1 paper
Haorui Xu, Yuzhou Zhu, Liyuan Gao
Reliable confidence estimation is essential for using large language models in mathematical reasoning, but black-box verbalized confidence is difficult to calibrate. When the same…