{"id":"52ba1194-7adf-42b4-90d3-f348f1251270","arxiv_id":"2506.20080","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":3.0,"correctness_risk":"low","formal_verification":"none","parameter_count":5,"one_line_summary":"Electroweak precision data set mass-dependent upper limits on dark photon mixing, and the Z width adds direct 95 percent bounds on its coupling to a 10 GeV dark fermion.","lead":"An electroweak precision fit using the latest Z and W data places updated 95 percent confidence limits on the dark photon mixing parameter and, for the first time, on the dark photon's coupling to a dark matter fermion. The CDF W boson mass case tightens limits below the Z mass but leaves a poor overall fit even at the best dark photon point.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The 95% exclusion in Eq. (8) is computed as χ²_AD−χ²_SM rather than relative to the dark-photon best fit; for the CDF W mass this changes the threshold at m_AD=200 GeV from 72.0 to 37.5, so the reported curves are not confidence intervals.","rationale":"The reader's weakest assumption concerns one-loop dark-photon corrections and correlation structure; that is a reasonable robustness worry, but the more immediate problem is internal to the statistical construction. The paper's Eq. (7) defines a standard χ², but Eq. (8) invokes a nonstandard reference point. For nested models the correct LRT for a fixed ε is against the best fit of the larger model, not against the SM. In the CDF scenario the SM is not just slightly worse; χ²_SM=68.2 vs χ²_AD,min=33.7 is a 34.5-unit improvement for one extra parameter. A threshold of 3.8 above 68.2 yields a 95% 'allowed' band that includes both the SM point and the best-fit point, which no single 95% confidence set for ε can do if the SM itself is excluded by the same data. This directly undermines the claimed CDF constraint, which the abstract foregrounds. I would not reject the paper outright: the PDG-based blue curve and the g_χ constraints may shift only mildly, because there the best-fit improvement over SM is small, and the formal setup is conventional. But the central statistical claim needs recomputation using profile-likelihood intervals. Hence I retain the reader's CONDITIONAL verdict, with the condition now being this recalculation rather than (or in addition to) the radiative-correction check. The reader's weakest_assumption did not identify this issue.","tokens_in":6547,"tokens_out":13093,"duration_ms":150288,"concrete_test":"Recompute the Fig. 1 curves with the same data and covariance matrices but with the profile-likelihood statistic Δχ²(ε)=χ²_AD(ε)−min_{ε′} χ²_AD(ε′) at fixed m_AD (and analogously in the (ε,g_χ) plane), instead of χ²_AD−χ²_SM. The paper's own numbers imply that the CDF red curve at m_AD=200 GeV will move from the χ²=72.0 level to the χ²=37.5 level; if it does, Eq. (8) must be replaced and the '95% CL exclusion' language revised.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"Equation (8) defines the 95% exclusion by χ²_AD(ε)−χ²_SM ≥ 3.8, i.e. by comparison with the Standard-Model minimum. A confidence interval on ε must instead use the minimum of the dark-photon model at the same m_AD: Δχ²(ε)=χ²_AD(ε)−min_{ε′} χ²_AD(ε′) ≥ 3.84. The two prescriptions are numerically very different in the CDF case. The paper reports χ²_SM=68.2 and, at (m_AD,ε)=(200 GeV,0.1001), χ²_AD=33.7. Equation (8) therefore excludes only ε with χ²_AD above 72.0, while the profile-likelihood 95% region excludes χ²_AD above 37.5. The red CDF curve is thus not a 95% confidence upper limit; it is a comparison against a SM fit that has χ²/dof=68.2/12 and is itself rejected at very high significance. The same construction enters the two-parameter Eq. (11) for g_χ. Since the mass-dependent tightening/relaxation in Fig. 1 and the abstract's exclusion statement are the central results, this statistical choice is load-bearing.","agreement_with_reader":"disagree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript revisits electroweak precision constraints on a kinetically mixed dark photon. The authors fit the Standard Model to a set of Z-pole observables, m_W, and Gamma_W, with m_W taken either from the PDG average or from the CDF measurement, obtaining chi^2_SM = 12.9 and 68.2, respectively. They then add a dark photon, using tree-level mass-matrix diagonalisation, and derive 95% exclusion contours for the kinetic mixing parameter epsilon (Eq. 8, Fig. 1), finding that the CDF value tightens the limits for m_AD < m_Z and relaxes them above m_Z. They extend the model to a dark Dirac fermion chi with m_chi = 10 GeV and derive limits in the (epsilon, g_chi) plane from the invisible Z width (Eqs. 9-11, Fig. 2). The central quantitative claim is that a dark photon with m_AD = 200 GeV and epsilon = 0.1001 improves the CDF fit to chi^2 = 33.7 and reduces the W-mass tension to 2.9 sigma.","tokens_in":6848,"tokens_out":5418,"duration_ms":55784,"significance":"If the reported constraints were correct, the paper would provide a useful update of dark photon EWPO limits and one of the direct bounds on the dark fermion coupling g_chi from Z-width data. The authors use a standard chi^2 minimisation, give explicit formulas for the mixing angle and partial width, and make the comparison against both PDG and CDF W-mass values; the cross-check against Ref. [20] is also a useful point of contact. The main statistical construction, however, means that the quoted CDF '95% exclusion' is not a confidence interval, so the headline results need substantial revision before their significance can be assessed.","major_comments":[{"comment":"The 95% exclusion is defined as chi^2_AD(epsilon) - chi^2_SM >= 3.8, i.e. as a comparison with the Standard-Model minimum rather than with the minimum of the dark-photon model at fixed m_AD. A confidence interval on epsilon must instead use the profile likelihood, Delta chi^2(epsilon) = chi^2_AD(epsilon) - min_{epsilon'} chi^2_AD(epsilon') >= 3.84. In the CDF case the two prescriptions differ materially: chi^2_SM = 68.2 and the best dark-photon point has chi^2_AD = 33.7, so Eq. (8) excludes epsilon with chi^2_AD above 72.0, whereas the profile-likelihood region excludes chi^2_AD above 37.5. The red CDF curve in Fig. 1 is therefore not a 95% confidence upper limit, and the same construction enters the two-parameter bound of Eq. (11). Because the abstract and Fig. 1 present these as exclusions at 95% CL, this statistical choice is load-bearing and must be corrected.","section":"Sec. 4.2, Eq. (8)"},{"comment":"The m_chi dependence is not treated as a parameter. The Z -> chi chi partial width in Eq. (10) depends on m_chi through the factor (1 + 2 m_chi^2 / m_Z^2) sqrt(1 - 4 m_chi^2 / m_Z^2), so the limits on g_chi weaken as m_chi approaches m_Z / 2. The paper fixes m_chi = 10 GeV, and the abstract claims 'first electroweak precision observable constraints' on the dark photon coupling to dark fermions; as presented, the claim applies to a single mass point and is not a constraint on the coupling model. A scan over m_chi, or at least an explicit statement of the mass range for which the limits apply, is needed.","section":"Sec. 4.3, Eq. (10) and Fig. 2"},{"comment":"The covariance matrix is not displayed and theory uncertainties are not propagated. The W-mass parametrisation of Ref. [15] and the W-width parametrisation of Ref. [16] are treated as exact theory predictions, although these predictions carry residual theoretical errors. Since the epsilon exclusions are driven by differences at the 10^-2 to 10^-3 GeV level in m_W and Gamma_W, neglecting these errors could overstate the limits. The authors should state the theory uncertainties and test their effect on the exclusion curves.","section":"Sec. 3, Eq. (7) and Table 1"}],"minor_comments":[{"comment":"There are typos: 'predications' in the Introduction and 'scenarious' in the Conclusions.","section":"Sec. 1 and Sec. 5"},{"comment":"The threshold 3.8 is a rounded form of the standard one-parameter 95% value 3.84; the paper should use the exact value for consistency with Eq. (11)'s 5.99.","section":"Sec. 4.2, Eq. (8)"},{"comment":"The sentence 'The covariance matrix ... is diagonal' is contradicted by the correlation matrices adopted from Refs. [18,19]; the experimental uncertainty matrix is diagonal, but the full covariance matrix is not.","section":"Sec. 3"},{"comment":"The grey region attributed to 'eigenmass repulsion' is not explained in the text; a short description of the branch choice in Eq. (3) would help the reader.","section":"Fig. 1"},{"comment":"The paper refers to 'the latest dataset found in [17]', but Ref. [17] is the 2022 PDG review; the dataset should be updated or the reference should be made precise.","section":"Table 1 and Ref. [17]"},{"comment":"The black dotted 'EWPO limit' taken from Ref. [6] is not described in the text; it should be stated which observables and which statistical definition that curve uses.","section":"Fig. 1"}],"recommendation":"major_revision","confidential_remarks":"The paper overlaps substantially with the authors' earlier paper (Ref. [11]) for the epsilon-only limits; the genuinely new elements are the CDF comparison and the g_chi constraints. The main concern, the use of Delta chi^2 relative to the SM minimum instead of the model minimum, is correctable in a revision. I would welcome a version that redefines the limits as profile-likelihood confidence regions and reports theory uncertainties."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Quick take: the stress-test note is correct and it matters. The authors define their 95% exclusions by chi2_AD - chi2_SM >= 3.8, i.e., how much worse the dark photon fit is than the SM fit. That is a model-comparison threshold, not a confidence interval on epsilon. For the CDF W mass the difference is enormous: their own numbers give chi2_SM = 68.2 and chi2_AD = 33.7 at the best fit (200 GeV, 0.1001), so their threshold is 72.0 while a profile-likelihood 95% threshold would be 37.5. The red CDF curve in Fig. 1 is therefore not an upper limit; it has a different shape and excludes almost nothing near the best fit. The same construction enters Eq. (11) for g_chi. That is load-bearing, because the mass-dependent tightening/relaxation and the 'first EWPO constraints on g_chi' are the two advertised results.\n\nWhat is good: the paper is cleanly written, uses a standard tree-level dark photon formalism, and reproduces the known epsilon constraints for the PDG W mass. The observation that constraints tighten below the Z pole and relax above is qualitatively consistent with earlier work. The tree-level Z->chi chi-bar width formula is simple and correctly identified. If the authors had used a profile likelihood, the g_chi bounds would likely survive in some form and could be a useful complement to relic-density and direct-detection limits. The references are appropriate, including the self-citation to Ref. [11], which is properly disclosed as the basis of this proceedings contribution.\n\nSoft spots besides the statistics: no covariance matrices or input tables are supplied, so the fit cannot be reproduced; theory uncertainties are not propagated; m_chi is fixed at 10 GeV without a scan; and one-loop dark photon corrections are not assessed. The 'first' claim also needs a clearer boundary against Ref. [11], since the paper says it is 'based on' that work.\n\nWho this is for: dark photon phenomenologists who want the updated CDF-W constraint should read Refs. [20-23] instead. The g_chi part is the only reason to look here, and it needs a corrected statistical treatment before it can be trusted.\n\nRecommendation: do not accept as is. The fix is straightforward — redo the exclusions with Delta chi2 relative to the dark-photon best fit, show the covariance matrix, and resubmit as a short note. A serious referee should see a corrected version, not this one.","headline":"The paper's advertised 95% exclusion curves are not confidence intervals — Eq. (8) compares against the SM minimum rather than the dark photon best fit — and the only genuinely new result (g_chi constraints) sits on that same flawed statistics.","tokens_in":7454,"tokens_out":5668,"would_cite":false,"duration_ms":63864,"reading_group":"no","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"A kinetically mixed dark photon tightens epsilon limits below the Z pole and can lower the CDF W-mass tension to 2.9 sigma.","keywords":["dark photon","kinetic mixing","electroweak precision observables","W boson mass","CDF W-mass anomaly","dark fermion coupling","Z invisible decay","95% confidence exclusions"],"falsifier":"A single future $W$-mass measurement with roughly 1 MeV uncertainty whose central value matches the current world average would falsify the CDF-motivated dark photon window, since that point predicts $m_W=80.4060$ GeV, about 29 MeV above the PDG value and far outside such an error bar. Independently, a per-mille measurement of the invisible $Z$ width would test the $g_\\chi$ plane directly through Eq. (10).","tokens_in":6313,"feed_emoji":"🌌","tokens_out":10302,"duration_ms":102622,"temperature":0.7,"pith_summary":"This paper argues that the full set of electroweak precision observables confines a kinetically mixed dark photon: for every dark photon mass $m_{A_D}$, the mixing parameter $\\epsilon$ is excluded above a 95% confidence curve, and that curve moves when the newer CDF $W$ mass is used. The exclusion tightens for $m_{A_D}m_Z$. It further claims the first electroweak-precision constraints on the dark photon's coupling $g_\\chi$ to a Dirac dark fermion, through the invisible decay $Z\\to\\chi\\bar\\chi$. If correct, the CDF anomaly is softened but not removed: the best dark photon point, $m_{A_D}=200$ GeV and $\\epsilon=0.1001$, predicts $m_W=80.4060$ GeV, reducing the discrepancy to $2.9\\sigma$ while the fit value $\\chi^2=33.7$ remains large.","feed_headline":"Dark photon eases W-mass anomaly to 2.9 sigma","feed_subtitle":"Electroweak precision fits also put the first direct limits on the dark photon's coupling to dark fermions.","key_machinery":"The machine that carries the argument is the tree-level mass-matrix diagonalisation of the dark photon and the Standard Model neutral gauge boson (Eqs. 2 to 4), which turns the kinetic mixing parameter $\\epsilon$ into a physical mixing angle $\\alpha$ and hence into shifts in the $Z$ couplings, including $C_{Z,\\chi\\bar\\chi}=g_\\chi\\sin\\alpha/\\sqrt{1-\\epsilon^2/\\cos^2\\theta_W}$. The Standard Model radiative corrections are kept in fixed parametrisations, so the dark photon enters only through these tree-level shifts. The $\\chi^2$ statistic with experimental covariance (Eq. 7), together with the exclusion thresholds of Eqs. (8) and (11), converts those shifts into 95% confidence regions. A distinctive feature is the grey exclusion region near $m_{A_D}=m_Z$, produced by eigenmass repulsion in the mixing-angle formula.","core_discovery":"The paper's central claim is that electroweak precision observables, collected into a $\\chi^2$ that compares theory with measurement, place tight limits on the dark photon. Adding the dark photon through tree-level kinetic mixing shifts the $Z$ mass and its couplings via a physical mixing angle, and the 95% exclusion on $\\epsilon$ is set by $\\chi^2_{A_D}-\\chi^2_{\\rm SM}\\ge3.8$. With the PDG world-average $W$ mass, the $\\epsilon$ exclusion reproduces earlier bounds; with the CDF $W$ mass, the exclusion strengthens for $m_{A_D}m_Z$. When the dark photon also couples to a Dirac dark fermion, the $Z$ inherits the invisible decay $Z\\to\\chi\\bar\\chi$, and the measured $Z$ width yields the first electroweak-precision upper limits on $g_\\chi$, set by the two-parameter condition $\\chi^2_{A_D}(\\epsilon,g_\\chi)-\\chi^2_{\\rm SM}\\ge5.99$.","pith_inferences":["Because $Z\\to\\chi\\bar\\chi$ is invisible, these $g_\\chi$ limits are equivalent to an upper bound on the invisible $Z$ width; a per-mille measurement of that width at a future collider would test the same parameter plane independently.","The CDF-motivated point with $\\epsilon\\simeq0.1$ predicts a $W$ mass about 27 MeV below the CDF value, so a future $\\sim$1 MeV measurement of $m_W$ landing near the world average would exclude the region that currently best fits CDF.","The paper adds the dark photon only at tree level; a full one-loop electroweak calculation including the dark photon would show whether the exclusion curves shift at the $10^{-3}$ level, which is a natural next test.","The new $g_\\chi$ bounds, when combined with relic-density and direct-detection constraints on $y=\\epsilon^2\\alpha_D(m_\\chi/m_{A'})^4$, should map the allowed region for light thermal dark matter more completely."],"forward_implications":["Below the $Z$ pole, any nonzero $\\epsilon$ worsens the electroweak fit relative to the Standard Model, so the 95% upper limit on $\\epsilon$ becomes stronger when the CDF $W$ mass is used.","The best dark photon point, $m_{A_D}=200$ GeV and $\\epsilon=0.1001$, raises the predicted $W$ mass to 80.4060 GeV and cuts the CDF discrepancy to $2.9\\sigma$, though the overall fit value remains large.","For a dark Dirac fermion with $m_\\chi