Analyzing Model Failure
A user gives a language model the prompt: 'Summarize the following text but do not use the letter E.' The model provides a fluent, well-written summary that includes the letter E. Based on this outcome, what fundamental capability is the model likely lacking, and why does this failure prevent the user's prompt from being effective?
0
1
Tags
Ch.2 Generative Models - Foundations of Large Language Models
Foundations of Large Language Models
Foundations of Large Language Models Course
Computing Sciences
Analysis in Bloom's Taxonomy
Cognitive Psychology
Psychology
Social Science
Empirical Science
Science
Related
A developer is building a tool that allows users to generate custom email responses by providing specific commands, such as 'Write a polite refusal to this meeting invitation.' They have two language models to choose from:
- Model X: A very large model trained on a vast library of books and websites, excelling at generating fluent, human-like text.
- Model Y: A smaller model specifically trained on a dataset of commands paired with their correct outputs.
Which model is the more suitable choice for this tool, and why?
Enabling Instruction Following via Pre-training
Diagnosing a Language Model's Output
Analyzing Model Failure