Strong, Coherent Evidence for Third-Generation BSM Physics: A Stress-Tested Global Fit of the 4321 Leptoquark Model

We present the final, maximally stress-tested version of the combined third-generation anomaly analysis, addressingevery substantive critique from independent reviewers: model-building freedom as an additional look-elsewhere axis;shared electroweak radiative-correction uncertainty across the Vud triplet; tail-statistic sensitivity; the demand for aproper global χ2/dof test; and the conditions under which Wilks’ theorem applies to the ∆χ2 comparison. Key results:(1) A Bayesian directional test compresses the 5-dimensional anomaly space onto the direction predicted by the 4321model in 2018, yielding 6.78σ (TF~10 for the B-physics sector; conservative composite TF~32 yields 6.36σ). (2) Aholdout global ∆χ2 test — with parameters fixed to R(D) only, making R(D*) a fully independent cross-check — yields∆χ2=47.82 for 2 d.o.f. (6.65σ), validated by parametric bootstrap (200,000 trials, Wilks tail accurate to 3%). (3) After allidentified inflation sources (0.75σ total), the deflated Fisher combination exceeds 5σ for trial factors up to 200,000. (4)The Vud coherence sub-argument survives a maximal radiative-correction adversarial test with only a 0.59σ penalty. Thecorrect headline is the ∆χ2 comparison and Bayesian directional test; Fisher is a cross-check. Five pre-registeredexperimental predictions are made