Author name: Miranda Muqing Miao, Lyle Ungar

arXiv:2603.25052v2 Announce Type: replace
Abstract: Large language models (LLMs) tend to verbalize confidence scores that are largely detached from their actual accuracy, yet the geometric relationship governing this behavior remain poorly understood….

arXiv:2603.25052v1 Announce Type: new
Abstract: Large language models (LLMs) tend to verbalize confidence scores that are largely detached from their actual accuracy, yet the geometric relationship governing this behavior remain poorly understood. In …