Learn Before
Prediction: The First Step of Self-Refinement
The initial phase in the self-refinement framework, known as 'Prediction,' involves using a Large Language Model to generate the first version of the output.
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
Prediction: The First Step of Self-Refinement
Feedback Collection: The Second Step of Self-Refinement
Refinement: The Third Step of Self-Refinement
Iterative Self-Refinement Process
Deliberate-then-Generate (DTG) Method
A common framework for improving a language model's output involves a cyclical process. Arrange the following stages of this process into the correct logical order, from start to finish.
A development team is improving a news-summarizing AI. Their process is as follows:
- The AI generates an initial summary of an article.
- A separate automated tool critiques the summary for conciseness and factual accuracy, producing a list of issues.
- The AI is then given the original article, its first summary, and the list of issues, and is prompted to write an improved version.
Which option correctly maps this process to the standard three-step self-refinement framework?
Analyzing a Flawed Self-Improvement Process
Learn After
A user is interacting with a language model to generate a short story. The user first provides the prompt: 'Write a story about a robot who discovers music.' The model then produces a three-paragraph story. After reading it, the user provides feedback: 'That's a good start, but make the robot's discovery more accidental.' The model then rewrites the story incorporating this feedback. Which part of this interaction represents the 'prediction' phase in a self-refinement cycle?
Role of the Initial Generation Step
In a self-refinement cycle for a language model, the 'prediction' phase is where the model analyzes feedback to generate an improved version of its initial output.