Verifier-Critic Loop

An evaluation pattern in which one model or prompted role produces an output while a second model, or the same model under different instructions, checks that output against rules, sources or criteria and requests a rewrite if needed. Sometimes called an AI review loop.