Prompting Techniques

Zero-Shot Prompting

No examples given; AI infers from instructions.

Example: "Classify this text as positive, negative, or neutral: 'I love this sunny day!'"

Expected Output: Positive.

One-Shot Prompting

Provide one example for guidance.

Example: "Convert to Roman numerals: 5 = V. Now: 10."

Expected Output: X.

Few-Shot Prompting

Give 2-5 examples to show patterns.

Example: "Q: What is the capital of France? A: Paris. Q: Capital of Japan? A: Tokyo. Q: Capital of Brazil?"

Expected Output: Brasília.

Chain-of-Thought (CoT)

Encourage step-by-step reasoning. Add "Think step by step."

Example: "Solve: If a bat and ball cost $1.10 total, and the bat costs $1 more than the ball, how much is the ball? Think step by step."

Expected Output: Ball = $0.05 (steps: Let ball = x; bat = x + 1; x + (x + 1) = 1.10 → 2x + 1 = 1.10 → x = 0.05).

Role-Playing

Assign a character for tailored responses.

Example: "You are a pirate captain. Explain photosynthesis in pirate lingo."

Expected Output: "Arrr, photosynthesis be like plants eatin' sunlight fer grub, turnin' it into sugar with their green sails!"

Self-Consistency

Generate multiple responses and pick the most common (run 3x, vote).

Example: "What is 15% of 200? Reason step by step. [Repeat 3 times and select majority answer.]"

Expected Output: 30 (consistent across runs).

Tree of Thoughts (ToT)

Explore branches of reasoning like a decision tree.

Example: "Plan a trip to Paris. Branch 1: Budget option. Branch 2: Luxury. For each, list 3 pros/cons and pick best."

Expected Output: Structured branches with reasoned choice.

Advanced Tips

• Use delimiters: Wrap inputs in ``` or quotes.

• Specify output format: "Respond in a table: Column1=Name, Column2=Age."

• Control length: "In 100 words max."

• Avoid bias: "Base on facts only."

• Pro Tip: Test variations—A/B your prompts.