窮人版智能體建模：在筆記型電腦上模擬大型LLM智能體社會

摘要

模擬由許多大型語言模型（LLM）智能體組成的社會代價高昂，但這類模擬通常要回答的是宏觀問題：相行為、典型化事實，以及隨智能體數量 N 的標度行為，而非單一智能體的認知。我們將統計物理學中的一個觀察轉化為方法：以低參數模型取代每個 LLM 智能體，該模型從數百到數千次廉價查詢中擬合而成，然後在筆記型電腦上以任意 N 運行該社會。這種方法是否有效，在模擬執行前就已決定，主要取決於每個智能體感知到的內容。我們引入「交互階數 × 記憶」分類法，將感知與記憶映射到有效理論，並預測替代誤差的 N 趨勢。我們在 LLM 宏觀經濟 EconAgent 的忠實重新實現以及其他七個具名 LLM 模擬上驗證了該分類法，其中智能體決策是從真實 LLM 誘發（主要為 DeepSeek）中克隆而來，成本僅數美元；預測的誤差趨勢逐格成立，而兩個被推翻的預測——皆涉及強烈飽和的響應，並追溯至其曲率——也由該理論在無自由參數的情況下定量匹配。

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.