Chain of Thought
Master transparent, step-by-step reasoning for more accurate and explainable AI agents
Your Progress
0 / 5 completedCommon CoT Patterns
There are several established patterns for implementing chain-of-thought reasoning, each with different trade-offs. The right approach depends on your problem complexity, token budget, and accuracy requirements.
Interactive: Compare Prompting Patterns
Select different patterns to see how they structure reasoning differently
Zero-Shot CoT
Simply add "Let's think step by step" to your prompt
Example Prompt:
Problem: What is 15% of 240? Let's think step by step:
AI Response:
Step 1: Convert percentage to decimal: 15% = 0.15 Step 2: Multiply: 0.15 × 240 Step 3: Calculate: 0.15 × 240 = 36 Answer: 36
✓ Advantages:
- •Simple to implement
- •No examples needed
- •Works on many problems
- •Low token cost
⚠ Limitations:
- •Less guidance
- •Quality varies
- •May skip important steps
Decision Guide: Which Pattern to Use?
🚀
Use Zero-Shot CoT When:
- • Quick prototyping or testing
- • Simple, well-defined problems
- • Token budget is limited
- • You don't have example chains to provide
🎯
Use Few-Shot CoT When:
- • Accuracy is critical
- • You have high-quality example chains
- • Problem domain is specific/technical
- • You need consistent formatting
🧩
Use Least-to-Most When:
- • Problem is very complex/multi-step
- • Subproblems can be solved independently
- • Building solutions incrementally makes sense
- • You need to parallelize sub-tasks
Advanced Techniques
🔄 Self-Consistency
Generate multiple reasoning chains for the same problem, then pick the most common answer. Dramatically improves accuracy (but uses more tokens).
🔍 Chain Verification
After generating a chain, ask the model to verify each step for logical consistency and catch errors.
🌳 Tree of Thoughts
Explore multiple reasoning paths simultaneously (like a search tree), then select the best branch. Great for creative or open-ended problems.