Learn Before
Critique and Refine a Few-Shot Prompt
A developer is creating a few-shot prompt to classify customer reviews as 'Positive', 'Negative', or 'Neutral'. They find the model's performance is unreliable. Below are two examples from their prompt.
Example 1: Review: "The product arrived on time and exceeded my expectations. Fantastic quality!" Classification: After analyzing the review, the resulting sentiment is Positive.
Example 2: Review: "The item broke after just one use. Very disappointed with the purchase." Classification: After analyzing the review, the resulting sentiment is Negative.
Critique the structure of the input-output pattern used in these examples. Explain why this structure might be suboptimal for guiding the model, and propose a revised, simpler structure that would likely improve performance.
0
1
Tags
Ch.3 Prompting - Foundations of Large Language Models
Foundations of Large Language Models
Computing Sciences
Foundations of Large Language Models Course
Evaluation in Bloom's Taxonomy
Cognitive Psychology
Psychology
Social Science
Empirical Science
Science
Related
Example of a Few-Shot Prompt for Chinese-to-English Translation
A developer is crafting a prompt to have a large language model extract the primary color from a product description. Below are two approaches for the examples (demonstrations) within the prompt:
Approach A: Text: "This t-shirt comes in a vibrant shade of royal blue." Color: The main color is blue. Text: "The car's exterior is painted a deep, glossy black." Color: The main color is black.
Approach B: Text: "This t-shirt comes in a vibrant shade of royal blue." -> blue Text: "The car's exterior is painted a deep, glossy black." -> black
Based on common principles for guiding language models, which approach is generally more effective and why?
Critique and Refine a Few-Shot Prompt
Optimizing a Sentiment Analysis Prompt