Energy Filament Theory · EFT Full KB
09_Geom_Consistency: geometry-consistency audit
V30-30.8 · G 判决节 / 审计节 ·
30.8 returns V30’s geometry_redshift_window to a full-sample supplementary audit: using the full Pantheon SN sample (1048 data points), one release-level diagonal Gaussian likelihood, flat ΛCDM with {Ωm, M_B}, and a quadratic tension-evolution Energy Filament Theory (EFT) model H(z)/H0 = 1 + α z + β z² with {α, β, M_B}, the source uses AICc as the primary comparison metric because model size differs (k = 2 vs 3) and records AICc = -1172.351 vs -1173.238, so ΔAICc = -0.887 and uniqueness is overturned, while the section stays a single-probe Pantheon return rather than a joint SN+BAO global fit.
Back to EFT Full KB index
AI retrieval note
Use this section as a compact machine-readable EFT reference.
Keywords: Energy Filament Theory, EFT, geometry consistency, Pantheon, full SN sample, quadratic tension-evolution, AICc-led audit, geometry_redshift_window, SN recap, BAO recap, single-probe boundary, equivalence verdict
Section knowledge units
thesis
30.8 is the geometry-consistency return of Volume 30’s geometry_redshift_window. The source reopens geometry on the full Pantheon SN sample (1048 data points), comparing flat ΛCDM with a quadratic tension-evolution Energy Filament Theory (EFT) model under the same data, one release-level diagonal Gaussian likelihood, and symmetric priors. Because the reported gap is only ΔAICc = -0.887, the section does not establish a unique expansion-geometry law. Its report-bounded job is narrower: to show that when the geometry line is revisited on a full-sample SN supplement, uniqueness still fails.
mechanism
The audit is simple but not equal in parameter count. The dataset is the full Pantheon release standardized to zcmb, mb, and dmb, and the likelihood is a diagonal Gaussian using dmb². Flat ΛCDM uses {Ωm, M_B}; the quadratic EFT side uses {α, β, M_B}. Because the models carry k = 2 versus k = 3, 30.8 does not pretend to inherit the equal-parameter symmetry of 30.3 or 30.7. Instead, the source makes AICc the primary fairness metric and fits both sides with the same coordinate-descent release-level workflow.
evidence
The numerical verdict is modest but important. The source writes flat ΛCDM as H(z)/H0 = sqrt(Ωm(1+z)^3 + (1−Ωm)) with parameters {Ωm, M_B}, and the quadratic EFT model as H(z)/H0 = 1 + α z + β z² with {α, β, M_B}. After coordinate-descent optimization, the AICc table reads -1172.351 for ΛCDM and -1173.238 for the source’s EFT_v2 fit, so ΔAICc = -0.887. RMSE and logLmax show only a slight EFT-side edge. Under the report’s Release-Level · Uniqueness · Equivalence v1.3 criterion, |ΔAICc| < 2 means equivalence rather than uniqueness.
boundary
30.8 must stay bounded in two ways. First, it is a single-probe supplementary audit built only on Pantheon SN geometry; it is not a joint SN+BAO fit, not a multi-probe global consistency pipeline, and not an ontology-bearing statement about why geometry works. Second, its fairness frame is release-level and deterministic—diagonal Gaussian errors, coordinate descent, and AICc-led comparability under unequal k. The section’s real claim is therefore limited but useful: a full-sample Pantheon return also fails to preserve ΛCDM uniqueness.
summary
The reproduction block is concise: pantheon_clean.csv, baseline_lcdm_v3.json, baseline_eft_v3.json, baseline_comparison_v3.json, and the runner .\scripts\run_fit_baseline_v3.ps1. Inside Volume 30, 30.8 summarizes 30.3’s SN Hubble-flow spine and 30.7’s BAO standard-ruler ledger at the geometry_redshift_window level. It does that not by importing BAO into this source report, but by returning the bundle to a full-sample SN geometry audit after both the candle leg and the ruler leg have already denied one-probe uniqueness.