将智能体流程编译进小模型权重，成本降百倍仍达前沿性能
Compiling Agentic Workflows into LLM Weights: Near-Frontier Quality at Two Orders of Magnitude Less Cost
- 把任务流程直接固化到小模型权重中，无需外部调度器
- 在3个复杂场景中实现接近前沿模型的准确率，推理成本降低90%以上
- 适合追求低成本、高隐私保护的工业级智能体部署
智能体编排框架已积累超过29万GitHub星标，均采用外部调度器+LLM的架构，每步注入指令并路由决策。最新研究显示，仅通过在前沿模型系统提示中写入流程即可完成程序化任务，但存在消耗上下文窗口、需为每轮对话调用前沿模型、且敏感流程易泄露给第三方等问题。将流程编译进小型微调模型权重（即‘地下智能体’）可解决上述问题。已有工作（SimpleTOD、FireAct、SynTOD、WorkflowLLM、Agent Lumos）验证该方法可行，但开发者仍普遍选择外部编排。本文通过实证分析旅行预订（14节点）、Zoom支持（14节点，含产品知识）、保险理赔（55节点，6个决策枢纽）三个场景，识别并克服了三项认知障碍，证明编译式智能体在保持近前沿性能的同时，实现成本降低两个数量级。
原文摘要 · Abstract (English)
Agent orchestration frameworks have proliferated, collectively exceeding 290,000 GitHub stars across LangGraph, CrewAI, Google ADK, OpenAI Agents SDK, Semantic Kernel, Strands, and LlamaIndex. All follow the same pattern: an external orchestrator above the LLM, injecting instructions and routing decisions every turn. Recent work has shown this architecture is dominated for procedural tasks by simply providing the procedure in a frontier model's system prompt [Dennis et al., 2026a], at the cost of consuming the context window, requiring a frontier model for every conversation, and exposing proprietary procedures to third-party providers. Compiling the procedure into the weights of a small fine-tuned model -- creating a subterranean agent -- should resolve all of these concerns, and prior work (SimpleTOD, FireAct, SynTOD, WorkflowLLM, Agent Lumos) has shown the technique works. Yet developer adoption has overwhelmingly favored orchestration. We identify three perceived barriers and address each empirically across travel booking (14 nodes), Zoom support (14 nodes, product-specific knowledge), and insurance claims (55 nodes, 6 decision hubs).
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。