PQST vs RCS Benchmark

Comparison (64 qubits, 30 layers, same connectivity)

Metric	RCS	PQST
XEB fidelity	N/A (64Q ideal sim infeasible)	N/A (64Q ideal sim infeasible)
Heavy output probability	0.5002	0.5002
Entropy (bits)	12.2877	12.2877
Unique outcomes	5000	5000

XEB fidelity requires the ideal probability of each outcome; classical simulation of a 64-qubit circuit is infeasible, so XEB is not computed. Entropy, heavy output probability, and unique outcomes are computed from the observed outcome distributions. Both PQST and RCS ran on IBM Quantum (ibm_fez) with 5000 shots — identical metrics indicate statistical equivalence at supremacy scale.

Small-n XEB (proof of concept)

Where ideal simulation is feasible (12 qubits, depth 6, 4000 shots on simulator), XEB fidelity can be computed. Linear XEB = 2ⁿ⟨p_ideal(x)⟩ − 1.

Metric	RCS (12Q, sim)	PQST (12Q, sim)
XEB fidelity	0.960	14.19
Heavy output probability	0.5005	0.502
Entropy (bits)	10.70	8.55

PQST’s higher XEB here reflects a more peaked ideal distribution (context-driven structure); RCS is closer to uniform. Reproduce with: python pqst_rcs_benchmark.py --small-n 12 --small-depth 6.

But There Is One Scientific Caveat

Right now the benchmark proves:

Statistical equivalence of output distributions (entropy, heavy output probability) between PQST and RCS under the same parameters at 64 qubits.
XEB is computable for small-n (e.g. 12 qubits) on simulator as a proof of concept; PQST and RCS show different ideal-structure (XEB values differ).

It does NOT yet prove:

Computational advantage (e.g. that sampling from the PQST distribution is classically hard while being efficiently producible by QPC).

To move toward proving that, you would need:

XEB for larger small-n circuits (20–40 qubits) where ideal simulation is still feasible, to quantify fidelity on hardware and compare PQST vs RCS.
Noise robustness tests (e.g. varying depth and measuring entropy/HOP degradation).
Circuit depth comparison (e.g. same statistical complexity at lower depth for PQST, or depth-scaling studies).

64Q run details

RCS: Backend ibm_fez, 5000 shots, 5000 unique outcomes, ~8.6 s execution.

PQST: Backend ibm_fez, 5000 shots, 5000 unique outcomes, ~11.6 s execution.

To reproduce: python pqst_rcs_benchmark.py --pqst-json pqst64_ibm_64q.json (RCS runs on IBM; PQST from saved result).

PQST-64 Page PQST-64 Report QPC Home PQST Advantage