Learn Before
Automated Task Creation for a Marketing Dataset
A research team is building a dataset to train a model for generating marketing slogans. They start with a small seed set of 10 human-written instructions. To expand this set, they provide a large language model with the following prompt:
'Here are some examples of instructions:
- Create a slogan for a new electric car.
- Write a tagline for a vegan restaurant.
- Invent a short, memorable phrase for a new fitness app.
Based on these, please generate a new and different instruction for a similar task.'
The model then outputs: 'Generate a catchy tagline for a new brand of eco-friendly coffee.'
Explain the underlying mechanism that enables the model to produce this new, relevant instruction based on the prompt it received.
0
1
Tags
Ch.4 Alignment - Foundations of Large Language Models
Foundations of Large Language Models
Foundations of Large Language Models Course
Computing Sciences
Application in Bloom's Taxonomy
Cognitive Psychology
Psychology
Social Science
Empirical Science
Science
Related
Instruction Sampling for Diversity in Self-Instruct
Example of a Prompt Template for Instruction Generation in Self-Instruct
Sample Generation in Self-Instruct
An AI development team provides a large language model with a prompt containing several existing task instructions, such as 'Translate this sentence into French' and 'Write a poem about the ocean.' The prompt then asks the model to generate a new, distinct instruction based on the examples provided. What is the primary function of including the existing instructions in the prompt?
Automated Task Creation for a Marketing Dataset
Analyzing a Flawed Prompt for Instruction Generation