The correct option is Prompt engineering.

The scenario describes crafting a more specific and contextual request in order to steer the model toward better outputs which is exactly what Prompt engineering does. By deliberately specifying the product details, audience, desired tone, and key attributes, the team shapes the input so the model can produce more relevant and engaging taglines. This is the practice of designing prompts with clear instructions and constraints to guide the model.

Reinforcement learning from human feedback or RLHF is a training approach where a model is optimized using a reward model built from human preferences. It is part of the model training pipeline rather than a technique for writing better user prompts at request time.

Model fine-tuning adjusts a model’s weights with additional domain or task specific data to change its behavior. It is not about how a single request is phrased during inference, so it does not match the scenario.

Data augmentation expands or transforms training datasets to improve model generalization. The scenario does not change training data and instead improves results by wording the input differently, so this option is not applicable.

Full AWS Practitioner Certification Question