Seed crossval
v306_seed_crossval.py · registry
Read the original description and evidence links
Leave-one-out cross-validation of the shared seed phi0: back-solving phi0 independently from the 5 seed-grammar observables (sin^2 theta12, sin^2 theta13, beta, Omega_b, lambda_C; NuFIT 6.0 / ACT DR6 / Planck / PDG centrals) and error-weighted-averaging reproduces the axiom seed phi0=1/(6pi)+48 c3^4=0.0531720 to 0.04% (one real seed, not a per-observable fit); leave-one-out predicts every held-out observable within 3 sigma out of sample (most-tensioned = sin^2 theta13, ~2 sigma); a data<->formula shuffle explodes the chi^2 by >100x (power control)
Read the original Python source- Linked claims