Submitted:
22 October 2026
Posted:
23 October 2026
You are already at the latest version
Abstract
The predictive organization previously identified in SPARC galaxy rotation-curve discrepancies [1] is tested here across a substantially larger and more distant population. We use 1,387 galaxies drawn from four independently acquired PROBES source surveys. SPARC is used only to help establish the stellar-disk normalization and an empirical radial gas-velocity reconstruction. For each galaxy we define the signed squared-velocity discrepancy, ΔV²(r) = Vobs²(r) − Vbar²(r), characterize its radial structure, and test whether population relationships learned from other galaxies predict the withheld galaxy. Strict leave-one-galaxy-out prediction gives a velocity RMSE of 47.34 km s⁻¹, compared with 50.86 km s⁻¹ for a population-average residual control, a 6.92% reduction across 75,722 held-out radial predictions. Each of the four constituent surveys independently shows positive error reduction. A leave-one-survey-out test, training on three surveys and predicting the fourth, gives a pooled 7.20% gain; the SCII survey is statistically unresolved. In survey-stratified grouped withholding, the gain remains 6.81% even when approximately half of the 1,387 galaxies are excluded from model construction. The composite sample has a median distance of 88.7 Mpc, approximately 5.2 times that of the original SPARC population. These results show that the same galaxy-level predictive architecture identified in SPARC can be recovered independently in a much larger, more distant, and less homogeneous four-survey population, even though its baryonic baseline must be reconstructed rather than supplied directly. Fixed SPARC coefficients expectedly do not transfer unchanged to PROBES, and direct near-to-far prediction is not established due to survey heterogeneity. Against a training-fitted radial-acceleration relation (RAR), the PRIM (Predictive Residual Informational Mapping) to PROBES is competitive overall but not uniformly superior; its physical origin remains unresolved.
Keywords:
galaxy rotation curves
; rotation-curve discrepancy
; predictive information
; cross-validation
; PROBES
; SPARC
; radial acceleration relation
Copyright: This open access article is published under a Creative Commons CC BY 4.0 license, which permit the free download, distribution, and reuse, provided that the author and preprint are cited in any reuse.