Choosing an In-Context Learning Strategy
A startup is developing a feature to summarize customer support tickets using a large, general-purpose language model. They have a very limited budget, no time for model fine-tuning, and only a small, unannotated dataset of 50 past tickets. Their immediate goal is to create a functional prototype as quickly as possible to demonstrate the concept.
Which in-context learning approach (Zero-shot, One-shot, or Few-shot) would be the most effective to start with for this initial prototype? Justify your choice by explaining its primary advantage given the startup's specific constraints.
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
Evaluation in Bloom's Taxonomy
Cognitive Psychology
Psychology
Social Science
Empirical Science
Science
Related
A user provides the following input to a large language model to get it to classify the sentiment of movie reviews. The model has not been specifically trained or fine-tuned on this particular task.
Classify the sentiment of the movie review. Review: "The plot was predictable and the acting was wooden." Sentiment: Negative Review: "An absolute masterpiece of modern cinema!" Sentiment:Which in-context learning approach does this prompt demonstrate?
Choosing an In-Context Learning Strategy
Analyzing Trade-offs in In-Context Learning Strategies