IQN (NewsDesk): A new paper posted to SSRN introduces Computationally Pumped Bath Noise, or CPBN, a noise model designed for integration into the PKTron quantum circuit simulator that allows simulated noise to depend on the sequence of operations that came before a given gate, rather than being fixed to that gate alone.
The 13 page paper, authored by Dr Zuhair Ahmed of the Centre of Excellence for Technology, Quantum and AI, departs from the conventional approach used in most quantum circuit simulators, where noise is typically modeled as stationary and tied to individual gates through a fixed error rate. Instead, CPBN represents the simulated environment as an explicit dynamical reservoir state that is driven by computational activity as it occurs, retains short term memory of recent operations, gradually relaxes back down over time when left undisturbed, and can propagate spatially between neighboring qubits. The practical implication is that, within this framework, the noise experienced by a given operation can depend on the computational history that preceded it, rather than being determined solely by which gate is being applied or a static error rate assigned to it in advance.
To validate the model, the paper describes a staged program of controlled software experiments establishing several key properties within the implemented toy model, including history dependence, gate order dependence, the effects of ablating the memory and pump mechanisms individually, environmental relaxation behavior, spatial propagation of noise between qubits, and adherence to completely positive trace preserving behavior, a mathematical requirement for any physically valid quantum noise model. Matched memoryless control conditions were used throughout to hold the marginal noise distribution constant, allowing the researchers to isolate genuine order effects from simple differences in noise rates. A further fixed endpoint adversarial design fixed the first and last operations in a given history while holding the multiset of middle gates constant, removing a possible confound in which the final context an operation experiences changes simply because the final history gate itself differs.
The paper’s most closely watched component is a matched history experiment run on IBM Quantum’s 156 qubit ibm_marrakesh processor, using 100 paired histories across 400 total circuits, each run for 4096 shots on fixed physical qubits with identical physical circuit depth and gate counts between paired histories. The hardware run produced a mean absolute paired order effect of approximately 0.0027, with a bootstrap 95 percent interval of 0.0023 to 0.0031. However, neither the paired Wilcoxon signed rank test, which returned a p value of 0.2046, nor the sign test, which returned a p value of 0.4168, reached statistical significance at the conventional threshold of 0.05.
The paper is notably direct about what this result does and does not show. The authors report the hardware observation strictly as evidence of a measurable, non zero absolute difference within this specific run, explicitly declining to characterize it as statistical confirmation that CPBN, or any history dependent physical mechanism, is responsible for the observed behavior on the hardware. The paper also includes a corrected treatment of an earlier permutation test the authors describe as statistically invalid, and outlines a preregistered, multi block replication program the authors say would be required before stronger causal or novelty claims could reasonably be supported.
