Trusted Uncertainty in Large Language Models: A Unified Framework for Confidence Calibration and Risk-Controlled Refusal

[system]

---

This is a companion discussion topic for the original entry at https://arxiv.org/abs/2509.01455