{"graphSha256":"f29d7b1eedeb88387238d2abda347f810d2179c5fdbead47c3ef48507ab20ca5","node":{"id":"GH078","type":"ficha","title":"L-09 — Remove accidental public-label leakage without rewriting source examples","url":"https://dakleyer.github.io/perspective-beyond/sources/l-09-remove-accidental-public-label-leakage-without-rewriting-source-examples-gh078/","original":"https://github.com/dakleyer/structural-awareness-contributions/blob/d41e2d4064dcdff4b4a2e80eb45b61c63f53e4dd/research/ecosystem-awareness/baseline/01I-artifacts/acc-auxiliary-model-v0.1/studies/conscientia-campaign-v0.1/L09_MATCHED_PUBLIC_VIEW_PROTOCOL.md","aliases":[],"clusters":["research"],"tags":["evidencia","experimentacion"],"summary":"","commit":"d41e2d4064dcdff4b4a2e80eb45b61c63f53e4dd","sourcePath":"research/ecosystem-awareness/baseline/01I-artifacts/acc-auxiliary-model-v0.1/studies/conscientia-campaign-v0.1/L09_MATCHED_PUBLIC_VIEW_PROTOCOL.md","commitUrl":"https://github.com/dakleyer/structural-awareness-contributions/commit/d41e2d4064dcdff4b4a2e80eb45b61c63f53e4dd","documentId":"doc-17f79bed02b7dcc294","tagWeights":{"evidencia":0.3333,"experimentacion":0.6667},"contextTokens":168},"relations":[["apoyo:GH078:C46","GH078","C46","documenta el concepto","",null,"apoyo",""],["apoyo:GH078:C124","GH078","C124","documenta el concepto","",null,"apoyo",""],["apoyo:GH078:C125","GH078","C125","documenta el concepto","",null,"apoyo",""],["apoyo:GH078:G17","GH078","G17","documenta el concepto","",null,"apoyo",""],["documento:G094","GH078","T2803","candidate transfer","Matched public-view counterworlds can test whether a reported acquisition benefit comes from new evidence rather than accidental fixture-label leakage.",null,"documento","La fuga es un confusor para políticas futuras; la fuente no afirma que los brazos fijos existentes la exploten."],["documento:G100","GH077","GH078","documented use","A fallible audit mechanism and a public-label leakage control address different confounders; both must remain visible in a fair comparison.",null,"documento","Los dos PRs siguen abiertos en el corte; sus controles sintéticos no acreditan efectos en proveedores o auditorías reales."]]}
