Output priming is a prompting technique where the beginning of the model's response is supplied in advance so the model continues in the desired format, tone, or structure. It is also called assistant prefilling, response priming, or response prefixing.
For example, an application that needs JSON may provide a partial assistant message containing { after the user message. The model then continues from that opening token rather than starting with a conversational preamble.
Output priming is useful for:
- structured output such as JSON or XML;
- skipping phrases like "Sure, here is...";
- maintaining a fixed voice or style;
- starting a response at a required section; and
- reducing brittle post-processing.
It is not a substitute for a clear task description, schema validation, or structured output. Some models and APIs do not support assistant prefilling, and reasoning-oriented modes may restrict or change the behavior.
Anthropic describes prefilling as a way to start the model's response in its prompt engineering best practices guide.
The LLM Knowledge Base is a collection of bite-sized explanations for commonly used terms and abbreviations related to Large Language Models and Generative AI.
It's an educational resource that helps you stay up-to-date with the latest developments in AI research and its applications.