Optimizing Temperature for Language Models with Multi-Sample Inference | Read Paper on Bytez