Pith. sign in

REVIEW 1 major objections 5 minor 43 references

Local subtractions in renormalization-group flows lose their guaranteed monotonicity at second order.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

T0 review · deepseek-v4-flash

2026-08-04 07:07 UTC pith:4E52YIRO

load-bearing objection Strong, honest paper: an exact defect b-flow with a sharp monotonicity transition at Δ=1/2, a clean rank-matching framework, and a Sec. VI F-loss reversal that is conditional on a stated but unproven large-N assumption. the 1 major comments →

arxiv 2608.02447 v1 pith:4E52YIRO submitted 2026-08-03 hep-th

Rank matching in renormalization-group irreversibility: Exact defect and entropic tests

Francesco Scardino This is my paper
classification hep-th
keywords renormalization groupirreversibilityrank matchingconnected cumulantsdefect b-functionentanglement entropyF-theoremmonotonicity
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper proposes rank matching to explain why some renormalization-group monotones exist and others fail. Locality fixes a subtraction of degree r; one further scale derivative always exposes a connected response of rank r+1, while the positivity inputs used in irreversibility proofs—bilinear forms and positive second variations—can sign only quadratic data. Degree one is therefore the last subtraction whose derivative those inputs can sign directly; higher-order filters need extra dynamics or a different projection. Exact solvable tests confirm the pattern: massive scalars on odd spheres turn around, a surface-defect b-function is monotone only for defect dimension at least 1/2, and disk entropy and sphere free energy share endpoints while distributing F loss differently over scale.

Core claim

For a one-scale observable W(R), let s_r = P_r(D)W be a locally subtracted presentation with a degree-r polynomial filter. One further scale derivative D s_r contains a nonzero separated connected cumulant κ_{r+1}(X) of the integrated scale deformation X, plus lower-rank and local terms. Because the positive structures available—reflection-positive overlaps, positive coupling-space metrics, strong-subadditivity Hessians, spectral sums of squares—are quadratic forms, they can sign only rank-two data. Hence at r=1 the derivative is within reach of positivity once a Ward or entropic identity supplies the quadratic form, while for r≥2 same-observable rank-two positivity cannot by itself fix the

What carries the argument

Rank matching and the scale-cumulant tower. The integrated scale deformation X generates derivatives through ∂_t⟨Y⟩=⟨∂_t Y⟩−⟨Y X⟩_c; iterating gives ∂_t^n W = (−1)^n κ_n(X) plus mixed insertions of rank at most n−1 and local terms. For a degree-r filter s_r=P_r(D)W, the exact statement D s_r = a_{r+1} κ_{r+1}(X)+R_{≤r}+local, with a_{r+1}≠0, fixes which connected response the monotonicity test exposes. Combined with the observation that the positivity inputs are quadratic forms, this explains why the direct proof stops at degree one and why the defect b-flow has a sharp phase boundary at Δ=1/2.

Load-bearing premise

The rank-matching rule presupposes that the positivity inputs for a given presentation are exactly quadratic forms in the same scale deformation; if a positive higher-rank structure or a different projection applies—as the paper notes happens for the absorptive four-dilaton amplitude—the obstruction does not constrain the sign, so the central claim is a statement about a class of proofs rather than a universal property of all RG functions.

What would settle it

Evaluate the running b-function of the four-dimensional monodromy defect at ν=2 by computing the Dirichlet sum in Eq. (40) numerically: the paper predicts DS_b<0 for all x>0 when ε_b≤1 and DS_b>0 near the infrared when ε_b>1. A computation that finds no overshoot for ε_b=1.4, or that finds a positive derivative for ε_b=0.8, would refute the phase diagram.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

If this is right

  • Any locally subtracted presentation of degree r≥2 has a scale derivative that generically involves connected three-point (or higher) data; sign-definiteness cannot be inferred from the same observable's two-point positivity alone.
  • The filtered free energy of a conformally coupled massive scalar on S^p turns around for every odd p≥3, so endpoint ordering does not imply running monotonicity of the sphere observable.
  • The natural surface-defect b-function is strictly decreasing for 1/2≤Δ<1, necessarily nonmonotone for 0<Δ<1/2, and satisfies b_UV≥b_IR on both sides; a local four-dimensional monodromy defect realizes the entire transition.
  • At the threshold Δ=1/2 the flow coefficient is completely monotone in the canonical spectral coordinate: derivatives of every order alternate in sign.
  • Disk entropy and filtered sphere free energy define two different positive loss measures with the same total F loss; the factor-of-two ultraviolet mismatch forces the sphere profile to lose more on a later finite range, giving a measure reversal (and, under continuity, a pointwise crossing).

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The rank-matching rule offers a practical diagnostic for any candidate running monotone: if the subtraction degree is two or higher and no positive spectral representation or alternative projection is supplied, a turnover is the default expectation.
  • The F-profile reversal implies that 'how fast F is lost' is not a universal datum even when endpoint charges agree; comparisons of RG loss rates between observables are presentation-dependent.
  • The threshold complete-monotonicity hierarchy suggests a testable conjecture: self-dual or threshold defect flows may be characterized by complete monotonicity of their canonical flow coefficients, which can be checked in other generalized-free defect families and in finite-multiplicity Gaussian realizations.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

1 major / 5 minor

Summary. The paper proposes a 'rank-matching' selection rule for renormalization-group irreversibility: a degree-r local subtraction produces, after one further scale derivative, a rank-(r+1) connected cumulant, whereas the standard positivity inputs (reflection-positive bilinear forms, Zamolodchikov/Fisher metrics, strong-subadditivity Hessians) control only quadratic/rank-two data. Hence degree one is the last subtraction whose monotonicity can be closed by such inputs alone. The paper then presents exact solvable tests: (i) massive scalars on odd spheres with p≥3 give nonmonotone filtered free energies; (ii) a generalized-free surface-defect b-function is strictly monotone for 1/2≤Δ<1, necessarily nonmonotone for 0<Δ<1/2, with a complete-monotonicity hierarchy at the threshold; (iii) a four-dimensional monodromy defect realizes the full b-flow phase; and (iv) under an explicitly stated large-N entropy assumption, disk entropy and filtered sphere free energy have equal total F loss but UV-ordering-reversed loss profiles. The paper distinguishes endpoint ordering, running monotonicity, and profile universality.

Significance. If the results hold, the paper gives a useful organizing principle for why some RG monotones close at quadratic order and others fail, backed by exact spectral derivations rather than numerical evidence. The generalized-free defect phase diagram, the complete-monotonicity hierarchy at Δ=1/2, and the four-dimensional monodromy parent are concrete, parameter-free results with independent anomaly cross-checks. The odd-sphere turnover derivation and the Dirichlet-form proof of the endpoint anomaly are detailed and internally consistent. The F-loss measure-reversal is novel but explicitly conditional on an unproven large-N entropy limit; this is the main weakness. Overall the exact b-flow and odd-sphere results are solid enough to justify publication if the conditional result is presented with its status made unambiguous.

major comments (1)
  1. [Sec. VI.C, Eq. (63)] The measure-reversal result rests on the equality μ_EE((0,∞))=μ_sph((0,∞))=ζ(3)/(8π²). This is not derived in the manuscript. The assumptions in Sec. VI.A assert that C_EE has an N→∞ limit at fixed x and approaches the expected UV and IR fixed-point coefficients, but strong subadditivity is a statement about the finite-N entropy; it does not by itself imply that the coefficientwise limit is nonincreasing, nor that no endpoint boundary layer exists. If an O(N^0) singlet entropy loss escapes into a boundary layer near x→0 or x→∞, the signed measure has nonzero total mass and the compensation argument in Eqs. (64)-(66) fails. Since the equal-total-loss premise is the load-bearing step that converts the local UV factor-of-two into a forced later compensation, this needs either a proof (for example, from a controlled 1/N expansion of the von Neumann entropy) or an explicit downgrade of the re
minor comments (5)
  1. [Sec. VI.A] The paragraph beginning 'Under this assumption, strong subadditivity passes to finite differences of the singlet coefficient' is duplicated verbatim on the first page of Sec. VI; one copy should be removed.
  2. [Throughout] The hat on \hat\Delta appears as a superscript 'b' in many places ('b∆'), presumably a macro/rendering problem. Please fix the notation so that the defect-primary dimension is typeset consistently.
  3. [Sec. V.A] The relation between the microscopic component number N, the multiplicity ν of generalized-free components, and the large-N 'parent' interpretation is easy to misread. The text eventually clarifies that ν is unrelated to N, but a single compact definition before Eq. (35) would improve readability.
  4. [Appendix C] The variables ξ and \tilde ξ in Eq. (C12) are adopted from Ref. [22] without a self-contained definition. A one-sentence explanation of their physical meaning (allowed mode content at the two endpoints) would help the reader verify the cross-check in Eq. (C13).
  5. [Sec. IV.A, Fig. 1] The caption says the curves are 'oriented by (-1)^{q+1}' and normalized to unit peak. It would be helpful to state explicitly that the raw derivative has alternating endpoint signs and that this orientation is what makes all curves cross zero in the same direction.

Circularity Check

0 steps flagged

No significant circularity: rank-matching rule and defect phase diagram are derived from explicit spectral sums; self-citations are not load-bearing.

full rationale

The central rank-matching selection rule is an exact cumulant-counting statement (Eq. (8)) plus a moment-cone no-implication example (Eq. (13)); it is not obtained by fitting a parameter or by assuming the conclusion. The defect b-flow phase diagram is derived from an absolutely convergent spectral Dirichlet-form sum (Eqs. (38), (40), Appendix D), with the endpoint anomaly Eq. (39) cross-checked against the independent monodromy-defect anomaly formula of Ref. [22]. The monotone/nonmonotone sides and the epsilon_b=1 threshold follow from explicit bounds (Eq. (42)) and infrared asymptotics (Eq. (43)), not from the definition of the filter. Section VI's F-loss reversal is explicitly conditional on a stated large-N entropy limit; the total-mass equality follows from shared conformal endpoints, and the ultraviolet ordering is computed in perturbation theory, so the reversal is a conditional theorem rather than a fitted prediction. Refs. [1,2] are self-citations, but they provide only the off-critical presentation vocabulary and the S3 example; the new claims (higher-odd spheres, b-phase diagram, monodromy parent, F-loss reversal) are derived in the text and appendices without relying on those citations as evidence. The minor self-citation warrants a small score, but I find no circular reduction of any central result to its inputs.

Axiom & Free-Parameter Ledger

0 free parameters · 5 axioms · 1 invented entities

No empirically fitted parameters: the parameters (m, f0, alpha, epsilon_b, nu, N) are physical variables of the solvable theories. The new content rests on standard QFT positivity assumptions, one explicitly stated large-N entropy limit, and a known monodromy-defect construction with an independent anomaly cross-check.

axioms (5)
  • domain assumption The positivity inputs considered are bilinear forms or positive second variations (reflection positivity, Fisher/Zamolodchikov metrics, strong-subadditivity Hessians, spectral representations).
    Sec. II B states this as the common origin of all cited positivity proofs; the rank-matching rule applies to this class of inputs, and Sec. VII C explicitly lists alternative routes beyond it.
  • ad hoc to paper The leading O(N^0) singlet coefficient of the disk von Neumann entropy has an N→∞ limit at fixed x and agrees with the UV/IR fixed-point coefficients.
    Sec. VI A and the abstract say 'Under a stated assumption'; this is required to pass strong subadditivity to finite differences and to equate endpoint losses in the F-profile comparison.
  • domain assumption The defect b-sum rule of Ref. [14] applies to the contact-subtracted noncoincident trace two-point correlator.
    Sec. V B and App. D use this sum rule to derive the Dirichlet form (38) and endpoint anomaly (39); the fixed-bulk caveat is noted in Sec. IV B.
  • domain assumption The four-dimensional monodromy defect with the alternate (singular) boundary condition is unitary and reflection positive.
    App. C cites Refs. [22,23] for the self-adjoint extension and reflection positivity; the bulk is free, so Wick factorization is exact.
  • standard math Spectral sums may be evaluated by zeta-function regularization and analytic continuation (e.g., zeta(-2k)=0, polynomial division of the spectral polynomial).
    App. B uses this to derive the exact resolvent (B7) for the odd-sphere witnesses; this is standard mathematical practice in QFT spectral analysis.
invented entities (1)
  • Four-dimensional monodromy (disorder) defect independent evidence
    purpose: Realizes the full 0 < hatDelta < 1 generalized-free defect-primary interval at finite multiplicity (nu = 2) in a local free bulk theory; provides the four-dimensional parent of the b-flow phase diagram.
    The alternate boundary condition and its defect anomaly bIR - bUV = -2 alpha^3 are matched to the independent defect-anomaly result in Ref. [22] (App. C), giving an external falsifiable handle. It is a known class of boundary condition reused as a parent, not a speculative new particle or force.

pith-pipeline@v1.3.0-daily-deepseek · 155 in / 15558 out tokens · 224602 ms · 2026-08-04T07:07:58.738550+00:00 · methodology

0 comments
read the original abstract

Why do local subtractions produce renormalization-group monotones in some settings but fail in others? We propose rank matching. Local counterterms fix how many scale derivatives scheme independence requires. Every derivative adds one connected insertion, while the available positivity inputs are bilinear forms or positive second variations and therefore control only quadratic data. Degree one is the last subtraction whose scale derivative stays within their direct reach. First-order subtractions can close when a Ward or entropic identity supplies a signed quadratic form. Higher orders need extra dynamics. Three exactly solvable tests exhibit both outcomes and show that the endpoint inequality can hold while running monotonicity fails. Massive scalars yield nonmonotone filtered free energies on every odd dimensional $p$-sphere with $p\geq3$. A generalized-free surface-defect $b$-function is strictly monotone when the defect-primary dimension $\widehat\Delta$ satisfies $1/2\leq\widehat\Delta<1$, and necessarily nonmonotone for $0<\widehat\Delta<1/2$. At the threshold, its flow coefficient is completely monotone in the canonical spectral coordinate, with derivatives of every order alternating in sign. A local four-dimensional monodromy defect realizes the full transition. Under a stated assumption on the large-component entropy limit, disk entropy and sphere free energy share endpoints and total $F$ loss but distribute it differently over scale. Rank matching separates endpoint ordering, running monotonicity, and the distribution of loss over scale.

Figures

Figures reproduced from arXiv: 2608.02447 by Francesco Scardino.

Figure 1
Figure 1. Figure 1: FIG. 1. Scale derivatives of the locally filtered scalar free energies on [PITH_FULL_IMAGE:figures/full_fig_p006_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: FIG. 2. The derivative [PITH_FULL_IMAGE:figures/full_fig_p008_2.png] view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

43 extracted references · 32 linked inside Pith

  1. [1]

    The monotone branch of the defect b family and the special sphereFprofile are examples

    Additional dynamics may control the higher con- nected variations of the same presentation. The monotone branch of the defect b family and the special sphereFprofile are examples

  2. [2]

    Disk entropy does this for F

    A different presentation of the same fixed-point charge may lower the filter degree and restore a quadratic identity. Disk entropy does this for F . In four dimensions the null-cone Markov construction instead uses a subtracted entropy observable to prove the endpointa-inequality [11, 12]

  3. [3]

    A different projection may reorganize higher-point information into quadratic positive data. The four- dilaton proof of the endpoint a-theorem projects onto the absorptive part of a forward amplitude, where unitarity produces a positive sum of squared transition amplitudes [4]. The routes need not be disjoint. The second and third change the positivity pr...

  4. [4]

    This is the standard tubular regulator of the self-adjoint extension

    Equation (C1) denotes the resulting renormalized quadratic form. This is the standard tubular regulator of the self-adjoint extension. The regular endpoint replaces +αby−αin Eq. (C3). The mode with transverse spin j⊥ = −α has two scale- invariant endpoint conditions, φ−α(r⊥, y)∼c −r−α ⊥ bΨ−(y),dim bΨ− = 1−α, φ−α(r⊥, y)∼c +rα ⊥ bΨ+(y),dim bΨ+ = 1 +α.(C4) T...

  5. [5]

    Dirichlet form, endpoint anomaly , and ultraviolet sign On the radius- R sphere, write the undeformed generalized-free two-point kernel as G0(Ω,Ω ′) = X ℓ,m λℓ(R)Yℓm(Ω)Y ∗ ℓm(Ω′), λℓ(R) =λ 0(R)qℓ.(D1) Gaussian resummation, equivalently the leading-large- N Hubbard–Stratonovich resummation in a parent theory, gives Gf =G 0(1 +f 0G0)−1, U(z) = ∞X ℓ=0 (2ℓ+ 1...

  6. [6]

    mX n=1 pn(α2)−ms # ≥sup m≥0

    Global monotonicity for0< ε b ≤1 Recall that α= εb 2 ,C α(x) = ∞X n=1 v3 n, 16 M(x) =u 2 0(1−u 0)2.(D16) Differentiating the telescoped identity (D12) gives xJ ′ εb (x) = 2εb [M(x) +R α(x)], Rα(x) = ∞X n=1 v3 n (1−u n−1 −u n).(D17) Since 0 < un < 1, one has Rα ≥ −Cα. It is therefore enough to bound the cubic mass of the spectral front. The recurrence give...

  7. [7]

    Define L = (cεb x)1/εb

    Infrared overshoot for1< ε b <2 To analyze the infrared endpoint, use the regularly varying spectrum qn ∼c εb n−εb , c εb = Γ(1 +ε b/2) Γ(1−ε b/2) .(D33) We abbreviate Bεb = B 3− 2 εb ,3 + 2 εb .(D34) Here B is the Euler beta function. Define L = (cεb x)1/εb . Uniformly on compact subsets of z >0, Stirling’s formula gives u⌊Lz⌋ − →U(z) = 1 1 +z εb , 17 Lv...

  8. [8]

    Moreover, R ∞ 0 ρ0(σ)dσ=ω(∞)−ω(0) = 1

    This function vanishes at the origin and has derivative ysinhy >0. Moreover, R ∞ 0 ρ0(σ)dσ=ω(∞)−ω(0) = 1. To obtain the flow coefficient, set g(τ) =τ 2Ib(τ), Bb(τ) = 2−3τ ψ1 τ+ 1 2 −τ 2ψ2 τ+ 1 2 , g′(τ) =τ Bb(τ).(E4) Letr(σ) =ω(σ)/σ 2 and H(σ) =−σ 3r′(σ) = 2ω(σ)−σω ′(σ).(E5) Then g′(τ) =τ 2 Z ∞ 0 dσ e−τ σH(σ).(E6) SinceH(0) = 0, a second integration by pa...

  9. [9]

    (1.2) of Ref

    Matching the disk and sphere normalizations For the disk deformation write δI=f Z d3y V(y), V= O2 2 .(F1) In the free-vector normalization introduced in the main text, ⟨O(y)O(0)⟩ c = 1√ 2|y| 2 , ⟨V(y)V(0)⟩ c = 1 4|y|4 .(F2) Using Eq. (1.2) of Ref. [33] in the canonical dimension-two normalization, whose two-point coefficient is CV = 1/π2, the boosted-SSA ...

  10. [10]

    Exact round-sphere loss profile For the ∆O = 1 singlet flow, define the running variable by the eigenvalues of the round-sphere determinant, ∆FS3 (x) = 1 2 ∞X n=1 n2 log 1 + x n , x= √ 2π 2f R, (F5) with its covariant local completion understood. Acting on one harmonic with Φ S3 = (1 −D )(1 −D/ 3)FS3 and H (0) sph = −DCsph/2, where D = x∂x, gives the nonl...

  11. [11]

    Santoni and F

    G. Santoni and F. Scardino,Nonadditivity in Quantum Field Theory: Replica Energies, Scaling Filters, and the Renormalization Group, arXiv:2606.12367

    Pith/arXiv arXiv

  12. [12]

    Santoni and F

    G. Santoni and F. Scardino, Phys. Rev. D113, L121701 (2026), arXiv:2603.09799

    Pith/arXiv arXiv 2026

  13. [13]

    Cuomo, Z

    G. Cuomo, Z. Komargodski, and A. Raviv-Moshe, Phys. Rev. Lett.128, 021603 (2022), arXiv:2108.01117

    Pith/arXiv arXiv 2022

  14. [14]

    Komargodski and A

    Z. Komargodski and A. Schwimmer, JHEP12(2011) 099, arXiv:1107.3987

    Pith/arXiv arXiv 2011

  15. [15]

    A. B. Zamolodchikov, JETP Lett.43, 730 (1986)

    1986

  16. [16]

    Nagar, A

    I. Nagar, A. Sever, and D.-L. Zhong, JHEP06(2024) 110, arXiv:2404.07290

    Pith/arXiv arXiv 2024

  17. [17]

    Casini, I

    H. Casini, I. Salazar Landea, and G. Torroba, JHEP10 (2016) 140, arXiv:1607.00390

    Pith/arXiv arXiv 2016

  18. [18]

    Casini and M

    H. Casini and M. Huerta, J. Phys. A40, 7031 (2007), arXiv:cond-mat/0610375

    Pith/arXiv arXiv 2007

  19. [19]

    Liu and M

    H. Liu and M. Mezei, JHEP04(2013) 162, arXiv:1202.2070

    Pith/arXiv arXiv 2013

  20. [20]

    Casini and M

    H. Casini and M. Huerta, Phys. Rev. D85, 125016 (2012), arXiv:1202.5650

    Pith/arXiv arXiv 2012

  21. [21]

    Casini, E

    H. Casini, E. Test´ e, and G. Torroba, Phys. Rev. Lett.118, 261602 (2017), arXiv:1704.01870

    Pith/arXiv arXiv 2017

  22. [22]

    Casini, E

    H. Casini, E. Test´ e, and G. Torroba, J. Phys. A50, 364001 (2017), arXiv:1703.10656

    Pith/arXiv arXiv 2017

  23. [23]

    Casini, M

    H. Casini, M. Huerta, and R. C. Myers, JHEP05(2011) 036, arXiv:1102.0440

    Pith/arXiv arXiv 2011

  24. [24]

    Shachar, R

    T. Shachar, R. Sinha, and M. Smolkin, SciPost Phys.15 (2023) 240, arXiv:2212.08081

    Pith/arXiv arXiv 2023

  25. [25]

    Jensen and A

    K. Jensen and A. O’Bannon, Phys. Rev. Lett.116, 091601 (2016), arXiv:1509.02160

    Pith/arXiv arXiv 2016

  26. [26]

    Wang, JHEP11(2021) 122, arXiv:2012.06574

    Y. Wang, JHEP11(2021) 122, arXiv:2012.06574

    Pith/arXiv arXiv 2021

  27. [27]

    Shachar, R

    T. Shachar, R. Sinha, and M. Smolkin, JHEP09(2024) 057, arXiv:2404.18403

    Pith/arXiv arXiv 2024

  28. [28]

    Casini, I

    H. Casini, I. Salazar Landea, and G. Torroba, JHEP07 (2023) 004, arXiv:2303.16935

    Pith/arXiv arXiv 2023

  29. [29]

    Nishioka and Y

    T. Nishioka and Y. Sato, JHEP05(2021) 074, arXiv:2101.02399

    Pith/arXiv arXiv 2021

  30. [30]

    Sato, JHEP05(2021) 202, arXiv:2102.11468

    Y. Sato, JHEP05(2021) 202, arXiv:2102.11468

    Pith/arXiv arXiv 2021

  31. [31]

    Kobayashi, T

    N. Kobayashi, T. Nishioka, Y. Sato, and K. Watanabe, JHEP 01(2019) 039, arXiv:1810.06995

    Pith/arXiv arXiv 2019

  32. [32]

    Bianchi, A

    L. Bianchi, A. Chalabi, V. Proch´ azka, B. Robinson, and J. Sisti, JHEP08(2021) 013, arXiv:2104.01220

    Pith/arXiv arXiv 2021

  33. [33]

    Giombi, E

    S. Giombi, E. Helfenberger, Z. Ji, and H. Khanchandani, JHEP02(2022) 041, arXiv:2102.11815

    Pith/arXiv arXiv 2022

  34. [34]

    S. S. Gubser and I. R. Klebanov, Nucl. Phys. B656, 23 (2003), arXiv:hep-th/0212138

    Pith/arXiv arXiv 2003

  35. [35]

    Giombi and B

    S. Giombi and B. Liu, JHEP12(2023) 004, arXiv:2305.11402

    Pith/arXiv arXiv 2023

  36. [36]

    Tr´ epanier, JHEP09(2023) 074, arXiv:2305.10486

    M. Tr´ epanier, JHEP09(2023) 074, arXiv:2305.10486

    Pith/arXiv arXiv 2023

  37. [37]

    Raviv-Moshe and S

    A. Raviv-Moshe and S. Zhong, JHEP08(2023) 143, arXiv:2305.11370

    Pith/arXiv arXiv 2023

  38. [38]

    I. R. Klebanov, T. Nishioka, S. S. Pufu, and B. R. Safdi, JHEP10(2012) 058, arXiv:1207.3360

    Pith/arXiv arXiv 2012

  39. [39]

    I. R. Klebanov, S. S. Pufu, and B. R. Safdi, JHEP10(2011) 038, arXiv:1105.4598

    Pith/arXiv arXiv 2011

  40. [40]

    Wang, JHEP02(2022) 061, arXiv:2101.12648

    Y. Wang, JHEP02(2022) 061, arXiv:2101.12648

    Pith/arXiv arXiv 2022

  41. [41]

    Komargodski, JHEP07(2012) 069, arXiv:1112.4538

    Z. Komargodski, JHEP07(2012) 069, arXiv:1112.4538

    Pith/arXiv arXiv 2012

  42. [42]

    Yonekura, JHEP04(2013) 011, arXiv:1212.3028

    K. Yonekura, JHEP04(2013) 011, arXiv:1212.3028

    Pith/arXiv arXiv 2013

  43. [43]

    Faulkner, JHEP05(2015) 033, arXiv:1412.5648

    T. Faulkner, JHEP05(2015) 033, arXiv:1412.5648

    Pith/arXiv arXiv 2015