推理-执行-合成器：具有静态O(1)上下文窗口的可扩展代理架构
Reasoner-Executor-Synthesizer: Scalable Agentic Architecture with Static O(1) Context Window
AI总结：
本文提出RES架构，通过分离意图解析、确定性数据检索与叙事生成，实现O(1)的token复杂度，有效解决LLM代理中上下文窗口过大导致的幻觉风险和token成本问题。
摘要：
大型语言模型（LLM）作为自主代理时通常使用检索增强生成（RAG），将检索到的文档输入上下文窗口，这导致两个问题：上下文长度增加会增加幻觉风险，且token成本与数据集规模线性增长。我们提出推理-执行-合成器（RES）架构，一种三层设计，严格分离意图解析（推理器）、确定性数据检索和聚合（执行器）以及叙述生成（合成器）。执行器使用零LLM token，并将固定大小的统计摘要传递给合成器。我们正式证明RES在数据集规模下实现O(1)的token复杂度，并在ScholarSearch上验证，该系统基于Crossref API（130M+篇文章）。在100次基准测试中，无论数据集包含42,000还是16.3 million篇文章，RES的平均token成本均为1,574 tokens。该架构通过构造消除数据幻觉：LLM从未看到原始记录。
英文摘要：
Large Language Models (LLMs) deployed as autonomous agents commonly use Retrieval-Augmented Generation (RAG), feeding retrieved documents into the context window, which creates two problems: the risk of hallucination grows with context length, and token cost scales linearly with dataset size. We propose the Reasoner-Executor-Synthesizer (RES) architecture, a three-layer design that strictly separates intent parsing (Reasoner), deterministic data retrieval and aggregation (Executor), and narrative generation (Synthesizer). The Executor uses zero LLM tokens and passes only fixed-size statistical summaries to the Synthesizer. We formally prove that RES achieves O(1) token complexity with respect to dataset size, and validate this on ScholarSearch, a scholarly research assistant backed by the Crossref API (130M+ articles). Across 100 benchmark runs, RES achieves a mean token cost of 1,574 tokens regardless of whether the dataset contains 42,000 or 16.3 million articles. The architecture eliminates data hallucination by construction: the LLM never sees raw records. KEYWORDS LLM agents; agentic architecture; hallucination elimination; token optimization; context window; retrieval-augmented generation; deterministic execution; scholarly metadata; Crossref API; O(1) complexity.
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。
