NCA-GENL: Generative AI and LLMsDemo

Question

In the context of fine-tuning LLMs, which metric below is most commonly used to evaluate a fine-tuned model’s performance?

A Number of layers
B Model size
C Accuracy on a validation set
D Training duration

Explanation

Accuracy measured on a held-out validation set evaluates how well the fine-tuned model generalizes to data not used for training. The number of layers, model size, and training duration are configuration or resource measures rather than direct performance metrics.

NCA-GENL: Generative AI and LLMsDemo

AstroTutor

Notes

Readiness

QuestionQ76

Community Discussion