Google

To improve the clarity and readability of prompts while reducing ambiguity, it is a common practice to format them with distinct fields. These fields are identified by labels (e.g., `SYSTEM`, `USER`, `Input:`, `Output:`) that structure the prompt and clarify the role of each piece of information for the language model.

Prompt Field Labeling for Clarity and Structure

You need a language model to act as a 'Sentiment Analyzer'. The model should take a customer review as input and output one of three sentiment labels: 'Positive', 'Negative', or 'Neutral'. Write a complete, structured prompt for this task. Your prompt must use distinct, labeled fields to clearly separate the instructions, the input data, and to guide the model towards the desired output format.

Constructing a Structured Prompt for a Classification Task

Based on the scenario described below, identify the primary reason for the language model's inconsistent performance and explain how restructuring the prompt with labeled fields would address the issue.

Diagnosing Inconsistent LLM Behavior

A developer is creating a prompt to have a language model summarize a news article. The initial prompt, shown below, often results in the model confusing the instructions with the article content itself.

Initial Prompt:
`Summarize the following text into three paragraphs. It's about a new AI chip. A new AI chip was announced today by a major tech company. It promises to be 10x faster than previous models...`

Which of the following revised prompts best uses distinct fields to resolve this ambig

Learn Before

Related