Reproducibility, Measurement Timing, and Structural Bias in a Discrete Probe-Geometry Backreaction Model
Paul Jarvis
We construct a minimal discrete model in which a quantum probe on a lattice is coupled to a single auxiliary two-level qubit representing local geometry, implemented three ways: continuous unitary evolution, unitary evolution interrupted by measurement on a fixed schedule, and unitary evolution interrupted by measurement triggered adaptively by the qubit's own entanglement entropy. We show continuous unitary evolution is bit-for-bit reproducible across runs, while both measurement-based implementations are reproducible only up to the first ambiguous measurement outcome, and the adaptive schedule introduces a further failure mode: runs can diverge in when subsequent measurements occur, not just their outcome. Using GPU-batched Monte Carlo ensembles of 200,000 trajectories, we confirm both effects are structural, not artifacts of particular seed choices (77.5% and 73% convergence rates respectively), and give an analytic argument for both, checked directly against exact simulation at every measurement window. The work does not resolve which class of backreaction describes any real physical system, but provides a concrete, checkable diagnostic for backreaction models.
•Sep 14, 2026•0 citations