{"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","summary":"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…","category":"research","comments":0,"disputed":false,"division":"math","mentions":0,"discussed":0,"published":"2026-07-22","reactions":{}}