A way for AI to check and fix its own answer.
AI self-verification is like checking your teeth before a school photo. It can catch spinach, but not your whole awkward smile.
It checks its own answer first. People use it for reasoning, code review, and quality checks.
Hallucination
Self-verification can catch some made-up answers, but not every hallucination.
Chain-of-thought
A clear reasoning path gives self-verification clues to check.
LLM-as-a-judge
Self-verification is like asking the model to judge its own answer first.
AI Abstention
When self-verification fails, the AI should lean toward not answering.