Evaluator-Optimizer Workflow

Analogous Procedural

How to study this: Anchor to the writer-editor analogy: generate -> critique -> refine in a loop.

Practice

Definition

A workflow where one LLM call generates a response and another provides evaluation and feedback in a loop, iteratively refining the output.

Analogous to a human writer iterating with editorial feedback.
When to use: when there are clear evaluation criteria AND "LLM responses can be demonstrably improved" through iterative refinement.

"Evaluator-optimizer" is a named workflow pattern; recognize the generator + evaluator loop and its precondition of clear, actionable evaluation criteria.

Start simple; only add agentic complexity when it demonstrably improves outcomes — the loop only pays off when feedback measurably improves responses.

🔗 Anchor: Anchor to the writer-editor analogy: generate -> critique -> refine in a loop.

Link this to something you already know:

Notes for this concept