A.6.6:5.2 - Episteme archetype: an evaluation result used as evidence

Tell. A report says that model M improved accuracy by 4%. The team points to EvalRun-2025-10-12, but that Work occurrence is neither the claim nor an evidence relation, and its log carrier does not become evidence merely by being attached.

Show. First recover the accuracy measure, reference condition or population, and percentage convention from the reported comparison. If those operands cannot be recovered, the exact 4% comparison claim remains unidentified; stop before relying on that numerical target. Once they are recovered, identify the result episteme that states the measured comparison and the target claim about the 4% improvement. State the exact A.2.4 evidence-use relation between that episteme and claim, including the relevant ClaimScope, polarity, window, and receiving use. If the decision also needs replayable source, carrier, provenance, currentness, or bounded-reliance information, use A.10 to cite the evaluation Work, its actual operation-result binding, the result episteme, the log carrier, and their independently obtaining direct relations.

Stop with the short evidence-use statement when it answers the question. No validatedBy(claim, Work) edge or scoped base-declaration record is required. If a project later needs a reusable evidence-relation declaration, that direct relation must first have its own participant meanings, predicate, applicability, and occurrence-identity rule.