N int64 15 247 | a int64 2 5 | r int64 4 60 | n int64 4 8 | t int64 8 16 | qubits int64 12 24 | factors listlengths 2 2 | tvd float64 0 0 | max_abs float64 0 0 | p_success_cf float64 0.26 0.5 | p_success_sim float64 0.26 0.5 | secs float64 0.05 58.4 |
|---|---|---|---|---|---|---|---|---|---|---|---|
15 | 2 | 4 | 4 | 8 | 12 | [
3,
5
] | 0 | 0 | 0.5 | 0.5 | 0.05 |
21 | 2 | 6 | 5 | 10 | 15 | [
3,
7
] | 0 | 0 | 0.322075 | 0.322075 | 0.14 |
33 | 5 | 10 | 6 | 12 | 18 | [
3,
11
] | 0 | 0 | 0.392775 | 0.392775 | 0.25 |
35 | 2 | 12 | 6 | 12 | 18 | [
5,
7
] | 0 | 0 | 0.324829 | 0.324829 | 0.6 |
39 | 2 | 12 | 6 | 12 | 18 | [
3,
13
] | 0 | 0 | 0.324068 | 0.324068 | 0.21 |
51 | 2 | 8 | 6 | 12 | 18 | [
3,
17
] | 0 | 0 | 0.5 | 0.5 | 0.14 |
55 | 2 | 20 | 6 | 12 | 18 | [
5,
11
] | 0 | 0 | 0.377295 | 0.377295 | 0.26 |
57 | 5 | 18 | 6 | 12 | 18 | [
3,
19
] | 0 | 0 | 0.317879 | 0.317879 | 0.22 |
65 | 3 | 12 | 7 | 14 | 21 | [
5,
13
] | 0 | 0 | 0.328801 | 0.328801 | 1.94 |
69 | 2 | 22 | 7 | 14 | 21 | [
3,
23
] | 0 | 0 | 0.446793 | 0.446793 | 1.91 |
77 | 2 | 30 | 7 | 14 | 21 | [
7,
11
] | 0 | 0 | 0.26157 | 0.26157 | 1.65 |
85 | 2 | 8 | 7 | 14 | 21 | [
5,
17
] | 0 | 0 | 0.5 | 0.5 | 0.54 |
87 | 2 | 28 | 7 | 14 | 21 | [
3,
29
] | 0 | 0 | 0.416127 | 0.416127 | 1.92 |
91 | 2 | 12 | 7 | 14 | 21 | [
7,
13
] | 0 | 0 | 0.327007 | 0.327007 | 1.67 |
93 | 2 | 10 | 7 | 14 | 21 | [
3,
31
] | 0 | 0 | 0.394458 | 0.394458 | 1.67 |
95 | 2 | 36 | 7 | 14 | 21 | [
5,
19
] | 0 | 0 | 0.322283 | 0.322283 | 1.84 |
111 | 2 | 36 | 7 | 14 | 21 | [
3,
37
] | 0 | 0 | 0.319477 | 0.319477 | 2 |
115 | 2 | 44 | 7 | 14 | 21 | [
5,
23
] | 0 | 0 | 0.429884 | 0.429884 | 1.79 |
119 | 2 | 24 | 7 | 14 | 21 | [
7,
17
] | 0 | 0 | 0.317118 | 0.317118 | 2.46 |
123 | 2 | 20 | 7 | 14 | 21 | [
3,
41
] | 0 | 0 | 0.385391 | 0.385391 | 2.29 |
143 | 2 | 60 | 8 | 16 | 24 | [
11,
13
] | 0 | 0 | 0.261699 | 0.261699 | 56.09 |
187 | 2 | 40 | 8 | 16 | 24 | [
11,
17
] | 0 | 0 | 0.389742 | 0.389742 | 51.67 |
221 | 2 | 24 | 8 | 16 | 24 | [
13,
17
] | 0 | 0 | 0.325519 | 0.325519 | 52.51 |
247 | 2 | 36 | 8 | 16 | 24 | [
13,
19
] | 0 | 0 | 0.324709 | 0.324709 | 58.41 |
exactbench
Exact-answer benchmarks for quantum cryptanalysis resource estimation (v1.0.0).
Source code and full reproduction: https://github.com/leanderdulac/exactbench
Every item is a triple (input, exact answer, tolerance), so a tool under test either reproduces the answer or it does not — in the spirit of the closed-form equilibria used to verify stellarator MHD codes.
This dataset mirrors the data released with the GitHub repository, which accompanies the preprint
L. França de Mello, Exact-Answer Benchmarks for Quantum Cryptanalysis Resource Estimation: Known-Key Instances, Published-Estimate Regression Targets, and Closed-Form Reference Values (2026). Paper link to be added once the preprint is public.
Contents
| Part | What it checks | Files |
|---|---|---|
| B1 | Known-answer Shor order finding: closed form for the full measurement distribution vs. exact Qiskit statevector simulation (24 instances, up to 24 qubits), plus an OpenQASM 3 export/re-import round trip for N = 15 | tables/factoring.jsonl, tables/qasm15_roundtrip.jsonl, results/qasm3/*.qasm |
| B2 | Known-key ECDLP on toy prime-order curves: exact law for the ideal Z_q variant; qubit variant cross-checked between a Qiskit circuit and an independent NumPy implementation | tables/ecdlp.jsonl |
| B3 | Regression targets transcribed from Roetteler et al. 2017 (arXiv:1706.06752), Gidney & Ekerå 2019 (arXiv:1905.09749) and Gidney 2025 (arXiv:2505.15917), with formula and arithmetic-consistency checks | results/results.json → regression |
| B4 | 50-digit reference values for Weibull and lognormal arrival-time models (illustrative test vectors, not forecasts) and known numerical pitfalls | results/results.json → arrival |
| Case study | Outputs of a closed-source estimator evaluated as a black box (recorded data only, no tool source code) | data/case_study_closed_source_estimator_978def8.json |
results/results.json is the complete output used in the paper (Python 3.13, Qiskit 2.6.2, Qiskit Aer 0.18.2, NumPy 2.5.4, SciPy 1.18.1, mpmath 1.4.1). The tables/*.jsonl files are flattened views of its factoring, ecdlp and qasm15 sections for the dataset viewer; no values were changed.
Main fields
- factoring:
N, basea, orderr, register sizesn/t, totalqubits,factors, total-variation distancetvdandmax_absbetween closed form and simulation, success probability from closed form (p_success_cf) and simulation (p_success_sim), runtimesecs. - ecdlp: curve
p,A,B, group orderq, generatorG, secret keyk, public pointQ, ideal Z_q variant TVD/success, and per-register-size (qubit_m) success probability, qubit count and Qiskit-vs-NumPy TVD. - qasm15_roundtrip: base
a, orderr, gate counts (u,cx), TVD of exported and re-imported circuits, success probability and SHA-256 of the.qasmfile.
Reproduce
git clone https://github.com/leanderdulac/exactbench && cd exactbench
python -m venv .venv && . .venv/bin/activate
pip install -r requirements.txt
python run_all.py # ~5 min on 8 cores -> results/results.json, results/qasm3/*.qasm
python make_tables.py
License
Data files are released under CC-BY-4.0 (as stated in the GitHub repository); the code in the GitHub repository is Apache-2.0 (LICENSE included here).
Citation
See CITATION.cff.
@software{francademello2026exactbench,
author = {França de Mello, Leandro},
title = {exactbench: exact-answer benchmarks for quantum cryptanalysis resource estimation},
version = {1.0.0},
year = {2026},
url = {https://github.com/leanderdulac/exactbench}
}
Author: Leandro França de Mello (IA Instituto, Brazil), ORCID 0009-0009-9304-3862.
- Downloads last month
- -