TRIAGE: Three-level Routing and Intelligent Agent Guidance for Efficient Execution
作者: Ruocan Wei
分类: cs.LG
发布日期: 2026-09-01
💡 一句话要点
提出TRIAGE框架以解决ReAct模型效率问题
🎯 匹配领域: 支柱九:具身大模型 (Embodied Foundation Models)
关键词: 大型语言模型 ReAct 效率优化 技能重用 自动化学习 安全监控 语义路由
📋 核心要点
- 现有的ReAct模型在处理每个查询时都需重新推理,导致效率低下,无法利用历史经验。
- TRIAGE框架通过将历史执行轨迹抽象为可重用技能,分类查询并重用历史信息,从而提高效率。
- 在实验中,TRIAGE实现了62.3%的令牌节省,且在ToolBench上验证了其跨领域的有效性和普适性。
📝 摘要(中文)
基于ReAct范式的大型语言模型(LLM)代理在工具使用和任务执行方面表现出色,但存在效率问题:每个查询都需从头开始推理,导致相似查询重复相同步骤而未能利用历史经验。为此,本文提出TRIAGE,一个三层路由框架,通过重用历史执行轨迹来减少令牌消耗。其核心创新是TaaS(Trajectory-as-a-Skill),将历史执行轨迹抽象为可重用技能,实现“经验即服务”。TRIAGE将查询分为三个层级:直接重用、技能替代和完整ReAct。大规模实验表明,TRIAGE在1007个安全监控查询中实现62.3%的令牌节省,且在ToolBench的跨领域验证中也取得了76.3%的令牌减少,验证了语义路由的普适性。
🔬 方法详解
问题定义:本文旨在解决ReAct模型在处理查询时的效率问题,现有方法每次查询都需从头推理,导致资源浪费和响应时间延长。
核心思路:TRIAGE框架通过重用历史执行轨迹,将其转化为可重用的技能,从而减少重复计算和令牌消耗。
技术框架:TRIAGE将查询分为三个层级:直接重用(相同查询)、技能替代(相似查询)和完整ReAct(新查询),并通过TaaS实现技能的提取与重用。
关键创新:最重要的创新在于TaaS的提出,它将历史轨迹转化为技能,允许在不同查询间共享经验,显著提高了执行效率。
关键设计:在设计中,TRIAGE采用了自动技能提取机制,能够从高频轨迹模式中提炼出确定性技能,并通过反馈循环不断提升系统的效率。
🖼️ 关键图片
📊 实验亮点
在1007个安全监控查询的实验中,TRIAGE实现了62.3%的令牌节省,其中56.0%的查询在技能替代层级执行,且无令牌消耗。跨领域验证在ToolBench上也取得了76.3%的令牌减少,显示出其强大的普适性和有效性。
🎯 应用场景
TRIAGE框架具有广泛的应用潜力,尤其在需要高效处理大量查询的领域,如安全监控、客户服务和智能助手等。通过减少计算资源的消耗,TRIAGE能够显著提升系统的响应速度和用户体验,未来可能在更多实际应用中得到推广。
📄 摘要(原文)
Large Language Model (LLM) agents based on the ReAct paradigm have demonstrated remarkable capabilities in tool use and task execution. However, ReAct suffers from a fundamental efficiency problem: every query triggers a complete reasoning loop from scratch, and similar queries repeat identical steps without leveraging historical experience. We propose TRIAGE,a three-level routing framework that reduces token consumption by reusing historical execution trajectories. Its core innovation is TaaS (Trajectory-as-a-Skill), which abstracts historical execution trajectories into reusable skills, realizing 'experience as a service'. TRIAGE classifies queries into three levels: (1) Direct Reuse-identical queries, 0 tokens; (2) Skill Substitution-similar queries, 0 tokens via deterministic parameter substitution; (3) Full ReAct-novel queries, automatically stored for future reuse. In large-scale experiments on 1,007 security monitoring queries, TRIAGE achieves 62.3% token savings, with 56.0% of queries at Level 2 and 5.5% at Level 1, both executing at zero cost. Cross-domain validation on ToolBench (15 domains, 345 queries) achieves 76.3% token reduction, confirming the generalizability of semantic routing. An online learning experiment demonstrates cold-start-to-mature evolution: the L2 hit rate rises from 0% to 57% within the first 100 queries, and the average token cost drops from 198 to 74.7. We also propose an automatic Skill extraction mechanism that distills high-frequency trajectory patterns into deterministic Skills, creating a positive feedback loop of 'the more you use it, the more efficient it becomes'.