How does critique-and-revise improve output quality?

It separates generation from evaluation. The first pass produces a draft, the critique pass scores it against your rubric and names specific weaknesses, and the revision pass fixes them. Forcing the model to evaluate its own work against explicit criteria catches errors a single pass would ship.

Is this the same as the reflexion or self-refine technique?

Yes, it implements the same idea — generate, reflect, refine — popularized as self-refine and reflexion. The key is a concrete rubric so the critique is grounded rather than a vague "looks fine."

How many revision rounds should I use?

One critique-revise pass captures most of the gain. Two rounds can help on complex tasks, but beyond that improvements taper and the model risks over-editing or drifting from the original intent.

Does this tool call an AI?

No. It builds the multi-stage prompt that you run yourself. It is fully client-side and needs no API key.

What makes a good rubric line?

Each line should be a checkable criterion, like "every claim has a source," "tone is formal," or "output is valid JSON matching the schema." Vague lines like "make it good" produce vague critiques.

Critique-and-Revise Loop Prompt Builder

Critique-and-revise loop prompt builder

A model’s first answer is a draft, not a final. The critique-and-revise pattern — also known as self-refine or reflexion — makes the model grade its own work against a rubric and then fix it in the same conversation. This builder turns your task and a set of criteria into a multi-stage prompt: generate, critique, revise. You control the rubric, how many rounds run, and the final output format.

How it works

You enter the task and a rubric (one criterion per line). The builder writes a prompt with explicit stages. Stage one produces a draft answer. Stage two critiques that draft against each rubric line, scoring it and naming concrete weaknesses — not “looks good” but “claim X has no source; tone is too casual.” Stage three rewrites the draft to address every weakness raised. If you choose multiple rounds, the critique and revision stages repeat on the latest version. The final output is formatted as you specified. Everything is assembled locally; you copy the prompt and run it anywhere.

Tips and notes

Write checkable rubric lines. “Every claim cites a source” beats “be accurate.”
One round captures most of the gain. Add a second only for genuinely complex outputs.
Watch for over-editing. Too many rounds can drift from your intent — keep the original task visible in the rubric.
Pairs well with a format requirement. Adding “output must be valid JSON” as a rubric line lets the critique stage catch malformed output before you do.