A team is developing a system to generate marketing slogans. The process involves two stages: first, an initial model generates 50 potential slogans, and second, a highly accurate scoring model selects the single best slogan from that set to display. The team observes that while the final selected slogans are grammatically perfect and on-topic, they are often generic and uninspired. They also notice that the initial 50 slogans generated in each batch are usually very similar to one another. Whic
0
1
Tags
Ch.5 Inference - Foundations of Large Language Models
Foundations of Large Language Models
Foundations of Large Language Models Course
Computing Sciences
Evaluation in Bloom's Taxonomy
Cognitive Psychology
Psychology
Social Science
Empirical Science
Science
Related
A team is developing a system to generate marketing slogans. The process involves two stages: first, an initial model generates 50 potential slogans, and second, a highly accurate scoring model selects the single best slogan from that set to display. The team observes that while the final selected slogans are grammatically perfect and on-topic, they are often generic and uninspired. They also notice that the initial 50 slogans generated in each batch are usually very similar to one another. Whic
Optimizing a Chatbot Response System
Evaluating System Design for Code Generation