More interesting part probably worth highlighting: The SAME model won't always return the same output when prompted with the same fact check.

You ask a human 1000 times a fact check question, they say the same answer 1000 times. You ask an LLM the same question a 1000 times, your results could vary significantly.

Humans work based on the Metamemory (knowing what they know), while LLMs are picking from statistical probability.