The curator engine parses and groups disparate, raw parametric memory statements across independent probing runs and model providers. It applies a structured schema pass to match semantically equivalent assertions against an active ledger of canonical claims, updating their wording when new evidence refines them while preserving version history. Each merged claim tracks cross-model consensus, recording which platforms recalled the fact, how frequently it appeared, and its average confidence score. This creates a consolidated inventory of persistent model beliefs, ready to be evaluated against research findings or validated through human editorial rulings.