What is Few-Shot Prompting?
Few-shot prompting involves providing AI with a few examples of the desired input-output pattern before asking it to perform the task. This technique helps ensure consistent formatting and style.
Why Use Few-Shot?
- ✅ Ensures consistent output format
- ✅ Shows exact style and tone
- ✅ Reduces ambiguity
- ✅ Improves first-try accuracy
Basic Few-Shot Example
Product Description Format
Create product descriptions following this format: **Product**: Wireless Headphones **Description**: Premium over-ear headphones with noise cancellation **Price**: $199 **Key Features**: 30-hour battery, quick charge, premium sound **Product**: Smart Watch **Description**: Fitness-focused smartwatch with health monitoring **Price**: $299 **Key Features**: Heart rate monitor, GPS, waterproof Now create a description for: Bluetooth SpeakerBest Practices
Example Quality
- • Use 2-5 high-quality examples
- • Show variety within consistency
- • Include edge cases if relevant
- • Make patterns clear
Format Guidelines
- • Keep same structure across examples
- • Clearly separate examples
- • Mark where examples end
- • Be explicit about the task
Common Use Cases
- Data formatting: Converting data to specific structures
- Content creation: Maintaining consistent style and tone
- Classification: Categorizing items consistently
- Email templates: Professional communication patterns
- Code generation: Following specific coding patterns
Email Classification Example
Classify emails as URGENT, IMPORTANT, or ROUTINE: Email: "Server down - all customers affected" Classification: URGENT Reason: Service outage impacting customers Email: "Quarterly review meeting next week" Classification: IMPORTANT Reason: Scheduled business meeting Email: "Office coffee machine maintenance" Classification: ROUTINE Reason: Regular facility maintenance Now classify: "Security breach detected"Tips for Success
- Quality over quantity: 3 great examples beat 10 poor ones
- Show edge cases: Include unusual but valid examples
- Be consistent: Follow the same pattern in all examples
- Clear boundaries: Make it obvious where examples end