Owner: [TENANT_CONFIGURATION_REQUIRED — AI Governance Lead] Status: Draft
Holds labeled/ground-truth datasets used by ai-evaluation-strategy.md: CV extraction ground truth, JD/candidate matching relevance judgments, fairness test profile sets (fairness/), and RAG groundedness question/answer sets (rag/).
All data in this directory must be fake/demo or properly de-identified — never real candidate/employee data (Section 7, root CLAUDE.md).
cv-extraction/ — labeled CV field-extraction ground truthmatching/ — JD + candidate batches with relevance judgmentsfairness/ — de-identified profile variants differing only in excluded characteristics (see fairness-and-bias-governance.md)rag/ — policy Q&A pairs with expected citationsNo datasets are populated yet in this documentation-only phase.