A Property-Weighted Preservation-Gated Residual-Response Screen for Galaxy Rotation Curves

This Version 4 record provides a journal-oriented, non-peer-reviewed repository/preprint version of the manuscript “A Property-Weighted Preservation-Gated Residual-Response Screen for Galaxy Rotation Curves”. This version reformulates the work as a conservative empirical state-selector screen. The objective is not to propose a replacement law of gravity, exclude dark matter, or claim a lensing-ready physical source model. Instead, the manuscript tests whether target-independent galaxy-state properties can predict when a RAR-preserving baseline should remain unchanged and when a bounded weak-dwarf residual-response branch may activate. The recommended Version-4 branch uses a property-weighted preservation gate and a 2% acceleration tolerance guard. In the internal SPARC replay, the Stage180 core selector reduces the equal-galaxy log-RMSE from 0.192667 for the RAR reference to 0.179855, corresponding to a change of -0.012813 dex. It gives 37 exact improvements, 136 exact ties/preservations, and 2 exact degradations; both exact degradations are within the small tolerance band, giving 37 practical improvements, 138 practical ties/preservations, and 0 practical degradations under a 2% acceleration tolerance. The record includes the manuscript and reproducibility materials. The supplementary package contains the formula-lock description, Stage180 score rules, summary tables, per-galaxy replay outputs, proxy/overlap check summaries, and a verification script reproducing the main reported values. The result should be interpreted as SPARC-internal and proxy-supported evidence for a falsifiable property-weighted state-selector architecture, not as independent external validation. The decisive next test remains a frozen no-retuning replay on fully decomposed non-SPARC rotation-curve data with pointwise R, Vobs, eVobs, Vgas, Vdisk/Vstar, and Vbul where applicable. This record is intended for transparency, reproducibility, and public dissemination of the submitted manuscript version and does not constitute peer-reviewed journal publication.