Recent Progress in Flavor

Model Building
Wolfgang Altmannshofer1 and Admir Greljo2
1
Department of Physics, University of California, Santa Cruz, and Santa Cruz
Institute for Particle Physics, Santa Cruz, CA 95064, USA;
email: [email protected]
2
Department of Physics, University of Basel, Klingelbergstrasse 82, CH 4056
arXiv:2412.04549v1 [hep-ph] 5 Dec 2024

Basel, Switzerland; email: [email protected]

Annu. Rev. Nucl. Part. Sci. 2025. 75:1–22 Keywords

https://doi.org/10.1146/annurev-nucl- quarks and leptons, flavor puzzles, beyond the SM, flavor physics
121423-100950

Copyright © 2025 by the author(s). Abstract

All rights reserved
The flavor puzzles remain among the most compelling open questions
in particle physics. The striking hierarchies observed in the masses
and mixing of charged fermions define the Standard Model (SM) flavor
puzzle, a profound structural enigma pointing to physics beyond the
SM. Simultaneously, the absence of deviations from SM predictions
in precision measurements of flavor-changing neutral currents imposes
severe constraints on new physics at the TeV scale, giving rise to the
new physics flavor puzzle. This review article provides an overview
of a selection of recent advancements in flavor model building, with a
particular focus on attempts to address one or both of these puzzles
within the quark sector.

1
Contents
1. INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
2. FLAVOR PUZZLES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.1. The Standard Model Flavor Puzzle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.2. The New Physics Flavor Puzzle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
3. OVERVIEW OF PROPOSED SOLUTIONS. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
3.1. Model Building for the Standard Model Flavor Hierarchies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
3.2. Model Building to Control New Physics Flavor Violation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
4. RECENT DEVELOPMENTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
4.1. Revisiting U (2) and SU (2) Symmetries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
4.2. Flavor Deconstruction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
4.3. Revisiting U (1) Symmetries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
4.4. Flavor symmetries in the SMEFT. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
4.5. Flavor Symmetries and Axions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
4.6. Flavor Clockwork Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
4.7. Modular and Eclectic Flavor Symmetries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
4.8. Flavor from Higher-Form and Non-Invertible Symmetries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
5. CONCLUSIONS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18

1. INTRODUCTION
Within the landscape of quantum field theories, the Standard Model (SM) of particle physics
is a prominent example of a chiral, spontaneously broken non-abelian gauge theory, wherein
anomaly cancellation imposes stringent constraints on the charge assignments of matter
fields within a single generation. However, nature often surprises us with unexpected com-
plexity, exemplified by the existence of three generations of particles—an apparent redun-
dancy that has no obvious reason. This replication of fermion gauge representations is
referred to as flavor. Understanding the origin of flavor remains one of the most compelling
and unresolved questions in particle physics.
Flavor physics is presently a highly dynamic field of research, driven by a substantial
ongoing experimental program and supported by a series of planned future experiments.
Currently, Bottom, Kaon, Charm, Tau, and Muon factories operate at full capacity, offering
a unique opportunity to deepen our understanding of the SM and uncover new physics.
As the Large Hadron Collider (LHC) achieves its design collision energy and transitions
into the high-luminosity era (1), the emphasis is shifting towards precision measurements.
Looking ahead, the prospect of an electron-positron circular collider emerges as the next
frontier (2, 3). In this era of precision, testing flavor and CP violation takes center stage,
offering the potential to discover new physics. This could involve either directly detecting
light new physics or observing indirect effects of heavy new physics, which may reveal
themselves through rare phenomena such as flavor-changing neutral currents (FCNC), and
electric dipole moments (EDMs). These rare and forbidden phenomena are particularly
sensitive to new physics at energy scales far beyond the TeV range, which are out of reach
for direct searches.
The ongoing, robust experimental program has revitalized interest in advancing theo-
retical approaches in flavor model building. The goal of flavor model building is twofold.
First, it aims to explain the observed hierarchical patterns of fermion masses and mixing

2 Altmannshofer and Greljo

angles, known as the SM flavor puzzle (section 2.1). Second, if new physics exists just be-
yond the electroweak scale, flavor model building seeks to explain why no signs of this new
physics have been detected in flavor-changing transitions, known as the new physics (NP)
flavor puzzle (section 2.2). This review will explore approaches and models that address
both flavor puzzles, evaluating their successes and limitations in the context of current and
future experimental data. Our emphasis is on the quark sector, but we will also comment
on the flavor of charged leptons and neutrinos where appropriate. After a comprehensive
review of traditional solutions to the flavor puzzles in section 3, the discussion will shift
towards an exploration of recent advancements in the field, which will be the central focus
of section 4.

2. FLAVOR PUZZLES
In this section, we define and discuss the flavor puzzles of the SM and NP separately.

2.1. The Standard Model Flavor Puzzle

The observed masses of quarks and leptons, along with their mixing under the weak nuclear
force, reveal a peculiar pattern. Up quarks, down quarks, and charged leptons exhibit a
distinct generational mass hierarchy, with each generation differing by approximately two
orders of magnitude. Additionally, the Cabibbo-Kobayashi-Maskawa (CKM) quark mixing
matrix is nearly a unit matrix, with hierarchically suppressed mixing between generations:
1 ≫ |V12 | ∼ λ ≫ |V23 | ∼ λ2 ≫ |V13 | ∼ λ3 , with λ ≃ 0.2. The SM fails to explain the
observed hierarchies, a deficiency referred to as the SM flavor puzzle.
The origin of masses and mixing in the SM is tied to the renormalizable interactions
with a single Higgs field H,

e r − Y pr q Hdr − Yepr ℓp Her ,

L ⊃ −Yupr q p Hu d p 1.

where qp , up , dp , ℓp , and ep are the five gauge representations of the chiral fermions, with
p, r = 1, 2, 3 as flavor indices. The core of the flavor puzzle lies in the following question:
Why is there a hierarchical structure in the input parameters Yfpr when they all enter the
theory in the same way, coupling to a single Higgs field? According to the singular value
decomposition theorem, Yf = Lf Ŷf Rf† , where Ŷf is a diagonal matrix with real and positive
entries, and Lf and Rf are unitary matrices. The masses are determined by the singular
values M̂f = Ŷf ⟨H⟩, while the quark mixing is given by the CKM matrix V = L†u Ld . The
experimentally determined values of the fermion masses and the CKM matrix elements,
shown in Figure 1, reveal a highly hierarchical structure in the Yukawa matrices.
It is important to note that the small values of the Yukawa couplings are ’technically
natural’ in the sense of t’Hooft’s naturalness criterion (7). Setting the Yukawa couplings to
zero enhances the symmetry of the theory, which ensures that radiative corrections to these
parameters are only logarithmically sensitive to the cutoff scale as long as no new significant
sources of flavor violation beyond the SM are introduced. This is in stark contrast to the
Higgs hierarchy problem, where the sensitivity to the cutoff scale is quadratic, motivating
a symmetry-based solution not far above the electroweak scale. Unfortunately, there is no
clear indication of a scale where the flavor puzzle is solved.
While the small Yukawa couplings are technically natural, they fall short of what one
might expect from Dirac’s naturalness, where dimensionless input parameters are of order

www.annualreviews.org • Recent Progress in Flavor Model Building 3

 Figure 1
Top: Masses of Standard Model particles in GeV/c2 , including uncertainties. Values for the Higgs
boson, gauge bosons, quarks, and leptons are taken from the PDG (4). Neutrino masses, several
orders of magnitude lighter, are not shown. Two neutrino mass differences
P have been measured
(5), and cosmological data constrain the sum of neutrino masses to mν ≲ 0.1 eV (6). The
lightest neutrino could potentially be massless. Bottom: Absolute values of CKM matrix elements
with uncertainties, derived from direct measurements as compiled by the PDG (4).

one. Even more perplexing is the specific hierarchical structure of these parameters. We
stress that the three 3 × 3 Yukawa matrices, Yu , Yd , and Ye , are entirely independent of each
other. Yet, they all show a similar hierarchical structure in their singular values and an
approximate alignment between Yu and Yd , as reflected by the CKM matrix. This peculiar
structure shouts for an explanation.
Compared to quarks and charged leptons, the neutrino sector behaves quite differently,
adding further complexity to the puzzle. First, neutrinos have extremely small masses, at
least six orders of magnitude smaller than the electron mass. Second, neutrino oscillation
experiments reveal large mixing angles (5), unlike the small and hierarchical mixing between
quarks. Additionally, the ratio of the two observed mass splittings in the neutrino sector
is not particularly large when compared to the pronounced mass hierarchies in the charged
fermion sector.
The renormalizable SM predicts that neutrinos are massless, calling for physics beyond
the SM to explain non-zero neutrino masses. However, when viewing the SM as an effective
field theory, the significant mass difference between neutrinos and charged fermions can be
explained by the leading higher-dimensional operator (8)

Yνpr
L⊃ ℓp ℓr HH . 2.
Λ

Here, Yνpr is a complex-symmetric matrix, diagonalized by Yν = UνT Ŷν Uν , while Λ represents

a high-energy scale. The operator breaks the lepton number by two units, and neutrinos

4 Altmannshofer and Greljo

are Majorana particles in this setup. A large energy scale Λ where the lepton number is
broken provides a compelling explanation for the smallness of the neutrino masses. The
see-saw mechanism with heavy right-handed Majorana neutrinos is arguably the simplest
realization of this scenario. The Pontecorvo-Maki-Nakagawa-Sakata (PMNS) matrix, the
leptonic counterpart to the CKM matrix, exhibits large and seemingly anarchic mixing,1
as one might expect if no special structure is imposed on Yνpr . This adds another layer of
complexity to the SM flavor puzzle, given that the CKM matrix behaves in a contrasting
manner.
Alternatively to the neutrino mass origin outlined above, the lepton number could re-
main unbroken, making neutrinos Dirac particles. In this scenario, their masses would arise
from renormalizable Yukawa interactions, similar to the charged fermions in Eq. (1). The
small neutrino masses would then result from extremely tiny Yukawa couplings, further
complicating the flavor puzzle.
Flavor model building attempts to explain the non-generic flavor structure of the SM
fermions described above. Although this review focuses on the quark sector, the lack of
flavor hierarchies in the neutrino sector offers essential insight into the problem.

2.2. The New Physics Flavor Puzzle

One practical implication of the SM flavor puzzle, even without knowing its solution, is the
existence of approximate accidental symmetries within the SM. The smallness of Yukawa
couplings and the CKM alignment leads to approximate flavor and CP conservation, result-
ing in highly suppressed FCNCs and EDMs. The resulting selection rules, exemplified by
the GIM mechanism (11), have been instrumental in designing precision tests of the SM that
are highly sensitive to new physics, which typically breaks these approximate symmetries.
So far, these tests have yielded negative results. However, the null findings have profound
consequences for the flavor structure of new physics if it exists near the TeV scale—a sit-
uation known as the new physics flavor puzzle. That is to say, new flavor parameters of a
TeV-scale extension of the SM are also not generic.
To illustrate this point, we employ the Standard Model Effective Field Theory (SMEFT)
as a framework to represent short-distance new physics effects (12). By augmenting the
P
SM Lagrangian with a tower of higher-dimensional local operators, L += O CO O, we
introduce new sources of flavor and CP violation that are absent at the renormalizable
level. Specifically, we focus on leading-order dimension-6 operators that preserve baryon
and lepton numbers. The Wilson coefficients of these operators are parameterized as CO =
cO Λ−2 , where Λ is the scale of a new physics completion.
Consider a flavor-anarchic scenario in which all dimensionless flavor parameters are of
order one, assuming no particular hierarchy or structure in the flavor sector of the new
physics. For a tree-level matching, this predicts cO ∼ 1. Under this assumption, many of
the dimension-6 operators contribute significantly to processes involving neutral meson os-
cillations (∆F = 2), charged lepton flavor violation (cLFV), EDMs, or strangeness-changing
decays (∆S = 1). The experimental absence of deviations from SM predictions in such pro-
cesses imposes stringent constraints on the scale Λ, often pushing it to values as high as six
orders of magnitude above the TeV scale. For a concrete illustration, Figure 1 of Ref. (13)

1 Interestingly, the PMNS matrix is approximately of the “tri-bimaximal” form (9) motivating

discrete flavor symmetries acting in the lepton sector, see e.g. (10) for a review.

www.annualreviews.org • Recent Progress in Flavor Model Building 5

 presents bounds on a selected set of operators in the Warsaw basis, explicitly showing the
flavor indices in the down-quark charged-lepton mass basis.
Let us stress that this creates a flavor puzzle only if we anticipate NP to emerge near the
TeV scale. Otherwise, a high cutoff scale Λ for the SM automatically resolves the tension
with experimental data.
The main motivation for expecting new physics at the TeV scale is the Higgs hierarchy
problem, which asks why the Higgs boson mass is so much lighter than the Planck scale
despite quantum corrections that should drive it higher. Symmetry-based solutions to this
problem—such as supersymmetry or composite Higgs models—imply that new physics is
expected not far above the EW scale to stabilize the Higgs mass. These theories often
introduce new particles and interactions with order-one couplings to the Higgs field. Con-
sequently, the flavor structure of such new physics cannot be generic; it must be hierarchical
and (or) aligned to avoid excessive violation of the SM’s approximate flavor and CP sym-
metries. This raises the possibility that the NP and the SM flavor structure might share a
common origin.
Secondly, from a purely experimental, bottom-up perspective, the TeV scale is directly
accessible to the operating LHC. Direct searches at this energy scale require some couplings
to be sizable to produce observable signals, bringing the flavor structure of new interactions
into sharp focus. For prompt signals in direct searches, if the new couplings to fermions are
not hierarchical or aligned, they could induce dangerous flavor-changing processes. There-
fore, understanding the flavor hierarchy in new physics is crucial for a consistent program
of direct searches at colliders.

3. OVERVIEW OF PROPOSED SOLUTIONS

The distinctive flavor structure observed in the SM and similarly non-generic flavor patterns
in the couplings of TeV-scale new physics demand an explanation. This section provides a
concise overview of the most prominent approaches developed to address the two puzzles.

3.1. Model Building for the Standard Model Flavor Hierarchies

The observed hierarchies in the quark and charged lepton masses and the CKM mixing
angles suggest an underlying mechanism beyond the SM. Here, we explore the two main
approaches that have been proposed to address this flavor puzzle: symmetry-based models
and geometry-based models.

3.1.1. Symmetry-Based Models. Symmetry-based models introduce additional symmetries,

known as horizontal or family symmetries, that act between different generations of
fermions. These symmetries constrain the Yukawa couplings, naturally leading to hierar-
chical mass and mixing patterns after symmetry breaking by small spurions. Depending on
the nature of the symmetry group, these models are classified into abelian and non-abelian
symmetries. They could be either gauged (fundamental) or global (accidental). The latter
can be approximate or (perturbatively) exact. The specifics of the flavor-symmetry-driven
mechanism often depend on other aspects of the BSM extension, such as the presence of
additional quark-lepton unification or the inclusion of supersymmetry.
The Froggatt-Nielsen (FN) mechanism is a seminal example utilizing an abelian hori-
zontal symmetry, typically a U (1) symmetry, to explain flavor hierarchies (14, 15, 16). In

6 Altmannshofer and Greljo

this framework, the SM fermions carry generation-dependent charges under the additional
U (1) symmetry. A scalar field, known as the FN (or flavon) field (Φ), acquires a vac-
uum expectation value (vev) ⟨Φ⟩, thus spontaneously breaking the flavor symmetry. The
vev ⟨Φ⟩ is assumed to be smaller than the cutoff scale Λ of the effective theory. Yukawa
couplings arise from higher-dimensional operators suppressed by powers of the small pa-
rameter ε = ⟨Φ⟩/Λ ≪ 1. By assigning appropriate U (1) charges, the FN mechanism can
qualitatively reproduce the observed mass hierarchies and mixing angles from a single ex-
pansion parameter ε. For instance, heavier fermions are assigned lower charges, resulting
in less suppression in their Yukawa couplings. The most straightforward UV completion
of the higher dimensional operators are chains of vector-like fermions that communicate
the symmetry breaking to the SM fermions. The FN setup contains a sufficient number
of free O(1) parameters to reproduce the observed fermion masses precisely. If the U (1)
symmetry is gauged, the models contain a Z ′ gauge boson with a mass of order ⟨Φ⟩. If
the U (1) is global, one instead expects a very light pseudo-Nambu-Goldstone boson. These
particles generically have flavor-changing couplings and can be effectively searched for with
flavor probes. Flavor models commonly feature a decoupling limit: by sending both ⟨Φ⟩
and Λ to high energies while keeping ε = ⟨Φ⟩/Λ constant, these models can address the SM
flavor puzzle without introducing excessive FCNCs at low energies. Current experimental
constraints impose a lower bound on the scale of flavor dynamics, while future experiments
could potentially reveal evidence for low-scale flavor models.

Non-abelian flavor symmetries, such as SU (3) or SU (2), organize the fermion gener-
ations into non-trivial group representations. By assigning fermions to multiplets, these
symmetries constrain the structure of the Yukawa couplings. The observed flavor struc-
ture typically arises through sequential and hierarchical symmetry breaking. Non-abelian
symmetries often lead to more restrictive textures than FN models. However, challenges in
these models include the complexity of the symmetry-breaking sector to reproduce realistic
parameters. Models based on SU (3) flavor symmetry have been proposed to generate re-
alistic mass hierarchies; see e.g. (17, 18, 19, 20). However, the sizable Yukawa coupling of
the top quark (yt ∼ 1) breaks the SU (3) symmetry, making SU (2) (or better U (2)) groups
more attractive (21, 22, 23, 24, 25, 26, 27). In this construction, the first two families form
a U (2) doublet, while the third family is a singlet. In the symmetric limit, only the third
generation acquires mass, providing a decent starting point. Symmetry breaking typically
occurs in two stages: first, the U (2) ≡ SU (2) × U (1) symmetry is broken to an intermediate
U (1), generating a small mass for the second generation; subsequently, the U (1) symmetry
is broken, yielding an even smaller mass for the first generation. For recent developments,
see section 4.1.

Radiative models propose that lighter SM fermion masses arise from loop corrections
rather than at the tree level (28). In these scenarios, mass hierarchies arise from the
suppression associated with loop factors, which intriguingly coincide with the two orders
of magnitude separating the masses of adjacent generations. The absence of tree-level
Yukawa couplings for the light generations can, for example, be imposed by using flavor
symmetries. The breaking of a flavor symmetry is then communicated from a separate
sector of the model to the SM fermions at the loop level. Additional fields and interactions
are introduced to facilitate the necessary loop diagrams. In contrast to FN models, their
masses can be at the same scale as flavor breaking, as the role of ε is effectively replaced

www.annualreviews.org • Recent Progress in Flavor Model Building 7

 by a loop factor. The wide range of possible quantum numbers and interactions for the
new fields gives rise to a plethora of models, both supersymmetric (29, 30, 31, 32) and
non-supersymmetric (33, 34, 35, 36).

3.1.2. Geometry-Based Models. Geometry-based models utilize extra spatial dimensions

to address flavor hierarchies. In these frameworks, the SM fields propagate in higher-
dimensional spaces and are localized at different positions in the extra dimension. The over-
lap of fermion wavefunctions with the Higgs field determines the effective four-dimensional
Yukawa couplings. This overlap can depend exponentially on the parameters of the model.
Adjusting localization parameters, therefore, allows one to generate the observed mass hi-
erarchies without the need for hierarchical fundamental couplings (37).
A concrete example in which this mechanism can be at work is the Randall-Sundrum
(RS) model (38) that introduces a warped extra dimension with a non-factorizable geometry,
where the metric depends exponentially on the extra-dimensional coordinate. The extra
dimension is bounded by an infrared (IR) brane and an ultraviolet (UV) brane. The warping
leads to exponential factors that naturally generate hierarchies. The Higgs naturalness
problem is addressed if the Higgs resides on the IR brane. Fermions that propagate in the
bulk have profiles that interpolate between the UV and IR branes and can be localized at
different positions along the extra dimension, resulting in hierarchical Yukawa couplings
with the Higgs. Heavier fermions, such as the top quark, are more localized towards the IR
brane, resulting in larger Yukawa couplings (39, 40, 41).
The flavor-dependent, extra-dimensional profiles of the SM fermions imply a rich flavor
phenomenology (42, 43). In generic setups, the constraints from meson mixing are very
stringent and push the new physics scale to O(10) TeV. Extra dimensional models can be
protected from large FCNCs by using bulk and brane flavor symmetries (44).

Partial compositeness proposes that SM fermions are mixtures of elementary and com-
posite states arising from a strongly coupled sector not far above the electroweak scale.
Models of partial compositeness were originally introduced to tame FCNCs in technicolor
models (45). The light fermions are mostly elementary, while the heavy top quark is mostly
composite. Per the AdS/CFT correspondence, models of partial compositeness map onto
models with warped extra dimensions, with the amount of compositeness given by the
localization of the fermions along the extra dimension.
Composite Higgs models (see (46, 47) for reviews) typically incorporate partial compos-
iteness to generate fermion masses. The Higgs field emerges as a composite state from the
strong sector, and the mixing between elementary fermions and composite operators leads
to effective Yukawa couplings after electroweak symmetry breaking. This framework natu-
rally explains the large top quark mass and addresses the hierarchy problem by protecting
the Higgs mass from large radiative corrections. The phenomenology closely follows one of
the RS models, and large FCNCs can be avoided using flavor symmetries (48, 49, 50).

3.2. Model Building to Control New Physics Flavor Violation

As discussed in section 2.2, generic BSM sources of flavor violation lead to prohibitively
large contributions to FCNCs unless the NP scale is far above the electroweak scale. Instead,
NP close to the electroweak scale can be viable if it exhibits a non-generic flavor structure
that does not excessively violate the approximate flavor symmetries of the SM.
From a bottom-up point of view, an attractive but fairly restrictive option is the as-

8 Altmannshofer and Greljo

sumption that there are no new sources of flavor (and CP) violation beyond the SM Yukawa
couplings. This is known as Minimal Flavor Violation (MFV) (51, 52, 53, 54). In an MFV
scenario, NP contributions are controlled by the same hierarchical structure that suppresses
FCNCs in the SM. To implement this idea systematically, the SM Yukawas are promoted to
spurion fields with appropriate transformation properties under the flavor symmetry to for-
mally restore flavor invariance of the Yukawa Lagrangian. The SM flavor symmetry group
is defined by the fermion kinetic terms

GF = U (3)q × U (3)u × U (3)d × U (3)ℓ × U (3)e , 3.

with the five U (3) factors corresponding to three copies of the left-handed quark doublets,
the right-handed up quarks, the right-handed down quarks, the left-handed lepton doublets,
and the right-handed charged leptons, respectively. The Yukawa couplings are interpreted
as bi-triplets of the relevant SU (3) factors Yu = (3q , 3̄u ), and so on. A BSM setup is said
to be minimal flavor violating if all new physics interactions are formally invariant under
the SM flavor group. This implies that they are either completely flavor-blind or must be
expressed as appropriate powers of Yukawa couplings. In addition, prohibiting new sources
of CP violation is required to avoid stringent limits from EDMs. Extensions to the lepton
sector are also possible (55).
In an MFV framework, constraints from meson mixing and other flavor observables are
significantly eased, making new physics near the TeV scale a feasible possibility. Further-
more, MFV predicts correlations between b → s, b → d, and s → d transitions, with the
relative magnitudes of NP effects governed by the corresponding CKM factors. It is impor-
tant to emphasize that MFV does not aim to resolve the flavor puzzle of the SM. Instead, it
leverages the hierarchical flavor structure of the Yukawa couplings to constrain NP effects
without providing an explanation for the origin of these hierarchies.

Attractive alternatives to MFV that, in addition, can also partially address the SM
flavor puzzle are models based on the U (2)5 flavor symmetry. This symmetry (or its sub-
groups) could be responsible for the hierarchical pattern of fermion masses and mixings
(see section 4.1 below) and, at the same time, suppress FCNCs in a sufficient way. The
first and second generations of fermions fa (a = 1, 2) transform as a U (2)f doublet, while
the third generation is assigned as a U (2)f singlet. The symmetric limit predicts that only
the third generation acquires mass, leaving the first two generations massless. This serves
as a plausible starting point for understanding the hierarchical structure of the Yukawa
couplings. Furthermore, the U (2)5 symmetry provides protection for flavor transitions be-
tween the first and second generations, which are subject to the most stringent experimental
constraints. Given that yt ∼ 1, U (2)5 provides a much better approximate symmetry than
U (3)5 . The realistic mass spectrum of the first and second generations, along with the CKM
mixing, arises from a small breaking of U (2)5 , introducing small but testable deviations in
flavor-changing processes.
Models based on the U (2) flavor symmetries have been originally proposed in the context
of SUSY (21, 22) and later formalized in a general context (56, 57). The framework of
“general MFV” discussed in (26) shares many similar aspects. From a phenomenological
point of view, U (2)5 models are less restrictive than MFV. While details depend on which
spurions break the symmetry, one typically still finds strong correlations between b → s
and b → d transitions.

www.annualreviews.org • Recent Progress in Flavor Model Building 9

 Yet another option to control NP effects in flavor changing processes is known as flavor
alignment. The idea behind flavor alignment is that BSM interactions are approximately
flavor diagonal, but not necessarily flavor universal, in the fermion mass eigenstate basis, i.e.,
in which the SM Yukawas are diagonal. This corresponds to an approximate U (1)6 flavor
symmetry in the quark sector, with each U (1) factor acting on one of the six quark flavors.
Most famously, in the context of supersymmetric models, the approximate alignment of
quark Yukawa couplings and soft SUSY breaking squark masses to avoid large FCNCs has
been proposed in (58).
While flavor alignment suppresses flavor off-diagonal couplings, one can expect O(1)
flavor non-universality. It is important to note that flavor alignment cannot be exact for
new physics couplings of left-handed quark doublets unless the couplings are entirely flavor
universal. Already in the SM, there is some misalignment between the left-handed up quarks
and left-handed down quarks, as parameterized by the CKM matrix. A new non-universal
flavor coupling can be aligned either with the left-handed up quarks or with the left-handed
down quarks, but not with both of them simultaneously. An irreducible amount of flavor
mixing of an order of the Cabibbo angle is predicted either for strange-to-down or charm-
to-up transitions. Combining information from neutral kaon mixing and neutral D meson
mixing typically gives the strongest constraints on models with flavor alignment (59).

4. RECENT DEVELOPMENTS
In the previous section, we reviewed traditional approaches to addressing the SM and NP
flavor puzzles. Now, we turn our discussion to a representative selection of recent develop-
ments in the field.

4.1. Revisiting U (2) and SU (2) Symmetries

Models based on a single approximate global U (2) flavor symmetry (21, 22, 23, 24) offer an
elegant framework to explain flavor hierarchies by distinguishing the third family from the
lighter two. While the main concept was outlined in section 3.1.1, recent advancements have
unveiled elegant realizations involving different charge assignments and symmetry-breaking
patterns.
For concreteness, consider a U (2)q flavor symmetry acting on the left-handed quarks (60,
13). Imposing this symmetry leads to an extended symmetry of the Yukawa sector, U (2)q ×
U (2)u × U (2)d (25, 26, 56), where the latter two factors are accidental. To break U (2)q ,
introduce two orthogonal spurion doublets, V = (0, a)T and V ′ = (b, 0)T . This breaking
structure generates quark Yukawa matrices Yu and Yd , with entries in the first row of O(b),
second row of O(a), and third row of O(1).
Applying singular value decomposition, Yf = Lf Ŷf Rf† , the right-handed rotations Ru,d
are matrices with O(1) elements that transform Yu,d into an upper triangular form. In
contrast, the left-handed rotations Lu,d are perturbative matrices with mixing angles of
O(b/a), O(a), and O(b) for the 1-2, 2-3, and 1-3 sectors, respectively. The CKM mixing
matrix V is then given by the product of left-handed rotations, V = L†u Ld . The singular
values of both Ŷu and Ŷd are hierarchically ordered as {b, a, 1}. By choosing parameters
such that 1 ≫ a ≫ b ≫ a2 , this grossly reproduces the observed hierarchies in quark
masses between generations as well as the CKM mixing hierarchy. The biggest outlier to
this picture is the bottom Yukawa coupling, which has a value of yb ∼ 10−2 .

10 Altmannshofer and Greljo

 This framework can be refined by also charging the right-handed up quarks uc , where
c denotes charge conjugation, resulting in a U (2)q+uc symmetry (13). In this setup, the
up-sector Yukawa matrix undergoes double suppression, where both left-handed and right-
handed rotations are perturbative and of similar structure to Lu,d discussed earlier. Conse-
quently, the singular values for the up-sector Yukawa matrix become Ŷu ∼ {b2 , a2 , 1}, while
those for the down-sector remain {b, a, 1}. A flavor-universal Z2 symmetry is introduced
whose breaking induces an overall downward shift in the down-quark mass spectrum by
a factor of ∼ 10−2 (e.g., via a two-Higgs doublet model setup). This setup provides a
better quantitative description and elegantly addresses the relative disparity between the
up- and down-sector hierarchies, as the down-sector hierarchy is more compressed than the
up-sector, and the third family is shifted downwards.
Leptons can naturally be included in both pictures. Since charged lepton mass hierar-
chies closely follow the down quarks, therefore, charging either left or right-handed leptons
produces Ŷe ∼ {b, a, 1}. When left-handed leptons are charged under the flavor symmetry,
the resulting perturbative left-handed rotations necessitate additional structure to account
for the large observed PMNS mixing (60). Alternatively, if the symmetry acts on right-
handed charged leptons (13), there are no constraints on the neutrino sector, naturally
allowing for the large PMNS mixing observed in experiments. Therefore, the most promis-
ing symmetries are U (2)q+e(c) and U (2)q+uc +ec . The latter, interestingly, is compatible
with SU (5) grand unification since 10 = q ⊕ uc ⊕ ec .
The IR features described above arise from specific UV dynamics. To illustrate one
possible UV completion, consider gauging an SU (2) flavor subgroup, which is anomaly-free
in certain cases, such as q +l. The SM gauge-singlet flavon field, Φ, is a flavor doublet whose
VEV breaks the gauged flavor symmetry at a high-energy scale ≳ PeV to comply with the
bounds from cLFV (60). The light Yukawa couplings emerge from a dimension-5 operator
in the symmetry-restored EFT, such as Λ1 q̄ΦHdp . When ⟨Φ⟩ = (0, vΦ )T , this operator gen-
erates the Yukawa coupling for one family, while insertions of Φ̃ = ϵΦ∗ provide masses for
the other family. With minimal field content completing the operator—for instance, a sin-
gle copy of a vectorlike quark weak doublet and a flavor singlet—the two operators become
proportional, leading to Yukawa matrices with rank two. Consequently, the minimal field
content enforces an accidental U (1) flavor symmetry for the first family at tree level, nat-
urally establishing the 1-2 hierarchy. A compelling mechanism to generate masses for the
first family is through radiative corrections (60). Rather than introducing new UV states,
new IR states S could run in the loop, with the first-family masses exhibiting only logarith-
mic sensitivity to their mass scale mS , provided mS ≲ mΨ . This approach fits well with
minimal quark-lepton unification, further enhancing the elegance of the mechanism (61).
This class of models, for sufficiently low flavor-breaking scales, predicts intriguing ex-
perimental signatures, including charged lepton flavor violation processes such as µ → e
conversion on heavy nuclei. By the end of this decade, upcoming experiments like Mu2e
and COMET are expected to make significant advancements, potentially probing the flavor-
breaking scale by another order of magnitude.

4.2. Flavor Deconstruction

Flavor deconstruction introduces a theoretical framework in which a gauge group G is ex-
tended to G3 in the UV, with one G factor assigned to each fermion family. The fermion
fields, fi (i = 1, 2, 3 for flavor), are each charged under their respective Gi factors, en-

www.annualreviews.org • Recent Progress in Flavor Model Building 11

 suring that anomaly cancellation occurs automatically within each family. The group G
may represent the full SM gauge group, a subgroup of it, or an extension incorporat-
ing partial or complete quark-lepton unification. This paradigm, involving non-universal
gauge symmetries, has gained momentum in recent years as a potential avenue for under-
standing flavor hierarchies as a consequence of multi-scale flavor dynamics, see for exam-
ple (62, 63, 64, 65, 66, 67, 68). The spontaneous symmetry breaking unfolds through a
two-step sequence. Initially, the extended gauge group G1 × G2 breaks to a diagonal sub-
group G12 , driven by a scalar field ϕ1 that acquires a vev, ⟨ϕ1 ⟩. Subsequently, the residual
G12 ×G3 symmetry breaks to the universal gauge group G via another scalar field ϕ2 , which
links G2 and G3 .
The spontaneous symmetry breaking to diagonal subgroups is naturally achieved using
scalar link fields in nontrivial representations under both gauge factors. To economize the
model, the fields are typically chosen so that ϕ1 f1 and f2 share the same gauge represen-
tation, as do ϕ2 f2 and f3 . The SM Higgs field H, in contrast, is charged only under G3 ;
therefore, the small Yukawas are generated via the link fields ϕ2 and ϕ1 . This configuration
offers a compelling mechanism for generating flavor hierarchies: the inter-family mass ratios
are dictated by the suppression factors ϵ1 = ⟨ϕ1 ⟩/Λ1 and ϵ2 = ⟨ϕ2 ⟩/Λ2 , where ϵi ∼ O(10−2 ).
Λi denotes the scale behind the effective operators responsible for the low-energy Yukawa
couplings.
This hierarchical structure, with ⟨ϕ1 ⟩ ≫ ⟨ϕ2 ⟩ > ⟨H⟩, effectively separates the dynamics
of the light and heavy families. The Higgs field is predominantly exposed to the third
family dynamics (69, 70, 71, 72) while being effectively shielded from the higher energy scales
associated with the lighter families. This framework allows for the TeV-scale NP with built-
in protection against FCNCs. Moreover, the multiscale structure preserves the stability of
the electroweak scale, ensuring the finite naturalness of the Higgs mass (62). In essence,
this setup elegantly connects the flavor puzzles of both the SM and NP to the electroweak
hierarchy problem. The IR phenomenology is governed by the U (2)5 flavor symmetry. These
models predict large effects in the third generation, specifically B-hadron and τ decays.
Direct LHC searches for new resonances coupling mostly to the third generation are less
constraining than those for the flavor-universal U (3)5 NP due to suppressed production
from valance parton distributions and difficulties detecting the third generation in the final
state (73, 74, 75, 76). This is most clearly exemplified by the constraints on semileptonic
interactions derived from high-mass Drell-Yan processes (77).
Furthermore, flavor deconstruction frameworks can be embedded within a deeper UV
context, such as warped extra dimensions (78), enriching their theoretical appeal. It has
also been proposed as a step toward electroweak flavor unification (79) or as a framework
capable of producing intriguing three-peaked stochastic gravitational wave signatures (80).
These would arise from a sequence of cosmological phase transitions in the early universe,
with the ratios of peak frequencies encoding the observed flavor hierarchies. Moreover, the
intrinsic multiscale structure that gives rise to quark and charged lepton hierarchies can
still accommodate the large (and seemingly anarchic) neutrino mixing in specific construc-
tions (63).

4.3. Revisiting U (1) Symmetries

As discussed in section 3.1.1, Froggatt-Nielsen models address the flavor problem by as-
signing flavor-dependent charges to the SM fermions under a U (1) flavor symmetry. This

12 Altmannshofer and Greljo

mechanism ensures that masses and mixing angles are parametrically determined by a
charge-dependent power of a small spurion, ϵn . A notable example of such a framework
was proposed long ago (16), using ϵ ≃ λ ≃ 0.2 to reproduce the observed hierarchical
masses and CKM mixing angles successfully. However, it is important to emphasize that
the choice of charge assignments is not unique. A wide range of viable configurations exists,
as free O(1) parameters can be tuned to accommodate observations with different sets of
charges or alternative values of the expansion parameter ϵ. As is often the case, there is a
trade-off between the simplicity of charge assignments and the degree of variation allowed
in the O(1) parameters. In (81, 82), numerous viable charge assignments have been iden-
tified, including some with remarkably minimal charges. Analogously, in the framework of
discrete FN models based on the ZN group, the simplest realizations occur for small values
of N ≥ 4 (83). These assignments produce distinct deviations in flavor-changing processes,
enabling future precision measurements to distinguish between different classes.
NP scenarios that account for the hierarchical flavor patterns of SM fermions often leave
distinct imprints on physics beyond the SM. This is well established in FN models, where the
horizontal U (1) symmetry dictates the flavor structure of NP couplings. Deviations from
the expected scaling may arise due to additional UV dynamics and can be systematically
classified and parameterized by higher powers of the flavor-breaking spurion (84). While
these extensions enhance the flexibility of FN models in fitting low-energy flavor data, their
size is constrained by self-consistency conditions.

Beyond their prominent role in FN models, flavor-dependent U (1) symmetries are sig-
nificant in other contexts as well. For instance, the accidental symmetries of the SM—the
three lepton family numbers Le , Lµ , Lτ , and baryon number—are observed to be exact
(perturbative) symmetries for vanishing neutrino masses. This could be because they arise
as global remnants of a spontaneously broken gauge symmetry. While B + L is anomalous,
many linear combinations of these symmetries can be consistently gauged if the SM fermion
content is extended by three right-handed neutrinos (85, 86).
A notable example is the difference between lepton flavor numbers, such as Lµ − Lτ ,
which remains anomaly-free even without right-handed neutrinos. A comprehensive catalog
of anomaly-free lepton-non-universal U (1) charges in the SM extended with +3νR was
developed in (86), building on the general classification framework of (87). In this scenario,
lepton flavor universality is generically violated, while cLFV is strongly suppressed (88,
89). These symmetries offer a compelling approach to (partially) address the NP flavor
puzzle, particularly by explaining the absence of cLFV, thereby permitting a lower NP
scale. Furthermore, certain chiral charge assignments enable radiative mass generation,
offering a partial solution to the SM flavor puzzle (86). As an additional application, a
class of lepton-flavored gauged U (1) models predicts exact proton stability, ensured by an
unbroken discrete remnant symmetry (90).
In phenomenological studies, lepton flavor universality violation in B-meson decays
naturally emerges in these models, either through a massive Z ′ gauge boson associated
with the spontaneously broken U (1) symmetry or via leptoquarks charged under U (1) (85,
91, 86). Additionally, a Z ′ boson with a mass in the range of 10–100 MeV could significantly
influence the anomalous magnetic moment of the muon (92).

www.annualreviews.org • Recent Progress in Flavor Model Building 13

 4.4. Flavor symmetries in the SMEFT
As already discussed in section 2.2, the SMEFT provides a valuable theoretical framework
to describe the low-energy effects of a generic high-energy theory beyond the SM. The
SMEFT Lagrangian consists of an infinite series of higher-dimensional local operators built
from SM fields and symmetries, with the details of the underlying theory encoded in the
Wilson coefficients. Investigating the correlations and patterns in these Wilson coefficients
helps map out the landscape of possible short-distance physics beyond the SM.
However, this broad approach comes with challenges. One of the main difficulties in
using SMEFT is the dramatic increase in independent Wilson coefficients, largely due to the
complexity introduced by flavor. For instance, when considering baryon-number conserving
dimension-six operators at leading order, the number of parameters surges from 59 with a
single generation to a staggering 2499 for three generations (93). Even in the renormalizable
SM, flavor-related parameters dominate—the Yukawa sector alone introduces 13 parame-
ters, compared to three in the gauge sector and two in the scalar potential. Understanding
the flavor structure is a critical open issue in the SMEFT.
Global flavor symmetries such as U (3)5 and U (2)5 and their breaking patterns offer an
effective way to structure the SMEFT, grouping theories beyond the SM into universality
classes (94, 95). Instead of considering the resulting model dependence as a limitation,
it should be seen as a valuable opportunity to gain systematic insights into UV physics
through experimental data.
Initial work has been done on developing a theoretical framework to classify flavor sym-
metries relevant to SMEFT. For example, in (95), the flavor structure of lepton and baryon
number-conserving dimension-6 operators within the SMEFT was investigated. Several
well-motivated flavor symmetries and symmetry-breaking patterns were proposed as com-
peting hypotheses regarding the ultraviolet dynamics beyond the SM. For each scenario,
independent operators were explicitly constructed and counted up to the first few orders in
the spurion expansion (see Fig. 1 in (95) for the leading order), offering setups ready for
phenomenological studies and global fits. Understanding the flavor structure of the SMEFT
renormalization group is essential (96), paving the way for phenomenological studies such
as those in (97, 98, 99).

4.5. Flavor Symmetries and Axions

As discussed in section 3.1.1, the SM flavor puzzle can be addressed by a spontaneously
broken horizontal U (1) flavor symmetry à la Froggatt-Nielsen. Such a dynamical struc-
ture implies observable consequences. In particular, if the U (1) symmetry is not gauged,
its spontaneous breaking gives a pseudo-Nambu-Goldstone boson, which is naturally light
and can be searched for in the experiment. Furthermore, if the U (1) flavor symmetry is
anomalous under QCD, the Goldstone can act as an axion and solve the strong CP prob-
lem (100). In this context, the pseudo-Nambu-Goldstone boson is sometimes referred to as
“axiflavon” (101) or “flaxion” (102), see also (103, 104, 105, 106). Generation-dependent
U (1) charges will, in general, give rise to flavor-changing couplings of the axion to fermions.
The couplings are proportional to the difference in U (1) charges and suppressed by the
axion decay constant fa .
In general, axion-like particles (ALPs) are well-motivated and naturally light extensions
of the SM, even if they do not address the strong CP problem. An example of an ALP
that is connected to a solution of the SM flavor puzzle is given in (83). The considered

14 Altmannshofer and Greljo

ALP is part of Froggatt-Nielsen models that are based on discrete ZN symmetries. In such
models, the ALP mass can be considerably heavier compared to continuous U (1) models or
the QCD axion. The Z4 and Z8 symmetries emerge as particularly promising examples.
An axion or ALP with flavor-changing couplings can give observable effects in a broad
range of flavor observables, including rare decays of mesons, neutral meson oscillations, rare
lepton decays, and electric and magnetic dipole moments of quarks and leptons (107, 108,
109, 110, 111, 112). Of particular interest are two-body decays of mesons and leptons that
can be parametrically enhanced compared to the standard weak decays of these particles.
For example, in the presence of a flavor changing strange-down coupling of an axion, the
branching ratio of the decay K → πa scales as BR(K → πa) ∝ m4W /(fa mK )2 . An analogous
enhancement is present in B → Ka or µ → ea. If the axion is light, ma ≲ 1 MeV, its
lifetime is macroscopic, and it will give a missing energy signature in detectors. Relevant
searches are therefore, for example, K → π + invisible, B → K + invisible, and µ →
e + invisible (113, 114, 115). Constraints from existing searches for K → π + invisible and
µ → e + invisible put bounds on the axion decay constant of around fa ≳ 1012 GeV (109)
and fa ≳ 109 GeV (113), respectively, if the couplings are flavor anarchic. Heavier ALPs
can give visible decay signatures, for example, a → γγ or a → e+ e− , that can be either
prompt or displaced and that can be effectively searched for at flavor factories.
Bounds derived from flavor observables are often competitive with those from collider
and beam dump experiments and astrophysical observations covering complementary pa-
rameter space regions. They are particularly significant in the MeV to GeV mass range,
where other probes typically provide weaker constraints.

4.6. Flavor Clockwork Models

The clockwork mechanism (116) is a fairly recent idea to generate exponentially small
interactions in theories without small parameters at the fundamental level. A fermionic
clockwork starts with 2N + 1 chiral fermions that are chained together by mass parameters
chosen such that a single chiral symmetry is shared among all the fermions. This gives
N massive Dirac fermions and a single massless chiral mode. The massless mode overlaps
with the N th original fermion that is suppressed by a clockwork factor 1/q N . If the free
parameter q is larger than one, q > 1, exponentially small couplings of the zero mode are
thus generated. The clockwork mechanism has been successfully applied to explain the
hierarchical structure of the quark, lepton, and neutrino masses (117, 118, 119, 120, 121,
122, 123). In these scenarios, indirect constraints from low-energy flavor observables, such
as rare decays and meson oscillations, can be effectively controlled. Furthermore, the new
degrees of freedom predicted by the fermionic flavor clockwork may be within reach of
collider experiments.
Flavor clockwork models share some similarities with FN setups. The role of the U (1)
charges is played by the length N of the clockwork chains, while the symmetry-breaking
spurion ⟨ϕ⟩/M ≪ 1 is replaced by the clockwork gear ratio 1/q. In fact, the clockwork
mechanism has motivated the construction of “inverted” Froggatt-Nielsen models, which
have the expansion parameter M/⟨ϕ⟩ (124). Interestingly, such constructions are anomaly-
free, and the horizontal U (1) symmetries can be gauged. In principle, the corresponding Z ′
gauge bosons can be light and can be searched for using precision flavor, astrophysics, and
beam dump experiments.
Supersymmetric versions of flavor clockwork models have been constructed in (125).

www.annualreviews.org • Recent Progress in Flavor Model Building 15

 The zero modes of the clockwork are identified with the fermions and sfermions of the
minimal supersymmetric SM. In addition to generating a hierarchical fermion spectrum,
the clockwork also predicts a specific flavor structure for the soft SUSY-breaking sfermion
masses. Interestingly, the simplest setup predicts large flavor mixing among first and second-
generation squarks, akin to the simplest supersymmetric U (1) FN models. Constraints from
kaon oscillations require the masses of either squarks or gluinos to be at the PeV scale or
above (126).

4.7. Modular and Eclectic Flavor Symmetries

Modular flavor symmetries were originally considered in the context of neutrino mass model
building (127, 128). While discrete flavor symmetries can successfully describe the mixing
pattern in the neutrino sector at zeroth order, it is challenging to obtain neutrino masses
and mixings that precisely match the observed values in minimal models. This challenge
has motivated research into a class of supersymmetric models that generalize discrete sym-
metries using so-called modular forms.2
In practice, Yukawa couplings are promoted to modular forms Y (τ ) that depend on
the complex modulus τ . If the vev of the modulus is close to a symmetry point (132,
133, 134), the Yukawa couplings can show a hierarchical pattern that is determined by
the “weight” and “level” of the modular form. In contrast to Froggatt-Nielsen models,
the achievable hierarchy patterns are more constrained, enhancing their predictive power.
These are determined by finite modular groups, which, in the simplest cases, are isomorphic
to permutation groups.
Models based on modular flavor symmetries usually have a much smaller field content
than FN models, which contain one or more flavon fields for spontaneous flavor breaking,
as well as a fairly large sector of vector-like matter that mediates the flavor breaking to
the SM fermions. Simple models based on modular flavor symmetries do not require such
ingredients. Interestingly, the most economical modular flavor model predicts the mass
splittings of neutrinos and the three mixing angles in the lepton sector in terms of a single
complex parameter, the vev of the modulus τ (127). Despite the challenge of reproducing
five known observables using only one complex parameter, this model performs relatively
well.
To obtain a hierarchical quark and charged lepton spectrum, as well as a hierarchical
CKM matrix more readily, one can add a “weighton” field, which provides suppression of
Yukawa couplings according to the chosen modular weights of the SM fermions (135). To
get a full agreement with observed masses and mixing angles, one typically needs to add
free parameters by writing the Yukawa couplings as linear combinations of more than one
modular form. Additional free parameters can also come from the Kähler potential (136).
Models that achieve fully realistic fermion masses and mixing can be found, for example,
in (137, 138, 139).
Modular flavor symmetries can also directly be combined with a U (1) flavor symmetry,
which is responsible for hierarchies among the three generations of quarks and lepton masses
and CKM angles. Such a setup has been constructed, for example, in (140) in the context of

2 Interestingly, the holomorphicity of the superpotential is an important ingredient to ensure

consistent modular transformation properties. Exploration of non-supersymmetric models with

modular invariance has only just started (129, 130, 131).

16 Altmannshofer and Greljo

a SU (5) grand unified theory. In another recent development involving modular symmetries,
an intriguing connection between the flavor puzzle and the strong CP problem has been
proposed (141).
Further developments in this direction of flavor model building are eclectic flavor symme-
tries (142, 143) that combine in a non-trivial way modular flavor symmetries, “traditional”
discrete flavor symmetries and CP-like symmetries in top-down string motivated models.

4.8. Flavor from Higher-Form and Non-Invertible Symmetries

Generalized symmetries extend the traditional notion of symmetries (144). So-called higher-
form global symmetries act not just on point-like objects like particles but on higher-
dimensional structures, such as lines, surfaces, or volumes. Non-invertible global symmetries
are permitted to have a product operation that goes beyond that of a group. In particular,
the inverse under the product operation does not need to exist.
Higher-form global symmetries and non-invertible global symmetries retain many fea-
tures and properties of “ordinary” global symmetries, and they lead, for example, to se-
lection rules on amplitudes. While the implications of these generalized symmetries are
still under active investigation, they provide new powerful tools to analyze the structure of
quantum field theories. They are also being applied to the study of particle physics phe-
nomenology. For a recent review on the topic of generalized symmetries in particle physics,
see (145).
Of particular interest to flavor model building is the result that non-invertible chiral
symmetries can naturally lead to exponential hierarchies. As shown in (146), symmetries
that suffer from abelian Adler-Bell-Jackiw anomalies can survive as non-invertible chiral
symmetries. These non-invertible symmetries imply selection rules for couplings that can
be broken by non-perturbative effects and give technically natural exponential hierarchies.
The authors of (146) apply this idea to generate exponentially small fermion masses and
speculate that this mechanism may find applications in particle physics model building.
In the related work (147), the authors show that gauging the familiar Lµ −Lτ symmetry
can result in non-invertible symmetries that protect neutrino masses. Non-zero (but expo-
nentially suppressed) neutrino masses can then arise non-perturbatively from instantons.
It needs to be seen if such constructions can provide a fully realistic neutrino spectrum,
including neutrino mixing, and if such a setup could be extended to the quark sector as
well.
In (148, 149), the authors show how certain string-motivated non-invertible symmetries
can be used to arrive at various textures of Yukawa matrices, including many examples with
so-called texture zeros both in the quark and lepton sector, which can be good starting
points for further flavor model building.
Finally, Ref. (150) initiates the study of the generalized global flavor symmetries of the
SM and their connection with models of gauge unification. They observe that the ordinary
SM flavor symmetries are intertwined with a one-form magnetic hypercharge symmetry,
forming a so-called two-group structure. They identify all possible vertical unification pat-
terns that are compatible with this structure. It might be interesting to extend such a study
to horizontal symmetries and gauge-flavor unification.

www.annualreviews.org • Recent Progress in Flavor Model Building 17

 5. CONCLUSIONS
The origin of flavor remains one of the key open questions in particle physics, driving efforts
to extend the SM. This challenge is tied to two interconnected puzzles. First, the observed
hierarchical patterns of SM fermion masses and mixing angles are highly non-generic, col-
lectively referred to as the SM flavor puzzle, suggesting the need for an explanation beyond
the SM. Second, NP near the electroweak scale typically induces unacceptably large con-
tributions to flavor-changing transitions, introducing the so-called NP flavor puzzle. This
puzzle suggests that NP must possess a non-generic flavor structure, particularly in sce-
narios addressing the Higgs naturalness problem, which often predicts new states around
the TeV scale with significant couplings to SM particles. Flavor model building seeks to
address both puzzles by constructing mechanisms to generate hierarchical flavor couplings,
integrating them into NP models, and deriving testable phenomenological predictions.
In this review, we have provided a concise summary of the two major flavor puzzles
in Section 2 and discussed traditional solutions in Section 3, including approaches that
use flavor symmetries to generate hierarchies, as well as geometric approaches. Recent
developments are reviewed in section 4.
Flavor models that are based on the popular U (2) or SU (2) flavor symmetries, as well
as U (1) models à la FN, continue to evolve and undergo refinement. Flavor models with a
spontaneously broken global U (1) symmetry predict the existence of light pseudo-Nambu-
Goldstone bosons, such as axions or axion-like particles, characterized by a distinctive
pattern of flavor-violating couplings. Developing innovative search strategies for these light,
weakly coupled particles remains a highly active area of research.
The concept of flavor deconstruction has recently garnered significant interest. This
framework proposes that gauge interactions are non-universal in the UV, with the SM
flavor hierarchies emerging from hierarchical scales associated with each family. This setup
also addresses the NP flavor puzzle, allowing for the TeV-scale physics, as anticipated by
the Higgs hierarchy problem.
Flavor symmetries have been systematically applied to the SMEFT, offering a robust
bottom-up framework to explore flavorful new physics. This approach enables a structured
investigation of how flavor dynamics can manifest in extensions of the SM. Among recent
developments, flavor models based on the clockwork mechanism proposed as an alternative
to FN models generate hierarchies through a structured chain of interactions. Additionally,
modular flavor symmetries have emerged as a powerful tool in model building, particularly in
the neutrino sector, where they offer elegant explanations for the observed flavor patterns by
linking flavor symmetries to the geometry of extra dimensions. Finally, innovative concepts
such as higher-form symmetries and non-invertible symmetries are beginning to shape the
landscape of flavor model building.
As precision flavor experiments push the boundaries of our understanding, flavor model
building remains indispensable, serving as both a guide for interpreting new data and a
cornerstone for uncovering the deeper principles that govern the structure of matter and
the fundamental forces of nature.

DISCLOSURE STATEMENT
The authors are not aware of any affiliations, memberships, funding, or financial holdings
that might be perceived as affecting the objectivity of this review.

18 Altmannshofer and Greljo

ACKNOWLEDGMENTS
We thank Joe Davighi, Xavier Ponce Dı́az, and Miguel Levy for their useful comments.
The research of WA is supported by the U.S. Department of Energy grant number DE-
SC0010107. AG has received funding from the Swiss National Science Foundation (SNF)
through the Eccellenza Professorial Fellowship “Flavor Physics at the High Energy Fron-
tier,” project number 186866.

LITERATURE CITED
1. Cerri A, et al. CERN Yellow Rep. Monogr. 7:867–1158 (2019)
2. Abada A, et al. Eur. Phys. J. ST 228(2):261–623 (2019)
3. Dong M, et al. arXiv:1811.10545 [hep-ex] (2018)
4. Navas S, et al. Phys. Rev. D 110(3):030001 (2024)
5. Esteban I, Gonzalez-Garcia MC, Maltoni M, Schwetz T, Zhou A. JHEP 09:178 (2020)
6. Aghanim N, et al. Astron. Astrophys. 641:A6 (2020), [Erratum: Astron.Astrophys. 652, C4
(2021)]
7. ’t Hooft G. NATO Sci. Ser. B 59:135–157 (1980)
8. Weinberg S. Phys. Rev. Lett. 43:1566–1570 (1979)
9. Harrison PF, Perkins DH, Scott WG. Phys. Lett. B 530:167 (2002)
10. Altarelli G, Feruglio F. Rev. Mod. Phys. 82:2701–2729 (2010)
11. Glashow SL, Iliopoulos J, Maiani L. Phys. Rev. D 2:1285–1292 (1970)
12. Grzadkowski B, Iskrzynski M, Misiak M, Rosiek J. JHEP 10:085 (2010)
13. Antusch S, Greljo A, Stefanek BA, Thomsen AE. Phys. Rev. Lett. 132(15):151802 (2024)
14. Froggatt CD, Nielsen HB. Nucl. Phys. B 147:277–298 (1979)
15. Leurer M, Nir Y, Seiberg N. Nucl. Phys. B 398:319–342 (1993)
16. Leurer M, Nir Y, Seiberg N. Nucl. Phys. B 420:468–504 (1994)
17. King SF, Ross GG. Phys. Lett. B 520:243–253 (2001)
18. Antusch S, King SF, Malinsky M, Ross GG. Phys. Lett. B 670:383–389 (2009)
19. Nardi E. Phys. Rev. D 84:036008 (2011)
20. Alonso R, Gavela MB, Merlo L, Rigolin S. JHEP 07:012 (2011)
21. Pomarol A, Tommasini D. Nucl. Phys. B 466:3–24 (1996)
22. Barbieri R, Dvali GR, Hall LJ. Phys. Lett. B 377:76–82 (1996)
23. Barbieri R, Hall LJ, Raby S, Romanino A. Nucl. Phys. B 493:3–26 (1997)
24. Barbieri R, Hall LJ, Romanino A. Phys. Lett. B 401:47–53 (1997)
25. Feldmann T, Mannel T. Phys. Rev. Lett. 100:171601 (2008)
26. Kagan AL, Perez G, Volansky T, Zupan J. Phys. Rev. D 80:076002 (2009)
27. Barbieri R, Buttazzo D, Sala F, Straub DM. JHEP 07:181 (2012)
28. Weinberg S. Phys. Rev. Lett. 29:388–392 (1972)
29. Banks T. Nucl. Phys. B 303:172–188 (1988)
30. Arkani-Hamed N, Cheng HC, Hall LJ. Phys. Rev. D 54:2242–2260 (1996)
31. Borzumati F, Farrar GR, Polonsky N, Thomas SD. Nucl. Phys. B 555:53–115 (1999)
32. Altmannshofer W, Frugiuele C, Harnik R. JHEP 12:180 (2014)
33. Barr SM. Phys. Rev. D 21:1424 (1980)
34. Balakrishna BS, Kagan AL, Mohapatra RN. Phys. Lett. B 205:345–352 (1988)
35. Dobrescu BA, Fox PJ. JHEP 08:100 (2008)
36. Graham PW, Rajendran S. Phys. Rev. D 81:033002 (2010)
37. Arkani-Hamed N, Schmaltz M. Phys. Rev. D 61:033005 (2000)
38. Randall L, Sundrum R. Phys. Rev. Lett. 83:3370–3373 (1999)
39. Grossman Y, Neubert M. Phys. Lett. B 474:361–371 (2000)
40. Gherghetta T, Pomarol A. Nucl. Phys. B 586:141–162 (2000)

www.annualreviews.org • Recent Progress in Flavor Model Building 19

 41. Huber SJ, Shafi Q. Phys. Lett. B 498:256–262 (2001)
42. Agashe K, Perez G, Soni A. Phys. Rev. D 71:016002 (2005)
43. Blanke M, Buras AJ, Duling B, Gori S, Weiler A. JHEP 03:001 (2009)
44. Cacciapaglia G, Csaki C, Galloway J, Marandella G, Terning J, Weiler A. JHEP 04:006 (2008)
45. Kaplan DB. Nucl. Phys. B 365:259–278 (1991)
46. Bellazzini B, Csáki C, Serra J. Eur. Phys. J. C 74(5):2766 (2014)
47. Panico G, Wulzer A. Lect. Notes Phys 913:1–316 (2016)
48. Csaki C, Falkowski A, Weiler A. JHEP 09:008 (2008)
49. Redi M, Weiler A. JHEP 11:108 (2011)
50. Glioti A, Rattazzi R, Ricci L, Vecchi L. arXiv:2402.09503 [hep-ph] (2024)
51. Chivukula RS, Georgi H. Phys. Lett. B 188:99–104 (1987)
52. Hall LJ, Randall L. Phys. Rev. Lett. 65:2939–2942 (1990)
53. Buras AJ, Gambino P, Gorbahn M, Jager S, Silvestrini L. Phys. Lett. B 500:161–167 (2001)
54. D’Ambrosio G, Giudice GF, Isidori G, Strumia A. Nucl. Phys. B 645:155–187 (2002)
55. Cirigliano V, Grinstein B, Isidori G, Wise MB. Nucl. Phys. B 728:121–134 (2005)
56. Barbieri R, Isidori G, Jones-Perez J, Lodone P, Straub DM. Eur. Phys. J. C 71:1725 (2011)
57. Fuentes-Martı́n J, Isidori G, Pagès J, Yamamoto K. Phys. Lett. B 800:135080 (2020)
58. Nir Y, Seiberg N. Phys. Lett. B 309:337–343 (1993)
59. Blum K, Grossman Y, Nir Y, Perez G. Phys. Rev. Lett. 102:211802 (2009)
60. Greljo A, Thomsen AE. Eur. Phys. J. C 84(2):213 (2024)
61. Greljo A, Thomsen AE, Tiblom H. JHEP 08:143 (2024)
62. Davighi J, Isidori G. JHEP 07:147 (2023)
63. Greljo A, Isidori G. Phys. Lett. B 856:138900 (2024)
64. Bordone M, Cornella C, Fuentes-Martin J, Isidori G. Phys. Lett. B 779:317–323 (2018)
65. Allwicher L, Isidori G, Thomsen AE. JHEP 01:191 (2021)
66. Fernández Navarro M, King SF. JHEP 08:020 (2023)
67. Barbieri R, Isidori G. JHEP 05:033 (2024)
68. Fuentes-Martı́n J, Lizana JM. JHEP 07:117 (2024)
69. Greljo A, Stefanek BA. Phys. Lett. B 782:131–138 (2018)
70. Allanach BC, Davighi J. JHEP 12:075 (2018)
71. Fuentes-Martı́n J, Stangl P. Phys. Lett. B 811:135953 (2020)
72. Covone S, Davighi J, Isidori G, Pesut M. arXiv:2407.10950 [hep-ph] (2024)
73. Davighi J, Stefanek BA. JHEP 11:100 (2023)
74. Davighi J, Gosnay A, Miller DJ, Renner S. JHEP 05:085 (2024)
75. Capdevila B, Crivellin A, Lizana JM, Pokorski S. JHEP 08:031 (2024)
76. Fernández Navarro M, King SF, Vicente A. JHEP 07:147 (2024)
77. Greljo A, Marzocca D. Eur. Phys. J. C 77(8):548 (2017)
78. Fuentes-Martin J, Isidori G, Lizana JM, Selimovic N, Stefanek BA. Phys. Lett. B 834:137382
(2022)
79. Davighi J, Tooby-Smith J. JHEP 09:193 (2022)
80. Greljo A, Opferkuch T, Stefanek BA. Phys. Rev. Lett. 124(17):171802 (2020)
81. Fedele M, Mastroddi A, Valli M. JHEP 03:135 (2021)
82. Cornella C, Curtin D, Neil ET, Thompson JO. arXiv:2306.08026 [hep-ph] (2023)
83. Greljo A, Smolkovič A, Valenti A. JHEP 09:174 (2024)
84. Asadi P, Bhattacharya A, Fraser K, Homiller S, Parikh A. JHEP 10:069 (2023)
85. Altmannshofer W, Davighi J, Nardecchia M. Phys. Rev. D 101(1):015004 (2020)
86. Greljo A, Soreq Y, Stangl P, Thomsen AE, Zupan J. JHEP 04:151 (2022)
87. Allanach BC, Davighi J, Melville S. JHEP 02:082 (2019), [Erratum: JHEP 08, 064 (2019)]
88. Altmannshofer W, Gori S, Pospelov M, Yavin I. Phys. Rev. D 89:095033 (2014)
89. Alonso R, Grinstein B, Martin Camalich J. JHEP 10:184 (2015)
90. Davighi J, Greljo A, Thomsen AE. Phys. Lett. B 833:137310 (2022)

20 Altmannshofer and Greljo

 91. Davighi J, Kirk M, Nardecchia M. JHEP 12:111 (2020)
92. Greljo A, Stangl P, Thomsen AE, Zupan J. JHEP 07:098 (2022)
93. Alonso R, Jenkins EE, Manohar AV, Trott M. JHEP 04:159 (2014)
94. Faroughy DA, Isidori G, Wilsch F, Yamamoto K. JHEP 08:166 (2020)
95. Greljo A, Palavrić A, Thomsen AE. JHEP 10:010 (2022)
96. Machado CS, Renner S, Sutherland D. JHEP 03:226 (2023)
97. Allwicher L, Cornella C, Isidori G, Stefanek BA. JHEP 03:049 (2024)
98. Grunwald C, Hiller G, Kröninger K, Nollen L. JHEP 11:110 (2023)
99. Greljo A, Palavrić A, Smolkovič A. Phys. Rev. D 109(7):075033 (2024)
100. Wilczek F. Phys. Rev. Lett. 49:1549–1552 (1982)
101. Calibbi L, Goertz F, Redigolo D, Ziegler R, Zupan J. Phys. Rev. D 95(9):095009 (2017)
102. Ema Y, Hamaguchi K, Moroi T, Nakayama K. JHEP 01:096 (2017)
103. Arias-Aragon F, Merlo L. JHEP 10:168 (2017), [Erratum: JHEP 11, 152 (2019)]
104. Björkeroth F, Di Luzio L, Mescia F, Nardi E. JHEP 02:133 (2019)
105. Bonnefoy Q, Dudas E, Pokorski S. JHEP 01:191 (2020)
106. Di Luzio L, Guerrera AWM, Dı́az XP, Rigolin S. JHEP 06:046 (2023)
107. Choi K, Im SH, Park CB, Yun S. JHEP 11:070 (2017)
108. Björkeroth F, Chun EJ, King SF. JHEP 08:117 (2018)
109. Martin Camalich J, Pospelov M, Vuong PNH, Ziegler R, Zupan J. Phys. Rev. D 102(1):015023
(2020)
110. Bauer M, Neubert M, Renner S, Schnubel M, Thamm A. JHEP 04:063 (2021)
111. Carmona A, Scherb C, Schwaller P. JHEP 08:121 (2021)
112. Bauer M, Neubert M, Renner S, Schnubel M, Thamm A. JHEP 09:056 (2022)
113. Calibbi L, Redigolo D, Ziegler R, Zupan J. JHEP 09:173 (2021)
114. Jho Y, Knapen S, Redigolo D. JHEP 10:029 (2022)
115. Knapen S, Langhoff K, Opferkuch T, Redigolo D. arXiv:2311.17915 [hep-ph] (2023)
116. Giudice GF, McCullough M. JHEP 02:036 (2017)
117. Patel KM. Phys. Rev. D 96(11):115013 (2017)
118. Ibarra A, Kushwaha A, Vempati SK. Phys. Lett. B 780:86–92 (2018)
119. Alonso R, Carmona A, Dillon BM, Kamenik JF, Martin Camalich J, Zupan J. JHEP 10:099
(2018)
120. Hong S, Kurup G, Perelstein M. JHEP 10:073 (2019)
121. Abreu de Souza F, von Gersdorff G. JHEP 02:186 (2020)
122. Kang YJ, Kim S, Lee HM. JHEP 09:005 (2020)
123. Babu KS, Saad S. Phys. Rev. D 103(1):015009 (2021)
124. Smolkovič A, Tammaro M, Zupan J. JHEP 10:188 (2019), [Erratum: JHEP 02, 033 (2022)]
125. Altmannshofer W, Gadam SA. Phys. Rev. D 104(3):035030 (2021)
126. Altmannshofer W, Harnik R, Zupan J. JHEP 11:202 (2013)
127. Feruglio F. arXiv:1706.08749 [hep-ph] (2019)
128. Ding GJ, King SF. Rept. Prog. Phys. 87(8):084201 (2024)
129. Ding GJ, Feruglio F, Liu XG. JHEP 01:037 (2021)
130. Qu BY, Ding GJ. JHEP 08:136 (2024)
131. Ding GJ, Lu JN, Petcov ST, Qu BY. arXiv:2408.15988 [hep-ph] (2024)
132. Okada H, Tanimoto M. Phys. Rev. D 103(1):015005 (2021)
133. Feruglio F, Gherardi V, Romanino A, Titov A. JHEP 05:242 (2021)
134. Novichkov PP, Penedo JT, Petcov ST. JHEP 04:206 (2021)
135. King SJD, King SF. JHEP 09:043 (2020)
136. Chen MC, Ramos-Sánchez S, Ratz M. Phys. Lett. B 801:135153 (2020)
137. Lu JN, Liu XG, Ding GJ. Phys. Rev. D 101(11):115020 (2020)
138. Okada H, Tanimoto M. Eur. Phys. J. C 81(1):52 (2021)
139. Liu XG, Yao CY, Ding GJ. Phys. Rev. D 103(5):056013 (2021)

www.annualreviews.org • Recent Progress in Flavor Model Building 21

 140. de Anda FJ, King SF, Perdomo E. Phys. Rev. D 101(1):015028 (2020)
141. Feruglio F, Parriciatu M, Strumia A, Titov A. JHEP 08:214 (2024)
142. Baur A, Nilles HP, Trautner A, Vaudrevange PKS. Phys. Lett. B 795:7–14 (2019)
143. Nilles HP, Ramos-Sánchez S, Vaudrevange PKS. JHEP 02:045 (2020)
144. Gaiotto D, Kapustin A, Seiberg N, Willett B. JHEP 02:172 (2015)
145. Brennan TD, Hong S. arXiv:2306.00912 [hep-ph] (2023)
146. Cordova C, Ohmori K. Phys. Rev. X 13(1):011034 (2023)
147. Cordova C, Hong S, Koren S, Ohmori K. Phys. Rev. X 14(3):031033 (2024)
148. Kobayashi T, Otsuka H. JHEP 11:120 (2024)
149. Kobayashi T, Otsuka H, Tanimoto M. arXiv:2409.05270 [hep-ph] (2024)
150. Cordova C, Koren S. Annalen Phys. 535(8):2300031 (2023)

22 Altmannshofer and Greljo