Google

One-shot learning is a technique that enhances a prompt by adding a single, complete example of a task, known as a demonstration. For instance, in grammar correction, this would involve showing how a sentence is corrected. This added demonstration allows the Large Language Model to learn from the provided experience and apply that learning to new inputs.

One-Shot Learning

A one-shot prompt for grammar correction enhances a basic prompt by including a single, complete example of the task, known as a demonstration. This allows the language model to learn from the provided context. The structure typically involves defining the model's role and task, presenting one correct input-output pair, and then providing the new input for the model to process. For example:

**DEMO**
*Input:* There is many reasons to celebrate.
*Output:* There are many reasons to celebrate.

**USER**
*Input:* She don’t like going to the park.
*Output:*

Example of a One-Shot Prompt for Grammar Correction

A developer is creating a prompt to instruct a language model to summarize a block of text into a single, concise sentence. Which of the following prompt structures best demonstrates the one-shot learning technique for this task?

Based on the scenario, describe how you would modify the developer's prompt by applying a specific prompting technique that involves providing one complete example. Explain why this modification would likely lead to more consistent and correctly formatted outputs from the model.

Improving Prompt Reliability for Date Formatting

A developer is trying to build a prompt that instructs a language model to extract a `product_name` and `sentiment` from user feedback. They have constructed the following prompt, which includes a single demonstration before the final user request:

---
**PROMPT START**

Extract the product name and the user's sentiment from the following text.

**Demonstration:**
*Input:* "The new Lumina-5 camera is absolutely fantastic! I love the picture quality."
*Output:* {"product_name": "Lumina-5 camera"}

**User Request:**
*Input:* "I'm very disappointed with the performance of the Aero-Drone. It stopped working after one flight."
*Output:*

**PROMPT END**
---

Analyze the provided prompt. Explain why the single demonstration is poorly constructed for the stated task and how this flaw is likely to cause the model to produce an incomplete or incorrect output for the 'User Request' section.

Learn Before

Related