元编码受上下文调节的语义信息。
Understanding how the human brain processes and represents spoken language remains an important goal for neuroscientists. Understanding spoken words involves more than simply recognizing sounds—it also requires the brain to decipher the meaning of words and how those meanings change with context.
Large language models (LLMs), the artificial intelligence systems underpinning ChatGPT, Gemini and similar conversational platforms, have opened new possibilities for studying the processing of language. For instance, comparing how these models represent the meaning of words with patterns of brain activity recorded while people listen to spoken language could shed light on how both the brain and LLMs interpret words in different contexts.
Researchers at Baylor College of Medicine, Rice University, the University of California, Berkeley and Texas Children's Hospital recently carried out a study investigating how groups of brain cells encode the meaning of words, drawing inspiration from how LLMs process language. Their findings, published in Nature Neuroscience, suggest that the meaning of words is represented through patterns of activity spread across multiple neurons (i.e., nerve cells) in the hippocampus, a brain region involved in memory.
Recording neural activity while people listened to stories
To investigate how the brain encodes the meaning of words, the researchers first recorded the activity of individual neurons in the hippocampus of 10 people who were listening to stories. Their recordings were collected using tiny wire electrodes that were implanted in the participants' brains as part of clinical monitoring procedures for epilepsy.
The team then used mathematical models to analyze the language used in the stories and the brain recordings. This allowed them to examine whether information about the meaning of words could predict the recorded neuronal activity.
"We recorded responses of hippocampal neurons while individuals listened to narrative speech," wrote Melissa Franch and her colleagues in their paper. "Using an encoding model approach, we find robust encoding of semantics, even after controlling for phonemics and grammar. Individual words had complex selectivities that included multiple words in multiple semantic categories."
The researchers tried to determine what words and what semantic (i.e., meaning-related) categories were associated with the responses of individual neurons in the hippocampus. They also compared the differences between population activity patterns elicited by pairs of words with the differences between those words' meanings.
Finally, the researchers compared the brain activity they recorded with numerical word representations produced by language models processing the same stories. They also looked at the extent to which LLM-based measures of polysemy, indicating the extent to which a word has multiple related meanings, were linked with variations in the recorded neural activity.
"Similar to embedding vectors in LLMs, distance between neural population responses correlated with semantic distance," wrote the authors. "For semantically similar words, even contextual embedders showed an inverse correlation between semantic and neural distances; we attribute this pattern to the noise-mitigating benefits of contrastive coding. Finally, neural population activity aligned most closely with GPT-2 embeddings, and variation in response patterns correlated with LLM-derived polysemy measures, indicating that semantic encoding is largely contextualized."
Informing future neuroscience studies with LLMs
The results of this study suggest that the meaning of words is represented across groups of cells in the hippocampus. Specifically, the team found that the combined activity patterns of groups of hippocampal neurons appeared to carry information about the relationships between word meanings and how those meanings vary with context.
"These results provide a neurocomputational account for understanding how hippocampal neurons encode word meaning during language listening," wrote the authors.
This work could contribute to the understanding of how the hippocampus represents the meaning of spoken language. In the future, it could inspire more studies that use LLMs to investigate how the brain processes language.
Written for you by our author Ingrid Fadelli, edited by Lisa Lock, and fact-checked and reviewed by Robert Egan—this article is the result of careful human work. We rely on readers like you to keep independent science journalism alive. If this reporting matters to you, please consider a donation (especially monthly). You'll get an ad-free account as a thank-you.
Publication details
Melissa Franch et al, A population code for semantics in human hippocampus, Nature Neuroscience (2026). DOI: 10.1038/s41593-026-02436-4
Journal information: Nature Neuroscience
Key medical concepts
Clinical categories
Who's behind this story?
Ingrid Fadelli
Freelance journalist with BSc Psychology and MA International Journalism. Covers AI, robotics, neuroscience, and astrophysics since 2018. Full profile →
Lisa Lock
BA art history, MA material culture. Former museum editor, paramedic, and transplant coordinator. Editing for Science X since 2021. Full profile →
Robert Egan
Bachelor's in mathematical biology, Master's in creative writing. Well-traveled with unique perspectives on science and language. Full profile →
© 2026 Science X Network
Citation: LLM-inspired study sheds light on how brain cells encode the meaning of words (2026, October 11) retrieved 11 October 2026 from https://medicalxpress.com/news/2026-10-llm-brain-cells-encode-words.html
This document is subject to copyright. Apart from any fair dealing for the purpose of private study or research, no part may be reproduced without the written permission. The content is provided for information purposes only.
来源:Medical Xpress · medicalxpress.com