Pattern
Prefilling the assistant turn
Steer the model's output — its format and its first tokens — by seeding the beginning of the assistant's reply in the request itself, rather than relying on instructions alone to produce the desired shape.
Classification — Steering output format and opening tokens by seeding the start of the assistant's response.
Full entry
The complete treatment — problem, how it works, when to use it, sources, and examples — lives in the volume below.
Browse the full pattern index or the catalog.