arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2602.03695cs.MAcs.AIcs.CL
收藏

Agent Primitives: 面向多智能体系统的可复用潜在构建模块

Agent Primitives: Reusable Latent Building Blocks for Multi-Agent Systems

School of Information Sciences, University of Illinois at Urbana-Champaign, IL, USA(伊利诺伊大学厄巴纳-香槟分校信息科学学院);Siebel School of Computing and Data Science, University of Illinois at Urbana-Champaign, IL, USA(伊利诺伊大学厄巴纳-香槟分校Siebel计算与数据科学学院)

Haibo Jin, Peng Kuang, Ye Yu, Xiaopeng Yuan, Haohan Wang

AI总结:

提出Agent Primitives,一组可复用的潜在构建模块,通过KV缓存内部通信和自动组合,提升多智能体系统的鲁棒性、效率和跨任务复用性。

摘要:

虽然现有的多智能体系统(MAS)可以通过多个智能体之间的协作处理复杂问题,但它们通常高度任务特定,依赖手动设计的智能体角色和交互提示,导致架构复杂性增加且跨任务复用性有限。此外,大多数MAS主要通过自然语言进行通信,使得它们在内部智能体历史中的长上下文、多阶段交互中容易受到错误累积和不稳定性的影响。在这项工作中,我们提出了 extbf{Agent Primitives},一组用于基于LLM的MAS的可复用潜在构建模块。受神经网络设计的启发,其中复杂模型由可复用组件构建,我们观察到许多现有的MAS架构可以分解为少量重复出现的内部计算模式。基于这一观察,我们实例化了三个原语:Review、Voting and Selection以及Planning and Execution。所有原语通过键值(KV)缓存进行内部通信,通过减轻多阶段交互中的信息退化来提高鲁棒性和效率。为了实现自动系统构建,一个Organizer智能体根据每个查询选择并组合原语,由先前成功配置的轻量级知识库引导,形成基于原语的MAS。实验表明,基于原语的MAS相比单智能体基线平均准确率提高12.0-16.5%,与基于文本的MAS相比,令牌使用量和推理延迟减少约3-4倍,同时相对于单智能体推理仅产生1.3-1.6倍的开销,并在不同模型骨干上提供更稳定的性能。

英文摘要:

While existing multi-agent systems (MAS) can handle complex problems by enabling collaboration among multiple agents, they are often highly task-specific, relying on manually crafted agent roles and interaction prompts, which leads to increased architectural complexity and limited reusability across tasks. Moreover, most MAS communicate primarily through natural language, making them vulnerable to error accumulation and instability in long-context, multi-stage interactions within internal agent histories. In this work, we propose \textbf{Agent Primitives}, a set of reusable latent building blocks for LLM-based MAS. Inspired by neural network design, where complex models are built from reusable components, we observe that many existing MAS architectures can be decomposed into a small number of recurring internal computation patterns. Based on this observation, we instantiate three primitives: Review, Voting and Selection, and Planning and Execution. All primitives communicate internally via key-value (KV) cache, which improves both robustness and efficiency by mitigating information degradation across multi-stage interactions. To enable automatic system construction, an Organizer agent selects and composes primitives for each query, guided by a lightweight knowledge pool of previously successful configurations, forming a primitive-based MAS. Experiments show that primitives-based MAS improve average accuracy by 12.0-16.5\% over single-agent baselines, reduce token usage and inference latency by approximately 3$\times$-4$\times$ compared to text-based MAS, while incurring only 1.3$\times$-1.6$\times$ overhead relative to single-agent inference and providing more stable performance across model backbones.

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。

↑