Title: Quantum state inference from coarse-grained descriptions: analysis and an application to quantum thermodynamics

Abstract: The characterization of physical systems relies on the observable properties which are measured, and how such measurements are performed. Here we analyze two ways of assigning a description to a quantum system assuming that we only have access to coarse-grained properties. More specifically, we compare the Maximum Entropy Principle method, with the Bayesian-inspired recently proposed Average Assignment Map method [P. S. Correia et al, Phys. Rev. A 103, 052210 (2021)]. Despite the fact that the assigned descriptions by both methods respect the measured constraints, and that they share the same conceptual foundations, the descriptions differ in scenarios that go beyond the traditional system-environment structure. The Average Assignment Map is thus shown to be a more sensible choice for the ever more prevalent scenario of complex quantum systems. We discuss the physics behind such a difference, and further exploit it in a quantum thermodynamics process.