What is a common purpose of fine-tuning an LLM into an instruct model?

• A. To improve its ability to follow user instructions.
• B. To decrease its size.
• C. To ignore user prompts.
• D. To reduce training data requirements.

Answer: (A)

Tuning an LLM to become an instruct model centers on teaching it to follow user instructions more reliably. By training on datasets that pair diverse prompts with preferred, carefully shaped responses, the model learns patterns for identifying what a user wants and producing outputs that meet those expectations. This alignment makes the model more predictable, helpful, and safe in real-world interactions because it focuses on executing the instruction rather than just generating plausible text.

That’s why the best choice is to improve its ability to follow user instructions. The other aims—reducing the model’s size, ignoring prompts, or lowering training data needs—are not the primary purpose of instruction tuning: size reduction is a separate objective, ignoring prompts runs counter to the goal, and data efficiency isn’t the central aim of this process.