Feedback neural network

Feedback neural networks are neural networks with the ability to provide bottom-up and top-down design feedback to their input or previous layers, based on their outputs or subsequent layers. This is notably used in large language models, specifically in reasoning language models (RLM). This process is designed to mimic self-assessment and internal deliberation, aiming to minimize errors (like hallucinations) and increase interpretability. This reflection is a form of "test-time compute", where additional computational resources are used during inference.