Zero-, one-, and few-shot learning: Worked Example — Prompt Engineering (NVIDIA-Certified Professional: Generative AI LLMs)

Understanding Zero-, One-, and Few-Shot Learning in Prompt Engineering

Zero-, one-, and few-shot learning are critical techniques in prompt engineering that enable large language models (LLMs) to perform tasks with minimal or no task-specific training data. These approaches leverage the model's pre-trained knowledge and adapt it dynamically through carefully crafted prompts.

Definitions

• Zero-shot learning: The model performs a task without any examples provided in the prompt.
• One-shot learning: The model is given exactly one example in the prompt to guide its response.
• Few-shot learning: The model is provided with a small number of examples (typically 2-5) in the prompt to illustrate the task.

Worked Example: Classifying Customer Feedback Sentiment

Suppose you are designing a prompt to classify customer feedback as positive, neutral, or negative. You want to test zero-, one-, and few-shot prompting techniques with an LLM certified under the NVIDIA-Certified Professional: Generative AI LLMs program.

Step 1: Zero-Shot Prompting

In zero-shot prompting, you provide the task description without any examples.

Prompt:

Classify the sentiment of the following customer feedback as positive, neutral, or negative: "The delivery was late and the package was damaged." Sentiment:

Model Reasoning: The feedback mentions negative experiences (late delivery, damaged package), so the sentiment is likely negative.

Expected Output: Negative

Step 2: One-Shot Prompting

Here, you provide one example to demonstrate the task format.

Prompt:

Classify the sentiment of the following customer feedback as positive, neutral, or negative.Example:"The product quality is excellent and I am very happy." Sentiment: PositiveNow classify:"The delivery was late and the package was damaged." Sentiment:

Model Reasoning: The example shows how to label positive feedback. The new feedback describes negative experiences, so the model should label it negative.

Expected Output: Negative

Step 3: Few-Shot Prompting

In few-shot prompting, multiple examples illustrate the task more comprehensively.

Prompt:

Classify the sentiment of the following customer feedback as positive, neutral, or negative.Examples:"The product quality is excellent and I am very happy." Sentiment: Positive"The delivery was on time but the packaging was average." Sentiment: Neutral"The customer service was unhelpful and rude." Sentiment: NegativeNow classify:"The delivery was late and the package was damaged." Sentiment:

Model Reasoning: The examples cover all sentiment classes, helping the model to better understand subtle differences. The new feedback clearly matches the negative sentiment example.

Expected Output: Negative

Summary of Steps and Benefits

• Zero-shot: Quick and requires no examples, but may be less accurate.
• One-shot: Provides a single example to clarify the task format, improving accuracy.
• Few-shot: Multiple examples help the model generalize better, often yielding the best performance.

By mastering these prompting techniques, NVIDIA-Certified Professionals can effectively control and optimize LLM outputs for diverse applications.

More in this topic

Zero-, one-, and few-shot learning: Quick Reference — Prompt Engineering (NVIDIA-Certified Professional: Generative AI LLMs)Prompt Engineering — NVIDIA-Certified Professional: Generative AI LLMsZero-, one-, and few-shot learning: Common Mistakes — Prompt Engineering (NVIDIA-Certified Professional: Generative AI LLMs)Controlling model output — Prompt Engineering (NVIDIA-Certified Professional: Generative AI LLMs)Chain-of-thought prompting — Prompt Engineering (NVIDIA-Certified Professional: Generative AI LLMs)Domain adaptation via prompting — Prompt Engineering (NVIDIA-Certified Professional: Generative AI LLMs)Zero-, one-, and few-shot learning — Prompt Engineering (NVIDIA-Certified Professional: Generative AI LLMs)Zero-, one-, and few-shot learning: Practice Questions — Prompt Engineering (NVIDIA-Certified Professional: Generative AI LLMs)