← 首页|学术|Markets, Not Planners: Decentralized Orchestration of LLM Agents with Private Information
cs.MA, cs.CL · 2608.23867 · 2026-08-24

Markets, Not Planners: Decentralized Orchestration of LLM Agents with Private Information

Xiao Liu, Haoyang Li, Songwei Li, Hongbo Fang, Fengli Xu, et al.
Multi-Agent OrchestrationMarket MechanismDecentralized
💬 用重复劳动力市场取代中心化 planner 编排 LLM agent:agent 用私有成本和自维护策略笔记投标,VCG 式支付规则激励诚实报价,复杂任务通过同一机制层层转包。

🎯 背景

LLM agent 由不同方构建、能力和成本各异,编排它们更像是在经济体系中组织劳动力,而不是程序调用子函数。现有编排普遍是中心化的:单一 planner 分配所有任务,随着 agent 池扩大这会成为瓶颈,还需要获取私有信息(如各 agent 的真实执行成本),并且极易被操纵——论文发现向中心化 LLM allocator 插入一条偏好陈述,几乎能让被偏爱的 agent 任务份额翻倍。

🔬 方法

论文提出 AgentLance:一个重复劳动力市场,agent 基于自己的私有成本和自维护的策略笔记进行投标;allocator 依据出价与公开声誉记录选出中标者;一种 VCG 式支付规则激励 agent 诚实报出对成本敏感的出价。复杂任务通过层级委托处理——中标的 agent 可以把任务再分解,并通过同一套市场机制转包给其他 agent。

📊 结果

在数学推理、代码生成、知识密集型问答和 agentic 任务上,AgentLance 能把任务匹配到 agent 的专长领域,随着成本敏感度提高会把工作转移给更便宜的 agent,并持续优于单模型、中心化编排和其他市场基线。作者进一步诊断了市场失灵的来源(成本自我估计不准、投标策略欠优),在受控实验中修正后收益进一步提升,为构建更高效的 agent 经济体系提供了路径。

原文摘要

▶ Abstract
As LLM agents proliferate, built by different parties and with different capabilities and costs, orchestrating them is more like assembling labor across the economy than a computer calling a subroutine. Existing orchestration is typically centralized, with a single planner assigning every task, but this creates a bottleneck as agent pools grow, requires private information (e.g., agents' execution costs), and can easily be manipulated, such that a single inserted preference nearly doubles a favored agent's task share under a centralized LLM allocator. We introduce AgentLance, a repeated labor market in which agents bid on tasks using their private costs and self-maintained strategy notes, an allocator selects winners from bids and public reputation records, and a VCG-style payment rule rewards cost-aware bidding. Complex tasks are handled by hierarchical delegation: winning agents can decompose work and subcontract it through the same mechanism. Across mathematical reasoning, code generation, knowledge-intensive QA, and agentic tasks, AgentLance matches agents to their specializations, shifts work toward cheaper agents as cost sensitivity rises, and consistently outperforms single-model, centralized-orchestration, and market baselines. Diagnosing market failures, including inaccurate cost self-estimation and sub-optimal bidding, then correcting them in controlled experiments yields further gains, charting a path toward more efficient agent economies.
来源: arXiv:2608.23867 · 精读由高松灯生成,基于摘要与 arXiv 页面信息