Evaluating the Source of Zero-Shot Capabilities
A colleague argues that a large language model's ability to perform a completely new task (e.g., 'Translate this English sentence into pirate slang') without specific training is simply because it has memorized a vast number of similar translation examples. Evaluate this argument. In your response, explain the more fundamental capability that enables this performance and how it is developed.
0
1
Tags
Ch.1 Pre-training - 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 large language model, pre-trained on a diverse corpus of internet text, is given the following prompt for the first time: "Rephrase the following sentence in the style of a 1920s radio announcer: 'The new smartphone will be available next week.'" The model successfully responds with something like: "Hark, good citizens! The magnificent new pocket telephone contraption shall grace our stores in but a week's time!" Which statement best analyzes the primary reason the model can successfully complete this novel task?
LLM Feature Development Strategy
Evaluating the Source of Zero-Shot Capabilities