Deterministic evaluation demonstrates evidence recovery improves downstream task completion across 944 configurations in the Orchard framework, highlighting tradeoffs in summary compression.
A summary may answer a current question correctly while losing distinctions required by a later responsibility. We study this relationship in two finite deterministic campaigns within the Orchard Cognitive Framework. Interlayer Observability 01 (IO-P1) crosses measurement geometry with the representation passed between processes. Complementary acquisition and absolute responses jointly reconstruct all 336 early maps per full route; either ablation leaves ambiguity while preserving local answers. Representation and Recovery 01 (RR-P1) prospectively tests whether a receiver can recognise insufficient representation, recover eligible evidence and recheck current authority. Across 944 configurations per arm, declared-family compression with recovery completes 532 later tasks, versus 320 without recovery and 488 with current-question retention. Equally informed Orchard and conventional implementations agree. Full retention completes 576 later tasks with less inquiry, preserving a substantive limit on the compression policy. A public ledger reduces duplicated histories while retaining scoped decisions; runtime, source storage, reconstruction and native owner growth are accounted for separately. Elementary fibre and factorisation conditions explain the information boundary. The contribution is an audited finite integration of representation loss, recovery opportunity and receiving authority, not a new sufficiency theorem or a general solution to alignment.
No takes yet. Share an insight, caveat, or question.
KIMBERLEY LAVERNE ASHER (2026) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: