Credal Transformer: A Principled Approach for Quantifying and Mitigating Hallucinations in Large Language Models Paper β’ 2510.12137 β’ Published Oct 14, 2025 β’ 2