Learn Before
Input-Output Patterns in Few-Shot Learning
In few-shot learning, the provided demonstrations create a pattern that explicitly maps example inputs to their corresponding outputs. This pattern serves as a template that the Large Language Model uses to infer the task and generate predictions for new, unseen inputs.
0
1
Tags
Ch.3 Prompting - Foundations of Large Language Models
Foundations of Large Language Models
Foundations of Large Language Models Course
Computing Sciences
Related
Examples of Few-Shot Learning Applications in NLP
Enabling Few-Shot Learning with Multiple Demonstrations
Input-Output Patterns in Few-Shot Learning
Sufficiency of Demonstrations in Few-Shot Learning
Applying Few-Shot Learning to Complex Reasoning Tasks
A user provides the following text to a large language model to get it to classify movie reviews:
Review: The plot was predictable and the acting was wooden. I was bored the entire time. Sentiment: Negative
Review: An absolute masterpiece! The cinematography was stunning and the story was deeply moving. Sentiment: Positive
Review: It was a decent film. Not the best I've seen this year, but it had some good moments. Sentiment: Neutral
Review: I couldn't stop laughing from beginning to end. A brilliant comedy. Sentiment:
The model correctly responds with "Positive". Which statement best analyzes the primary reason for the model's successful performance on this task?
Constructing a Few-Shot Prompt for a Novel Task
Critiquing a Prompt for a Custom Extraction Task
Example of a Few-Shot Prompt for Polarity Classification
Learn After
Use of Simpler Patterns in Few-Shot Learning
A user provides a language model with the following examples to teach it a new task:
Input: Apple -> Output: AInput: Banana -> Output: BInput: Cherry -> Output: CWhen the user then provides the new input
Input: Grape, the model responds withOutput: G. The user was expecting the output to be the full wordGrape.Which of the following best explains why the model produced an unexpected result?
Constructing an Effective Input-Output Pattern
Predicting Model Behavior from a Few-Shot Pattern