Whose Probability?

Why a diffusion model gives a bound on its answer’s probability, and what RL does with it