Quantum Datasets Hub for Reproducible Benchmarks and QML (aqora.io)

Aqora launched a Quantum Datasets Hub: a centralized registry for publishing, discovering and reusing datasets tailored to quantum benchmarking and quantum machine learning (QML). The hub provides immutable, versioned datasets (publisher/name@version or aqora://publisher/name/vX), supports public or private team visibility, and enforces clean metadata (SPDX license, tags, size). It ships with quantum-focused collections—graph families for QAOA/MaxCut, molecular Hamiltonians for VQE, OpenQASM 3 circuit corpora—and classical companions used in hybrid workflows. Sample offerings include HamLib Binary Optimization and MNISQ; datasets are downloadable as Parquet/CSV/JSON and loadable directly into pandas or polars with pushdown filtering for efficient transfer. For researchers this matters because reproducible benchmarks depend on stable inputs and clear lineage. Aqora’s immutable versions plus simple Python load snippets, CLI/UI upload, and automatic Parquet conversion for directories make it easy to pin dataset bytes, log dataset versions alongside code commits, and rerun experiments across machines. The hub also enables private iteration before public release, citation-stable URIs in papers (DOI support planned), and straightforward integration with Qiskit/PennyLane/Cirq. Practically, teams can seed randomness, pin datasets, and share exact data artifacts—reducing friction in comparative QAOA/VQE evaluations and QML model training while improving reproducibility and discoverability.