穷人版智能体建模：在笔记本电脑上模拟大型LLM智能体社会

摘要

模拟由众多大型语言模型（LLM）智能体组成的社会成本高昂，然而对此类模拟提出的问题通常是宏观层面的：相行为、典型化事实，以及随智能体数量 N 的标度行为，而非单个智能体的认知。我们将一个统计物理学观察转化为一种方法：用低参数模型替代每个 LLM 智能体，该模型从几百到几千次廉价查询中拟合得到，然后即可在笔记本电脑上以任意 N 运行该社会模拟。该方法是否有效，在模拟运行之前便已确定，主要取决于每个智能体的感知内容。我们引入了一种「交互阶数 × 记忆」分类法，将感知和记忆映射为一种有效理论，并预测替代模型误差随 N 变化的趋势。我们在对 LLM 宏观经济学模型 EconAgent 的忠实重新实现，以及其他七个已命名的 LLM 模拟上验证了该分类法，智能体决策克隆自真实 LLM 的引出（主要为 DeepSeek），成本仅数美元；预测的误差趋势逐格成立，而两个被推翻的预测——均涉及强饱和响应，且归因于其曲率——也被该理论在无自由参数的情况下定量匹配。

English

Simulating societies of many large language model (LLM) agents is expensive, yet the questions asked of such simulations are usually macroscopic: phase behaviour, stylised facts, and scaling with the number of agents N, not the cognition of any single agent. We turn a statistical-physics observation into a method: replace each LLM agent by a low-parameter model fitted from a few hundred to a few thousand cheap queries, then run the society at any N on a laptop. Whether this works is decided before the simulation runs, chiefly by what each agent perceives. We introduce an [interaction order x memory] taxonomy that maps perception and memory to an effective theory and a predicted N-trend of the surrogate error. We validate it on a faithful reimplementation of the LLM macroeconomy EconAgent and seven further named LLM simulations, with agent decisions cloned from genuine LLM elicitations (primarily DeepSeek) for a few dollars; the predicted error trends hold cell by cell, and the two refuted predictions, both on a strongly saturating response and traced to its curvature, are themselves matched quantitatively by the theory with no free parameters.