What are some use cases for fine-tuning of a model? Note: There are 2 correct answers to this question.

Question

Options

• ATo introduce new knowledge to a model in a resource-efficient way
• BTo quickly create iterations on a new use case
• CTo sanitize model outputs
• DTo customize outputs for specific types of inputs

• A
  89% (16)
• B
  6% (1)
• C
  6% (1)

Explanation

Fine-tuning adapts a pre-trained model on a domain-specific dataset, which efficiently injects new knowledge (A) without training from scratch - making it highly resource-efficient compared to pre-training. It also excels at customizing outputs for specific input types (D), such as producing formal legal language, JSON-structured responses, or a particular tone for a given domain.

Why the distractors are wrong:

• B is incorrect because fine-tuning is slow and resource-intensive; rapid iteration on new use cases is better suited to prompt engineering.
• C is incorrect because output sanitization is a separate concern handled by content filters, RLHF/RLAIF alignment, or output validation layers - not fine-tuning itself.

Memory tip: Think of fine-tuning as "teaching old knowledge new tricks" - it's about what the model knows (new domain knowledge) and how it responds (output style/format), not about speed or safety filtering.

No community discussion yet for this question.