The paper changes the audit unit from one answer to a query family.
Its main distinctions are:
- destructive erasure: an attribution atom is absent and not recoverable within the audit budget;
- indexical erasure: the missing atom can be recovered from surviving material through a bounded number of queries;
- attribution sharding: no single rendering carries complete attribution, but the family union does.
The metric suite includes:
- DD — Dereference Depth
- RR — Recoverability Ratio
- Shard Coverage
- ACP — Atomic Co-presence
- FC — Family Coverage
- ASI — Attribution Sharding Index
- CC — Complementarity Coefficient
- Family Erasure Skew
- CSC — Claim–Source Convergence
The worked Empty Bracket pair is described as:
- full family coverage;
- incomplete best-single-rendering coverage;
- positive sharding;
- one-step recoverability for the missing atoms.
The protocol requires:
- an ex-ante query family;
- externally anchored attribution atoms;
- fresh-session capture discipline;
- bounded dereference rules;
- separate reporting of total completeness and fragmentation.
The absorption hypothesis proposes that extremely faithful compression of a source’s self-description may cause the source to disappear as a separately cited object. The paper explicitly states that one specimen does not establish the mechanism.