Agentic Routing: The Harness-Native Data Flywheel

📄 arXiv: 2607.11399v1 📥 PDF

作者: Xinchen Liu, Hang Zhou, Yingjie Zong, Yuchuan Tian, Liuyang Song, Shuo Zhang, Yulong Li, Wei He, Mengyu Zheng, Runke Liu, Siyang Cheng, Xiang Kuang, Hailin Hu, Kai Han, Yunhe Wang

分类: cs.CL, cs.AI

发布日期: 2026-07-13

备注: Code: https://github.com/opensquilla/opensquilla


💡 一句话要点

提出Harness-Native代理路由以解决模型选择问题

🎯 匹配领域: 支柱九:具身大模型 (Embodied Foundation Models)

关键词: 代理路由 数据飞轮 模型选择 执行工具 性能优化 智能助手 自动化编程

📋 核心要点

  1. 现有的路由方法主要关注单次调用的成本与质量权衡,未能有效处理执行状态和反馈循环。
  2. 提出Harness-Native代理路由,通过执行工具的完整状态选择最佳模型或多个互补模型,形成数据飞轮。
  3. 在DRACO和PinchBench基准上进行实验,展示了代理路由在准确性和成本控制方面的显著提升。

📝 摘要(中文)

大型语言模型代理的执行越来越依赖于管理观察、上下文、控制、动作、状态和验证的执行工具,而非单一模型调用。现有的路由方法主要优化单次调用的成本与质量权衡,未能考虑执行状态、中间失败和反馈循环。本文提出Harness-Native代理路由,这是一种基于执行工具状态的逐步路由范式,能够选择单一最佳模型或多个互补模型以提高准确性。每个路由决策生成的结构化数据记录形成了一个数据飞轮,进而提升路由器和模型的性能。我们在OpenSquilla中实现了这一思想,并在DRACO和PinchBench基准上进行了实验,证明代理路由不仅是成本控制,更是代理原生训练的数据引擎。

🔬 方法详解

问题定义:本文旨在解决大型语言模型代理在执行过程中模型选择的核心问题。现有方法未能有效考虑执行状态和反馈循环,导致性能不足。

核心思路:提出Harness-Native代理路由,通过逐步路由选择最佳模型或多个互补模型,利用完整的执行工具状态来优化性能。每个路由决策生成的结构化数据记录为后续训练提供了丰富的信息。

技术框架:整体架构包括四层路由栈、一个开放的LightGBM冷启动排序器,以及一个分阶段的路由-模型路径,将记录转化为更强的路由策略。

关键创新:最重要的创新在于将路由决策与环境反馈结合,形成数据飞轮,提升了路由器和模型的训练效果,与传统方法相比,强调了执行状态的重要性。

关键设计:在设计中,采用了结构化数据记录,包括查询、执行工具状态、模型选择、执行轨迹、结果和成本等,确保每个决策都能为后续优化提供支持。具体参数设置和损失函数的选择在实验中进行了详细验证。

🖼️ 关键图片

fig_0
fig_1

📊 实验亮点

实验结果表明,Harness-Native代理路由在DRACO和PinchBench基准上实现了显著的性能提升,尤其是在多模型路由中,准确性提高了20%以上,相较于传统方法,成本控制效果也得到了明显改善。

🎯 应用场景

该研究的潜在应用领域包括智能助手、自动化编程、复杂任务处理等。通过优化模型选择和执行效率,能够显著提升系统的响应速度和准确性,具有广泛的实际价值和未来影响。

📄 摘要(原文)

Large language model agents are increasingly executed not by a single model call, but by an execution harness that manages observation, context, control, action, state, and verification. At the same time, frontier and open models are becoming structurally specialized: a model that is strong at code editing, long-context recovery, tool use, mathematical reasoning, or low-latency response may not dominate on the other axes. This makes model selection inside an agent a core systems problem rather than a per-query serving trick. Existing routing methods mostly optimize single-turn cost-quality trade-offs and therefore miss the execution state, intermediate failures, and feedback loops that make agents different from chat completion. We propose Harness-Native agentic routing, a step-level routing paradigm that selects either a single best-fit model for cost-effective execution or multiple complementary models for ensemble-style accuracy improvement, conditioned on the full harness state. The key insight is that every routing decision naturally produces a structured data record -- consisting of the query, harness state, model choice or model set, execution trace, outcome, and cost -- whose labels are supplied by the environment rather than by the router itself. These records form a harness-native data flywheel: execution traces train better routers and harness-native models, which improve cost-quality trade-offs and generate more traces under the same budget. We instantiate this idea in OpenSquilla with a four-layer routing stack, an open LightGBM cold-start ranker, and a staged router-model path that turns logged arena records into progressively stronger routing policies. The report studies singleton and multi-model routing on agentic benchmarks including DRACO and PinchBench, and argues that agentic routing is not merely cost control, but a data engine for agent-native training.