Analyzing a Flawed Multi-Step Prompt
A user is solving a multi-step problem. In the first step, they ask a language model: 'What is the capital of France?' and receive the answer: 'Paris.' For the second step, they immediately ask a new, separate question: 'What is its population?' The model provides a generic or incorrect response. Explain the fundamental reason for the model's failure in the second step and describe how the user should have structured the second prompt to ensure a correct and relevant answer.
0
1
Tags
Ch.3 Prompting - Foundations of Large Language Models
Foundations of Large Language Models
Foundations of Large Language Models Course
Computing Sciences
Analysis in Bloom's Taxonomy
Cognitive Psychology
Psychology
Social Science
Empirical Science
Science
Related
Example of Sequential Sub-Problem Solving: Environmental Study Duration
A problem is being solved in sequential steps. The overall goal is to 'Calculate the total profit from selling 200 widgets.' The first step was to ask, 'What is the revenue from selling 200 widgets if each widget sells for $15?', to which a model answered, 'The revenue is $3000.' Now, to solve the next part of the problem, you need to ask the model, 'What is the total cost of producing 200 widgets if the production cost per widget is $8?'. Which of the following is the most effective and complete prompt to send to the model for this second step?
Constructing a Contextual Prompt for a Multi-Step Problem
Analyzing a Flawed Multi-Step Prompt