Week 2: Advanced Reasoning & Planning
This week focuses on reasoning patterns that make agents smarter and more reliable. You'll learn how to make LLMs think step-by-step, critique their own outputs, and correct mistakes automatically.
Learning Objectives
By the end of this week, you will:
- ✅ Understand Chain of Thought (CoT) prompting for step-by-step reasoning
- ✅ Build Plan-and-Execute agents that plan before acting
- ✅ Implement Reflection patterns for self-improvement
- ✅ Use Tree of Thoughts for exploring multiple reasoning paths
- ✅ Create Self-RAG systems that detect and fix their own errors
Sessions
01. Planning Agents & CoT02. Reflection Agents04. Structured Reasoning05. Advanced Self-RAGWeekend Project
Week Overview
Key Concepts
| Concept | Description | When to Use |
|---|---|---|
| Chain of Thought | "Think step by step" | Complex reasoning tasks |
| Plan-and-Execute | Plan first, then act | Multi-step problems |
| Reflection | Generate, critique, refine | Quality improvement |
| Tree of Thoughts | Explore multiple paths | Creative/strategic tasks |
| Self-RAG | Verify and correct | RAG reliability |
The Big Insight
Smaller models + reasoning patterns can outperform larger models!
A GPT-4o-mini agent with self-correction often beats a single GPT-4o call—and costs 10× less.
Architecture Preview
Prerequisites
Before starting Week 2, make sure you've completed:
- Week 1: Agent Foundations (ReAct, Tool Use)
- Basic understanding of LLM API calls
Notebooks
| Session | English | Korean |
|---|---|---|
| 01. Planning Agents | 📘 Notebook (opens in a new tab) | 📘 노트북 (opens in a new tab) |
| 02. Reflection Agents | 📘 Notebook (opens in a new tab) | 📘 노트북 (opens in a new tab) |
| 03. Structured Reasoning | 📘 Notebook (opens in a new tab) | 📘 노트북 (opens in a new tab) |
| 04. LangGraph Planning | 📘 Notebook (opens in a new tab) | 📘 노트북 (opens in a new tab) |
| 05. Self-RAG | 📘 Notebook (opens in a new tab) | 📘 노트북 (opens in a new tab) |
| Weekend Project | 📘 Project Page | 📘 프로젝트 페이지 |