研究多智能体大模型通信如何让输出趋同或分化。
BOUNDARY_SYNC: Measuring Communication-Induced Representational Coupling in Multi-Agent LLM Systems
- 用耦合放大因子量化通信导致的表示趋同
- 文本通信使输出显著趋同（CAF=0.803），图像通信也类似
- 组规模和提示设计可控制趋同方向，适合系统设计参考
随着大语言模型作为通信智能体部署，智能体间通信是否导致输出趋同？我们提出 BOUNDARY_SYNC 协议，通过耦合放大因子（CAF = JSD_cond / JSD_baseline）测量表示耦合，CAF < 1 表示同质化，CAF > 1 表示多样化。在受控的 GPT-4o 实验中（N=30，约 9,900 次 API 调用），测量了文本与图像通信下的耦合情况。关键发现：(1) 文本通信导致显著同质化（CAF=0.803 [0.740, 0.873]，d=1.30，p<0.001），经无通信对照与提示扰动验证；(2) 图像通信在同模态基线中也呈现同质化（CAF=0.834 [0.811, 0.858]），效应比例相当；(3) 群体规模调节耦合方向——K=5 时同质化，而 K=3 时 CAF>1.0（点估计值 1.14 和 1.06，置信区间待定），显示向多样化的转变趋势；(4) 跨模型复现结果差异极大（CAF 0.034–0.803），DeepSeek 受格式伪影主导；(5) 耦合为无状态过程，由提示上下文驱动而非累积更新，持续共识引发单调收敛。这些结果确立了大模型智能体耦合的真实存在、可测量性及提示层面的可控性，对多智能体系统设计有直接意义。
原文摘要 · Abstract (English)
As large language models (LLMs) are deployed as communicating agents, does inter-agent communication cause outputs to converge? We introduce BOUNDARY_SYNC, a protocol measuring representational coupling via the Coupling Amplification Factor (CAF = JSD_cond / JSD_baseline), where CAF < 1 indicates homogenization and CAF > 1 indicates diversification. In controlled GPT-4o experiments (N=30, ~9,900 API calls), we measure coupling in text and image communication. Key findings: (1) text communication causes significant homogenization (CAF=0.803 [0.740, 0.873], d=1.30, p<0.001), confirmed by no-communication ablation and prompt-perturbation controls; (2) image communication also homogenizes under within-modality baselines (CAF=0.834 [0.811, 0.858]), with comparable proportional effect; (3) group size moderates coupling direction -- K=5 produces homogenization while K=3 yields CAF > 1.0 (point estimates 1.14 and 1.06, CI pending), suggesting a directional shift toward diversification; (4) cross-model replication shows extreme variation (CAF 0.034-0.803), with DeepSeek dominated by format artifacts; (5) coupling is stateless -- driven by prompt context rather than cumulative updating, with continuous consensus producing monotonic convergence. These results establish LLM agent coupling as real, measurable, and controllable at the prompt level, with direct implications for multi-agent system design.
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。