Agent¶

Reference · 速查

RAG 论文¶

Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks (2020, paper - Lewis et al. / Meta) —— RAG 概念奠基论文。
Self-RAG: Learning to Retrieve, Generate, and Critique through Self-Reflection (2023, paper) —— LLM 自决策何时检索 + 自我评估。
Corrective RAG (CRAG) (2024, paper) —— 检索后纠正的范式。
Anthropic - Contextual Retrieval (2024, blog) —— Contextual chunk pre-prompting，工业可复现 35% 提升。
Lost in the Middle: How Language Models Use Long Contexts (2023, paper - Stanford) —— LLM 长上下文中间内容易失，RAG chunk 顺序设计的依据。

ReAct: Synergizing Reasoning and Acting in Language Models (2022, paper) —— Agent 推理-行动循环奠基。
Toolformer: Language Models Can Teach Themselves to Use Tools (2023, paper) —— LLM 工具使用。
Anthropic - Building Effective Agents (2024, blog) —— 5 类 Agent 模式 + 何时用 workflow vs agent。工业建议必读。
Model Context Protocol (MCP) Specification (2024+, official-doc - Anthropic) —— Anthropic 提出的 LLM-工具开放协议。
Anthropic Claude Engineering Blog (blog) —— Constitutional AI / Claude API / Agent SDK 等深度内容。

LangChain Documentation (official-doc) —— Models / Agents / Tools / LangGraph / Deep Agents 等模块。
LlamaIndex Documentation (official-doc) —— Component Guides 含 Indexing/Loading/Storing/Querying/Evaluating/Observability/MCP；5 workflow 模式 (Agents/Workflows/Structured Data/Query/Chat)。与本 wiki ai-workloads/ 的"应用 / Runtime 桥 / 工程纪律"三层组织部分对齐。
OpenAI Cookbook (official-doc) —— 实践代码模式。

RAGAS: Automated Evaluation of Retrieval Augmented Generation (2023, paper) —— RAG 评估指标 (faithfulness / answer relevancy / context precision / recall)。
Evaluating LLM Applications with TruLens (official-doc) —— TruLens RAG triad evaluation。
OpenAI Evals (official-doc) —— 评估框架。

A Survey on RAG Meeting LLMs: Towards Retrieval-Augmented Large Language Models (2024, survey) —— 工业综述。
The Rise and Potential of Large Language Model Based Agents (2023, survey) —— Agent 系统综述。

待补：2025-2026 RAG / Agent 演进；vLLM Agent serving；Agent Memory / Long-term context 最新研究