CUF Persistence Collapse Tests Across Six Structural System Classes Under Invariant Protocol and Transformed Observables

This record contains the locked pre-registration for an empirical benchmark testing persistence-collapse behavior in dynamical systems under an invariant protocol. The benchmark evaluates six structural system classes: Standard Map, Arnold Cat Map, Baker Map, Tent Map, Lozi Map, and Lorenz 63. The primary endpoint is hard persistence, defined as uninterrupted corridor retention across the full post burn measurement window; the secondary endpoint is soft persistence, defined as the fraction of measurement blocks in which identity remains intact. The benchmark also evaluates operator-dependent perturbation effects using baseline, noise, sign limited, saturation-limited, piecewise rule-switch, and reversal operators. Pre-registered predictions concern the existence and ordering of persistence collapse transitions, divergence between hard and soft persistence near the cliff, cross system structural differences between conservative and dissipative regimes, and benchmark failure conditions that count against universality under the locked protocol. This upload is intended as a pre-results record: protocol parameters, predictions, outcome classifications, and falsification conditions are fixed prior to inspection of production results. The record supports reproducibility, auditability, and priority for the benchmark design before release of code outputs and empirical result packages.