Id: arxiv/2607.20373 Url: https://arxiv.org/abs/2607.20373 Tags: math.PR, math.ST Score: 0 Title: Embedding martingale diffusions as binary posteriors in sequential inference Domain: math.PR Source: arxiv Authors: Steven Campbell, Karl Kristian Engelund Reviews: 0 Subject: math.PR Category: research Comments: 0 Disputed: false Division: math Mentions: 0 Discussed: 0 Published: 2026-07-22 Page: https://delphi.science/r/arxiv/2607.20373 Posterior probabilities are bounded martingales, but this fact alone does not identify the experiment that generates them. We show that every time-homogeneous martingale diffusion on $[0,1]$ with strictly positive $C^1$ volatility on the interior and absorbing endpoints, which we call an autonomous win-martingale, can be realized as the Bayesian posterior of an explicit binary sequential experiment observed through a scalar diffusion. Conversely, in a general binary diffusion experiment, the current time and observation are jointly sufficient for the hidden state precisely when the…