{"id":"74caffc5-852e-4028-956a-16b40b268e87","arxiv_id":"2510.12396","paper_version":2,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"First dedicated CMS search for bb + dilepton + missing transverse momentum in the 2HDM+a model finds no signal and sets 95% CL cross-section limits of 10^-2 to 10^-3 pb for mH = 400-2000 GeV.","lead":"The CMS experiment searched 138 fb^-1 of 13 TeV proton-proton collisions for dark matter produced alongside bottom quarks and a pair of electrons or muons, in a model where a new heavy Higgs boson decays to a Z boson and a dark-matter mediator. No excess was found, and new 95% confidence upper limits on the production rate range from about 10^-2 pb to 10^-3 pb.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"DY pTmiss recoil-correction extrapolation is the least secure step: derived at pTmiss<65 GeV, validated only in a no-b-jet region, yet it shapes the SR (Nb>=1, large pTmiss) templates used for the limits.","rationale":"The reader identified the same weakest point: the DY pTmiss correction is extrapolated from a low-pTmiss sideband and verified in a no-b-jet region, not in the SR phase space. I agree this is the most load-bearing assumption because the MLP discriminant is built from pTmiss and related variables, and DY is the dominant background for low-mH signals. However, I do not consider this a demonstrated flaw: the recoil-correction method is standard, the paper explicitly acknowledges the limitation, and the existing control-region and post-fit checks provide substantial support in the phase space that is covered. The proposed test would settle whether the extrapolation actually fails in the SR corner; until then, the concern does not justify changing the reader's ACCEPT verdict, but it does warrant a clear caveat. I also note that the tt analytic-solution veto is another unvalidated modeling step, but it is secondary to the DY shape correction because the veto is applied to a smaller background component and the MLP deliberately excludes jet variables to mitigate jet-mismodeling effects. Overall, the paper is internally consistent, uses a standard CR strategy, and provides a falsifiable null result with limits; the DY correction extrapolation is the appropriate focus for a stress-test.","tokens_in":46799,"tokens_out":8495,"duration_ms":79846,"concrete_test":"Derive the DY recoil correction from a dedicated DY-enriched sideband that matches the SR in the b-jet requirement: exactly two OSSF leptons with |mll-mZ|<25 GeV, Nb>=1, no tt analytic solution, and 65<pTmiss<300 GeV. Apply this alternatively derived correction to the DY simulation, re-run the full background-only fit and the CLs limit setting, and compare the resulting 95% CL upper limits in Fig. 7 for the mH=400 GeV row to the nominal ones. If the observed limits shift by more than the systematic uncertainty assigned to the nominal pTmiss scale/resolution correction (or by more than ~20% in the high-MLP4 bins), the extrapolation does not close in the SR phase space and the limits should be re-derived with a larger shape uncertainty.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central null result and the 95% CL upper limits depend on the accuracy of the background prediction in the SR. The dominant background in the low-to-mid MLP4 bins is DY, and its pTmiss shape is corrected with a data-driven recoil correction derived in a sideband with pTmiss<65 GeV and then extrapolated to the SR (Section 6). The only stated validation of this extrapolation is a region with no b jets and pTmiss<300 GeV. This does not cover the SR phase space with Nb>=1 and pTmiss extending well above 300 GeV, nor does it validate the correction in events with a b-tagged jet, where the jet flavor and the loose b-tag requirement alter the event composition relative to the no-b-jet validation region. Moreover, the DY control region used in the fit requires Nb=0 and pTmiss<140 GeV, so the fit constrains only the DY normalization, not the pTmiss shape in the SR phase space. Because the MLP discriminant uses pTmiss and mT(ell,ell,ptmiss) as leading inputs, a miscalibrated DY tail would directly reshape the SR templates and shift the derived limits, especially for low mH signals that populate the lower MLP4 bins. The paper itself flags this limitation by stating the extrapolation is verified only in the no-b-jet, pTmiss<300 GeV region. This is a genuine validation gap rather than an observed discrepancy; the post-fit CR plots show agreement within the covered phase space, but they do not close the gap.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper presents a search for dark matter produced in association with b quarks and a lepton pair in 138 fb^-1 of 13 TeV CMS data, using the 2HDM+a model with the process pp -> bbH, H -> Za, Z -> ll, a -> chichi. A multivariate MLP discriminant is trained on simulated signal and background events, and a profile-likelihood fit combines a 17-bin signal region with four control regions (DY, tt, WZ, ZZ) to constrain background normalizations. A data-driven correction to the DY pTmiss scale and resolution is derived in a low-pTmiss sideband and applied to the signal region. The observations are consistent with the standard-model expectation, and 95% CL upper limits are set on sigma(pp->bbH) B(H->Za) B(Z->ll) B(a->chichi) in the range 10^-2 to 10^-3 pb for mH = 400-2000 GeV. The results are interpreted as constraints on the 2HDM+a parameter space and compared with the region favored by the relic-density calculation.","tokens_in":47200,"tokens_out":2849,"duration_ms":27584,"significance":"If the result holds, this is the first dedicated experimental search for this specific bbZ(ll)+pTmiss signature, and it probes regions of the 2HDM+a parameter space (in particular high tan(beta)) that are motivated by the gamma-ray galactic center excess and are not directly covered by earlier mono-Z or mono-H searches. The analysis is careful and thorough: it uses the full Run 2 dataset, applies a multivariate discriminant, enumerates systematic uncertainties in detail, provides control-region validation plots, and makes tabulated results available in HEPData. The main strengths are the explicit signal model, the use of data-driven background normalizations with cross-checked control regions, and the public record of the results.","major_comments":[{"comment":"The data-driven correction to the DY pTmiss scale and resolution is derived in a sideband with pTmiss < 65 GeV and then extrapolated to the signal region. The only stated verification is in a region with no b jets and pTmiss < 300 GeV (Section 6). This does not directly validate the correction in the SR phase space, which requires Nb >= 1 and extends to pTmiss values well above 300 GeV. Because the MLP discriminant uses pTmiss and mT(ell,ell,pTmiss) as leading inputs (Section 5.3), a miscalibration of the DY tail would directly reshape the SR templates and thereby shift the derived limits, especially for low-mH signals that populate the lower MLP4 bins. Please either provide a validation in a b-enriched region with higher pTmiss, or quantify the uncertainty from this extrapolation and show its effect on the final limits.","section":"Section 6"},{"comment":"The background-only fit changes the WZ and ZZ normalizations by +101% and +113%, respectively. These processes populate the high-MLP4 (signal-like) bins, and the paper states that their normalization uncertainties are among the most important systematics. The WZ and ZZ control regions are single-bin, so they constrain the overall normalization but not the shape of the MLP4 distribution in the SR. Given the very large correction factors, additional validation is needed to demonstrate that the shape modeling in the b-enriched phase space is reliable. Please show pre-fit and post-fit comparisons in finer-grained b-enriched CRs, or otherwise discuss how the large k-factors are constrained by the data.","section":"Section 8"}],"minor_comments":[{"comment":"The definition of the MLP4 score in Eq. (4) appears to contain a typographical issue: the floor function is not explicitly written. Please clarify the transformation.","section":"Section 5.3"},{"comment":"The y-axis of Figure 7 is described as the upper limit on sigmaB but the scaling by arbitrary factors (x10^-n) is only noted in the caption. It would be clearer to label each panel with the absolute scale or indicate the scaling directly on the axis.","section":"Section 8.1 / Figure 7"},{"comment":"The benchmark parameters in Eq. (1) are cited to Refs. [3,52], but readers may benefit from a direct reference to the specific 2HDM+a parameter scan in which this benchmark was first used. Please check whether a more precise citation is available.","section":"References"},{"comment":"The validation of the DY Nb=0/Nb>=1 normalization consistency is described only in words. A figure or table showing the compatibility of the ratios in the two categories would strengthen this cross-check.","section":"Section 6"}],"recommendation":"major_revision","confidential_remarks":"The two major comments concern validation gaps in the background modeling that are directly relevant to the central null result and the derived limits. They are not fatal, but they require additional supporting material or analysis before the paper can be accepted. The large WZ/ZZ normalization factors deserve particular scrutiny in the internal review, even though the paper explains them plausibly."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Bottom line: this is the first dedicated LHC search for the bbH, H->Za, a->chi chi channel with Z->ll, and it is a careful piece of work. The channel is well motivated by the galactic-center excess and relic-density calculations, and the limits exclude part of the 2HDM+a parameter space favored by those calculations at high tan(beta). That is a genuinely new result, even if it is an incremental step within the broader DM search program.\n\nWhat the paper does well: the analysis is methodologically clean. A profile-likelihood fit combines four control regions with a 17-bin MLP discriminant in the signal region. The tt veto via the analytical reconstruction is sensible and removes a large fraction of the dominant background. Systematic uncertainties are enumerated and mostly constrained by the fit. The MLP deliberately excludes jet variables to avoid propagating jet mismodeling, which shows care. The post-fit agreement in the control regions is good, and the model interpretation uses benchmarks from the LHC DM working group without any circular fitting to the data.\n\nThe main soft spot is the DY pTmiss correction. It is derived in a sideband with pTmiss < 65 GeV and extrapolated to the signal region, where the MLP relies heavily on pTmiss and mT(ell,ell,ptmiss). The validation is done only in a no-b-jet region with pTmiss < 300 GeV, which does not cover the actual SR phase space with Nb>=1 and higher pTmiss. The paper itself flags this limitation, and it is a genuine validation gap rather than a demonstrated failure: the control-region fits show agreement, there is no observed discrepancy, and this is a null result. Still, because DY is the dominant background in the low-to-mid MLP bins, a miscalibrated tail would directly shift the limits. I would ask the authors to add a validation in a b-tagged region or otherwise comment on the phase-space coverage, but I would not reject over it.\n\nOne smaller caveat: the WZ and ZZ normalizations are fitted to data and increase by 101% and 113%. The authors explain this as a phase-space effect, and the fit accommodates it, but it would be worth seeing a bit more cross-checking on those large pulls.\n\nWho this is for: people working on 2HDM+a phenomenology, LHC dark matter searches, and the galactic-center excess. It deserves a serious referee. I would send it to peer review and expect it to survive with reasonable revision.","headline":"First dedicated CMS probe of the bb+Z(ll)+MET channel: a solid, careful search whose main soft spot is a validation gap in the DY pTmiss correction.","tokens_in":47664,"tokens_out":2180,"would_cite":true,"duration_ms":21597,"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 first dedicated search for dark matter produced with bottom quarks and a Z boson at a hadron collider finds no signal and sets cross-section limits down to 10^-3 pb.","keywords":["dark matter","pseudoscalar mediator","2HDM+a","missing transverse momentum","bottom-quark jets","Z boson","Higgs boson","collider search"],"falsifier":"The central null result would be weakened if, in a dedicated validation region requiring at least one b-tagged jet, a Z-like dilepton pair, no extra leptons, and 300 < pT^miss < 500 GeV, the ratio of observed events to the corrected prediction deviated from unity by more than the quoted uncertainty.","tokens_in":46689,"feed_emoji":"🌌","tokens_out":5787,"duration_ms":54750,"temperature":0.7,"pith_summary":"This paper argues that a new dark-matter production channel—bottom-quark pairs recoiling against a leptonically decaying Z boson and large missing transverse momentum, with the missing momentum originating from a pseudoscalar mediator decaying into dark matter—can be isolated from standard model backgrounds, and that no excess is observed in the full 138 fb^-1 dataset. The channel matters because it reaches a region of the 2HDM+a parameter space, especially large tan-beta, that other dark-matter and flavor searches cannot reach, including the region favored by a dark-matter interpretation of the galactic-center gamma-ray excess. The paper's central result is the first experimental constraint on the product sigma(pp->bbH) x B(H->Za) x B(Z->ll) x B(a->chi chi), with observed 95% confidence-level upper limits falling from about 10^-2 pb for a 400 GeV heavy scalar to about 10^-3 pb for a 2000 GeV heavy scalar. These limits exclude heavy-scalar masses up to roughly 900 GeV for small mediator masses in the benchmark model, and they exclude a substantial part of the parameter space preferred by the observed dark-matter relic density.","feed_headline":"No dark matter signal in bottom-quark plus Z search","feed_subtitle":"Full 138 fb^-1 dataset agrees with standard model; limits reach 10^-2 to 10^-3 pb.","key_machinery":"The central machinery is the event-selection plus multivariate discriminant: a loose requirement of at least one b-tagged jet is combined with a Z-mass window on an opposite-sign same-flavor dilepton pair and a missing-transverse-momentum threshold, and then a fully connected neural network (the MLP, binned into 17 MLP4 score intervals) combines kinematic variables such as pT^miss, the transverse mass mT, and mT2 to separate signal from background. Backgrounds are controlled with four single-bin control regions for Drell-Yan, top-quark pair, WZ, and ZZ production, an analytical kinematic solver vetoes dileptonic top-quark events, and the Drell-Yan missing-momentum scale and resolution are co","core_discovery":"The discovery claim is a null result: in proton-proton collisions at 13 TeV, events with an opposite-sign same-flavor lepton pair near the Z mass, at least one b-tagged jet, and large missing transverse momentum agree with standard model predictions. The paper establishes the first dedicated collider search for this final state and reports observed 95% CL upper limits on sigma(pp->bbH) B(H->Za) B(Z->ll) B(a->chi chi) ranging from about 10^-2 pb at mH = 400 GeV to about 10^-3 pb at mH = 2000 GeV. In the benchmark 2HDM+a interpretation, the results exclude heavy scalar masses up to about 900 GeV for small pseudoscalar masses, and up to about 1.1 TeV for tan-beta near 25, covering a meaningful","pith_inferences":["An extension the paper leaves implicit is that including final states where the Z decays to tau leptons, or where the dilepton pair arises from a non-resonant spectrum, could cover configurations where the present lepton selection loses efficiency.","Because the dominant limitations come from the normalization uncertainties of the WZ and ZZ backgrounds, a future analysis could improve sensitivity with a dedicated diboson-enriched control region or by exploiting jet substructure for more boosted topologies.","The tabulated model-independent results allow reinterpretation in other pseudoscalar-mediator or two-Higgs-doublet scenarios beyond the benchmark, which is a natural follow-up given the cosmological motivation."],"forward_implications":["The quoted upper limits provide the first model-independent cross-section bounds for the bb + Z + missing-momentum final state, applicable to any model with a heavy scalar decaying to Z plus invisible particles in association with bottom quarks.","In the benchmark 2HDM+a scenario, heavy scalar masses up to about 900 GeV are excluded for small pseudoscalar masses, closing part of the parameter space favored by dark-matter relic-density calculations.","The analysis excludes low mediator masses over a broad range of the mixing angle sin(theta), leaving only very small and very large values of sin(theta) uncovered.","The search is most sensitive to semi-boosted topologies with mH - ma near 1 TeV, while compressed mass configurations and configurations with very large mass splittings remain less constrained."],"fun_headline_variants":["Null result in first dark matter search with bottom quarks and Z","CMS sees no dark matter in bottom-quark and Z boson events","First LHC hunt for dark matter via bb+Z finds no signal","Bottom-quark plus Z: no dark matter excess in CMS data"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The analysis assumes that the data-driven correction to the Drell-Yan missing-momentum scale and resolution, derived at low missing momentum and checked in a region without b jets, remains valid in the signal region where missing momentum is large and at least one b jet is present.","fun_headline_variants_meta":{"raw":{"variants":["Null result in first dark matter search with bottom quarks and Z","CMS sees no dark matter in bottom-quark and Z boson events","First LHC hunt for dark matter via bb+Z finds no signal","Bottom-quark plus Z: no dark matter excess in CMS data"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000601,"raw_usage":{"total_tokens":2686,"prompt_tokens":825,"completion_tokens":1861,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":569,"completion_tokens_details":{"reasoning_tokens":1784}},"tokens_in":569,"tokens_out":1861,"duration_ms":14957,"temperature":1.0,"reasoning_tokens":1784,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-04T09:55:21.497519+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"The central null result would be weakened if, in a dedicated validation region requiring at least one b-tagged jet, a Z-like dilepton pair, no extra leptons, and 300 < pT^miss < 500 GeV, the ratio of observed events to the corrected prediction deviated from unity by more than the quoted uncertainty.","supporting_citations":[],"review_version":1}