How Qlro proves itself
Every claim on this site is backed by one of four artifacts: the peer-facing paper, the reference implementation, real hardware validation, or the live accuracy dashboard. Any auditor can reproduce any number here from these sources.
Real hardware results across IQM, Rigetti, IonQ: Pearson r = 0.96 on the IQM+IonQ stable subset, r = 0.69 on the drifted Rigetti Cepheus subset. The two-point r(τ) observation lives here.
Public accuracy dashboard: community-submitted (predicted, observed) fidelity pairs, rolling Pearson r and RMSE, monthly snapshot minted to Zenodo with its own DOI.
Workload-Conditioned Physical Projection framework. Peer-facing revision. Proves the locality axiom (adding/removing a vendor cannot change any other vendor's score). Appendix D.1 reports a failed ZNE extension — we publish negative results alongside positive ones.
Full SDK under Apache 2.0. pip install qlro. Reproduce every scoring step, every benchmark claim, and every entry in the accuracy dashboard locally.