--- title: "RAG patterns — RAG 模式" source: "https://systems-analysis.info/int/RAG_patterns_%E2%80%94_RAG_%E6%A8%A1%E5%BC%8F" wiki: "systems-analysis.info/int" article: "RAG_patterns_—_RAG_模式" language: "zh" categories: - "Category:Chinese" - "Category:Large language models" - "Category:Prompt engineering" revision_id: 6118 wiki_created_at: 2026-09-06T23:58:35Z wiki_modified_at: 2026-09-06T23:58:35Z downloaded_at: 2026-09-07T23:12:09Z --- # RAG patterns — RAG 模式 **RAG 模式**（英语：*RAG Patterns*）是一套用于构建**检索增强生成**（Retrieval-Augmented Generation, RAG）系统的架构和方法论。这些模式旨在通过将大型语言模型（LLM）与外部动态可访问的数据源集成，解决 LLM 的一些基本问题，如幻觉、知识过时和领域特异性不足^{[\[1\]](https://systems-analysis.info/int/RAG_patterns_%E2%80%94_RAG_%E6%A8%A1%E5%BC%8F#cite_note-lewis2020-1)}。RAG 的发展已从简单的线性流程演变为复杂的模块化和代理系统^{[\[2\]](https://systems-analysis.info/int/RAG_patterns_%E2%80%94_RAG_%E6%A8%A1%E5%BC%8F#cite_note-survey2024-2)}。 ## 主要 RAG 模式随着技术的发展，出现了多种 RAG 模式，每种模式都旨在解决特定问题，并在质量、速度和成本之间进行权衡。 - **Classic RAG（经典 RAG）** — 这是一种基础方法，将用户查询向量化，在向量数据库中搜索相关片段（chunk）；然后将找到的 chunk 连同问题一起输入 LLM 以生成答案^{[\[1\]](https://systems-analysis.info/int/RAG_patterns_%E2%80%94_RAG_%E6%A8%A1%E5%BC%8F#cite_note-lewis2020-1)}。 - **Multi-Query RAG（多查询 RAG）** — LLM 生成原始查询的多个改写或细化版本；对所有版本进行搜索，并将结果合并，从而提高召回率（*recall*）^{[\[3\]](https://systems-analysis.info/int/RAG_patterns_%E2%80%94_RAG_%E6%A8%A1%E5%BC%8F#cite_note-langchain-multiquery-3)}。 - **HyDE (Hypothetical Document Expansion)** — 用于克服简短查询与长文档之间的“语义鸿沟”。LLM 首先生成一个“假设性”的文档答案，然后使用其嵌入进行搜索，这通常能提高检索质量^{[\[4\]](https://systems-analysis.info/int/RAG_patterns_%E2%80%94_RAG_%E6%A8%A1%E5%BC%8F#cite_note-hyde-4)}。 - **Hybrid Retrieval（混合检索）** — 结合了语义（向量）和词法（BM25）搜索。混合方案已成为生产系统的标准：向量搜索覆盖语义匹配，而 BM25 则能找到精确的术语、ID 或缩略词；最后通过融合（fusion）合并结果^{[\[5\]](https://systems-analysis.info/int/RAG_patterns_%E2%80%94_RAG_%E6%A8%A1%E5%BC%8F#cite_note-weaviate-hybrid-5)[\[6\]](https://systems-analysis.info/int/RAG_patterns_%E2%80%94_RAG_%E6%A8%A1%E5%BC%8F#cite_note-qdrant-hybrid-6)[\[7\]](https://systems-analysis.info/int/RAG_patterns_%E2%80%94_RAG_%E6%A8%A1%E5%BC%8F#cite_note-milvus-fulltext-7)}。 - **Re-ranking（重排序）** — 一个两阶段过程：快速检索器返回一组候选结果（例如，前 100 个），然后由交叉编码器（cross-encoder）或其他重排序器重新计算相关性，并选出最佳结果（例如，前 5 个）提供给 LLM^{[\[8\]](https://systems-analysis.info/int/RAG_patterns_%E2%80%94_RAG_%E6%A8%A1%E5%BC%8F#cite_note-nogueira2019-8)[\[9\]](https://systems-analysis.info/int/RAG_patterns_%E2%80%94_RAG_%E6%A8%A1%E5%BC%8F#cite_note-cohere-rerank-9)}。 - **Query Routing（查询路由）** — 在包含多个异构数据源（不同的索引、数据库、API）的系统中，查询通过路由器（基于 LLM 的选择器或分类器）被导向最佳数据源；通常包含回退（fallback）策略^{[\[10\]](https://systems-analysis.info/int/RAG_patterns_%E2%80%94_RAG_%E6%A8%A1%E5%BC%8F#cite_note-llama-router-10)}。 - **Agentic/Web RAG（代理式 RAG）** — LLM 充当代理：它分解复杂问题，规划迭代步骤，并利用工具（向量搜索、网页搜索）进行反馈驱动的操作。典型的实现是 ReAct 范式^{[\[11\]](https://systems-analysis.info/int/RAG_patterns_%E2%80%94_RAG_%E6%A8%A1%E5%BC%8F#cite_note-react-11)}；对于面向网页信息收集且要求强制引用的场景，可参考 WebGPT^{[\[12\]](https://systems-analysis.info/int/RAG_patterns_%E2%80%94_RAG_%E6%A8%A1%E5%BC%8F#cite_note-webgpt-12)}。 ### 相关及新兴范式 - **GraphRAG（图 RAG）** — 使用知识图谱作为数据源和上下文选择机制；搜索沿着实体间的连接结构和文本内容进行，从而提高了多跳（multi-hop）问题的可解释性和答案质量^{[\[13\]](https://systems-analysis.info/int/RAG_patterns_%E2%80%94_RAG_%E6%A8%A1%E5%BC%8F#cite_note-graphrag-13)[\[14\]](https://systems-analysis.info/int/RAG_patterns_%E2%80%94_RAG_%E6%A8%A1%E5%BC%8F#cite_note-graphrag-project-14)}。 - **MM-RAG（多模态 RAG）** — 处理文本和视觉信息源（扫描件、图表、表格）。例如，VisRAG 展示了在多模态文档上进行面向视觉语言模型（VLM）的检索和生成^{[\[15\]](https://systems-analysis.info/int/RAG_patterns_%E2%80%94_RAG_%E6%A8%A1%E5%BC%8F#cite_note-visrag-15)}。 - **Packaging & Context Handling（上下文打包）** — 将检索到的 chunk 集成到提示词中的方法，包括：*Stuff*、*Map-Reduce*、*Refine*、*Tree-of-Chunks (RAPTOR)*^{[\[16\]](https://systems-analysis.info/int/RAG_patterns_%E2%80%94_RAG_%E6%A8%A1%E5%BC%8F#cite_note-raptor-16)}。 ## 模式对比表 | 模式 | 应用场景 | 对质量的影响 | 成本/延迟 | 风险与限制 | |----------------------|------------------------------------------------------------|-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|-----------|--------------------------------------------------------| | **Classic RAG** | PoC 和基于同构数据库的简单问答 | 基线水平；严重依赖嵌入质量^{[\[1\]](https://systems-analysis.info/int/RAG_patterns_%E2%80%94_RAG_%E6%A8%A1%E5%BC%8F#cite_note-lewis2020-1)} | 低 | 对措辞敏感；存在上下文不相关的风险 | | **Hybrid Retrieval** | 大多数生产场景；尤其适用于包含大量代码、缩略词或 ID 的数据 | 提高召回率；能覆盖精确术语^{[\[5\]](https://systems-analysis.info/int/RAG_patterns_%E2%80%94_RAG_%E6%A8%A1%E5%BC%8F#cite_note-weaviate-hybrid-5)[\[6\]](https://systems-analysis.info/int/RAG_patterns_%E2%80%94_RAG_%E6%A8%A1%E5%BC%8F#cite_note-qdrant-hybrid-6)[\[7\]](https://systems-analysis.info/int/RAG_patterns_%E2%80%94_RAG_%E6%A8%A1%E5%BC%8F#cite_note-milvus-fulltext-7)} | 低/中 | 需要调整融合权重；管理两个索引 | | **Re-ranking** | 对精度要求极高的场景 | 显著提升 top-k 结果的精确率^{[\[8\]](https://systems-analysis.info/int/RAG_patterns_%E2%80%94_RAG_%E6%A8%A1%E5%BC%8F#cite_note-nogueira2019-8)[\[9\]](https://systems-analysis.info/int/RAG_patterns_%E2%80%94_RAG_%E6%A8%A1%E5%BC%8F#cite_note-cohere-rerank-9)} | 中/高 | 增加额外的延迟和成本 | | **Multi-Query** | 简短或涉及多个方面（multi-aspect）的查询 | 提高召回率^{[\[3\]](https://systems-analysis.info/int/RAG_patterns_%E2%80%94_RAG_%E6%A8%A1%E5%BC%8F#cite_note-langchain-multiquery-3)} | 中 | 可能产生多余或带噪声的改写 | | **HyDE** | 简短或模糊且存在较大“语义鸿沟”的查询 | 提升零样本（*zero-shot*）检索质量^{[\[4\]](https://systems-analysis.info/int/RAG_patterns_%E2%80%94_RAG_%E6%A8%A1%E5%BC%8F#cite_note-hyde-4)} | 中 | 效果依赖于“假设性”文本的质量 | | **Query Routing** | 存在多个数据源（文档库、SQL、API、网页）的场景 | 通过选择正确的数据源来提高相关性^{[\[10\]](https://systems-analysis.info/int/RAG_patterns_%E2%80%94_RAG_%E6%A8%A1%E5%BC%8F#cite_note-llama-router-10)} | 中 | 路由错误会导致检索失败 | | **Agentic/Web RAG** | 复杂、探索性、多步骤的查询 | 解决超出线性流程能力范围的任务^{[\[11\]](https://systems-analysis.info/int/RAG_patterns_%E2%80%94_RAG_%E6%A8%A1%E5%BC%8F#cite_note-react-11)[\[12\]](https://systems-analysis.info/int/RAG_patterns_%E2%80%94_RAG_%E6%A8%A1%E5%BC%8F#cite_note-webgpt-12)} | 高 | 实现复杂，可能陷入循环；需要设置安全护栏（guardrails） | 关键 RAG 模式对比 ## 实践与架构 ### 实施阶段 1. **Proof of Concept (PoC)：** 从 **Classic RAG** 开始，在一个有限但有代表性的数据集上验证嵌入质量和基本检索效果^{[\[1\]](https://systems-analysis.info/int/RAG_patterns_%E2%80%94_RAG_%E6%A8%A1%E5%BC%8F#cite_note-lewis2020-1)}。 2. **Minimum Viable Product (MVP)：** 实施 **Hybrid Retrieval** 和 **Re-ranking**，这是投入产出比最高的组合^{[\[5\]](https://systems-analysis.info/int/RAG_patterns_%E2%80%94_RAG_%E6%A8%A1%E5%BC%8F#cite_note-weaviate-hybrid-5)[\[8\]](https://systems-analysis.info/int/RAG_patterns_%E2%80%94_RAG_%E6%A8%A1%E5%BC%8F#cite_note-nogueira2019-8)}。 3. **Production（生产环境）：** 添加查询转换（**HyDE**、**Multi-Query**），并在必要时引入 **Query Routing**；配置可观测性（记录检索、重排序和响应日志）并进行 A/B 测试^{[\[3\]](https://systems-analysis.info/int/RAG_patterns_%E2%80%94_RAG_%E6%A8%A1%E5%BC%8F#cite_note-langchain-multiquery-3)[\[10\]](https://systems-analysis.info/int/RAG_patterns_%E2%80%94_RAG_%E6%A8%A1%E5%BC%8F#cite_note-llama-router-10)}。 ### 关键组件 - **分块（Chunking）：** 影响质量最关键的因素之一。简单的固定大小分块常会破坏语义单元。推荐使用结构化（基于标记）或递归（段落→句子→词）的分割器^{[\[17\]](https://systems-analysis.info/int/RAG_patterns_%E2%80%94_RAG_%E6%A8%A1%E5%BC%8F#cite_note-rcsplit-17)[\[18\]](https://systems-analysis.info/int/RAG_patterns_%E2%80%94_RAG_%E6%A8%A1%E5%BC%8F#cite_note-llama-hier-18)}。 - **嵌入与元数据：** 为每个 chunk 存储 document_id、页面/章节、标题、日期等元数据；这对于过滤和正确引用来源至关重要。 - **混合检索与重排序：** 使用 BM25+向量搜索并进行融合（或 RRF），然后对一小部分候选结果使用交叉编码器进行重排序^{[\[5\]](https://systems-analysis.info/int/RAG_patterns_%E2%80%94_RAG_%E6%A8%A1%E5%BC%8F#cite_note-weaviate-hybrid-5)[\[6\]](https://systems-analysis.info/int/RAG_patterns_%E2%80%94_RAG_%E6%A8%A1%E5%BC%8F#cite_note-qdrant-hybrid-6)[\[8\]](https://systems-analysis.info/int/RAG_patterns_%E2%80%94_RAG_%E6%A8%A1%E5%BC%8F#cite_note-nogueira2019-8)}。 - **上下文打包：** 对于大型语料库，可选择 *Map-Reduce*、*Refine* 或 *Tree-of-Chunks* 等策略^{[\[16\]](https://systems-analysis.info/int/RAG_patterns_%E2%80%94_RAG_%E6%A8%A1%E5%BC%8F#cite_note-raptor-16)[\[18\]](https://systems-analysis.info/int/RAG_patterns_%E2%80%94_RAG_%E6%A8%A1%E5%BC%8F#cite_note-llama-hier-18)}。 ### 常见错误（反模式） - **仅使用向量搜索**而忽略 BM25 → 在处理代码/ID/缩略词时失败^{[\[5\]](https://systems-analysis.info/int/RAG_patterns_%E2%80%94_RAG_%E6%A8%A1%E5%BC%8F#cite_note-weaviate-hybrid-5)[\[7\]](https://systems-analysis.info/int/RAG_patterns_%E2%80%94_RAG_%E6%A8%A1%E5%BC%8F#cite_note-milvus-fulltext-7)}。 - **chunk 过大或过小** → 导致上下文丢失或嵌入“模糊化”^{[\[17\]](https://systems-analysis.info/int/RAG_patterns_%E2%80%94_RAG_%E6%A8%A1%E5%BC%8F#cite_note-rcsplit-17)}。 - **生产环境中缺少重排序** → LLM 接收到充满噪声的上下文^{[\[8\]](https://systems-analysis.info/int/RAG_patterns_%E2%80%94_RAG_%E6%A8%A1%E5%BC%8F#cite_note-nogueira2019-8)}。 - **缺少可观测性**和来源追踪 → 无法分析错误原因（参见 RAG 评估）。 ## 质量评估与指标评估在检索层面（离线）和端到端生成层面进行。 ### 检索器指标 - **Hit Rate, Recall@k, MRR** — 衡量相关文档的覆盖率和排名位置。 - **Context Precision & Recall** — 衡量检索到的上下文在多大程度上不含“垃圾信息”并覆盖了所有必要内容（已在 RAGAS 中实现）^{[\[19\]](https://systems-analysis.info/int/RAG_patterns_%E2%80%94_RAG_%E6%A8%A1%E5%BC%8F#cite_note-ragas-19)}。 ### 生成器指标（端到端） - **Faithfulness / Groundedness** — 衡量答案与所提供上下文的一致性。 - **Answer Relevancy（答案相关性）** — 衡量答案与原始问题的一致性。为了自动化评估，可使用开源框架，如：**RAGAS**、**TruLens**（*RAG triad*：上下文相关性、依据性和答案相关性）、**DeepEval**^{[\[20\]](https://systems-analysis.info/int/RAG_patterns_%E2%80%94_RAG_%E6%A8%A1%E5%BC%8F#cite_note-trulens-20)[\[21\]](https://systems-analysis.info/int/RAG_patterns_%E2%80%94_RAG_%E6%A8%A1%E5%BC%8F#cite_note-deepeval-21)}。 ## 参见 - Retrieval-Augmented Generation (RAG) - 向量数据库 - 嵌入 - AI 代理 - GraphRAG - MM-RAG - LLM 评估与基准测试 ## 参考文献 - Lewis, P., Perez, E., et al. (2020). *Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks*. NeurIPS. arXiv:2005.11401. - Fan, W., Ding, Y., et al. (2024). *A Survey on RAG Meeting LLMs: Towards Retrieval-Augmented Large Language Models*. KDD. DOI:10.1145/3637528.3671470; arXiv:2405.06211. - Gao, L., Ma, X., Lin, J., Callan, J. (2023). *Precise Zero‑Shot Dense Retrieval without Relevance Labels (HyDE)*. ACL 2023. ACL Anthology; arXiv:2212.10496. - Nogueira, R., Cho, K. (2019). *Passage Re‑ranking with BERT*. arXiv:1901.04085. - Weaviate Docs. *Hybrid search (BM25+Vector)*. [1]. - Qdrant Docs. *Hybrid Queries*. [2]. - Milvus Docs. *Full‑Text Search* / *Hybrid Search*. [3] / [4]. - LangChain Docs. *MultiQueryRetriever*. [5]. - Cohere Docs. *Rerank — best practices*. [6]. - LlamaIndex Docs. *Routing (query routers/selectors)*. [7]. - Yao, S., et al. (2023). *ReAct: Synergizing Reasoning and Acting in Language Models*. ICLR. arXiv:2210.03629. - Nakano, R., et al. (2021). *WebGPT: Browser‑assisted question‑answering with human feedback*. arXiv:2112.09332. - Microsoft Research Blog. *GraphRAG: Unlocking LLM discovery on narrative private data*. (2024). [8]. - Microsoft Research. *Project GraphRAG*. (2024). [9]. - Yu, S., et al. (2024). *VisRAG: Vision‑based Retrieval‑augmented Generation on Multi‑modality Documents*. arXiv:2410.10594. - Sarthi, P., et al. (2024). *RAPTOR: Recursive Abstractive Processing for Tree‑Organized Retrieval*. arXiv:2401.18059. - Es, S., et al. (2024). *RAGAs: Automated Evaluation of Retrieval Augmented Generation*. EACL (Demo). [10]. - TruLens Docs. *RAG Triad*. [11]. - DeepEval (GitHub). *The LLM Evaluation Framework*. [12]. - LangChain Docs. *RecursiveCharacterTextSplitter*. [13]. - LlamaIndex Docs. *HierarchicalNodeParser*; *Response Synthesis (Tree/Refine)*. [14]; [15]. ## 注释 1. ↑ ^{[1.0](https://systems-analysis.info/int/RAG_patterns_%E2%80%94_RAG_%E6%A8%A1%E5%BC%8F#cite_ref-lewis2020_1-0)} ^{[1.1](https://systems-analysis.info/int/RAG_patterns_%E2%80%94_RAG_%E6%A8%A1%E5%BC%8F#cite_ref-lewis2020_1-1)} ^{[1.2](https://systems-analysis.info/int/RAG_patterns_%E2%80%94_RAG_%E6%A8%A1%E5%BC%8F#cite_ref-lewis2020_1-2)} ^{[1.3](https://systems-analysis.info/int/RAG_patterns_%E2%80%94_RAG_%E6%A8%A1%E5%BC%8F#cite_ref-lewis2020_1-3)} Lewis, P., Perez, E., et al. (2020). *Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks*. NeurIPS. arXiv:2005.11401. 2. [↑](https://systems-analysis.info/int/RAG_patterns_%E2%80%94_RAG_%E6%A8%A1%E5%BC%8F#cite_ref-survey2024_2-0) Fan, W., Ding, Y., et al. (2024). *A Survey on RAG Meeting LLMs: Towards Retrieval-Augmented Large Language Models*. KDD. DOI:10.1145/3637528.3671470; arXiv:2405.06211. 3. ↑ ^{[3.0](https://systems-analysis.info/int/RAG_patterns_%E2%80%94_RAG_%E6%A8%A1%E5%BC%8F#cite_ref-langchain-multiquery_3-0)} ^{[3.1](https://systems-analysis.info/int/RAG_patterns_%E2%80%94_RAG_%E6%A8%A1%E5%BC%8F#cite_ref-langchain-multiquery_3-1)} ^{[3.2](https://systems-analysis.info/int/RAG_patterns_%E2%80%94_RAG_%E6%A8%A1%E5%BC%8F#cite_ref-langchain-multiquery_3-2)} LangChain Docs. *MultiQueryRetriever*. https://python.langchain.com/docs/how_to/MultiQueryRetriever/ 4. ↑ ^{[4.0](https://systems-analysis.info/int/RAG_patterns_%E2%80%94_RAG_%E6%A8%A1%E5%BC%8F#cite_ref-hyde_4-0)} ^{[4.1](https://systems-analysis.info/int/RAG_patterns_%E2%80%94_RAG_%E6%A8%A1%E5%BC%8F#cite_ref-hyde_4-1)} Gao, L., Ma, X., Lin, J., Callan, J. (2023). *Precise Zero‑Shot Dense Retrieval without Relevance Labels*. ACL 2023. arXiv:2212.10496; ACL Anthology: 2023.acl‑long.99. 5. ↑ ^{[5.0](https://systems-analysis.info/int/RAG_patterns_%E2%80%94_RAG_%E6%A8%A1%E5%BC%8F#cite_ref-weaviate-hybrid_5-0)} ^{[5.1](https://systems-analysis.info/int/RAG_patterns_%E2%80%94_RAG_%E6%A8%A1%E5%BC%8F#cite_ref-weaviate-hybrid_5-1)} ^{[5.2](https://systems-analysis.info/int/RAG_patterns_%E2%80%94_RAG_%E6%A8%A1%E5%BC%8F#cite_ref-weaviate-hybrid_5-2)} ^{[5.3](https://systems-analysis.info/int/RAG_patterns_%E2%80%94_RAG_%E6%A8%A1%E5%BC%8F#cite_ref-weaviate-hybrid_5-3)} ^{[5.4](https://systems-analysis.info/int/RAG_patterns_%E2%80%94_RAG_%E6%A8%A1%E5%BC%8F#cite_ref-weaviate-hybrid_5-4)} Weaviate Docs. *Hybrid search (BM25+Vector)*. https://docs.weaviate.io/weaviate/concepts/search/hybrid-search 6. ↑ ^{[6.0](https://systems-analysis.info/int/RAG_patterns_%E2%80%94_RAG_%E6%A8%A1%E5%BC%8F#cite_ref-qdrant-hybrid_6-0)} ^{[6.1](https://systems-analysis.info/int/RAG_patterns_%E2%80%94_RAG_%E6%A8%A1%E5%BC%8F#cite_ref-qdrant-hybrid_6-1)} ^{[6.2](https://systems-analysis.info/int/RAG_patterns_%E2%80%94_RAG_%E6%A8%A1%E5%BC%8F#cite_ref-qdrant-hybrid_6-2)} Qdrant Docs. *Hybrid Queries*. https://qdrant.tech/documentation/concepts/hybrid-queries/ 7. ↑ ^{[7.0](https://systems-analysis.info/int/RAG_patterns_%E2%80%94_RAG_%E6%A8%A1%E5%BC%8F#cite_ref-milvus-fulltext_7-0)} ^{[7.1](https://systems-analysis.info/int/RAG_patterns_%E2%80%94_RAG_%E6%A8%A1%E5%BC%8F#cite_ref-milvus-fulltext_7-1)} ^{[7.2](https://systems-analysis.info/int/RAG_patterns_%E2%80%94_RAG_%E6%A8%A1%E5%BC%8F#cite_ref-milvus-fulltext_7-2)} Milvus Docs. *Full‑Text Search* 和 *Hybrid Search*. https://milvus.io/docs/full-text-search.md; https://milvus.io/docs/hybrid_search_with_milvus.md 8. ↑ ^{[8.0](https://systems-analysis.info/int/RAG_patterns_%E2%80%94_RAG_%E6%A8%A1%E5%BC%8F#cite_ref-nogueira2019_8-0)} ^{[8.1](https://systems-analysis.info/int/RAG_patterns_%E2%80%94_RAG_%E6%A8%A1%E5%BC%8F#cite_ref-nogueira2019_8-1)} ^{[8.2](https://systems-analysis.info/int/RAG_patterns_%E2%80%94_RAG_%E6%A8%A1%E5%BC%8F#cite_ref-nogueira2019_8-2)} ^{[8.3](https://systems-analysis.info/int/RAG_patterns_%E2%80%94_RAG_%E6%A8%A1%E5%BC%8F#cite_ref-nogueira2019_8-3)} ^{[8.4](https://systems-analysis.info/int/RAG_patterns_%E2%80%94_RAG_%E6%A8%A1%E5%BC%8F#cite_ref-nogueira2019_8-4)} Nogueira, R., Cho, K. (2019). *Passage Re‑ranking with BERT*. arXiv:1901.04085. 9. ↑ ^{[9.0](https://systems-analysis.info/int/RAG_patterns_%E2%80%94_RAG_%E6%A8%A1%E5%BC%8F#cite_ref-cohere-rerank_9-0)} ^{[9.1](https://systems-analysis.info/int/RAG_patterns_%E2%80%94_RAG_%E6%A8%A1%E5%BC%8F#cite_ref-cohere-rerank_9-1)} Cohere Docs. *Rerank — best practices*. https://docs.cohere.com/docs/reranking-best-practices 10. ↑ ^{[10.0](https://systems-analysis.info/int/RAG_patterns_%E2%80%94_RAG_%E6%A8%A1%E5%BC%8F#cite_ref-llama-router_10-0)} ^{[10.1](https://systems-analysis.info/int/RAG_patterns_%E2%80%94_RAG_%E6%A8%A1%E5%BC%8F#cite_ref-llama-router_10-1)} ^{[10.2](https://systems-analysis.info/int/RAG_patterns_%E2%80%94_RAG_%E6%A8%A1%E5%BC%8F#cite_ref-llama-router_10-2)} LlamaIndex Docs. *Routing (query routers/selectors)*. https://docs.llamaindex.ai/en/stable/module_guides/querying/router/ 11. ↑ ^{[11.0](https://systems-analysis.info/int/RAG_patterns_%E2%80%94_RAG_%E6%A8%A1%E5%BC%8F#cite_ref-react_11-0)} ^{[11.1](https://systems-analysis.info/int/RAG_patterns_%E2%80%94_RAG_%E6%A8%A1%E5%BC%8F#cite_ref-react_11-1)} Yao, S., et al. (2023). *ReAct: Synergizing Reasoning and Acting in Language Models*. ICLR 2023. arXiv:2210.03629. 12. ↑ ^{[12.0](https://systems-analysis.info/int/RAG_patterns_%E2%80%94_RAG_%E6%A8%A1%E5%BC%8F#cite_ref-webgpt_12-0)} ^{[12.1](https://systems-analysis.info/int/RAG_patterns_%E2%80%94_RAG_%E6%A8%A1%E5%BC%8F#cite_ref-webgpt_12-1)} Nakano, R., et al. (2021). *WebGPT: Browser‑assisted question‑answering with human feedback*. arXiv:2112.09332. 13. [↑](https://systems-analysis.info/int/RAG_patterns_%E2%80%94_RAG_%E6%A8%A1%E5%BC%8F#cite_ref-graphrag_13-0) Microsoft Research Blog. *GraphRAG: Unlocking LLM discovery on narrative private data*. 2024. https://www.microsoft.com/en-us/research/blog/graphrag-unlocking-llm-discovery-on-narrative-private-data/ 14. [↑](https://systems-analysis.info/int/RAG_patterns_%E2%80%94_RAG_%E6%A8%A1%E5%BC%8F#cite_ref-graphrag-project_14-0) Microsoft Research. *Project GraphRAG*. https://www.microsoft.com/en-us/research/project/graphrag/ 15. [↑](https://systems-analysis.info/int/RAG_patterns_%E2%80%94_RAG_%E6%A8%A1%E5%BC%8F#cite_ref-visrag_15-0) Yu, S., et al. (2024). *VisRAG: Vision‑based Retrieval‑augmented Generation on Multi‑modality Documents*. arXiv:2410.10594; OpenReview: zG459X3Xge. 16. ↑ ^{[16.0](https://systems-analysis.info/int/RAG_patterns_%E2%80%94_RAG_%E6%A8%A1%E5%BC%8F#cite_ref-raptor_16-0)} ^{[16.1](https://systems-analysis.info/int/RAG_patterns_%E2%80%94_RAG_%E6%A8%A1%E5%BC%8F#cite_ref-raptor_16-1)} Sarthi, P., et al. (2024). *RAPTOR: Recursive Abstractive Processing for Tree‑Organized Retrieval*. arXiv:2401.18059. 17. ↑ ^{[17.0](https://systems-analysis.info/int/RAG_patterns_%E2%80%94_RAG_%E6%A8%A1%E5%BC%8F#cite_ref-rcsplit_17-0)} ^{[17.1](https://systems-analysis.info/int/RAG_patterns_%E2%80%94_RAG_%E6%A8%A1%E5%BC%8F#cite_ref-rcsplit_17-1)} LangChain Docs. *RecursiveCharacterTextSplitter*. https://python.langchain.com/docs/how_to/recursive_text_splitter/ 18. ↑ ^{[18.0](https://systems-analysis.info/int/RAG_patterns_%E2%80%94_RAG_%E6%A8%A1%E5%BC%8F#cite_ref-llama-hier_18-0)} ^{[18.1](https://systems-analysis.info/int/RAG_patterns_%E2%80%94_RAG_%E6%A8%A1%E5%BC%8F#cite_ref-llama-hier_18-1)} LlamaIndex Docs. *HierarchicalNodeParser* 和 *Tree Summarization*. https://docs.llamaindex.ai/en/stable/api/llama_index.core.node_parser.HierarchicalNodeParser.html; https://docs.llamaindex.ai/en/stable/examples/low_level/response_synthesis/ 19. [↑](https://systems-analysis.info/int/RAG_patterns_%E2%80%94_RAG_%E6%A8%A1%E5%BC%8F#cite_ref-ragas_19-0) Es, S., et al. (2024). *RAGAs: Automated Evaluation of Retrieval Augmented Generation*. EACL (Demo). https://aclanthology.org/2024.eacl-demo.16/ 20. [↑](https://systems-analysis.info/int/RAG_patterns_%E2%80%94_RAG_%E6%A8%A1%E5%BC%8F#cite_ref-trulens_20-0) TruLens Docs. *RAG Triad*. https://www.trulens.org/getting_started/core_concepts/rag_triad/ 21. [↑](https://systems-analysis.info/int/RAG_patterns_%E2%80%94_RAG_%E6%A8%A1%E5%BC%8F#cite_ref-deepeval_21-0) DeepEval (GitHub). https://github.com/confident-ai/deepeval