Native-language eminence calibration corrects cultural, gender, and domain blind spots in historiometric neurodivergence databases: evidence from 480,078 Chinese-Wikipedia biographies

Background. Retrospective (historiometric) neurodivergence (ND) studies derive eminence and biographical coverage almost exclusively from English-language knowledge graphs (Wikidata/DBpedia/English Wikipedia), which are vulnerable to WEIRD sampling bias. Methods. (1) The English-graph failure is shown directly: DBpedia types 579,388 athletes and 3,026 Chinese athletes, but adding an English-abstract condition reduces the Chinese-athlete count to zero. (2) The October 2026 Chinese-Wikipedia SQL dump is parsed via the new normalised schema (cl_target_id -> linktarget), yielding 480,078 named persons with a native-language eminence proxy (page_len), birth year, gender, occupation and cultural circle. Four hypotheses are tested (Fisher exact; Mann-Whitney with rank-biserial effect). Results. For the five East Asian circles where the name-matching channel is verified at 100% (1,231/1,231 existing cards resolve), only 2.2% (55/2,500) of the top-500 native-eminence figures are present; pooled non-Western coverage is 1.9% vs 5.6% for Western controls (OR=0.329, p=3.0e-6). Omitted figures sit at the 98.5th percentile of native eminence (vs 99.6th included; rank-biserial r=.32) - the deficit removes the top of the distribution, not the tail. The gap is compounded for women (1.6% vs 7.0%, OR=0.22, p=2.4e-5) and sports (0.4% vs 2.0%, OR=0.19, p=.015). A 4,042-person remediation pool (702 women) is released. Conclusions. Native-language calibration is a prerequisite for cross-cultural historiometric ND research. No living person is diagnosed (Goldwater Rule); labels are most-plausible hypotheses. Files: manuscript PDF/docx; data+scripts in the zip. Source: Chinese-Wikipedia 2026-10-01 (CC-BY-SA). Single author, CC-BY-4.0.