Low-Redundant Optimization for Large Language Model Alignment