# Open Quantum Evidence Atlas — Analysis Report **Version:** 1.2.0 **Snapshot:** 2026-07-29T00:13:14Z **Author:** Takahiro Tsuchiya **Interactive Atlas:** https://atlas.tsuchiyalab.com ## Decision Proceed with the broad Open Quantum Evidence Atlas and use Q‑NEKO as a prospective observability benchmark. A Q‑NEKO-only product would currently be an empty graph; the 645-record EU27–Japan corpus supplies a defensible baseline while Q‑NEKO becomes a live test of policy-to-graph lag. ## Measured finding Only 17 of 645 EU27–Japan quantum publications (2.6%) expose a complete project–funding–publication–dataset–software chain. The supporting pattern is a visibility cliff: 392 (60.8%) connect to a project and funder, 179 (27.8%; Wilson 95% reference band 24.4–31.3) to a dataset, and 48 (7.4%; reference band 5.7–9.7) to software. A stricter 87-record title-literal sensitivity set preserves the pattern: 51.7% project-linked, 24.1% dataset-linked and 8.0% software-linked. “Connected” means at least one explicit OpenAIRE edge. Dataset and software edges are Graph observability signals; they do not by themselves establish access, licensing, documentation quality or actual reuse. An absent edge means the relationship was not observable in the audited graph, not that the underlying output does not exist. ## Method 1. Discover publications with eight quantum search phrases and deduplicate the Japan-query union. 2. Restrict the corpus to 2020–2026 publications with both Japan and EU27 affiliations. 3. Audit all 645 records for project and funder objects in research-product links. 4. Audit Scholix relations for all 645 observed records, including all 87 title-literal records. 5. Classify dataset and software edges and report Wilson 95% binomial reference bands for subset comparison. 6. Test Q‑NEKO names and identifiers separately as a prospective project-visibility preflight. ## Interpretation The current graph makes investment and project structures substantially easier to observe than reusable-output links. The 17 complete chains and the 60.8% to 7.4% fall—an 8.2× visibility gap—are therefore treated as an evidence-infrastructure gap, not as a claim that researchers failed to produce or reuse software. The operational response is to make grant, DOI, repository and Software Heritage identifiers part of project delivery and to re-audit their visibility over time. ## Alien / OpenAIRE MCP cross-check An authenticated single-record run through the official Alien/OpenAIRE demo resolved the featured DOI to the same OpenAIRE record and confirmed EC grant 101102140 (QIA-Phase1). Its typed dataset/software calls reported non-zero totals but returned no rows on pages 1 or 2. A controlled follow-up identified the cause: the official `/v1/researchProducts/links` operation defaults to page 0, where direct calls returned the one dataset and two software rows, while Alien's `ResearchLinksInput` rejected page 0 before any upstream request. The discrepancy is therefore a reproducible MCP schema/offset defect, not an OpenAIRE null. It is published as an interoperability diagnostic and is not merged into the 645-record denominator. ## Live integrity recheck On 29 July 2026, an uncached bounded recheck repeated the eight discovery-count queries, the featured DOI and grant lookup, the page-0/page-1 link contract, and all four Q‑NEKO aliases. Six of six integrity checks passed, all eight discovery counts matched the snapshot, and Q‑NEKO remained at zero projects and zero research products. A fresh full-corpus execution then exactly reproduced all 645 corpus IDs, all 645 dataset/software link states, and all four connection rates; the comparison is published as machine-readable JSON. ## Falsification and limits The full-corpus audit removes sampling error within the observed boundary, but does not turn that boundary into all EU–Japan quantum activity. Wilson intervals are binomial reference bands for comparing subsets; they do not quantify search-boundary, metadata or coverage uncertainty. Publication-year slices preserve the project-to-software gap in every complete year from 2020 through 2025; the 2026 cohort is partial through 5 May and is not interpreted as a trend. The “software cliff” would weaken if additional identifiers, repositories or improved OpenAIRE classification lifted software visibility toward the observed dataset rate. Search-phrase coverage, affiliation resolution, graph classification and snapshot timing still bound the inference. ## Q‑NEKO and GQSO Q‑NEKO is a watchlist case, not the observed corpus. Its absence at the snapshot date is interpreted as observation lag, not research failure. The Global Quantum Statecraft Observatory (GQSO) connection is prospective. The Atlas is a standalone OpenAIRE audit today. Its evidence-chain metrics can later be mapped into GQSO jurisdiction lanes, but this release does not claim a current partnership or completed technical integration. ## Reproducibility and licensing The executed notebook, evidence snapshot, connection-rate table, observed corpus, Scholix audit, Alien/OpenAIRE MCP cross-check and extraction pipeline are distributed beside this report. OpenAIRE Graph is acknowledged as the source under its CC BY terms; the snapshot declares the discovery, deduplication, affiliation-filtering and link-classification transformations. Code is MIT licensed. Derived data, analysis and documentation are CC BY 4.0.