arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2603.23626cs.LGcond-mat.stat-mechcs.AIcs.CLnlin.AO
收藏

大语言模型信息易感性理论

A Theory of LLM Information Susceptibility

School of Physics, Peking University(北京大学物理学院);Center for High Energy Physics, Peking University(北京大学高能物理中心)

Zhuo-Yang Song, Hua Xing Zhu

AI总结:

本文提出大语言模型信息易感性理论,探讨计算资源充足时固定LLM对策略集预算敏感性的影响,通过多变量效用函数框架分析多预算通道架构,并验证了共缩放架构在提升性能方面的有效性。

摘要:

大语言模型(LLMs)正越来越多地作为优化模块应用于智能体系统中,但此类LLM介导改进的基本限制仍不明确。本文提出大语言模型信息易感性理论,核心假设是当计算资源足够大时,固定LLM的干预不会增加策略集对预算的性能敏感性。我们开发了一个多变量效用函数框架,将这一假设推广到具有多个相互关联预算通道的架构中,并讨论了在何种条件下共缩放可以超越敏感性界限。我们通过结构多样、模型规模跨越一个数量级的领域进行了实证验证,展示了嵌套、共缩放架构打开的响应通道无法被固定配置所获得。这些结果阐明了何时LLM干预有助于提升性能,何时不会,证明了统计物理工具可以为AI系统设计提供预测约束。如果敏感性假说普遍成立,该理论表明嵌套架构可能是开放性智能体自我改进的必要结构条件。

英文摘要:

Large language models (LLMs) are increasingly deployed as optimization modules in agentic systems, yet the fundamental limits of such LLM-mediated improvement remain poorly understood. Here we propose a theory of LLM information susceptibility, centred on the hypothesis that when computational resources are sufficiently large, the intervention of a fixed LLM does not increase the performance susceptibility of a strategy set with respect to budget. We develop a multi-variable utility-function framework that generalizes this hypothesis to architectures with multiple co-varying budget channels, and discuss the conditions under which co-scaling can exceed the susceptibility bound. We validate the theory empirically across structurally diverse domains and model scales spanning an order of magnitude, and show that nested, co-scaling architectures open response channels unavailable to fixed configurations. These results clarify when LLM intervention helps and when it does not, demonstrating that tools from statistical physics can provide predictive constraints for the design of AI systems. If the susceptibility hypothesis holds generally, the theory suggests that nested architectures may be a necessary structural condition for open-ended agentic self-improvement.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑