Overview This dataset provides the complete pre-computed oracle databases for the Scale-Dependent Gravitational Field Theory (SDGFT) Machine-Learning Toolkit (sdgft-ml-toolkit). The two Parquet files encode a dense, high-resolution mapping from the two free SDGFT parameters (Δ, δg) to 37 physical observables spanning cosmology, particle physics, gravitational-wave astronomy, and astrophysics. These files serve as the ground-truth reference for the GNN surrogate model, the CVAE inverter, and the experimental-validation pipeline described in the accompanying code repository. Files File Rows Columns Size Description oracle_db.parquet 1,000,000 39 ~3.2 GB Full parameter sweep (1000 × 1000 grid). Δ ∈ [0.01, 0.50], δg ∈ [0.001, 0.100]. Snappy-compressed. oracle_gold.parquet 10,000 39 ~1.9 GB Gold-standard subset: Quadruple precision (128-bit) with adaptive Gauss–Kronrod quadrature. Used for validation. Parameter Space The SDGFT is a two-parameter extension of General Relativity that promotes Newton's constant to a scale-dependent running coupling G(k). Anomalous dimension (Δ): [0.01, 0.50] — Controls the power-law running of G(k) in the deep UV. Graviton mass gap (δg): [0.001, 0.100] — Dimensionless IR deformation parameter. Controls late-universe deviations from ΛCDM. Axiom Point: (Δ* = 0.2083, δg* = 0.0417) is the unique fixed point where all 37 observables simultaneously agree with experimental data within 2σ. Observable Catalogue (37 columns) 1. Cosmological Observables (10) H0, Omega_m, Omega_Lambda, Omega_k, sigma_8, n_s, r_tensor, S_8, z_eq, t_universe 2. Particle Physics Observables (8) m_higgs, m_top, m_W, m_Z, alpha_s_MZ, sin2_theta_W, m_electron, alpha_em_MZ 3. Gravitational-Wave Observables (7) f_gw_peak, Omega_gw, h_c_nHz, dephasing_BBH, delta_v_gw, f_ring_BH, tau_ring_BH 4. Astrophysical Observables (7) M_TOV, R_14, Lambda_tidal, v_rot_flat, M_BH_shadow, Gamma_PPN, Beta_PPN 5. Quantum-Gravity Signatures (5) d_spectral, S_BH_correction, E_trans_LIV, tau_proton, G_Newton_eff Data Format & Usage Format: Apache Parquet (v2.6), Snappy compression. All values are stored as float64. import pandas as pd # Load the database df = pd.read_parquet("oracle_db.parquet") # Access axiom point axiom = df.query("abs(Delta - 0.2083) < 0.0005 and abs(delta_g - 0.0417) < 0.00005") Relation to Code Repository These data files are consumed by the sdgft-ml-toolkit Python package for GNN training and CVAE inversion. Repository: https://github.com/cosmologicmind/sdgft-ml-toolkit