Quite unexpectedly, I opened Nature of 16 October 2025 only to realise it featured a paper by Chi-Fang Chen et al. on a valid MCMC algorithm for quantum computing. The motivation is for simulating quantum multi-body physics, with details that escape me. The paper claims this is the first valid quantum MCMC quantum thermal simulation, with further application similar to the MCMC revolution in Bayesian inference. (An earlier proposal by Davies in 1974 cannot be implemented for many-bodies problems.)
“Although early proposals for simulating thermalization directly considered simulating the global system–bath Hamiltonian evolution, more tractable approaches use a master equation, or Lindbladian, to capture the effect of a large bath on the small system using a continuous-time Markovian process. However, previous approaches inherited issues of the prototypical Davies generator, which worked well in quantum optics but has unphysical features in noncommuting many-body systems because of the exponentially small level spacing. Our nature-inspired algorithm takes precisely the form of a Lindbladian, inherits the locality from the physical model, exactly satisfies detailed balance and resembles the interactions we expect from weak coupling to a Markovian thermal bath.”
The setup is one of a continuous-time quantum Markov chain (or Lindbladian) over a finite set of configurations. The arguments for validating the algorithm (into a completely spelled out theorem!) involve quantum detailed balance (illustrated by the above cartoon) and locality (which I understand as a form of Markovianity in the updating rule). The transition attached to the algorithm however remains incomprehensible to me, involving a Fourier transform of a Hermitian jump operator (both notions escaping me in this context). And the illustration on an Ising model does not seem particularly quantum related, albeit the “Pauli operators” may be unrelated with the original spin binaries… There is also no MCMC reject step.
“Detailed balance (…) gives a coherent Gibbs sampler, in which we prepare the purified Gibbs state by a natural adiabatic path parametrized by the inverse temperature β. Of course, the purified Gibbs state reduces to the (mixed) Gibbs state once we trace out the purifying replica system, but the purification opens doors to advanced quantum algorithmic tools, such as verification (for example, through the swap test or measuring the energy of the parent Hamiltonian) or faster mean estimation.”
If I understand correctly the above “Gibbs sampler” relates to the Gibbs distribution as in the original 1984 paper of Geman and Geman! It may thus be that the current paper is similarly annunciating a new era in scientific computing, although I remain in the dark as to how to link it with a practical problem.
The running shoe company 
