Self-correction means asking a model to check its own work and try again: draft an answer, critique it against the task, then revise. Techniques such as Self-Refine, Reflexion and CRITIC differ mainly in where the critique comes from. The model can judge itself, keep notes from earlier failed attempts, or check claims with an outside tool such as a search engine or a code runner.
Self-critique without outside feedback often adds cost without fixing errors, so measure whether it helps on your task. The guides here start with prompt-level self-correction, then move to agents: reflection as a reasoning pattern, an agent that learns from its own tool failures across runs, and an agent allowed to change its own setup within limits you set.
The library lists all guides with filters for topic, level, category and reading time.