"The core of intelligence lies in the ability of 'self-validation and self-correction', while current large models are merely ...