将大模型搜索与推理解耦，实现低成本高可控的智能检索。
Decoupling Search from Reasoning: A Vendor-Agnostic Grounding Architecture for LLM Agents
- 通过MCP兼容网关将搜索独立于推理模型，暴露路由、缓存等控制权
- 在SimpleQA上以91%更低成本接近原生精度（86.1% vs 87.7%）
- 适合需严格输出控制和多模型复用的生产级智能体系统
生产环境中的大模型智能体越来越依赖实时搜索，但原生搜索将检索策略、服务商选择、证据注入、成本、延迟和生成行为耦合在单一模型-服务边界内。这种耦合使得接地机制难以检查、调优、复用或迁移，并可能引发搜索导致的冗长输出，破坏严格的输出契约。我们提出解耦搜索接地（DSG），一种厂商无关的边界机制，通过MCP兼容网关将接地移出推理模型，显式暴露服务商路由、源感知上下文渲染、可配置回退、检索深度控制以及精确与语义缓存等一级控制。在五个前沿模型上对SimpleQA、FreshQA和HotpotQA的测试显示，原生搜索在时效敏感的FreshQA上表现更优，但当控制能力重要时，DSG展现出更强性能：在SimpleQA上以91%更低搜索成本接近原生准确率（86.1% vs 87.7%），保持简洁回答契约，达到99.4%的热缓存命中率且延迟降低68%。部署为大规模智能体工作负载的共享接地层，支持可互换模型，在电商查询理解（QIU）任务中保持或略微超过原生搜索准确率，同时将搜索成本降低超98%。实时接地应被视为可优化的接口边界，而非固定模型特性。
原文摘要 · Abstract (English)
Production LLM agents increasingly depend on real-time search, yet native search grounding bundles retrieval policy, provider choice, evidence injection, cost, latency, and generation behavior behind a single model-provider boundary. This coupling makes grounding hard to inspect, tune, reuse, or port, and can trigger Search-Induced Verbosity that breaks strict output contracts. We present Decoupled Search Grounding (DSG), a vendor-agnostic boundary that moves grounding outside the reasoning model through an MCP-compatible gateway, exposing provider routing, source-aware context rendering, configured fallback, retrieval-depth control, and exact plus semantic caching as first-class controls. Across five frontier models on SimpleQA, FreshQA, and HotpotQA, native search leads on recency-sensitive FreshQA, but DSG exposes a stronger frontier when control matters: on SimpleQA it nearly matches native accuracy (86.1% vs. 87.7%) at 91% lower search cost, preserves concise answer contracts, and reaches a 99.4% warm-cache hit rate with 68% lower latency. Deployed as a shared production grounding layer for large-scale agentic workloads with interchangeable models, DSG matches or slightly exceeds native-search accuracy on an e-commerce query-understanding (QIU) workload while cutting search cost by over 98%. Real-time grounding is best treated as an optimizable interface boundary, not a fixed model feature.
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。