Linear Preference Optimization: Decoupled Gradient Control via Absolute Regularization

Open in new window