What does “Gradient Descent” mean?

A training method that fine-tunes models directly on preference pairs without needing a separate reward model.

A vector showing how much each parameter should change to reduce loss.

The core training algorithm that updates model weights by propagating prediction errors backward through the network.

A consistent pattern of error or unfairness in data or model behavior.

A regularization method that softens hard labels to improve generalization.

A training hyperparameter controlling how much parameters change each update step.