Empirical Benchmarks, Accuracy, and Decoding Throughput Datasets for qector-decoder-v3 (v0.6.8) on Low-Resource Edge Hardware

Empirical Benchmarks, Accuracy, and Decoding Throughput Datasets for QECTOR Decoder v3 (v0.6.8) on Low-Resource Edge Hardware Guillaume Lessard (qector.store) Software Platform: qector-decoder-v3 (Rust/Python Quantum Error Correction Decoding Platform) Resource Type: Dataset / Software Benchmark Artifacts Licensing & Availability: Benchmark Datasets & Reports: Creative Commons Attribution 4.0 International (CC-BY-4.0) Software Core (qector-decoder-v3): Source-Available (Free for non-commercial research, academic, and personal use. Commercial license required for commercial/enterprise deployment.) Abstract This record contains the complete empirical dataset, raw benchmark files, high-resolution vector figures, and the scientific PDF report for the release of QECTOR Decoder v3 (v0.6.8) by Guillaume Lessard. The primary objective of this benchmarking campaign was to evaluate the algorithmic accuracy, decoding throughput, and memory scaling of Minimum Weight Perfect Matching (MWPM) and Union-Find (UF) quantum error correction decoders operating in a resource-constrained edge environment. All empirical tests were conducted on an edge machine featuring an HP Dual-Core x86_64 CPU, 3.1 GB RAM, and no local hard drive (booted live from an AntiX Linux USB drive without swap memory). Key Benchmark Findings Exact Mathematical Parity: The compiled C++/Rust qector_blossom_weighted decoder achieves exact parity in Logical Error Rate (LER) and failure counts when compared against PyMatching (v2.4.0) across all tested physical noise rates ($p \in [0.002, 0.008]$) and distances $d \in \{3, 5, 7, 9\}$. 100% Algorithmic Faithfulness: All CPU decoders (FastUnionFind, UnionFind, Blossom, SparseBlossom, BatchDecoder, and CPUBatchDecoder) achieved 100% parity verification ($H \cdot \hat{e} \pmod 2 = s$) across odd distances $d \in [3, 19]$. Topological Speed Advantage at High Distance: While exact Blossom MWPM scales with graph complexity ($1.77 \times 10^4$ shots/s at $d=9$), qector_union_find maintains $1.62 \times 10^5$ shots/s at $d=9$ ($9.1\times$ faster), demonstrating the speed utility of topological decoders on low-power host CPUs. Complete File Manifest & Checksums Filename Size MD5 Checksum Description qector_decoder_v068_report.pdf 158.07 KB 8722ed2deab761f1ad01aa9966b54277 Compiled 3-page scientific PDF benchmark report. accuracy_circuit.json 3.54 KB 6a892a304cffaadb2568cd89a2aded2e Circuit-level LER datasets ($p=0.003$, 100k shots). faithfulness.json 7.99 KB 32b728d09c6bf238fff8cd2ff7de0621 Parity checks across distances $d=3 \dots 19$. manifest.json 2.03 KB f674dcf8c2816022be549bbdfcb639fb Machine architecture, dependencies, and wheel SHA256 hashes. throughput.json 29.67 KB d5fee43963f0a7922d8cf72e6952305c Phenomenological syndrome throughput sweeps ($N=1\text{k}\dots200\text{k}$). throughput_circuit.json 2.15 KB 7ead0ec636b3cc17544554aff9101ce3 Circuit-level memory decoding throughput dataset. ler_vs_distance.svg 62.05 KB 6268ab777313ac38457af93ae1789035 Vector graphic: Logical Error Rate vs Distance $d$. throughput_scaling_d5.svg 61.01 KB f6740958b7dcf1c3a725881c8189a3fa Vector graphic: Throughput scaling at code distance $d=5$. throughput_vs_distance_50000.svg 56.23 KB 96b38ae4c4158c33da37cb6832bc57b9 Vector graphic: Throughput vs distance at 50,000 shots. throughput_vs_distance_200000.svg 54.44 KB 70a4a7ffed6140d4f605cdc5f83def70 Vector graphic: Throughput vs distance at 200,000 shots. us_per_shot_200000.svg 53.98 KB 9b319c1e62f8bccd7082afcdfa0fcd65 Vector graphic: Latency ($\mu\text{s/shot}$) at 200,000 shots. bench_main.py 21.98 KB 0882f49a4ef39a337dee1344ace6b8a7 Main phenomenological benchmark suite script. bench_circuit.py 8.92 KB 75fb3a983de9220a9bdb0030f94f90e4 Stim-integrated circuit-level benchmark script. calibrate.py 4.32 KB eeb5b06a9ba3be1030f01a0471e208fc Quick execution calibration script. probe.py 1.76 KB 66324b75f34526c510ea0fb8e92bf04e Native extension C++/Rust bindings diagnostic script. stim_probe.py 1.80 KB 474f935b61b423aa228db1b9107b4e5a Stim DEM integration diagnostic script. qload.py 3.33 KB df6768750f70680db3fa2006e37e7a69 Python helper module to load and parse JSON datasets. BENCHMARK.md 2.37 KB 9b80add4ec5d03f2cf27b57ac116a936 User guide for reproducing benchmark runs. ddr2.txt 24.88 KB 19e9ad34f8d4829f5d98c029d1875deb Terminal execution log and console output capture. Environment & Software Specifications CPU: HP Dual-Core x86_64 (2 logical threads) RAM: 3.1 GB Available Operating System: Linux 5.10.240-antix.1-amd64-smp-x86_64 (Live USB Boot) Python Version: 3.13.5 Software Version: qector-decoder-v3 v0.6.8 Key Dependencies: pymatching==2.4.0, stim==1.16.0, sinter==1.16.0, numpy==2.2.6, scipy==1.18.0, ldpc==2.4.1, beliefmatching==0.2.0 Reprodicibility Steps Extract the repository archive files. Initialize a clean virtual environment: Bash python3 -m venv qector_env source qector_env/bin/activate pip install qector-decoder-v3==0.6.8 pymatching==2.4.0 stim==1.16.0 numpy scipy Run the analysis scripts: Bash python qload.py python bench_main.py python bench_circuit.py Official Software & Dataset Citations Software Citation @software{lessard2026qector, author = {Guillaume Lessard}, title = {{QECTOR Decoder v3}: Rust/Python Quantum Error Correction Decoding Platform}, year = {2026}, version = {0.6.8}, url = {https://www.qector.store}, note = {Source-available. Commercial license required for commercial use.} } Dataset Citation @dataset{lessard2026qector_benchmarks, author = {Guillaume Lessard}, title = {{Empirical Benchmarks, Accuracy, and Decoding Throughput Datasets for QECTOR Decoder v3 (v0.6.8) on Low-Resource Edge Hardware}}, month = jul, year = 2026, publisher = {Zenodo}, doi = {10.5281/zenodo.21501377}, url = {https://doi.org/10.5281/zenodo.21501377}} Licensing This record contains two categories of material, licensed under different terms. Data, reports, and documentation (datasets, JSON and CSV artifacts, figures, PDF reports, certificates, Markdown and text reports): made available under the Creative Commons Attribution 4.0 International License (CC-BY-4.0). You may share and adapt this material, including for commercial purposes, provided you give appropriate credit to Guillaume Lessard / iD01t Productions and indicate if changes were made. Code (all .py files, .pyi stubs, notebooks, and scripts): licensed under the PolyForm Noncommercial License 1.0.0. You may use, copy, and modify the code for noncommercial purposes only. Any commercial use, including use in or to develop a commercial product or service, requires a separate commercial license from iD01t Productions. The full license text is included in this record (LICENSE-POLYFORM-NONCOMMERCIAL-1.0.0) and at https://polyformproject.org/licenses/noncommercial/1.0.0/ . Copyright (c) 2026 Guillaume Lessard / iD01t Productions. All rights reserved except as expressly granted above. Note on prior versions: earlier versions of this record were published with a CC-BY-4.0 metadata license. That grant is irrevocable for copies already obtained under it. The terms above apply to this version and all future versions.