PIR-JEPA SPARC: Automated Hidden Physics Detection in Galaxy Rotation Curves This paper applies PIR-JEPA (Physics Intermediate Representation with JEPA-based Langevin diffusion expression priors) to the SPARC galaxy rotation curve dataset (Lelli, McGaugh NGC 0024 at 2/3. Mean Delta-s = 0. 042 across confirmed galaxies. The Delta-s hidden physics signal is genuine, not a pipeline artifact. Stage 3 — JEPA candidate scoring (v3): Direct scoring of JEPA-generated candidate expressions across all 12 top galaxies: log (|vbar|+1) wins 10/12 galaxies (mean MAE = 0. 093) log (|r|+1) wins 2/12 galaxies (CamB, UGC07323 — gas-dominated dwarfs with vbar ≈ 0) Log candidates win 12/12 galaxies in total Grammar baseline 1/r²: mean MAE = 0. 753; vbar·r^ (-1/2): mean MAE = 0. 317 Physical interpretation: The dominance of log (|vbar|+1) is consistent with the Radial Acceleration Relation (McGaugh et al. 2016): the hidden correction depends on baryonic velocity, not radius alone. The two log (r) outliers are physically explained by near-zero baryonic velocity in their outer regions, where the correction must become spatial. This is the first data-driven, grammar-free recovery of a RAR-consistent symbolic correction term from the SPARC dataset, without prior knowledge of MOND, NFW profiles, or the RAR itself. Related work: PIR-JEPA v2: https: //doi. org/10. 5281/zenodo. 19477508 PIR Architecture v3: https: //doi. org/10. 5281/zenodo. 19428230 PIR-JEPA LIGO v2: https: //doi. org/10. 5281/zenodo. 19586642
Hanif Muhammad (Fri,) studied this question.