Conformal Tail Risk Control for Large Language Model Alignment | Read Paper on Bytez