Exploring the SMEFT landscape: Bayesian Model Selection for indirect discovery
Pith reviewed 2026-05-22 08:56 UTC · model grok-4.3
The pith
Treating SMEFT as a space of competing operator-subset hypotheses rather than one high-dimensional model enables Bayesian selection that finds no significant new physics while sharpening Wilson coefficient posteriors.
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
Core claim
Bayesian inference performed at the level of the discrete space of SMEFT operator subsets, navigated efficiently by a genetic algorithm and with evidence approximated by the Bayesian Information Criterion, assigns high posterior probability to the Standard Model and no statistically significant probability to any non-Standard Model operator subset when confronted with LEP and LHC Run 2 data; the resulting Bayesian model average posteriors on Wilson coefficients nevertheless exhibit substantially improved characterization potential compared with conventional global fits.
What carries the argument
Bayesian model selection over the discrete space of SMEFT operator subsets, which produces an operator correlation matrix that encodes the relational structure among high-posterior models.
Load-bearing premise
The Bayesian Information Criterion supplies a sufficiently accurate approximation to the true Bayesian evidence when comparing high-dimensional discrete operator-subset models.
What would settle it
A future measurement in the electroweak or top sector that produces a decisive preference for one particular non-Standard Model operator subset over the Standard Model would falsify the reported absence of significant evidence.
read the original abstract
We develop a framework for indirect discovery in the Standard Model Effective Field Theory (SMEFT) based on Bayesian model selection over operator subsets. We argue that SMEFT should be understood as a structured space of competing hypotheses rather than a single high-dimensional model, with each operator subset corresponding to a physically distinct low-energy realisation of new dynamics. Bayesian inference is applied at the level of model space itself, assigning posterior probabilities to operator subsets and marginal inclusion probabilities to individual operators. A genetic algorithm efficiently navigates the high-dimensional discrete model space, concentrating evaluations in the high-posterior region, while the Bayesian Information Criterion provides a tractable approximation to the Bayesian evidence. We apply this framework to a dataset comprising electroweak precision observables from LEP and Higgs, top-quark, and diboson measurements from LHC Run 2, at both linear and quadratic order in the Wilson coefficients, with one-loop renormalisation group evolution systematically included. The analysis finds no statistically significant evidence for any departure from the SM, and demonstrates that Bayesian Model Average posteriors on Wilson coefficients carry substantially improved characterisation potential compared to traditional global fits. The operator correlation matrix encodes the relational structure of the model posterior, identifying operator pairs that co-appear in high-posterior models and flat directions where additional measurements would be most valuable. The sensitivity of all results to the choice of matching scale $\mu_0$ is assessed, and its promotion to a continuous parameter of inference is identified as a natural extension of the framework.
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The manuscript develops a Bayesian model selection framework for SMEFT in which operator subsets are treated as competing hypotheses. A genetic algorithm is used to explore the discrete model space, with the Bayesian Information Criterion (BIC) serving as a tractable proxy for the marginal likelihood. The method is applied to a global dataset of LEP electroweak precision observables together with LHC Run-2 Higgs, top-quark and diboson measurements, both at linear and quadratic order and with one-loop RGE evolution included. The analysis reports no statistically significant evidence for departure from the SM, claims that Bayesian model-averaged posteriors on Wilson coefficients provide substantially improved characterisation relative to conventional global fits, and introduces an operator correlation matrix to expose co-appearance patterns and flat directions.
Significance. If the BIC approximation remains reliable in the presence of the strong correlations that arise from RGE mixing and flat directions in electroweak, Higgs and top data, the work supplies a principled way to incorporate model uncertainty into indirect SMEFT searches. The reported improvement in posterior characterisation and the diagnostic value of the correlation matrix could usefully inform experimental priorities. The absence of quantitative validation of the BIC approximation or of convergence diagnostics for the genetic algorithm, however, leaves the quantitative claims only moderately supported at present.
major comments (3)
- [§3] §3 (BIC approximation): The central results on model probabilities, inclusion probabilities and BMA posteriors rest on treating BIC = log L_max − (k/2) log N as an accurate proxy for the marginal likelihood. No quantitative comparison to exact evidence (or to nested sampling) is presented for even a modest subset of models, nor is any error estimate given for the approximation in the presence of the operator correlations induced by one-loop RGE mixing and flat directions. This directly affects the headline claim that the data are consistent with the SM at the reported level.
- [§3.3] §3.3 (genetic algorithm): The efficiency and reliability of the genetic algorithm in concentrating evaluations on the high-posterior region of the 2^N model space is asserted without convergence diagnostics, multiple independent runs, or comparison against exhaustive enumeration on a reduced operator basis. Because the reported posterior model probabilities and the operator correlation matrix are obtained from these samples, the lack of validation undermines the claimed improvement in characterisation.
- [Results section] Results section (model probabilities): The statement that “no statistically significant evidence for any departure from the SM” is obtained is load-bearing for the paper’s conclusions, yet it is derived solely from BIC-based model probabilities. A direct sensitivity study replacing BIC with a more accurate evidence estimator on the highest-probability models would be required to confirm that the ranking is stable.
minor comments (3)
- [Results section] The definition of the operator correlation matrix (presumably Eq. (X) in the results section) should be accompanied by an explicit formula showing how it is computed from the posterior model probabilities.
- The sensitivity analysis with respect to the matching scale μ₀ is mentioned in the abstract but the corresponding figures or tables are not referenced in the text; a short paragraph summarising the quantitative variation would improve clarity.
- Notation for the Wilson coefficients (e.g., the distinction between linear and quadratic contributions) should be made uniform between the text and the tables of results.
Simulated Author's Rebuttal
We thank the referee for the thorough and constructive review. The comments correctly identify areas where additional validation would strengthen the manuscript. We address each major comment below, indicating planned revisions where appropriate.
read point-by-point responses
-
Referee: [§3] §3 (BIC approximation): The central results on model probabilities, inclusion probabilities and BMA posteriors rest on treating BIC = log L_max − (k/2) log N as an accurate proxy for the marginal likelihood. No quantitative comparison to exact evidence (or to nested sampling) is presented for even a modest subset of models, nor is any error estimate given for the approximation in the presence of the operator correlations induced by one-loop RGE mixing and flat directions. This directly affects the headline claim that the data are consistent with the SM at the reported level.
Authors: We agree that a direct quantitative comparison to exact evidence would provide stronger support for the BIC-based results. Computing nested sampling or other exact estimators for models in this high-dimensional space with RGE-induced correlations is computationally demanding. BIC remains a standard and asymptotically justified approximation for large N in the model selection literature. In the revised manuscript we will expand the discussion in §3 to include the known limitations of BIC under strong correlations, cite relevant validation studies, and provide a qualitative assessment of how RGE mixing affects the approximation in our dataset. revision: partial
-
Referee: [§3.3] §3.3 (genetic algorithm): The efficiency and reliability of the genetic algorithm in concentrating evaluations on the high-posterior region of the 2^N model space is asserted without convergence diagnostics, multiple independent runs, or comparison against exhaustive enumeration on a reduced operator basis. Because the reported posterior model probabilities and the operator correlation matrix are obtained from these samples, the lack of validation undermines the claimed improvement in characterisation.
Authors: The genetic algorithm was run with population-size and generation-count settings chosen to achieve stable best-fit values. To meet the referee’s request we will add, in the revised §3.3, explicit convergence diagnostics (fitness trajectories and population diversity metrics), results from at least three independent runs with different random seeds, and a direct comparison against exhaustive enumeration on a reduced operator basis (e.g., the electroweak-only subset) where 2^N remains tractable. revision: yes
-
Referee: [Results section] Results section (model probabilities): The statement that “no statistically significant evidence for any departure from the SM” is obtained is load-bearing for the paper’s conclusions, yet it is derived solely from BIC-based model probabilities. A direct sensitivity study replacing BIC with a more accurate evidence estimator on the highest-probability models would be required to confirm that the ranking is stable.
Authors: The SM model indeed receives the highest posterior probability under our BIC ranking, with substantial separation from the next models. Performing a full sensitivity study with alternative evidence estimators on the top-ranked models is feasible for a small number of cases and will be included in the revision. We will report the change (or stability) in model ranking when the highest-probability models are re-evaluated with a more accurate estimator, thereby directly addressing the robustness of the “no significant evidence” conclusion. revision: partial
- Full quantitative validation of the BIC approximation against exact marginal likelihoods across the entire model space, owing to prohibitive computational cost.
Circularity Check
No significant circularity detected in derivation chain
full rationale
The paper develops a Bayesian model selection framework for SMEFT operator subsets, using a genetic algorithm to explore the discrete model space and the BIC as a tractable approximation to the marginal likelihood when ranking subsets against external electroweak, Higgs, top, and diboson data from LEP and LHC Run 2. Posterior model probabilities, marginal inclusion probabilities, Bayesian model averages, and the operator correlation matrix are all computed from these data-driven likelihoods at linear and quadratic order with one-loop RGE evolution. No step reduces by the paper's own equations or self-citations to a quantity defined solely by internal fitted parameters; the headline conclusions (no significant SM departure, improved BMA characterisation) remain dependent on the input observables and are therefore self-contained against external benchmarks.
Axiom & Free-Parameter Ledger
free parameters (1)
- Model-space prior probabilities
axioms (2)
- domain assumption The Bayesian Information Criterion approximates the Bayesian evidence well enough for reliable model comparison among SMEFT operator subsets.
- domain assumption The genetic algorithm efficiently concentrates evaluations in the high-posterior region of the discrete model space without missing important models.
Reference graph
Works this paper leans on
-
[1]
M. Cirelli, A. Strumia and J. Zupan,Dark Matter,2406.01705
work page internal anchor Pith review Pith/arXiv arXiv
-
[2]
D. Bodeker and W. Buchmuller,Baryogenesis from the weak scale to the grand unification scale,Rev. Mod. Phys.93(2021) 035004 [2009.07294]
-
[3]
S.F. King,Neutrino mass models,Rept. Prog. Phys.67(2004) 107 [hep-ph/0310204]
work page internal anchor Pith review Pith/arXiv arXiv 2004
-
[4]
Craig,Naturalness: past, present, and future,Eur
N. Craig,Naturalness: past, present, and future,Eur. Phys. J. C83(2023) 825 [2205.05708]
-
[5]
Higgs Physics at the HL-LHC and HE-LHC
M. Cepeda et al.,Report from Working Group 2: Higgs Physics at the HL-LHC and HE-LHC,CERN Yellow Rep. Monogr.7(2019) 221 [1902.00134]
work page internal anchor Pith review Pith/arXiv arXiv 2019
-
[6]
P. Azzi et al.,Report from Working Group 1: Standard Model Physics at the HL-LHC and HE-LHC,CERN Yellow Rep. Monogr.7(2019) 1 [1902.04070]
-
[7]
G. Isidori, F. Wilsch and D. Wyler,The standard model effective field theory at work,Rev. Mod. Phys.96(2024) 015006 [2303.16922]
-
[8]
Dimension-Six Terms in the Standard Model Lagrangian
B. Grzadkowski, M. Iskrzynski, M. Misiak and J. Rosiek,Dimension-Six Terms in the Standard Model Lagrangian,JHEP10(2010) 085 [1008.4884]
work page internal anchor Pith review Pith/arXiv arXiv 2010
-
[9]
C.W. Murphy,Dimension-8 operators in the Standard Model Eective Field Theory,JHEP10 (2020) 174 [2005.00059]
-
[10]
R. Boughezal, Y. Huang and F. Petriello,Exploring the SMEFT at dimension eight with Drell-Yan transverse momentum measurements,Phys. Rev. D106(2022) 036020 [2207.01703]
- [11]
-
[12]
T. Corbett, J. Desai, O.J.P. Éboli, M.C. Gonzalez-Garcia, M. Martines and P. Reimitz, Impact of dimension-eight SMEFT operators in the electroweak precision observables and triple gauge couplings analysis in universal SMEFT,Phys. Rev. D107(2023) 115013 [2304.03305]
-
[13]
J. de Blas, D. Chowdhury, M. Ciuchini, A.M. Coutinho, O. Eberhardt, M. Fedele et al., Hepfit: a code for the combination of indirect and direct constraints on high energy physics models,The European Physical Journal C80(2020)
work page 2020
- [14]
- [15]
-
[16]
Constraining the top-Higgs sector of the Standard Model Effective Field Theory
V. Cirigliano, W. Dekens, J. de Vries and E. Mereghetti,Constraining the top-Higgs sector of the Standard Model Effective Field Theory,Phys. Rev. D94(2016) 034031 [1605.04311]. – 32 –
work page internal anchor Pith review Pith/arXiv arXiv 2016
-
[17]
V. Cirigliano, W. Dekens, J. de Vries, E. Mereghetti and T. Tong,Anomalies in global SMEFT analyses. A case study of first-row CKM unitarity,JHEP03(2024) 033 [2311.00021]
- [18]
-
[19]
S. Bißmann, J. Erdmann, C. Grunwald, G. Hiller and K. Kröninger,Constraining top-quark couplings combining top-quark andBdecay observables,Eur. Phys. J. C80(2020) 136 [1909.13632]
-
[20]
S. Bißmann, C. Grunwald, G. Hiller and K. Kröninger,Top and Beauty synergies in SMEFT-fits at present and future colliders,JHEP06(2021) 010 [2012.10456]
-
[21]
L. Allwicher, C. Cornella, G. Isidori and B.A. Stefanek,New physics in the third generation. A comprehensive SMEFT analysis and future prospects,JHEP03(2024) 049 [2311.00020]
-
[22]
S. Bruggisser, R. Schäfer, D. van Dyk and S. Westhoff,The Flavor of UV Physics,JHEP05 (2021) 257 [2101.07273]
-
[23]
S. Bruggisser, D. van Dyk and S. Westhoff,Resolving the flavor structure in the MFV-SMEFT,JHEP02(2023) 225 [2212.02532]
-
[24]
L. Bellafronte, S. Dawson and P.P. Giardino,The importance of flavor in SMEFT Electroweak Precision Fits,JHEP05(2023) 208 [2304.00029]
-
[25]
C. Grunwald, G. Hiller, K. Kröninger and L. Nollen,More synergies from beauty, top, Z and Drell-Yan measurements in SMEFT,JHEP11(2023) 110 [2304.12837]
-
[26]
G. Hiller and D. Wendler,Missing energy plus jet in the SMEFT,JHEP09(2024) 009 [2403.17063]
- [27]
-
[28]
J. ter Hoeve, L. Mantani, J. Rojo, A.N. Rossia and E. Vryonidou,Higgs trilinear coupling in the standard model effective field theory at the high luminosity LHC and the FCC-ee,Phys. Rev. D112(2025) 013008 [2504.05974]
-
[29]
L. Mantani and V. Sanz,Probing the flavour-blind SMEFT: EFT validity and the interplay of energy scales,JHEP06(2025) 147 [2503.02935]
-
[30]
J. ter Hoeve, L. Mantani, J. Rojo, A.N. Rossia and E. Vryonidou,Connecting scales: RGE effects in the SMEFT at the LHC and future colliders,JHEP06(2025) 125 [2502.20453]
- [31]
-
[32]
V. Maura, B.A. Stefanek and T. You,The Higgs Self-Coupling at FCC-ee,Phys. Rev. Lett. 135(2025) 141802 [2503.13719]
-
[33]
J. de Blas, A. Goncalves, V. Miralles, L. Reina, L. Silvestrini and M. Valli,Constraining new physics effective interactions via a global fit of electroweak, Drell-Yan, Higgs, top, and flavour observables,2507.06191. [36]CMScollaboration,Combined effective field theory interpretation of Higgs boson, electroweak vector boson, top quark, and multi-jet measu...
-
[34]
New Physics Reach through Precision at Future Colliders: a Multi-Pronged Approach
T. Armadillo, E. Celada, J. ter Hoeve, F. Maltoni, L. Mantani, J. Rojo et al.,New Physics Reach through Precision at Future Colliders: a Multi-Pronged Approach,2604.16596
work page internal anchor Pith review Pith/arXiv arXiv
- [35]
-
[36]
E.E. Jenkins, A.V. Manohar and M. Trott,Renormalization Group Evolution of the Standard Model Dimension Six Operators II: Yukawa Dependence,JHEP01(2014) 035 [1310.4838]
-
[37]
E.E. Jenkins, A.V. Manohar and M. Trott,Renormalization Group Evolution of the Standard Model Dimension Six Operators I: Formalism and lambda Dependence,JHEP10(2013) 087 [1308.2627]
- [38]
- [39]
-
[40]
J. de Blas, J.C. Criado, M. Perez-Victoria and J. Santiago,Effective description of general extensions of the Standard Model: the complete tree-level dictionary,JHEP03(2018) 109 [1711.10391]
work page internal anchor Pith review Pith/arXiv arXiv 2018
- [41]
-
[42]
From the EFT to the UV: the complete SMEFT one-loop dictionary
G. Guedes and P. Olgoso,From the EFT to the UV: the complete SMEFT one-loop dictionary,SciPost Phys.20(2026) 074 [2412.14253]
work page internal anchor Pith review Pith/arXiv arXiv 2026
-
[43]
J.C. Criado and M. Pérez-Victoria,Field redefinitions in effective theories at higher orders, JHEP03(2019) 038 [1811.09413]
-
[44]
J. de Blas et al.,Physics Briefing Book: Input for the 2026 update of the European Strategy for Particle Physics,2511.03883
-
[45]
Bayes in the sky: Bayesian inference and model selection in cosmology
R. Trotta,Bayes in the sky: Bayesian inference and model selection in cosmology,Contemp. Phys.49(2008) 71 [0803.4089]
work page internal anchor Pith review Pith/arXiv arXiv 2008
-
[46]
Jaynes,Probability Theory: The Logic of Science, CUP (2003)
E.T. Jaynes,Probability Theory: The Logic of Science, CUP (2003)
work page 2003
-
[47]
Jeffreys,The Theory of Probability, Oxford Classic Texts in the Physical Sciences (1939)
H. Jeffreys,The Theory of Probability, Oxford Classic Texts in the Physical Sciences (1939)
work page 1939
- [48]
-
[49]
The effect of the two-loop SMEFT RGEs at future colliders
L. Mantani, P. Olgoso and A.N. Rossia,The effect of the two-loop SMEFT RGEs at future colliders,2604.16599
work page internal anchor Pith review Pith/arXiv arXiv
-
[50]
L. Born, A. Greljo and A.E. Thomsen,When Two Loops Matter: Electroweak Precision in the SMEFT,2604.26817
work page internal anchor Pith review Pith/arXiv arXiv
-
[51]
M. Battaglia, M. Grazzini, M. Spira and M. Wiesemann,Sensitivity to BSM effects in the Higgs pT spectrum within SMEFT,JHEP11(2021) 173 [2109.02987]
- [52]
-
[53]
S. Di Noi and R. Gröber,Renormalisation group running effects inpp→t¯thin the Standard Model Effective Field Theory,Eur. Phys. J. C84(2024) 403 [2312.11327]
-
[54]
R. Bartocci, A. Biekötter and T. Hurth,Renormalisation group evolution effects on global SMEFT analyses,2412.09674
-
[55]
F. Maltoni, G. Ventura and E. Vryonidou,Impact of SMEFT renormalisation group running on Higgs production at the LHC,JHEP12(2024) 183 [2406.06670]
- [56]
- [57]
-
[58]
L. Allwicher, M. McCullough and S. Renner,New physics at Tera-Z: precision renormalised, JHEP02(2025) 164 [2408.03992]
-
[59]
H. Akaike,A new look at the statistical model identification,IEEE Transactions on Automatic ControlAC-19(1974) 716
work page 1974
-
[60]
Schwarz,Estimating the Dimension of a Model,Annals Statist.6(1978) 461
G. Schwarz,Estimating the Dimension of a Model,Annals Statist.6(1978) 461
work page 1978
-
[61]
J.H. Holland,Adaptation in Natural and Artificial Systems, University of Michigan Press, Ann Arbor, MI (1975)
work page 1975
-
[62]
D.E. Goldberg,Genetic Algorithms in Search, Optimization, and Machine Learning, Addison-Wesley, New York (1989)
work page 1989
-
[63]
Electroweak Measurements in Electron-Positron Collisions at W-Boson-Pair Energies at LEP
A.F. Gad,Pygad: An intuitive genetic algorithm python library, 2021. [68]ALEPH, DELPHI, L3, OPAL, LEP Electroweakcollaboration,Electroweak Measurements in Electron-Positron Collisions at W-Boson-Pair Energies at LEP,Phys. Rept.532(2013) 119 [1302.3415]
work page internal anchor Pith review Pith/arXiv arXiv 2021
- [64]
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.