Pith. sign in

REVIEW 3 major objections 4 minor 7 references

In a closure test with an injected W' signal, a Standard Model-only PDF fit absorbs the new physics into the large-x antiquark PDFs, and using those biased PDFs in a subsequent SMEFT fit excludes the true theory and looks compatible with th

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 →

T0 review · deepseek-v4-flash

2026-08-02 03:25 UTC pith:CRXGPOLY

load-bearing objection Worth hearing warning about PDF contamination, but the conservative energy-cut remedy is validated only with a post hoc Qmax choice; the simultaneous fit is the cleaner result. the 3 major comments →

arxiv 2607.13885 v1 pith:CRXGPOLY submitted 2026-07-15 hep-ph

Tailored PDFs for new physics searches

classification hep-ph
keywords parton distribution functionsSMEFTDrell-YanHL-LHCnew physicsclosure testSIMUnetlarge-x PDFs
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 demonstrates that a standard two-step analysis — first fit the proton's parton distribution functions assuming the Standard Model, then fit new-physics coefficients with those PDFs fixed — can actively hide new physics. In a simulated HL-LHC dataset with an injected W' boson, the first step absorbs the signal into the poorly constrained large-x antiquark PDFs; the second step then excludes the true new-physics value and appears compatible with the Standard Model. The paper proposes and validates two remedies: conservative PDF fits that cut data above an energy threshold, and a simultaneous PDF+SMEFT fit performed with the SIMUnet tool. Both recover the injected signal and exclude the Standard Model. The upshot is that how PDFs are produced matters for whether an indirect new-physics search succeeds or fails.

Core claim

The central claim is that the large-x parton distribution functions, which are mostly pinned down by high-energy tails of collider observables, can soak up unaccounted energy-growing new physics without worsening the fit. In the authors' closure test, a heavy W' boson parametrized by the SMEFT coefficient Ŵ=8×10−5 is injected into synthetic high-mass Drell-Yan data at the HL-LHC. A PDF fit performed under the Standard Model assumption yields a BSM-biased PDF set whose large-x antiquark PDFs absorb the signal; a SMEFT-only fit using these biased PDFs then excludes the true underlying theory and appears compatible with the Standard Model. The paper further claims two robust strategies — a cons

What carries the argument

The machinery is a closure test: NNPDF4.0 is taken as the true underlying PDF set, synthetic high-mass Drell-Yan data at 14 TeV are generated including the SMEFT effect of a heavy W', and fits are compared. The key mechanism is the degeneracy between new-physics contributions growing with energy and the large-x PDF degrees of freedom — the fit can trade one for the other because both shape the high-energy tail. The two mitigation tools are an energy cut Qmax, which removes the BSM-sensitive high-energy data from the PDF fit, and SIMUnet, a neural-network PDF-fitting framework augmented with an SMEFT layer that fits PDFs and Wilson coefficients in a single simultaneous fit.

Load-bearing premise

The mitigation strategies assume new-physics signals grow steadily with energy, so cutting data above Qmax cleanly removes the contamination; if the signal sits in an intermediate mass window or appears at threshold, the cut cannot be chosen a priori and the conservative approach can fail.

What would settle it

Repeat the closure test with a lighter W' (e.g., MW' around 2 TeV, Ŵ larger) so the new-physics effect peaks in the intermediate-invariant-mass region rather than growing through the entire tail, and check whether the Qmax=500 GeV conservative fit still recovers the injected signal; a failure there would show that the proposed mitigation is tied to the energy-growth assumption.

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

If this is right

  • If a PDF set is fitted under the Standard Model assumption, a later SMEFT-only analysis using that set can falsely exclude the true new-physics parameter point and look Standard-Model-compatible.
  • Removing high-energy data above an aggressive cut (Qmax=500 GeV) before fitting PDFs prevents absorption and yields SMEFT bounds centered on the true injected coefficient.
  • A simultaneous PDF+SMEFT fit recovers the true PDF luminosity and excludes the Standard Model in the (Ŵ, Ŷ) plane without needing an energy cut.
  • Varying Qmax and checking for systematic drift in PDFs and SMEFT bounds is a model-agnostic diagnostic for BSM contamination; SM-only data yields cut-independent results.
  • The same qualitative recovery holds in a heavy color-octet scenario affecting the ttbar high-mass tail, where the simultaneous fit gives the tighter bound.

Where Pith is reading between the lines

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

  • A direct corollary the paper leaves implicit: global PDF fits that do not include SMEFT degrees of freedom should be re-checked against simultaneous fits whenever the observables extend into kinematic regions where new physics would be largest.
  • The absorption mechanism implies that the size of large-x PDF shifts between SM-only and simultaneous fits could itself serve as a data-driven alarm for possible new physics.
  • A testable extension: repeat the closure test with new physics that is not monotonic in energy, such as a narrow resonance or threshold production, to map exactly where the conservative energy-cut strategy fails; the paper's cut value has no a priori justification.
  • The Qmax-cut strategy trades away high-energy constraining power; an alternative worth testing is profiling or penalizing large-x PDF directions instead of cutting data, which may retain more information.

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

3 major / 4 minor

Summary. This paper presents a closure-test study of the risk that standard sequential PDF-then-SMEFT analyses can hide new physics. Using NNPDF4.0 as the underlying truth, the author generates synthetic HL-LHC high-mass Drell-Yan data that include a W'-induced SMEFT signal (W_hat = 8e-5) and then fits PDFs under the SM assumption. A subsequent SMEFT-only fit with these BSM-biased PDFs excludes the true injected W_hat and appears compatible with the SM, demonstrating that the signal can be absorbed into the large-x antiquark PDFs. The paper then compares two mitigation strategies: conservative fits that impose an energy cut Qmax on the PDF fit, and simultaneous PDF+SMEFT fits using SIMUnet. The conservative fit with Qmax = 500 GeV recovers the injected signal, while Qmax = 1500 GeV does not; the simultaneous fit recovers the signal in a two-parameter W_hat-Y_hat scenario and reproduces the true PDF luminosity. The paper also asserts, without showing results, that the same conclusions hold for a colour-octet scenario in t-tbar production. The full numerical details are delegated to Ref. [1].

Significance. If the central claim holds, the paper provides an important quantitative caution: global PDF fits performed under the SM assumption may absorb energy-growing BSM effects, and a subsequent SMEFT-only analysis using those PDFs can then exclude the true theory with high apparent significance. The closure-test construction is appropriate, the injected parameters are stated explicitly, and the simultaneous-fit strategy with SIMUnet is a credible and useful alternative that avoids the need for an energy cut. The paper also benefits from using established public tools and a reproducible methodology, which strengthens the credibility of the main negative result. However, the validation of the conservative energy-cut strategy is weakened by the post hoc choice of Qmax, and the generalization to t-tbar is asserted without quantitative support. The simultaneous-fit result is the more robust and original contribution.

major comments (3)
  1. [Section 3, Fig. 2] The demonstration that the conservative strategy works relies on a post hoc choice of Qmax = 500 GeV. The paper shows that Qmax = 1500 GeV fails and explicitly admits 'there is no a priori way of choosing an optimal Qmax'. Since the successful cut is selected after seeing the injected signal, the claim that the conservative approach 'successfully recovers the injected new physics' as a model-agnostic mitigation strategy is not established. The suggested diagnostic of varying Qmax and checking stability is not demonstrated to converge to the truth: a comparison between, e.g., 1500 GeV and 2000 GeV could appear mutually consistent while both PDFs are already biased. The paper should either provide an a priori criterion for selecting Qmax, or explicitly weaken the claim to an illustration that a sufficiently low cut can remove the contamination when the NP scale is already known.
  2. [Section 4, paragraph 3] The colour-octet/t-tbar scenario is asserted without any figure, table, or quantitative fit result: 'We find the same qualitative outcome in an independent scenario...'. Since the abstract and Section 4 present this as evidence that both mitigation strategies generalize beyond the W' Drell-Yan case, the claim is unsupported in this manuscript. If the details are contained in Ref. [1], the text should include at least a summary plot or a quantitative comparison; otherwise this generalization should be marked as a pointer to future work, not as a result of this paper.
  3. [General (Sections 2-4)] The manuscript is explicitly a summary of Ref. [1] and delegates all numerical methods, fit settings, replica counts, and validation to that reference. As a standalone journal submission, this makes the closure test impossible to verify from the text alone. The reader cannot assess the statistical robustness of the fits (e.g., number of replicas, neural-network architecture, minimization settings, or the validation procedure). At minimum, the paper should state the key hyperparameters and the number of replicas, and clarify what new material it adds beyond Ref. [1]. This is acceptable for a proceedings contribution but is a substantive deficiency for a standard journal article.
minor comments (4)
  1. [Equation (1)] The notation 'O(6) i' should be written as O_i^{(6)} or similar for clarity; the superscript is currently ambiguous.
  2. [Figure 1] The y-axis is labeled 'Number of replicas' but the total number of replicas is not stated in the text. Please specify the replica count and indicate whether the histograms are normalized.
  3. [Abstract and Section 3] The abstract states that 'Both approaches are shown to successfully recover the injected new physics', but the conservative approach's success is conditional on a Qmax value that the text itself says cannot be chosen a priori. The wording should be qualified to reflect this limitation.
  4. [Section 2] The phrase 'BSM-biased PDFs obtained by fitting the same data under the SM assumption' is slightly confusing: the fit is to the synthetic data, but 'the same data' could be misread as the SM-only assumption being part of the data. Rephrase for clarity.

Circularity Check

0 steps flagged

No significant circularity: the central closure test is explicit and self-contained; the conservative-cut Qmax choice is an acknowledged post-hoc limitation rather than a disguised input, and the self-citations are not load-bearing for the Drell-Yan result.

full rationale

The paper is an explicit closure-test summary rather than a first-principles derivation. In Sect. 2, the BSM-biased PDF set is constructed by fitting the same synthetic data under the SM assumption, so finding that a subsequent SMEFT-only fit with those PDFs returns W_hat≈0 is an expected consequence of the construction; however the paper labels this as a closure test and does not present the biased fit as an independent prediction. The positive control with true PDFs (Fig. 1) shows the SMEFT fit itself can recover the injection. No equation in the paper makes the result equivalent to its input. The conservative strategy in Sect. 3 uses Qmax=500 GeV, and the paper explicitly states 'there is no a priori way of choosing an optimal Qmax'; this is a post-hoc selection and a validity caveat, but not a circular step because the paper admits it and the energy-growth argument is independent. The t-tbar colour-octet generalisation in Sect. 4 is deferred to Ref. 1, a self-authored companion, which is an omitted proof in this contribution; but the central Drell-Yan demonstration is displayed in this paper and uses open-source SIMUnet/NNPDF code, so it is not a closed self-citation loop. Overall, the circularity score is low; the caveats concern external validity and completeness, not definitional circularity.

Axiom & Free-Parameter Ledger

3 free parameters · 4 axioms · 0 invented entities

No new entities are introduced. The inputs are standard SMEFT benchmark scenarios and an existing PDF set; the free parameters are scenario injections and the post hoc chosen energy cut.

free parameters (3)
  • Q_max (energy cut) = 500 GeV (1500 GeV also tested)
    Chosen by hand after comparing outcomes; the 500 GeV cut is the one that 'fully removes' the BSM distortion, while 1500 GeV fails. The paper admits there is no a priori way to choose the optimal cut.
  • Injected Wilson coefficient W-hat = 8e-5 (M_W' = 13.8 TeV)
    A scenario parameter chosen large enough to be detectable; the closure test's success depends on the injection being sizable.
  • Injected Wilson coefficient Y-hat = 1.5e-4 (M_Z' = 18.7 TeV)
    Second correlated SMEFT direction injected only in the simultaneous-fit test; arbitrary scenario choice.
axioms (4)
  • domain assumption The heavy W' effect is adequately parametrised by a single SMEFT oblique coefficient W-hat (and Y-hat by Z')
    Used to generate synthetic data and define the fitted Wilson coefficients. If real new physics involves additional operators, the recovery demonstration does not transfer.
  • domain assumption SMEFT effects grow monotonically with energy
    Invoked in Section 3 to justify the Qmax cut strategy. Not all BSM scenarios satisfy this; the conservative approach depends on this property.
  • domain assumption NNPDF4.0 Monte Carlo replicas provide the true underlying PDF and uncertainty law
    Closure-test setup in Section 2: synthetic data are generated from NNPDF4.0 as the 'true' PDF. The demonstration is only as valid as this benchmark.
  • domain assumption The PDF fit under the SM assumption has sufficient flexibility to absorb the injected signal
    The hiding effect relies on NNPDF neural-network PDFs being able to reabsorb the distortion into large-x antiquark PDFs; a less flexible parametrisation could reduce the effect.

pith-pipeline@v1.3.0-alltime-deepseek · 3724 in / 15386 out tokens · 152220 ms · 2026-08-02T03:25:00.139239+00:00 · methodology

0 comments
read the original abstract

Parton Distribution Functions (PDFs) at large Bjorken-$x$ are mostly constrained by high-energy measurements, and thus risk to absorb the energy-growing effects of unaccounted new physics (NP) if it were present in the high-energy tails of hadron-collider observables used in global PDF fits. If undetected, such contamination biases the resulting PDFs and can potentially hide the absorbed NP signals from subsequent searches. We illustrate this effect in a Standard Model Effective Field Theory (SMEFT) risk-assessment scenario affecting high-mass Drell-Yan production at the High-Luminosity LHC, and compare two strategies to obtain robust PDFs and SMEFT bounds: conservative fits that exclude data above an energy cut, and simultaneous fits of PDFs and Wilson coefficients performed with the {\tt SIMUnet} tool. Both approaches are shown to successfully recover the injected new physics that is otherwise absorbed and hidden by the PDFs.

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

7 extracted references · 7 linked inside Pith

  1. [1]

    E. Cole, M. N. Costantini, E. Hammou, L. Mantani, F. Merlotti, M. Morales-Alvarado and M. Ubiali, arXiv:2602.20235 [hep-ph]

  2. [2]

    Hammou, Z

    E. Hammou, Z. Kassabov, M. Madigan, M. L. Mangano, L. Mantani, J. Moore, M. Morales Alvarado and M. Ubiali, JHEP11, 090 (2023), arXiv:2307.10370 [hep-ph]

  3. [3]

    Greljo, S

    A. Greljo, S. Iranipour, Z. Kassabov, M. Madigan, J. Moore, J. Rojo, M. Ubiali and C. Voisey, JHEP07, 122 (2021), arXiv:2104.02723 [hep-ph]

  4. [4]

    M. N. Costantini, E. Hammou, Z. Kassabov, M. Madigan, L. Mantani, M. Morales Alvarado, J. M. Moore and M. Ubiali (PBSP), Eur. Phys. J. C84(2024) 805, arXiv:2402.03308 [hep- ph]

  5. [5]

    Iranipour and M

    S. Iranipour and M. Ubiali, JHEP05, 032 (2022), arXiv:2201.07240 [hep-ph]

  6. [6]

    R. D. Ballet al.(NNPDF), Eur. Phys. J. C82(2022) 428, arXiv:2109.02653 [hep-ph]

  7. [7]

    Hammou and M

    E. Hammou and M. Ubiali, Phys. Rev. D111(2025) 095028, arXiv:2410.00963 [hep-ph]