📄 中文摘要
神经科学揭示了我们社会本质的基本机制：在许多涉及互动的社会情境中，人类大脑活动会与他人同步。尽管大型语言模型（LLMs）被广泛接受为人类行为的强大近似，且多LLM系统的探索也在不断深入，但它们是否能与人类社会心智进行有意义的比较仍存在争议。研究通过实证证据探讨了社会互动LLMs之间的神经同步，提出在社会模拟中引入神经同步作为一种新颖的代理指标，以此为这一争论提供支持。该研究为理解语言模型在社会互动中的表现提供了新的视角，并为未来的研究奠定了基础。
社会互动语言模型之间的神经同步
出处: Neural Synchrony Between Socially Interacting Language Models
发布: 2026年2月23日
📄 English Summary
Neural Synchrony Between Socially Interacting Language Models
Neuroscience has revealed a fundamental mechanism of our social nature: human brain activity synchronizes with others in various social contexts involving interaction. While large language models (LLMs) are widely recognized as powerful approximations of human behavior, and multi-LLM systems are extensively explored to enhance their capabilities, the question of whether they can be meaningfully compared to human social minds remains contentious. This study investigates neural synchrony between socially interacting LLMs as empirical evidence for this debate. It introduces neural synchrony during social simulations as a novel proxy for analyzing the interactions, providing new insights into the performance of language models in social contexts and laying the groundwork for future research.