AI Dictionary › Prompting
Prompt di Auto-correzione
Self-correction prompting is a technique in which the model is explicitly asked to identify and correct its own errors within the same response or in an immediately following turn, without the intervention of external verification criteria or a second model.
The prompt instructs the model to first produce a response, then reread it with a critical eye actively looking for logical errors, inaccuracies or inconsistencies, and finally produce a corrected version. Unlike generic verification, the emphasis here is on immediate, direct correction rather than merely flagging problems, often within a single exchange.
This technique is useful in tasks where careless errors or small inaccuracies are common, such as arithmetic calculations, consistency across different parts of a long text, or precise adherence to numerical instructions: asking the model to double-check its own calculations before delivering the final result, or to verify that every constraint listed in the prompt was actually respected.
The concept belongs to the family of techniques that exploit the model's ability to critically evaluate its own output, a research area that grew alongside the observation that models, when explicitly instructed to do so, often manage to spot their own errors that they had not noticed during the initial generation.
Grace evaluates whether you explicitly ask the model to recheck and correct its own calculations or claims before considering the answer complete.
From our network
Kaimaki Web: Websites That Win Customers
Custom websites, web apps and digital marketing for growing businesses.
Visit kaimakiweb.com →From the Agora Intelligence blog
📱 Download the Android app (beta) iOS coming soon
Say what you mean. Get what you need.
Grace Certified, the AI coach that trains and certifies your prompt engineering, by Agora Intelligence.