不同的状态编码导致语言模型代理同步行为差异，编码方式影响集体动态结果。
Same physical state, different collective dynamics: state encodings select synchronization outcomes in language-model agents
- 用圆周矩或直方图编码邻居相位信息，改变集体同步效果。
- GPT中矩编码使6/6种子实现同步，直方图编码则0/6；Claude中效果相反。
- 编码影响模型决策概率，揭示其作为有效相互作用定律的一部分。
语言模型代理依赖环境状态编码进行决策，但这些编码常被视为可互换接口。我们使用预训练语言模型设计了一个环形同步实验，在保持物理系统不变的前提下，对状态编码进行干预：每个代理仅看到邻居相对相位的摘要，并选择前进、停留或后退。将状态编码为低阶圆周矩而非直方图，导致了不同的集体动态结果。在GPT中，矩编码使种群在6/6种子下实现同步，而直方图编码则无一成功；该现象在Claude中重现但方向相反。重放相同场数据时，各代理前进/停留/后退的概率变化远超同一编码内的重复波动，且在GPT、Claude和Gemini中均成立；在GPT中，仅呈现方式改变即可调整操作者，即使矩值固定。因此，状态编码构成模型依赖的有效相互作用律，而非中性接口。
原文摘要 · Abstract (English)
Language-model agents act on state encodings of their environment, yet these are treated as interchangeable interfaces. Using pretrained language models, we designed a circular-synchronization experiment applying a state-encoding intervention while holding the physical system fixed: each agent sees only a summary of its neighbours' relative phases and chooses to advance, stay or retard. Encoding that state as low-order circular moments rather than as a histogram selected different collective outcomes. In GPT the moment encoding synchronized the population in 6/6 seeds and the histogram encodings in 0/6; the effect replicated in Claude but reversed direction. Replaying identical fields shifted each agent's advance/stay/retard probabilities far beyond within-encoding repeat variation, in GPT, Claude and Gemini; in GPT, presentation alone shifted the operator with the moment values fixed. State encodings therefore form part of a model-dependent effective interaction law, not a neutral interface.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。