Technique 0New

Zero-Shot Prompting

No examples needed

Zero-Shot Prompting — No Examples

Zero-shot is when you give the model a task without any examples. The model uses only its pre-trained knowledge. This is the simplest, but not always the most reliable approach.

Task Examples:
Zero-shot Prompt:
Determine the sentiment of this review:

"Great product, fast delivery, highly recommend!"

Sentiment:
Simple classification task — model understands without examples
Model Response:
Click "Show"

Zero-shot Prompt Patterns:

Direct Question
[Question]?
What is machine learning?
Classification
Determine [what] in text: "[text]"
Determine the language: "Bonjour monde"
Transformation
[Action] the text: "[text]"
Fix errors in text: "Helo wrold"
Generation
Write [what] about [topic]
Write a slogan for a coffee shop
When to Use Zero-shot
  • • Simple classification tasks
  • • Translations
  • • Basic questions
  • • When format is not critical
When to Use Other Techniques
  • • Multi-step reasoning → CoT
  • • Need exact format → Few-shot
  • • Complex calculations → PoT
  • • Need accuracy → Self-Consistency
Key Insight

Zero-shot is the starting point. Always begin with a simple prompt and add complexity only when needed: add examples (Few-shot), ask for step-by-step thinking (CoT), or use more advanced techniques.