docs/workspace/compare-evaluation-analysis/structured-compare-calibration/latest/synthetic-hiring-replica-semantic-instability/scenario.md