← Back to Glossary
Reflection & Self-Correction
A workflow where an agent evaluates its own generated output, identifies errors, and revises its answer before finalizing.
Why it exists
Improves code generation and reasoning accuracy by catching bugs and hallucinated claims internally.
How it works
The agent runs a self-critique prompt (or compiler check), parses errors, and generates an updated output draft.
Related terms
Where you'll use this
roadmap