LLM 智能体是递归推理还是统计外推？多智能体相互依赖决策中的 in-context learning 研究
Recursive Reasoning or Statistical Extrapolation? In-Context Learning in Multi-Agent Interdependent Decision-Making
AI 导读
一项 arXiv 研究通过多智能体不完全信息公共物品博弈检验 LLM 智能体的 in-context learning（ICL）究竟源自递归信念推理还是统计外推。实验发现，当历史反馈的统计结构被扰乱时，更长上下文带来的决策收益基本消失，决策质量退化到与无上下文基线相当，且策略相互依赖性越强退化越明显。作者提出用理性预期均衡（REE）基准区分推理与外推，并结论称此环境下 ICL 更接近统计外推。
正文 · 原文
Abstract:In-context learning (ICL) enables large language model (LLM) agents to improve decisions using interaction history, yet it remains unclear whether such improvement reflects refined internal reasoning or mere extrapolation of statistical patterns. To disentangle these mechanisms, we study LLM agents in multi-agent incomplete-information games that require recursive belief reasoning. By constructing a public goods game and manipulating the statistical structure of historical feedback, we evaluate decision quality against a history-independent rational expectations equilibrium (REE) benchmark. Our experiments reveal that when historical statistical patterns are disrupted, the benefits of longer context largely vanish, degrading decision quality to the no-context baseline in a way sharply amplified by stronger strategic interdependence. These results suggest that, in such strategic environments, ICL behavior is more consistent with statistical extrapolation than with strategic reasoning. Our work extends the mechanistic study of ICL to strategic multi-agent settings, introduces REE as a diagnostic tool for distinguishing reasoning from extrapolation, and provides a reusable framework for probing the boundaries of LLM reasoning in recursive belief tasks.
| Comments: | Findings of the Association for Computational Linguistics: EMNLP 2026 |
| Subjects: | Artificial Intelligence (cs.AI); General Economics (econ.GN) |
| Cite as: | arXiv:2609.18591 [cs.AI] |
| (or arXiv:2609.18591v2 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2609.18591 arXiv-issued DOI via DataCite |
Submission history
From: Yu Liu [view email]
[v1]
Wed, 16 Sep 2026 12:49:19 UTC (9,400 KB)
[v2]
Sat, 24 Oct 2026 07:42:51 UTC (9,400 KB)
来源：arXiv cs.AI · arxiv.org