Bridge Paper: WCM-PL × Quantum Optimization A First QAOA-versus-Classical Cross-Check of WCM-PL's Vertex Assignment on Real Earthquake Data, via blueqat's Quantum Computing MCP Server
Masanori Watabe, Claude Sonnet 5 Kurado · Zenodo (CERN European Organization for Nuclear Research) · 2026
WCM-PL's core operation -- unit-normalize a point and assign it to its nearest reference-polytope vertex byargmax dot product -- is a deterministic, classical nearest-neighbor rule. This paper asks a narrow,concrete question: does a quantum optimization algorithm, given the SAME assignment problemreformulated as a QUBO (maximize alignment, subject to a one-hot vertex-selection constraint), recoverthe same answer? Using blueqat's quantum-computing MCP server (a Model Context Protocol interfaceexposing gate-circuit simulation, QAOA, VQE, and real quantum hardware access), 20 points from aCell5Sensor-embedded window of the JMA earthquake catalog around the 2016 Kumamoto M7.3mainshock were each formulated as an independent 5-qubit, 15-term QUBO (sized to fit the service'sfree-tier resource limits) and solved via QAOA (steps=2, shots=256, the free tier's maximum). QAOA'sbest-sampled bitstring matched WCM-PL's classical assignment in 18 of 20 points (90%). The twomismatches were investigated directly rather than left unexplained: re-running both showednon-reproducible answers (a different wrong vertex on repeat for one point, and the correct vertex onrepeat for the other, despite an unambiguous 0.71-margin gap to the second-best vertex) -- evidence ofclassical-optimizer non-convergence at this shallow QAOA depth and shot count, not of QAOA discoveringa genuine alternative structure. This paper reports the result at face value: a first, small-scale but concretedemonstration that WCM-PL's classical assignment rule and a quantum variational optimizer agree on realdata far more often than they disagree, alongside an honest, checked account of where and why they partways. ※ The License for codes is AGPL 3.0-or Later