Improving Model Alignment Through Collective Intelligence of Open-Source LLMS