{"id":"fec2243f-3a1d-4b3e-a318-d37c4352a005","arxiv_id":"2507.12598","paper_version":2,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"low","formal_verification":"none","parameter_count":5,"one_line_summary":"ATLAS measures the Higgs to Z and photon signal yield as 1.3 +0.6 -0.5 times the Standard Model prediction in the combined 305 inverse femtobarn dataset, with 2.5 sigma observed significance.","lead":"ATLAS searched for a rare decay of the Higgs boson into a Z boson and a photon using 165 inverse femtobarns of new LHC collision data from 2022 to 2024, combined with an earlier 140 inverse femtobarn dataset. The combined measurement matches the Standard Model prediction, giving 2.5 sigma observed significance and the best expected sensitivity to date for this decay.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Background spurious-signal systematic may under-cover template bias: the SS criterion tolerates |SS| near 0.2ΔS+σ and only three template variants are tested, so a biased Zγ/Z+jets shape could shift μ by more than the quoted 0.11 impact.","rationale":"The paper's central claim—an observed combined signal strength μ=1.3+0.6−0.5 with 2.5σ significance, consistent with the Standard Model and yielding the most stringent expected sensitivity—is well supported by the internal consistency of the statistical framework, the detailed systematic accounting, and the explicit spurious-signal controls. The reader's weakest assumption correctly identifies background modelling as the softest point. I agree and sharpen it: the SS systematic is estimated from a single template with only three composition variants, and the acceptance criterion is lenient for large SS uncertainties, so the coverage of a genuinely biased template is not demonstrated. The impact of this systematic (0.11 on μ) is small relative to the statistical uncertainty (0.5), and even a factor-of-two underestimate would leave the conclusion 'consistent with the SM' intact. The expected-sensitivity claim is stated as the analysis's expected performance under the adopted background model, and the observed excess is not used to claim discovery. The proposed check—an alternative-template SS study—would settle whether the systematic is under-covered. Because the concern is a caveat rather than a demonstrated flaw, and the central physics conclusion is robust to plausible variations in the SS systematic, the reader's ACCEPT verdict stands unchanged.","tokens_in":56444,"tokens_out":16703,"duration_ms":188004,"concrete_test":"For each of the 13 categories, rebuild the background template using (a) a statistically independent Zγ sample with full Geant4 simulation instead of AtlFast3, and (b) a Z+jets control region defined by inverted photon isolation rather than inverted identification. Re-run the SS study with the same function families, fit ranges, and mH scan (120–130 GeV). If the maximum |SS| for the same selected function exceeds the assigned systematic in any category, or if any function-range choice changes, the systematic is under-covered and the quoted μ and its 0.11 SS impact should be revised.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Section 6 selects the analytic background function via a spurious-signal (SS) study on a template combining fast-simulation Zγ (AtlFast3) with a data-driven Z+jets control region. The acceptance criterion requires one of |SS−σ|, SS, or |SS+σ| to be below 0.2ΔS; this is satisfied whenever the 1σ lower edge of SS falls below threshold, so a function with a 1.5σ upward fluctuation passes. The assigned systematic is the maximum |SS| from only three template variants (nominal and the Zγ-fraction ± uncertainties), which does not probe errors in the Zγ fast-simulation mass resolution or in the Z+jets shape extrapolation to the signal region. Table 3 shows observed yields exceeding S+B by 20–40% in several categories (inclusively N68=44380 vs S+B=36080); the fit absorbs the normalization, but a correlated shape error in the 120–130 GeV window would bias μ directly. The quoted SS impact is 0.11 on μ; if the true template bias is larger, the combined observed significance of 2.5σ could shift materially.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper reports a search for the Higgs boson decay H→Zγ in the ℓℓγ final state using 165 fb−1 of √s=13.6 TeV pp collision data recorded by ATLAS in 2022–2024, and combines the result with the earlier 140 fb−1 Run-2 search. The Run-3-only fit gives μ=0.9+0.7−0.6 (expected 1.0±0.7), corresponding to an observed (expected) significance of 1.4σ (1.5σ). The combined Run-2+Run-3 fit gives μ=1.3+0.6−0.5 (expected 1.0+0.6−0.5), with observed (expected) significance 2.5σ (1.9σ); the authors state this is the most stringent expected sensitivity to date for this decay. The analysis uses 13 exclusive categories, XGBoost-based BDT classifiers, analytic background models selected via spurious-signal studies, and a simultaneous unbinned maximum-likelihood fit.","tokens_in":56639,"tokens_out":8045,"duration_ms":100160,"significance":"If correct, the combined measurement is a meaningful step for this rare Higgs decay: it improves the expected sensitivity over the ATLAS+CMS Run-2 combination and yields a measurement consistent with the Standard Model within current precision. The paper is careful in several respects: it reports a detailed systematic breakdown (Table 4), performs spurious-signal studies over multiple function families and fit ranges, applies Wald tests to avoid overfitting, checks category consistency (p=0.37), and documents the Run-2/Run-3 correlation scheme. The statistical uncertainty dominates, and the reported spurious-signal impact is small (0.11 on μ in Run-3). The main caveat is the limited coverage of the background-template systematic, discussed below.","major_comments":[{"comment":"The background-model systematic assigned in the spurious-signal study is the maximum |SS| from only three template variants (nominal and the Zγ-fraction ±1σ variations), and the acceptance criterion requires only one of |SS−σ|, SS, or |SS+σ| to be below 0.2ΔS, so a template with a 1.5σ upward SS fluctuation can pass. This procedure does not directly cover errors in the AtlFast3 Zγ mass resolution or in the extrapolation of the data-driven Z+jets shape into the 120–130 GeV signal window. Since the combined observed (expected) significance of 2.5σ (1.9σ) and the claim of the most stringent expected sensitivity to date rest on the background model, I request a stress test in which the background template is rebuilt with alternative Zγ simulation (e.g., full Geant4 instead of AtlFast3), alternative Z+jets control-region choices or smoothing orders, and with the signal-region m(Zγ) resolution varied; the resulting maximum |SS| and the shift in μ should be reported. The closure test in Section 6 (fitting a template built by adding the SM signal to the background template) uses the same template and therefore does not address this concern.","section":"Section 6 and Section 8.2"}],"minor_comments":[{"comment":"The notation is inconsistent: |SS| is defined as the absolute value of the maximum fitted signal yield, but the acceptance criterion then refers to 'SS' in 'one of |SS−σ|, SS, or |SS+σ|'; please clarify whether SS is signed and define the quantity entering each term.","section":"Section 6"},{"comment":"The axis labels in Figure 2 appear malformed ('ATLAS-1 = 13.6 TeV, 165 fbs'); this is presumably a rendering artifact, but the labels should be corrected to show the proper ATLAS label and fb−1 units.","section":"Figure 2"},{"comment":"After the Gaussian Process smoothing the text states that 'The |SS| is then re-evaluated', but does not specify whether the re-evaluated value replaces the assigned systematic or is only used to confirm that the choice of function and range is unchanged; please state this explicitly.","section":"Section 6"}],"recommendation":"major_revision","confidential_remarks":"This is a standard ATLAS experimental search with no obvious scope or attribution issues. The requested additional validation of the spurious-signal systematic is feasible within a revision; if the authors can show that the background-template systematic is robust under alternative template constructions, the paper should be accepted."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThe ATLAS H→Zγ search is a solid, workmanlike update: Run-3 data at 13.6 TeV, 165/fb, combined with the earlier Run-2 result. The combined μ = 1.3 ± 0.6 is consistent with the SM; observed significance is 2.5σ versus 1.9 expected. Not a discovery, and the paper doesn't claim one.\n\nWhat's new: the Run-3 analysis itself — new luminosity, relaxed photon and muon pT thresholds, a dedicated multi-lepton category, and XGBoost-based categorization — plus the combination. The analysis is thorough on the things that matter: spurious-signal studies over multiple function families and fit ranges, Wald tests for overfitting, a clear systematics breakdown, and a documented correlation scheme for the combination. Statistical uncertainty dominates; systematic is 0.17 on μ, with spurious signal at 0.11.\n\nSoft spots: the background model is the main thing to probe. The spurious-signal systematic is evaluated on only three template variants (nominal and Zγ-fraction ±1σ), so it doesn't vary the fast-simulation Zγ mass resolution or the Z+jets shape extrapolation. That could under-cover a template bias in the 120–130 GeV window. Also, the acceptance criterion — one of |SS±σ| below 0.2ΔS — does in principle allow a function with a 1.5σ upward fluctuation to pass, as the stress-test notes. But this is a minor caveat, not a load-bearing flaw: the SS impact is 0.11 on μ, and even doubling it would leave systematics well below the 0.5–0.6 statistical uncertainty. The central result would only shift materially under a large, category-correlated template bias, and there is no evidence of that.\n\nThe math and data handling look honest and internally consistent. The citation pattern is unremarkable for a collaboration paper; the Run-2 analysis is properly cited and extended. I would treat this as the current best single-experiment sensitivity for H→Zγ and a useful input to global Higgs coupling fits.\n\nWho it's for: Higgs physics specialists, people doing global coupling fits, and anyone tracking rare decay searches. It deserves a serious referee — likely one round of minor comments, then acceptance.\n\nRecommendation: send it to peer review, don't desk-reject.","headline":"ATLAS's Run-3 H→Zγ search is a careful, internally consistent update that delivers the most sensitive expected limit to date; the 2.5σ observed excess is SM-consistent and the only real soft spot is the narrow coverage of the background-template systematics.","tokens_in":57236,"tokens_out":3367,"would_cite":true,"duration_ms":41887,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper reports a search for the rare Higgs boson decay $H\\to Z\\gamma$ in the dilepton-plus-photon final state, and claims a combined measurement consistent with the Standard Model with the most precise expected sensitivity to date.","keywords":["Higgs boson","Z gamma decay","rare decay","signal strength","maximum-likelihood fit","double-sided Crystal Ball","13.6 TeV proton-proton collisions","Standard Model consistency"],"falsifier":"Repeat the spurious-signal study with several times larger simulated background samples and replace the smoothed background templates with the unsmoothed templates; if the best-fit signal strength in the 120-130 GeV window shifts by more than the assigned spurious-signal uncertainty, the background-modelling assumption behind the measured excess would be falsified.","tokens_in":56171,"feed_emoji":"⚛️","tokens_out":9394,"duration_ms":92925,"temperature":0.7,"pith_summary":"The paper searches for the rare, loop-induced Higgs boson decay $H\\to Z\\gamma$ in events where the $Z$ decays to an electron or muon pair and a photon. Using 165 fb$^{-1}$ of 13.6 TeV proton-proton collisions, and combining this with the earlier 140 fb$^{-1}$ 13 TeV search, it measures a signal strength of $\\mu=1.3^{+0.6}_{-0.5}$ with an observed significance of $2.5\\sigma$, close to the $1.9\\sigma$ expected. This is consistent with the Standard Model prediction and provides the most stringent expected sensitivity to date for this decay. The result matters because new charged particles in the loop could shift the $H\\to Z\\gamma$ rate relative to $H\\to\\gamma\\gamma$, so a precise measurement is a sensitive test of physics beyond the Standard Model.","feed_headline":"Rare Higgs decay to Z and photon seen at 2.5 sigma","feed_subtitle":"Combined 13 and 13.6 TeV proton-proton data give a signal strength consistent with the Standard Model.","key_machinery":"The central object is the reconstructed invariant mass $m_{Z\\gamma}$ of the lepton-pair-plus-photon system. The signal is modelled in each of 13 exclusive event categories by a double-sided Crystal Ball function, while the background is described by analytic functions selected through a spurious-signal study applied to templates built from simulated $Z\\gamma$ events and data-driven $Z$+jets events. Twelve of the categories are defined with XGBoost-based classifiers, and one dedicated multi-lepton category targets Higgs production in association with a vector boson or a top-quark pair. A simultaneous unbinned extended maximum-likelihood fit to the $m_{Z\\gamma}$ distributions extracts the signal strength.","core_discovery":"This paper claims that the decay $H\\to Z\\gamma$, searched in the $\\ell\\ell\\gamma$ final state with $\\ell=e,\\mu$, is present in the combined 13 TeV and 13.6 TeV datasets at a rate consistent with the Standard Model. The combined best-fit signal strength is $\\mu=1.3^{+0.6}_{-0.5}$, with an observed (expected) significance of $2.5\\sigma$ ($1.9\\sigma$); the standalone Run-3 measurement is $\\mu=0.9^{+0.7}_{-0.6}$, corresponding to observed (expected) significance of $1.4\\sigma$ ($1.5\\sigma$). The combination is stated to provide the most stringent expected sensitivity to date for this decay. Under the Standard Model production cross-sections, it corresponds to an observed branching fraction of $(2.0^{+0.9}_{-0.8})\\times10^{-3}$, compared with the predicted $(1.54^{+0.10}_{-0.11})\\times10^{-3}$.","pith_inferences":["Editorial extension: if the central value remains above the Standard Model as more 13.6 TeV data accumulate, the ratio of $H\\to Z\\gamma$ to $H\\to\\gamma\\gamma$ branching fractions would become the most discriminating single observable for charged new particles in the loop.","Editorial extension: combining this measurement with the other experiment's Run-3 search could push the combined significance past the evidence threshold, provided the background-modelling systematics scale down with the larger sample.","Editorial extension: the new multi-lepton category isolates Higgs production with a vector boson or a top-quark pair, so with additional data it could be subdivided to measure production-mode-specific $H\\to Z\\gamma$ couplings, a sharper new-physics test than the inclusive rate."],"forward_implications":["The measured signal strength $\\mu=1.3^{+0.6}_{-0.5}$ is compatible with the Standard Model value $\\mu=1.0$, so the data do not establish a new-physics contribution to this decay.","The observed significance of $2.5\\sigma$ is below the $5\\sigma$ evidence threshold, so the decay is not yet definitively observed in this dataset alone.","Assuming Standard Model production cross-sections, the measurement translates to a branching fraction $(2.0^{+0.9}_{-0.8})\\times10^{-3}$, consistent with the predicted $(1.54^{+0.10}_{-0.11})\\times10^{-3}$.","The combined expected sensitivity surpasses the previous two-experiment combination of Run-2 data, making this the most precise expected constraint on the $H\\to Z\\gamma$ rate to date."],"supporting_citations":[{"why":"Supplies the Standard Model Higgs branching ratios and production cross-sections used to define the signal strength and the expected sensitivity.","marker":"[12]"},{"why":"Provides the Run-2 measurement and the methodological template (kinematic fit, spurious-signal study) that the Run-3 search extends and with which it is combined.","marker":"[23]"},{"why":"Provides the other experiment's Run-2 search, quoted as context for the previous signal-strength measurements in this channel.","marker":"[24]"},{"why":"Provides the previous two-experiment Run-2 combination whose expected sensitivity the new combination is claimed to surpass.","marker":"[25]"},{"why":"Supplies the XGBoost classifier used to build the twelve multivariate event categories that drive most of the sensitivity gain.","marker":"[26]"},{"why":"Supplies the asymptotic likelihood-ratio formulas used to compute the quoted significances and signal-strength uncertainties.","marker":"[116]"}],"fun_headline_variants":["Rare Higgs decay to Z and photon at 2.5 sigma in ATLAS","ATLAS combined 13+13.6 TeV data: H to Z gamma at 2.5 sigma","H to Z gamma: ATLAS sees 2.5 sigma signal consistent with SM","Rare decay H to Z gamma: most stringent search yet, 2.5 sigma"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The search assumes that the fitted background curves in the 120-130 GeV window, together with the size of the spurious-signal uncertainty estimated from limited simulation samples, fully cover any residual mismodelling of the $m_{Z\\gamma}$ spectrum.","fun_headline_variants_meta":{"raw":{"variants":["Rare Higgs decay to Z and photon at 2.5 sigma in ATLAS","ATLAS combined 13+13.6 TeV data: H to Z gamma at 2.5 sigma","H to Z gamma: ATLAS sees 2.5 sigma signal consistent with SM","Rare decay H to Z gamma: most stringent search yet, 2.5 sigma"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000775,"raw_usage":{"total_tokens":3484,"prompt_tokens":1057,"completion_tokens":2427,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":673,"completion_tokens_details":{"reasoning_tokens":2341}},"tokens_in":673,"tokens_out":2427,"duration_ms":22127,"temperature":1.0,"reasoning_tokens":2341,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T16:43:28.801844+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Repeat the spurious-signal study with several times larger simulated background samples and replace the smoothed background templates with the unsmoothed templates; if the best-fit signal strength in the 120-130 GeV window shifts by more than the assigned spurious-signal uncertainty, the background-modelling assumption behind the measured excess would be falsified.","supporting_citations":[],"review_version":1}