Neuro-Core

Detect temporal dependence in quantum hardware even when outcome histograms remain Born-compliant.

Quick Demo

pip install numpy
python demo_smoke.py
Decision: FLAG

Category-Defining Run-Quality Audit Primitive for Quantum Hardware

Neuro-Core is a vendor-agnostic, trial-resolved quantum execution audit engine designed to detect schedule-coupled structure under Born-stable marginals.

It operates beyond histogram-level validation and provides an information-theoretic witness for temporal dependence in quantum runs.

Why Neuro-Core?

Most validation pipelines stop at marginal statistics. Neuro-Core operates at the trial-resolved layer.

Neuro-Core addresses a deeper operational question:

Can schedule-coupled structure exist even when outcome histograms remain Born-compliant?

Yes — and Neuro-Core detects it.

Core Architecture

Neuro-Core enforces a strict two-layer validation logic.

1️⃣ Born Gate (Sanity Layer)

Verifies histogram indistinguishability across counter-balanced schedules.

If marginals differ → reject at gate

If marginals match → proceed to witness layer

2️⃣ Witness Engine (Trial-Resolved Layer)

Computes a schedule-coupled information-theoretic witness using:

Conditional Mutual Information (CMI)

Quantile-based memory floors

Matched permutation null family

Trial-order sensitivity

Drift-robust evaluation

Output

W (bits)

p-value

Floor comparison

QA flag

Structured JSON audit report

What This Is Not

Neuro-Core is not:

Error mitigation

Noise modeling

Decoder optimization

Calibration tooling

It is a:

Trial-resolved run-quality audit primitive for detecting memory signatures invisible to marginal checks.

Minimal Example

from neurocore_audit.engine import run_audit

result = run_audit("AuditReport.json")

print(result["W_bits"])
print(result["p_value"])
print(result["flag"])

Output Structure

Neuro-Core produces a structured audit report containing:

Born gate result

Witness value (bits)

Permutation null statistics

Quantile floor estimate

Decision rule

Severity label

Designed for integration into CI or production QA pipelines.

Intended Users

Quantum hardware teams

Control-stack developers

Validation engineers

Benchmarking groups

Research labs studying temporal correlations

Category Definition

Neuro-Core defines a new validation layer:

Trial-resolved schedule-coupled run audit under Born-stable marginals.

This repository contains the v0.1.1 reference implementation.

Installation

pip install -e .

Version

Current version: v0.1.1.

Citation

If you use Neuro-Core in research, please cite:

See CITATION.cff

License

See LICENSE.txt

Roadmap

Public benchmark dataset

CI-ready audit wrapper

Hardware validation cases

Extended null families