Multiverse Parameter Dataset: 2000 Self-Consistent Universes Generated by Neural-Guided Symbolic Regression

Candidate universes were generated via Monte Carlo sampling with parameter ranges spanning up to 6 orders of magnitude. Each candidate universe was subjected to an 8-layer physical self-consistency filter, including flatness, cosmic age, Big Bang nucleosynthesis, structure formation, causality, no-ghost condition, quantum stability, and user-defined constraints. A total of 2,000 candidate universes passed all eight layers and were retained in the final dataset.Data Content:Each universe is represented by a set of 37 physical constants, including but not limited to: the gravitational constant GG, the speed of light cc, the reduced Planck constant ℏℏ, the Boltzmann constant kBkB​, the Hubble constant H0H0​, the dark energy density ΩΛΩΛ​, the string coupling gsgs​, its dual g~sg~​s​, the string mass MsMs​, the string length ℓsℓs​, the compactification radius RR, the dual radius R~R~, the holographic dimension NN, the 10-dimensional gravitational constant G10G10​, and the cosmological constant ΛΛ.

The dataset also provides the coefficient matrix for 66 power-law relations discovered via symbolic regression, as well as the detailed pass/fail status for each of the 8 filtering layers.Data Format:All data files are provided in CSV format (Unicode text, .csv), which can be accessed using standard data processing software such as Excel, Python Pandas, R, or MATLAB.Reuse Potential:This dataset can be used to validate power-law relations among physical constants, test the effectiveness of the self-consistency filtering criteria, and explore the structural characteristics of the multiverse parameter space.