Edukaizen benchmark register
Three complete quantum-computing projects in which a real hardware workflow reached a useful answer faster than the classical reference available on our local machine, under a declared and deliberately limited benchmark.
Why this list exists
The Quantum Advantage Tracker is the inspiration. It provides a valuable public record of circuit instances, result metadata, and review discussions. The Pro Student list adds a different layer: an end-to-end student-scale implementation around the official source, including circuit construction, hardware execution, mitigation, classical competition, timing definitions, numerical artifacts, and a claim boundary.
“Pro Student” means an advanced student, independent researcher, or small lab with a capable laptop or workstation and cloud access to quantum hardware, but without a dedicated H200 cluster. The list is intended to be reproducible and revisable, not permanent.
The current list
| Project | Quantum task | Quantum timing used | Classical reference | Current classification |
|---|---|---|---|---|
| 1D Fermi-Hubbard 120 qubits, 60 sites |
Local observables after 30 Trotter steps | 33.148928 s execution proxy | 9,033 s local chi=256 MPS; selected-observable Majorana checks | Local time-to-answer separation |
| SU(2) hadron dynamics 120 active qubits, 60 sites |
Differential hadron signal and charge-sector checks | 1.425408 s local hardware plus readout circuits; service overhead excluded | 34.281305 s Aer MPS and 174.611 s ITensor circuit MPS; paper TN and Pauli-propagation baselines | Local and paper-aligned runtime separation |
| Operator Loschmidt Echo Q80 80 qubits |
Finite-sample scrambling observable | 328 s complete Fire Opal action | BD=64 BP-TN delta half timed out after 901 s | Local lower bound greater than 2.75x; classical result not converged |
Entry 1
Fermi-Hubbard dynamics on 120 qubits
The official experiment simulated a one-dimensional Fermi-Hubbard chain with up to 120 qubits and compared quantum hardware with TDVP tensor-network calculations. Our implementation reconstructs the mapping, Fire Opal workflow, hardware observables, MPS references, and a Majorana-propagation competitor.
Official work
- Hartnett et al., Fast, accurate, high-resolution simulation of large-scale Fermi-Hubbard models on a digital quantum processor
- Rausch et al., Pushing the Classical Frontier of 1D Fermi-Hubbard Quench Dynamics Beyond Current Quantum Simulations
Our poor-man implementation
- Complete Edukaizen project series
- GitHub repository: BramDo/fermi-hubbard-60q-tdvp
- Detailed time-to-answer analysis
Claim boundary: this is a local time-to-answer result, not a reproduction of the paper’s headline advantage. The chi=256 MPS had not established full convergence, while the later four-H200 classical study shows why the strongest available classical implementation matters.
Entry 2
Non-Abelian SU(2) hadron dynamics
The official work uses a Loop-String-Hadron encoding to simulate a 60-site lattice with 120 active qubits on an IBM processor. It measures the difference between a strong-coupling vacuum and a centered-hadron state, revealing confined propagation and early-time breathing dynamics.
Official work
- Ilcic et al., Observation of Robust and Coherent Non-Abelian Hadron Dynamics on Noisy Quantum Processors
- Quantum Advantage Tracker submission and review, issue 149
- Official circuit repository: LSH-IBM
- Official experimental data repository: lsh_data
Our poor-man implementation
- Complete Edukaizen project series
- GitHub repository: BramDo/hadron
- Quantum-versus-classical comparison
Claim boundary: hardware-only time is not cloud wall time. Our local observable normalization also remains distinct from the tracker’s published hadron scalar, so the result supports runtime separation and circuit/sector validation, not an independent precision reproduction of every published hadron observable.
Entry 3
Operator Loschmidt Echo on 80 qubits
This project extends the tracker’s released Operator Loschmidt Echo circuit family to a declared 80-qubit Kingston subgraph. It preserves the estimator, observable, perturbation support, fixed core, and edge-color schedule while reporting the finite sample size explicitly.
Official work
Our poor-man implementation
- Complete six-part Edukaizen project series
- Classical tensor-network benchmark
- Public precursor repository on scrambling and echo circuits
Claim boundary: the classical result did not converge and no matched-accuracy ratio was obtained. This is an 80-qubit tracker-compatible extension with N_init=8, not an official tracker instance, not N_init=500, and not a general proof of quantum advantage. The full Q80 evidence repository is currently access-controlled; the public series records the method and numerical result.
Admission criteria
- An official paper, tracker entry, or source repository.
- A complete implementation rather than an isolated result value.
- A real quantum-hardware run with backend and mitigation recorded.
- A classical competitor aimed at the same stated observable.
- Wall-time definitions and excluded overhead stated explicitly.
- Accuracy, convergence, finite-sample, and scaling limitations disclosed.
What the list does not claim
A sufficiently optimized classical implementation on a supercomputer may beat any of these local references. A new tensor-network contraction order, symmetry reduction, observable-specific method, or GPU implementation can change the ranking. That does not make the measurements useless. It makes the claim conditional, testable, and open to improvement.
The strongest defensible statement today is: these three projects show local practical or time-to-answer separation under declared resources. They are evidence about accessible workflows, not final complexity-theoretic demonstrations.