---
title: "Lilian Weng - 信源与标准 - AI·外脑 | 思行致远"
canonical: "https://siyuanailab.com/knowledge/sources/lilian-weng/"
dateModified: "2026-07-07"
datePublished: "2026-07-07"
---

打开 AI·外脑目录  

 

AI·外脑

 [返回专栏首页](https://siyuanailab.com/knowledge/)

 ×

 

站内搜索 

⌕

   

主要栏目

 

- [AI 核心概念](https://siyuanailab.com/knowledge/ai-core/)

- [工作流](https://siyuanailab.com/knowledge/workflows/)

- [工具配置与选型](https://siyuanailab.com/knowledge/toolkits/)

- [信源与标准](https://siyuanailab.com/knowledge/sources/)

- [场景案例 待开发](https://siyuanailab.com/knowledge/scenarios/)

- [职业与能力](https://siyuanailab.com/knowledge/career/)

 

专题

 

- [从这里开始](https://siyuanailab.com/knowledge/start/)

- [获取与安装](https://siyuanailab.com/knowledge/install/)

- [建设日志](https://siyuanailab.com/knowledge/ops/)

  

[信源与标准子目录](https://siyuanailab.com/knowledge/sources/) 

信源分级与框架（1）

 

- [AI 学习信源：如何分级与使用](https://siyuanailab.com/knowledge/sources/ai-source-principle/)

官方文档与产品更新（6）

 

- [Google AI 官方动态](https://siyuanailab.com/knowledge/sources/google-ai-official-updates/)

- [MCP 官方文档](https://siyuanailab.com/knowledge/sources/mcp-official-docs/)

- [Midjourney 官方文档](https://siyuanailab.com/knowledge/sources/midjourney-official-docs/)

- [OpenAI 官方文档](https://siyuanailab.com/knowledge/sources/openai-official-docs/)

- [Anthropic 官方文档](https://siyuanailab.com/knowledge/sources/anthropic-official-docs/)

- [AI 官方文档与产品更新](https://siyuanailab.com/knowledge/sources/ai-official-docs-directory/)

研究论文与预印本（6）

 

- [Anthropic Research](https://siyuanailab.com/knowledge/sources/anthropic-research/)

- [arXiv](https://siyuanailab.com/knowledge/sources/arxiv/)

- [Google Research Blog](https://siyuanailab.com/knowledge/sources/google-research-blog/)

- [Lilian Weng](https://siyuanailab.com/knowledge/sources/lilian-weng/)

- [OpenAI Research](https://siyuanailab.com/knowledge/sources/openai-research/)

- [AI 研究论文与预印本](https://siyuanailab.com/knowledge/sources/ai-research-papers-directory/)

权威媒体与机构洞察（7）

 

- [a16z AI](https://siyuanailab.com/knowledge/sources/a16z-ai/)

- [MIT Technology Review AI](https://siyuanailab.com/knowledge/sources/mit-technology-review-ai/)

- [Sequoia AI](https://siyuanailab.com/knowledge/sources/sequoia-ai/)

- [TechCrunch AI](https://siyuanailab.com/knowledge/sources/techcrunch-ai/)

- [The Verge AI](https://siyuanailab.com/knowledge/sources/the-verge-ai/)

- [Y Combinator Blog](https://siyuanailab.com/knowledge/sources/y-combinator-blog/)

- [AI 权威媒体与机构洞察](https://siyuanailab.com/knowledge/sources/ai-media-insights-directory/)

深度播客与视频（6）

 

- [AI & I](https://siyuanailab.com/knowledge/sources/ai-and-i/)

- [Andrej Karpathy（YouTube）](https://siyuanailab.com/knowledge/sources/andrej-karpathy-youtube/)

- [Google DeepMind Podcast](https://siyuanailab.com/knowledge/sources/google-deepmind-podcast/)

- [Latent Space](https://siyuanailab.com/knowledge/sources/latent-space/)

- [Lenny's Podcast](https://siyuanailab.com/knowledge/sources/lennys-podcast/)

- [AI 深度播客与视频](https://siyuanailab.com/knowledge/sources/ai-podcasts-videos-directory/)

Newsletter 与策展者（7）

 

- [AI Valley](https://siyuanailab.com/knowledge/sources/ai-valley/)

- [AINews by smol.ai](https://siyuanailab.com/knowledge/sources/ainews-by-smol-ai/)

- [Ben's Bites](https://siyuanailab.com/knowledge/sources/bens-bites/)

- [Every](https://siyuanailab.com/knowledge/sources/every/)

- [Peter Yang (Behind the Craft)](https://siyuanailab.com/knowledge/sources/peter-yang/)

- [Zara's AI learning library](https://siyuanailab.com/knowledge/sources/zaras-ai-learning-library/)

- [AI Newsletter 与策展者](https://siyuanailab.com/knowledge/sources/ai-newsletters-curators-directory/)

社交媒体账号（6）

 

- [Andrej Karpathy（X）](https://siyuanailab.com/knowledge/sources/andrej-karpathy-x/)

- [Greg Isenberg（X）](https://siyuanailab.com/knowledge/sources/greg-isenberg-x/)

- [Justine Moore（X）](https://siyuanailab.com/knowledge/sources/justine-moore-x/)

- [Lenny Rachitsky（X）](https://siyuanailab.com/knowledge/sources/lenny-rachitsky-x/)

- [Swyx（X）](https://siyuanailab.com/knowledge/sources/swyx/)

- [AI 社交媒体账号](https://siyuanailab.com/knowledge/sources/ai-social-accounts-directory/)

  

信源与标准 · 入门

 

# Lilian Weng

 

杨思远 / 思行致远 · 2026/7/7 · 置信度：高

 

Lilian-Weng博客强化学习提示词工程AI安全

 

OpenAI 研究员 Lilian Weng 的个人博客，以高质量长文把 RLHF、Agent、安全等复杂主题讲清楚。

 

# Lilian Weng

 

> OpenAI 研究员 Lilian Weng 的个人博客，以高质量长文把 RLHF、Agent、安全等复杂主题讲清楚。

 

---

 

## 一、是什么

 

Lilian Weng 的博客（lilianweng.github.io）是 AI 领域公认的高质量中文/英文技术博客之一。作者现任 OpenAI 研究员，文章覆盖强化学习、大语言模型、AI Agent、奖励建模、AI 安全等主题。

 

博客特点：

 

- **长文深度解析**：单篇常过万词，系统梳理一个主题。
 
- **引用丰富**：每篇文章附大量论文链接，可直接延伸阅读。
 
- **表达清晰**：把研究概念转化为工程和产品从业者也能理解的叙述。

 

## 二、为什么存在

 

学术论文分散且门槛高，Lilian Weng 的博客充当「研究转译器」：把多个相关论文的核心思想整合成一篇可读性强的综述，帮助读者快速建立对一个子领域的认知框架。

 

## 三、核心机制

 

### 3.1 内容生产方式

 

- 个人撰写，更新频率不高，但每篇质量稳定。
 
- 主题选择紧跟 AI 研究前沿和工程实践热点。

 

### 3.2 典型主题

 

- LLM Powered Autonomous Agents
 
- Reward Hacking in Reinforcement Learning
 
- Prompt Engineering
 
- Reward Modeling / RLHF

 

### 3.3 使用路径

 

- **入门一个复杂主题** → 先读她的长文，再追文中引用。
 
- **准备技术分享或文章** → 用她的结构作为大纲参考。

 

## 四、常见误解

 

| 误解 | 真相 |
| --- | --- |
| 这是 OpenAI 官方文档 | 是个人博客，不代表 OpenAI 官方立场 |
| 更新很频繁 | 更新节奏慢，可能数月一篇 |
| 只适合研究者 | 工程师、产品经理、内容创作者都能从中受益 |
| 内容永远最新 | 研究领域进展快，部分细节需结合最新论文更新 |

 

## 五、怎么用

 

### Level 0：收藏博客

 

把 [https://lilianweng.github.io/](https://lilianweng.github.io/) 加入书签。

 

### Level 1：按主题阅读

 

遇到 RLHF、Agent、安全等概念时，先搜她的博客看是否有相关长文。

 

### Level 2：提取引用清单

 

把文章末尾的 References 当作该主题的必读论文列表。

 

### Level 3：与官方文档对照

 

她的解读是很好的起点，但产品功能和参数应以官方文档为准。

 

## 六、延伸阅读

 

- 目录卡：[AI研究论文与预印本](https://siyuanailab.com/knowledge/sources/ai-research-papers-directory/)
 
- 关联概念：`[[提示词工程]]`

 

### 代表作品/入口

 

| 作品 | 说明 | 链接 |
| --- | --- | --- |
| LLM Powered Autonomous Agents | LLM Agent 系统综述 | [https://lilianweng.github.io/posts/2023-06-23-agent/](https://lilianweng.github.io/posts/2023-06-23-agent/) |
| Reward Hacking in Reinforcement Learning | 强化学习中的奖励 hacking 问题 | [https://lilianweng.github.io/posts/2024-11-28-reward-hacking/](https://lilianweng.github.io/posts/2024-11-28-reward-hacking/) |
| Prompt Engineering | 提示工程实践指南 | [https://lilianweng.github.io/posts/2023-03-15-prompt-engineering/](https://lilianweng.github.io/posts/2023-03-15-prompt-engineering/) |
| 博客首页 | 全部文章索引 | [https://lilianweng.github.io/](https://lilianweng.github.io/) |

 

### 适用场景与行动

 

**场景：**需要判断「Lilian Weng」是否适合作为 AI 学习、研究或行业观察信源时。

 

**适合谁：**AI 学习者、研究跟踪者、需要追溯一手材料的产品与内容工作者。

 

**下一步：**把 Lilian Weng 加入或排除出自己的 AI 信源清单，并在关键判断前交叉验证。

 

### 相关卡片

 

- ai-research-papers-directory

 

这张卡片由杨思远基于真实工作流和实测经验整理。

 

如发现信息过时或有误，可通过 [/contact](https://siyuanailab.com/contact) 反馈。

  

[← 上一张：Lenny's Podcast](https://siyuanailab.com/knowledge/sources/lennys-podcast/) [下一张：OpenAI Research →](https://siyuanailab.com/knowledge/sources/openai-research/)

 [返回 AI·外脑入口](https://siyuanailab.com/knowledge/)