Self-refine

Self-refine is a technique in which a language model critiques its own output and then rewrites it based on that feedback, repeating the cycle to improve quality without additional training.

Self-refine sits in the Prompting & context part of the Agentik {OS} glossary, which defines the words used to build and run AI agent systems.

Also called self-refinement.