Doubly Robust Alignment for Large Language Models | Read Paper on Bytez