When Not to Use an LLM

1. Which of the following scenarios is best suited for a classical classifier instead of an LLM?

Classifying user reviews into sentiment categories Generating creative writing prompts Automating customer support chat responses Creating personalized recommendations

2. What key advantage do classical models provide in the context of explainability?

They generate more dynamic output They offer feature importances and stable behavior They are more flexible in handling various data types They require less training data

3. In what situation is it acceptable to use an LLM according to the lesson?

High-speed fraud checks Handling structured tabular data Generating open-ended language responses Categorizing a very narrow set of inputs

4. What hybrid approach is suggested for effectively using both classical and LLM models?

Using an LLM for all classifications Classifying all input with a classical model A classifier handles confident inputs, and LLM deals with low-confidence cases Training both models on the same data set

Continue: ML in the Enterprise → Review this lesson Retake quiz