Herding CATs: ALARA for Agent Harness Engineering in Portable Composable Multi-Agent Teams
Herding CATs: ALARA for Agent Harness Engineering in Portable Composable Multi-Agent Teams
AI总结:
本文提出ALARA原则应用于多智能体团队的代理 harness 工程,通过引入CAT数据层,使用户能够直接声明工具访问权限并修改代理使用的工具,从而提升智能体在各种任务中的表现。
摘要:
行业从业者和学术研究人员经常使用多智能体系统来加速他们的工作,但用户使用的应用系统并未提供一种简单统一的机制来可扩展地管理代理 harness 的关键组件。这种缺乏控制对个体人机交互的质量产生了负面影响,并且限制了从业者协调上下文工程努力的能力。定义此类系统中智能体可以执行的行为规范仍然分散在文本指令文件中(无法保证合规性)或框架内部配置中,使得这些规范在跨团队和项目共享、版本控制或协作维护时变得困难。应用辐射安全中的ALARA原则(暴露量应尽可能低),我们引入了一种通过相互关联的纯文本文件表达的上下文-智能体-工具(CAT)数据层,允许用户为每个智能体直接声明工具访问权限,并在处理时修改智能体使用的工具本身。我们通过使用命令行 shell(加载团队并执行智能体运行)-- npcsh -- 和评估22个本地托管的模型(从0.6B到35B参数)在115个实际任务中的表现(包括文件操作、网络搜索、多步骤脚本、工具链和多智能体委托)来展示CAT数据层的能力。我们还表征了哪些模型家族在某些任务类别中成功,以及在约2500次总执行中它们的失败点。
英文摘要:
Industry practitioners and academic researchers regularly use multi-agent systems to accelerate their work, but the applications through which users operate these systems do not provide a simple, unified mechanism for scalably managing critical components of the agent harness. This lack of control adversely impacts both the quality of individual human-agent interactions and reduces the capacity for practitioners to coordinate context engineering efforts. The behavioral specifications that define what agents in such systems can do remain fragmented across prose instruction files -- for which compliance cannot be guaranteed -- or framework-internal configurations, making these specifications difficult to share, version, or collaboratively maintain across teams and projects. Applying the ALARA principle from radiation safety (exposures kept as low as reasonably achievable) to context, we introduce a context-agent-tool (CAT) data layer expressed through interrelated plain-text files, allowing users to directly declare tool access for each agent and to modify the tools themselves that are used by the agents when processing. We demonstrate capability of this CAT data layer to enable real agentic usage by using a command-line shell that loads the team and executes agent runs -- \texttt{npcsh} -- and evaluating 22 locally-hosted models from 0.6B to 35B parameters across 115 practical tasks spanning file operations, web search, multi-step scripting, tool chaining, and multi-agent delegation. We characterize which model families succeed in certain task categories and where they break down across $\sim$2500 total executions.
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。