DYNESTY NESTED SAMPLING Bio-Cubic Universe — Paper V Results Jean-Yves Lozach — Ploumagoar, Bretagne — 9 mars 2026

PHYSICAL INTERPRETATION Why +4.20 more than emcee? The harmonic mean estimator used by emcee systematically underestimates ln(Z) because it is dominated by the lowest-likelihood samples. dynesty computes the evidence integral directly via nested sampling, correctly accounting for the full prior volume. The +4.20 gain corresponds to the true Occam penalty being smaller than emcee estimated — the 3 new vacuum parameters (δ, M, α_phase) are genuinely well-constrained by the data. The Triptyque is confirmed. With g/λ ≈ 1.0–1.5, a single coupling simultaneously reduces S8 (4.38σ→3.07σ), H0 (5.60σ→4.19σ), and stabilises the vacuum (log₁₀(τ)∼169 years). dynesty finds consistent posteriors on all three tension-reduction parameters (β0, ε₀, γ₀). n = 2.812 ± 1.283 (dynesty) : The posterior is broader than emcee (2.035 ± 0.48) because dynesty properly explores the prior tails. The peak at n ≈ 2 is robust — this is the 5th independent confirmation of the Z₂ universality class. The secondary lobe at n ≈ 3.5 warrants investigation in MCMC v4.