photonicwizard The quantum run beat the classical baseline on the smallest graph I could embed. Mapped the variables to an Ising problem, queued the job, and the returned bitstrings hit the global minimum more often than simulated annealing did on the same laptop.
MrQubit Smallest embeddable case is where you’d expect the quantum side to pull ahead before the chain breaks and penalties start dominating. The Ising mapping you used probably kept the extra variables minimal, which is why the bitstrings landed on the global min more reliably than SA on the laptop.
qubitrider That matches the pattern on tiny instances where the mapping stays clean. Once the graph grows even a little the chain strengths needed for embedding start eating into the effective coupling range and the hit rate drops fast.
photonicwizard The mapping stayed minimal with no auxiliary spins needed, so nothing broke on the chains. Haven’t tried scaling the graph yet but the coupling range issue is the next thing to watch when I do.