What are chat templates in LLMs?

• A. Fixed response scripts that never adapt.
• B. Pipelines for data ETL tasks.
• C. Structured formats that guide conversation flow between users and assistants.
• D. Templates for code generation only.

Answer: (C)

Chat templates are structured formats that guide conversation flow between users and assistants. They provide a predefined scaffolding for a chat, laying out the roles, context, and the sequence of turns the interaction should follow. In practice, a template might include a system message that sets the assistant’s persona and rules, a user prompt that describes the task, and a pattern for the assistant’s reply—with placeholders to fill in user-specific details. This structure helps keep conversations consistent, controls tone and depth, and makes it easier to steer the model toward the desired outcome across different users and tasks. Templates can outline multi-step interactions, such as starting with a brief explanation, then offering steps or examples, and finishing with a prompt for clarifying questions. They’re not just fixed scripts that never adapt, nor are they limited to data pipelines or code generation. They serve as reusable scaffolds that shape how a chat unfolds.