ALPS: Attention Localization and Pruning Strategy for Efficient Alignment of Large Language Models | Xiaol.x | Podwise