Learnable Multipliers: Freeing the Scale of Language Model Matrix Layers - Explained Simply | ArXiv Explained