Diagnosing Prompt Insufficiency
A developer is using a few-shot prompt to teach a large language model to solve geometry word problems. Each example in the prompt provides a problem description and the final calculated area (e.g., 'Problem: A rectangle has a length of 10cm and a width of 5cm. What is its area? Answer: 50 sq cm'). The model performs well on problems about rectangles but fails on problems about triangles. Based on the structure of the prompt examples, analyze the fundamental reason for the model's inability to generalize to a different shape.
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
LLM's Answer (7) to the Prompt for Calculating the Average of 2, 4, and 9
Creating a CoT Prompt by Incorporating Reasoning Steps
Diagnosing a Flawed Prompting Strategy
A developer is trying to get a large language model to solve two-step arithmetic word problems. They use a few-shot prompting strategy, providing several examples. Each example consists of a word problem followed only by its final numerical answer (e.g., 'Problem: ... Answer: 15'). The model consistently fails to solve new, slightly different word problems. What is the most likely reason for the model's poor performance?
Diagnosing Prompt Insufficiency