Prompt Engineering : Zero-Shot vs Few-Shot Prompting

This image was generated using FLUX.1 schnell running locally on M3 Pro, including the text. Device: 'cpu_core_count': 12, 'gpu_core_count': 18, , Total time: 72.083s Peak memory: 19.0GB

Introduction

Getting prompts just right achieving optimal result from the llm's model is a crucial and iterative process. While we mostly do write in simple english, the prompts can be a bit tricky to get right. In this article we will discuss the difference between zero-shot and few-shot prompting and how to use them effectively.

Zero-shot Prompting

Zero-shot prompting is when you provide a prompt to the model and expect it to generate the output without any additional examples. This is the simplest form of prompting and is used when you want to generate a response to a prompt without providing any additional context. No examples are provided to the model hence the name zero-shot.

Zero-shot Learning Example

Task: Translate English sentences to French.

Prompt:

Translate the following sentence to French:
"Where is the nearest restaurant?"

Model Output:

"Où est le restaurant le plus proche?"

Few-shot Prompting

In this style of prompting we provide a few examples to the model to help it understand the task better. This is useful when you want to generate a response to a prompt that requires some context. Few-shot prompting is more powerful than zero-shot prompting as it provides the model with some examples to learn from.

Few-shot Learning Example

Task: Translate English sentences to French.

Prompt:

Translate the following sentences to French:
1. "What time is it?" -> "Quelle heure est-il?"
2. "Can you help me?" -> "Pouvez-vous m'aider?"
3. "Where is the nearest restaurant?" -> 

Model Output:

"Où est le restaurant le plus proche?"

Caution: Few-shot prompting can be tempting and is the right choice in most cases but it can also lead to overfitting. If we give the model too many examples, it might start memorizing the examples and not generalize well to new examples.

In conclusion, zero-shot prompting is useful when you want to generate a response to a prompt without providing any additional context. Few-shot prompting is more powerful as it provides the model with some examples to learn from. It is important to strike the right balance between zero-shot and few-shot prompting to get the best results from the model.