Loading...
Loading...
Contributors:
Dataset URI:
10.5281/zenodo.23489995Tekmerion claim bundle for Spec 013 (red-team remediation), bundle version 0.2.0. This is the engineering truth record for one part of the Tekmerion research engine: the claims the spec makes, the test that could prove each claim wrong (its falsifier), the standing of each claim, and the assessment of how the evidence held up. Deposited 2026-10-20 so this record is anchored outside the code repository (red-team follow-up RTE-23 — the other spec bundles were anchored on 2026-10-13; this one was held back until its after-build assessment existed). The SHA-256 hashes below are the anchor values: any copy of these files, in the repository or elsewhere, can be checked against them. Files in this deposit: - aee-claims.json — SHA-256 ec2f1ac23a7b2a6a93d1f43f94808088cecc178e7859b6e3f4eb82c31160f0f4. The claim bundle: 47 claims (1 draft, 46 supported), each with its falsifier. - aee-assessment.json — SHA-256 e51875a1627deecdbf3c14ad7fa14f51546488afe8b44acb8c4309ca930d7ebe. The assessment: made on paper, before the build; its instruction was to work through the failure modes it found and reassess, and its summary reads: "Assessed 47 claim(s); found 222 failure mode(s); 0 claim(s) below the 0.70 confidence threshold.". - aee-assessment-after-implement.json — SHA-256 fba9b06819ab7f03f8bde135c737b44a62b1c9c5bc0682347660589ad31c999b. The after-build assessment: made after the build; its instruction was to clarify one claim's wording before the record could be closed, and its summary reads: "Assessed 47 claim(s); found 223 failure mode(s); 1 claim(s) below the 0.70 confidence threshold.". Standing, stated plainly: of the 47 claims, 46 stand as supported and checked on shipped evidence. One claim, CLM-013-25, remains a draft at a low score (0.250, below the program's 0.70 confidence threshold), and the after-build assessment's overall outcome is therefore CLARIFY — one claim's wording to settle, not a failed build. CLM-013-25's text says every evidence link must resolve. What was built is narrower: links in platform forms must resolve or the write is refused, while references to an organisation's own records are labeled operator-attested — recorded as stated by the operator, never presented as checked. Closing the claim needs a decision on its wording; this deposit anchors the record exactly as it stands, shortfall included. What this spec fixed, re-measured: the acceptance battery re-ran all 57 attacks the original red team had demonstrated against the system. 56 now fail, 1 is contained (it works only against a weaker check that the record itself warns about), and none succeeds. The compatibility measurement compared every frozen and previously deposited record before and after the fixes and found no silent changes — every earlier verdict and record stands exactly as issued.
Keyword(s):
anchoring; assessment; claim bundle; research integrity; tekmerion.
Grants and Funding:
No Grants or Funding Data Available
Associated Publication(s):
No Associated Publication Data Available
Supplementary Data:
No Supplementary Data Available
Related Dataset(s):
No Related Dataset Data Available
The National Library of Medicine (NLM) provides information about biomedical datasets. Inclusion on an NLM Web page does not imply endorsement of, or agreement with, the contents by NLM or the National Institutes of Health. Learn more: NLM Web Policies (nih.gov)