{"id":"d82099ef-531c-4bf8-900f-f93358a262b6","arxiv_id":"2603.22233","paper_version":2,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"low","formal_verification":"none","parameter_count":5,"one_line_summary":"A first dedicated search for highly Lorentz-boosted H->WW decays at the LHC finds mu = -0.19 +0.48/-0.46, consistent with no signal above background.","lead":"This CMS analysis searched for highly boosted Higgs bosons decaying to pairs of W bosons in 138 fb-1 of 13 TeV proton collisions, using a transformer-based jet tagger to catch the merged decay products. It found no signal: the measured rate relative to the standard model is -0.19 +0.48/-0.46, the first dedicated constraint in this channel.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"No significant objection identified; the 0l QCD transfer-function assumption is the main fragility but is adequately covered by the profiled uncertainties.","rationale":"The reader's weakest-assumption identification points to the same step I would single out if forced: the 0l QCD transfer-function extrapolation is the least protected element of the analysis. But I do not think it rises to a load-bearing concern that should change the verdict. The measured mu is consistent with both zero and the SM at roughly the 1-2 sigma level; the dominant systematic is explicitly attributed to this background model; and the data are only moderately sensitive (expected significance 1.86 sigma). Even a modest distortion of the QCD background shape would shift mu within its quoted uncertainty rather than overturn the central null result. The paper's internal consistency, the documented profiling of the TF coefficients, and the absence of anomalous post-fit pulls support acceptance. The proposed MC closure test would strengthen confidence further, but the absence of such a test in the preprint is not a fatal omission given the large statistical and systematic uncertainties already reported.","tokens_in":50955,"tokens_out":7204,"duration_ms":85506,"concrete_test":"Run a high-statistics MC closure test using the official QCD multijet simulation: apply the full 0l event selection, compute the true m* ratio between CR1/SR1 and CR2/SR2, and fit the degree-3/6 Bernstein polynomials over 50-250 GeV. Perform a chi-square/KS test in the 110-140 GeV window to check for narrow unmolded structure. Separately, inject a SM signal (mu=1) into pseudo-data and verify that a simultaneous fit with free TF coefficients recovers mu unbiased within its uncertainty. If either test reveals significant bias, the 0l background model needs strengthening; if not, the assumption is validated.","verdict_should_be":"UNCHANGED","load_bearing_attack":"No significant objection identified. The central claim is a null result (mu = -0.19 +0.48/-0.46), not a precise positive measurement, so the impact of any single background-modeling assumption is limited. The most fragile-looking step is the 0l QCD multijet transfer function of Eq. 1, where CR data are multiplied by a Bernstein polynomial to predict the SR shape. This is a legitimate concern because the polynomial coefficients are left unconstrained in the fit (Sec. 6.2.1), so a signal-like bump could in principle be absorbed into the background. However, several protections are present: (i) PART was explicitly mass-decorrelated during training; (ii) the polynomial order was chosen by a Fisher F-test, not ad hoc; (iii) the TF parameters are profiled nuisance parameters, and Sec. 8 states that QCD TF modeling is the dominant 0l systematic, meaning the quoted total uncertainty already includes this flexibility; and (iv) the post-fit pulls in Fig. 5 show no localized data excess near 125 GeV beyond statistical noise. A degree-6 Bernstein polynomial over the 50-250 GeV range can represent smooth ratios, and there is no identified physics source for a sharp CR/SR feature at the Higgs mass. The hypothesis that the true ratio has structure narrower than the polynomial resolution is testable but currently unsupported.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"This manuscript reports a search for highly Lorentz-boosted Higgs boson production in the H→WW decay channel using 138 fb^-1 of 13 TeV proton-proton collisions recorded by CMS. Events are classified into a 0-lepton channel (fully hadronic and semileptonic WW decays with nonisolated leptons) and a 1-lepton channel (semileptonic WW decays with an isolated lepton). The Higgs boson is reconstructed as a single large-radius jet, identified with the transformer-based PART tagger, with a fine-tuned variant for the 1-lepton channel and Lund-jet-plane reweighting for signal-efficiency calibration. Backgrounds are estimated via data-driven transfer functions for QCD multijet events in the 0-lepton channel and via simulation with control-region normalizations in the 1-lepton channel. The combined signal strength relative to the standard model expectation is μ = -0.19 +0.48/-0.46, giving no evidence of a signal; the expected significance is 1.86σ and the observed significance is 0.0σ. Results are also presented as simplified template cross sections.","tokens_in":51345,"tokens_out":15700,"duration_ms":148647,"significance":"If correct, this is the first dedicated measurement of highly boosted H→WW production, extending differential Higgs measurements at high transverse momentum to a previously unexplored final state. The analysis is carefully executed: the systematic uncertainty tables are detailed (Table 3), the LJP calibration uncertainty is decomposed into statistical, modeling, and extrapolation components, and the background estimates are cross-checked in data. The paper is honest in reporting a null result and the low expected sensitivity of the 0-lepton channel. The HEPData record supports future combinations. The central null result is robust to plausible uncertainties in the 0-lepton QCD background model because that channel contributes only 0.36σ to the expected combined significance; the 1-lepton channels drive the sensitivity.","major_comments":[],"minor_comments":[{"comment":"The notation 'TFia' is undefined and potentially confusing: clarify that the a-type transfer function is shared between TF1a and TF2a, while TF1b and TF2b are independent. In addition, because the transfer-function coefficients are unconstrained and fitted simultaneously with the signal, please state whether any signal-injection closure test was performed to check that the Bernstein polynomial cannot absorb a narrow H→WW peak near 125 GeV. The current text relies on the post-fit pulls and on the Fisher F-test, but a direct statement of the closure would strengthen the dominant 0-lepton systematic.","section":"Section 6.2.1, Eqs. (1)–(2)"},{"comment":"The extrapolation of LJP corrections measured on W→qq decays in data to 3- and 4-prong H→WW signal jets is a core assumption of the signal-efficiency calibration. The tt validation is described, but the paper should explicitly discuss the expected size of non-universality effects (e.g., color connection, subjet matching, or quark flavor) and justify that the quoted 9–27% uncertainty envelope covers these effects.","section":"Section 5, LJP calibration"},{"comment":"The statement that the observed result lies 2.1 standard deviations below the SM expectation should define the test statistic used (e.g., sqrt(t_1) with the uncapped negative-μhat region) and should be reconciled with the quoted asymmetric uncertainty of μ = -0.19 +0.48/-0.46. As written, this claim is easy to misinterpret and should be clarified or softened.","section":"Section 9"},{"comment":"The sentence 'Parton showering, fragmentation, and hadronization are modeled with PYTHIA8.230 ...' appears twice in Section 3. Also, the statement that electroweak V+2j production corresponds to O(alpha_S^4) seems inconsistent with an electroweak process; please check the coupling order.","section":"Section 3"},{"comment":"The multi-level column headers for signal strength and significance are difficult to parse. Consider splitting the observed/expected signal strength and the observed/expected significance into separate tables or using clearer column spans.","section":"Table 4"},{"comment":"When more than one AK8 jet satisfies P(H1l)>0.75, the H-candidate jet is not uniquely defined. Please state the selection rule explicitly (presumably the jet with the highest PART score or the one closest to the lepton).","section":"Section 7.1"},{"comment":"The claim that this is the 'first dedicated study' of highly Lorentz-boosted H→WW decays should be substantiated with a brief statement confirming the absence of prior dedicated measurements in this exact final state, since the cited differential H→WW measurement [15] does not use the single-jet topology.","section":"Introduction"}],"recommendation":"minor_revision","confidential_remarks":"This is a solid CMS search with an honest null result. The reader's concern about the 0-lepton QCD transfer function is real but not fatal, because the 0-lepton channel contributes only 0.36σ expected significance to the combination; the 1-lepton channels drive the result. I recommend minor revision, primarily asking for a brief closure-test statement in Section 6.2.1 and clarification of a few notation points. No load-bearing technical errors were found."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"First thing to know: this is a genuine first measurement. CMS has searched for boosted H→WW in merged large-radius jets with 0 or 1 isolated lepton, using 138 fb−1, and finds no signal. The combined signal strength is μ = −0.19 +0.48/−0.46, expected significance 1.86σ, observed 0.0σ. That is a null result with large uncertainties. It does not settle anything dramatic, but it fills a real gap in the high-pT Higgs program.\n\nWhat the paper does well: the analysis is methodologically serious. The PART tagger is mass-decorrelated, the 1ℓ channel uses a fine-tuned version that improves signal efficiency by about 60%, and the Lund jet plane calibration is a reasonable way to get data-driven efficiency corrections for a final state with no clean SM counterpart. The calibration is validated in top-enriched regions, and the scale factors are quoted with 9–27% uncertainties, honestly. The background estimates are standard and the systematic tables are detailed. I do not see a circular step: signal templates come from external SM simulation, tagger calibration uses W→qq data, not the signal region. References are appropriate.\n\nThe soft spots, in proportion. The 0ℓ QCD background is the most fragile piece: the CR-to-SR transfer function is a floating Bernstein polynomial whose coefficients are fit with the signal regions included. In principle a Higgs-mass bump could be absorbed into the background shape. The paper mitigates this with mass decorrelation, an F-test for polynomial order, and profiled nuisance parameters that are themselves the dominant systematic. I think the stress test is right: this is the right thing to worry about, but not a demonstrated flaw. The other thing I would push on is the VH top-quark normalization: the fit wants about 2.0 times simulation for the tt background in the VH category. That is a large pull and suggests the VH background modeling is less comfortable than the rest. The category is low-statistics and the central result is null, so it does not change the conclusion, but a referee should ask for more detail on that CR. The negative combined μ, and the 2.1σ downward deviation when allowing negative signals, looks like statistical noise; I would not read anything into it.\n\nThis paper is for LHC experimenters working on boosted jets and Higgs differential measurements, and for theorists who use high-pT Higgs constraints. It deserves a serious referee. I would accept it for review and expect a usable publication after normal vetting.","headline":"A solid, first-of-its-kind null search for boosted H→WW; the analysis is careful, the main risk is the 0ℓ background transfer function, and the VH top normalization deserves a hard look.","tokens_in":51764,"tokens_out":3346,"would_cite":true,"duration_ms":36199,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"A dedicated search for highly Lorentz-boosted Higgs bosons decaying to W boson pairs finds no excess over background, with the measured signal strength at −0.19 +0.48/−0.46 times the standard-model expectation.","keywords":["Higgs boson","WW decay","boosted regime","jet substructure","transformer tagger","Lund jet plane","signal strength","simplified template cross sections"],"falsifier":"In the public data release, compare the post-fit background prediction with the observed data in the 110–140 GeV window of each 0-lepton signal region, and repeat the combined fit with the Bernstein polynomial order increased by one: if a systematic residual appears in several bins around 125 GeV, or if the signal strength shifts by more than roughly 0.3 relative to the published value, the background-only conclusion would not be robust.","tokens_in":50825,"feed_emoji":"⚛️","tokens_out":7053,"duration_ms":79349,"temperature":0.7,"pith_summary":"The paper aims to measure Higgs boson production at large transverse momentum in the WW decay channel, a regime that had not been isolated in this final state before. Because the two W bosons merge into a single wide jet, the search identifies candidate jets with a transformer-based jet tagger, calibrates the tagger with Lund-jet-plane reweighting, and looks for a signal bump in the reconstructed jet-mass spectrum. Using 138 inverse femtobarns of 13 TeV proton-proton collisions, the authors find a best-fit signal strength of −0.19 +0.48/−0.46 relative to the standard model, with an observed significance of 0.0 sigma against an expected 1.86 sigma. In plain terms, the data are fully compatible with background alone, and the paper presents the first dedicated constraint on boosted H to WW production, along with unfolded cross-section measurements for direct theory comparison.","feed_headline":"First boosted Higgs-to-WW search finds no signal","feed_subtitle":"Measured rate is −0.19 +0.48/−0.46 times the standard-model expectation; observed significance 0σ vs 1.86σ expected.","key_machinery":"The load-bearing object is the single large-radius jet that contains the merged decay products of the two W bosons. To recognize the H to WW substructure, the analysis uses PART, a transformer-based self-attention tagger that classifies jets into categories by quark flavor and prong multiplicity; in the 1-lepton channel the tagger is fine-tuned on the lepton-in-jet topology. Signal efficiency is calibrated with Lund-jet-plane reweighting, which rescales simulated jet-splitting densities using W-to-qq data. For semileptonic decays, the candidate mass is reconstructed by adding the missing transverse momentum to the jet under the assumption that the neutrino is collinear with the jet. The domi","core_discovery":"The central claim is that, in the high-transverse-momentum regime where both W bosons fit inside a single large-radius jet, the H to WW production rate is consistent with the standard-model background prediction. The combined fit to the 0-lepton and 1-lepton channels gives a signal strength of −0.19 +0.48/−0.46, corresponding to an observed significance of 0.0 sigma and an expected significance of 1.86 sigma; when negative signal strengths are allowed, the observed result lies 2.1 standard deviations below the standard-model expectation. The paper thus establishes that no excess of boosted H to WW events appears in the data, and interprets the measurement as the first dedicated study of this","pith_inferences":["The authors do not stress it, but the negative central signal strength, while consistent with background, is also consistent with a mild suppression of the Higgs rate at high transverse momentum; combining this channel with existing boosted H to bb, tau tau, and gamma gamma measurements in a common fit would sharpen that test.","The neutrino-collinear mass correction assumes the neutrino shares the jet's pseudorapidity; a closure study in semileptonic top-quark events, where the neutrino momentum can be constrained, would directly quantify any bias in the reconstructed Higgs mass peak near 125 GeV. The paper does not report such a dedicated test.","The Bernstein transfer functions are fit with unconstrained coefficients in the signal regions, so their fitted shapes can partially absorb smooth data fluctuations near the Higgs mass; an independent background estimate using a high-score sideband in data, rather than a polynomial extrapolation, would be a useful cross-check of the 0-lepton result.","A natural extension is to apply the same fine-tuned tagger and calibration procedure to the larger dataset now being collected; with more luminosity the expected 1.86 sigma sensitivity would grow, and the currently inclusive 0-lepton channel could be split by production process."],"forward_implications":["The measurement places the first dedicated constraint on boosted H to WW production; even without an excess, it demonstrates the channel's sensitivity and provides a benchmark for future searches in this topology.","The unfolded simplified-template cross sections in the gluon-fusion transverse-momentum bins and the vector-boson-fusion mass bin give generator-level comparisons for higher-order QCD predictions, which are especially uncertain in the boosted regime.","The transfer-learning recipe—pre-training a broad jet classifier, then fine-tuning on the narrow lepton-in-jet topology—yields roughly 60% higher signal efficiency at 1% background efficiency, a gain that directly improves the search's expected significance.","The 0-lepton channel's data-driven background method shows that fully hadronic and non-isolated-lepton WW decays can be included in a boosted-Higgs search, extending coverage beyond the single-isolated-lepton topology.","The use of the common simplified-template phase-space binning allows this measurement to be combined with other boosted Higgs channels and other decay modes in a joint high-transverse-momentum cross-section extraction."],"fun_headline_variants":["Boosted Higgs to WW decays: no signal in first search","High-momentum Higgs search: only background in WW","CMS finds no boosted Higgs in WW decay channel","First boosted H→WW search: consistent with background","No excess of boosted Higgs decays to WW"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The 0-lepton result assumes that the QCD multijet background shape in the signal regions is a smooth, low-order polynomial rescaling of the control-region shape, with no sharp or non-monotonic features near 125 GeV that the fitted transfer functions cannot capture.","fun_headline_variants_meta":{"raw":{"variants":["Boosted Higgs to WW decays: no signal in first search","High-momentum Higgs search: only background in WW","CMS finds no boosted Higgs in WW decay channel","First boosted H→WW search: consistent with background","No excess of boosted Higgs decays to WW"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000236,"raw_usage":{"total_tokens":1380,"prompt_tokens":820,"completion_tokens":560,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":564,"completion_tokens_details":{"reasoning_tokens":484}},"tokens_in":564,"tokens_out":560,"duration_ms":5878,"temperature":1.0,"reasoning_tokens":484,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-04T05:41:48.030154+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"In the public data release, compare the post-fit background prediction with the observed data in the 110–140 GeV window of each 0-lepton signal region, and repeat the combined fit with the Bernstein polynomial order increased by one: if a systematic residual appears in several bins around 125 GeV, or if the signal strength shifts by more than roughly 0.3 relative to the published value, the background-only conclusion would not be robust.","supporting_citations":[],"review_version":1}