Zero-shot and Few-shot Learning

Prompting Without Training

Zero-shot vs Few-shot

One of the most powerful capabilities of LLMs is their ability to perform tasks they were never explicitly trained on, using only instructions or examples in the prompt.

Zero-shot Prompting

Give the model a task description with no examples. Works well for common tasks.

Few-shot Prompting

Provide 2-5 examples of input-output pairs before your actual query. The model learns the pattern from examples and applies it.

Best Practices

• Use consistent formatting across all examples
• Include edge cases in your examples
• Order examples from simple to complex
• Match the diversity of examples to your expected input distribution

1from openai import OpenAI
2client = OpenAI()
3
4# Zero-shot
5zero_shot = client.chat.completions.create(
6    model="gpt-4o",
7    messages=[
8        {"role": "user", "content": "Classify the sentiment: 'This product is amazing!' -> "}
9    ]
10)
11
12# Few-shot
13few_shot = client.chat.completions.create(
14    model="gpt-4o",
15    messages=[
16        {"role": "user", "content": """Classify the sentiment as positive, negative, or neutral.
17
18Text: "I love this!" -> positive
19Text: "Terrible experience." -> negative  
20Text: "It's okay." -> neutral
21Text: "Best purchase I've ever made!" -> """}
22    ]
23)