Diagnosing and Correcting a Reasoning Error
A user gives a language model a complex, multi-step logic puzzle. The model provides a single, definitive answer, which turns out to be incorrect. Based on common failure modes for this type of task, explain the most likely reason for the model's error and describe a specific change to the user's instructions that would increase the probability of getting a correct answer.
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
Ch.3 Prompting - Foundations of Large Language Models
Analysis in Bloom's Taxonomy
Cognitive Psychology
Psychology
Social Science
Empirical Science
Science
Related
Analyzing Prompt Effectiveness for Logical Problems
A user wants a Large Language Model to solve the following word problem: 'A bakery starts the day with 480 cookies. They sell 150 in the morning and 95 in the afternoon. They then bake a new batch of 120 cookies. How many cookies do they have at the end of the day?' The user's initial prompt is simply the word problem itself. Which of the following revised prompts is most likely to guide the model to a correct and reliable answer?
Example of an Enhanced Role-Playing Prompt for Mathematical Reasoning
Diagnosing and Correcting a Reasoning Error