Empirical Falsification of the Persistence Collapse Sector of Conditional Unlocking Fields Across Six Structural System Class

This archive contains the preregistered V4 empirical benchmark package for testing the persistence collapse sector of the Conditional Unlocking Fields (CUF) framework across six structural system classes under a locked invariant protocol. The benchmark studies the standard map, Arnold cat map, baker map, tent map, Lozi map, and Lorenz-63 under six operator families: baseline, noise, sign, saturation, piecewise, and reversal. This study does not implement the full canonical CUF formalism and does not claim total validation of CUF. Instead, it operationalizes a specific empirical sector of the framework: persistence maintained under constraint, violation or overflow relative to a corridor, and collapse of persistence under controlled parameter variation. The primary preregistered observables are hard persistence, defined as no exit over the post burn window, and soft persistence, defined as occupancy fraction over the post burn window. Additional recorded diagnostics include first exit timing, overflow magnitude, recovery count, transition width, and related summary quantities. The purpose of this benchmark is to determine whether CUF style persistence collapse behavior is empirically detectable across structurally distinct systems under a single locked protocol, and to distinguish cases of strong support, partial support, preregistered null behavior, and protocol mismatch or failure behavior. The benchmark is therefore falsification-oriented and preserves mixed outcomes rather than forcing a universal success interpretation. This archive preserves the full benchmark evidence package, including the preregistration aligned code, configuration files, sweep level summary outputs, transition summaries, trajectory level metrics, trace bundles, warning logs, and generated figures. It should be read as a downstream empirical operationalization of the persistence collapse sector of CUF rather than as the canonical CUF source itself. Canonical CUF mathematics remains upstream of this archive; this record documents the empirical V4 benchmark layer.