Abstract page for arXiv paper 2609.02959: The Geometry of Ignorance: LLMs Know When to Temper Bayesian Priors