Improving Model Guidance for Multi-Step Problems
You are trying to guide a language model to solve multi-step word problems correctly. You provide it with an example to learn from, but find that the model still makes mistakes on new problems. Your task is to improve the example by breaking down the solution into a clear, step-by-step thought process. Given the initial example in the case study below, rewrite the 'Answer' portion to explicitly show the intermediate logical steps needed to reach the final solution.
0
1
Tags
Ch.2 Generative Models - Foundations of Large Language Models
Foundations of Large Language Models
Foundations of Large Language Models Course
Computing Sciences
Application in Bloom's Taxonomy
Cognitive Psychology
Psychology
Social Science
Empirical Science
Science
Related
One-Shot Chain-of-Thought (COT) Prompting
Improving Model Guidance for Multi-Step Problems
Example of a Multi-Step Arithmetic Word Problem (Tom's Marbles)
You are creating a demonstration to teach a large language model how to solve multi-step arithmetic word problems. The goal is to show the model a clear, step-by-step reasoning process that is easy to follow.
Problem: A library starts the day with 80 books on the 'new arrivals' shelf. During the day, 45 books are checked out. Later, librarians add a new shipment of 22 books to the shelf. How many books are on the shelf at the end of the day?
Which of the following demonstrations is structured most effectively to guide the model?
You are preparing a demonstration to show a language model how to solve a word problem. Arrange the following reasoning steps into the most logical and clear sequence to arrive at the correct answer for the problem below.
Problem: A bakery starts with 20 bags of flour. They use 5 bags for bread and 3 bags for cakes. Later in the day, they receive a delivery of 10 more bags. How many bags of flour does the bakery have at the end of the day?