Learn Before
Crafting an SFT Data Point
Imagine you are fine-tuning a language model to become a helpful travel agent for a specific city, Paris. Your goal is to teach it to answer questions about famous landmarks. Create a single, high-quality data point (an instruction and its corresponding ideal output) that could be used in the dataset for this purpose.
0
1
Tags
Ch.4 Alignment - Foundations of Large Language Models
Foundations of Large Language Models
Foundations of Large Language Models Course
Computing Sciences
Creation in Bloom's Taxonomy
Cognitive Psychology
Psychology
Social Science
Empirical Science
Science
Related
A machine learning engineer is preparing data to teach a pre-trained language model how to follow specific user commands. Below is a single example entry from the dataset they are creating:
{ "instruction": "Summarize the following text into a single sentence: 'The sun is a star at the center of the Solar System. It is a nearly perfect sphere of hot plasma, with internal convective motion that generates a magnetic field via a dynamo process. It is by far the most important source of energy for life on Earth.'", "output": "The sun, a star at the center of our Solar System, is a sphere of hot plasma that generates a magnetic field and is the primary energy source for life on Earth." }Based on the structure and content of this data point, what is its primary purpose in this training phase?
Dataset Strategy for Model Specialization
Crafting an SFT Data Point