{"id":"bacc004c-e2ec-4634-95ed-6b6cd6eef8d3","arxiv_id":"2505.05300","paper_version":2,"verdict":"ACCEPT","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"low","formal_verification":"none","parameter_count":5,"one_line_summary":"A search for dark matter produced with one or two top quarks finds no significant signal and excludes scalar and pseudoscalar mediator masses below 310 and 320 GeV at 95% confidence.","lead":"The CMS experiment searched for dark matter produced together with one or two top quarks in 138 inverse femtobarns of LHC proton-proton collisions. The search found no significant new physics and set improved limits on the masses and couplings of dark matter mediator particles.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Transfer-factor shape extrapolation in the AH/SL pTmiss tails is the central caveat; it is standard, internally flagged, and not shown to fail, so the ACCEPT verdict stands.","rationale":"The paper is a standard CMS Run-2 search with a transparent profile-likelihood analysis, HEPData tables, and an honest report of the 1.9 sigma excess; the null-result claim is well supported. The weakest point is the transfer-factor shape assumption in the AH/SL tails, and the paper flags it itself in Section 7. This is a legitimate modeling caveat common to this class of searches, not an internal inconsistency, and no demonstrated bias or closure failure is presented by the external evidence. A signal-injection test is the direct way to determine whether the per-bin transfer factors can absorb a real signal; absent evidence that they do, the ACCEPT verdict is not moved. I agree with the reader's weakest_assumption.","tokens_in":48873,"tokens_out":10577,"duration_ms":118920,"concrete_test":"Perform a signal-injection closure test with the published COMBINE model for the m_a = 150 GeV pseudoscalar hypothesis: generate pseudo-data from the background-only post-fit model plus a signal at the observed best-fit signal strength, refit with the full Section 6 transfer-factor parametrization, and compare the fitted signal strength and local significance to the injected values; repeat with the per-bin tt transfer factors replaced by a single smooth transfer function per channel. If the free per-bin parameters bias the fitted signal strength downward by more than one ensemble standard deviation, or reduce the median local significance by more than 0.5, the transfer-factor flexibility can absorb signal and the exclusion boundaries are not robust; if the signal is recovered, the concern is resolved.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The load-bearing assumption is in Section 6: an unconstrained multiplicative parameter is assigned to each background in each pTmiss bin of the AH and SL channels, shared across CRs and SRs, with the transfer of shape and composition left to MC simulation (Ref. [79]). The AH tt CR (Section 4.4.1) selects one lepton with mT<140 GeV while the AH SRs have zero leptons and mbT>140 GeV; the SL tt CR (Section 4.4.2) selects two leptons with mllT2<80 GeV while the SL SRs require mT>140, mWT2>180, mbT>140, and a topness split. At high pTmiss the CR yield falls, so the per-bin parameter is less tightly anchored and can be pulled by SR data. This is exactly the region where the excess in the AH 2b SR appears and where the 310/320 GeV exclusion boundaries are set. Section 7 explicitly names the per-bin background estimation parameters, especially for tt, among the leading uncertainties. The concern is not that the paper misuses the method; it is that the null result and the limits rely on a shape-transfer assumption that is not directly validated in the paper.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This CMS paper presents a search for dark matter produced in association with a single top quark or a top quark pair using the full Run 2 proton-proton collision dataset at sqrt(s) = 13 TeV, corresponding to 138 inverse femtobarns. Events are classified into all-hadronic (0 lepton), single-lepton (1 lepton), and dileptonic (2 lepton) channels, with signal regions binned in missing transverse momentum for the first two channels and in a neural network discriminant for the dilepton channel. Background estimation combines Monte Carlo simulation with data control regions, using per-bin unconstrained transfer factors for the dominant backgrounds in the hadronic and single-lepton channels. A simultaneous profile likelihood fit is performed across all regions and channels. No significant deviations from the standard model are observed, with the largest local excess being 1.9 standard deviations for a 150 GeV pseudoscalar mediator. The results are interpreted in a simplified scalar/pseudoscalar mediator model and, for the first time in this final state, translated into limits on axion-like-particle couplings to top quarks. Expected (observed) 95% CL exclusion limits on the mediator mass are 410 (310) GeV for the scalar and 380 (320) GeV for the pseudoscalar benchmark.","tokens_in":49030,"tokens_out":2852,"duration_ms":32671,"significance":"If the results hold, this analysis provides the most sensitive LHC constraint to date on spin-0 mediators coupling to top quarks and dark matter in the combined t+DM and tt+DM topologies, improving on the previous CMS search by about 40%. The simultaneous optimization for both production modes across three lepton multiplicities and the inclusion of the dileptonic channel are genuine advances. The paper is internally consistent: the selection tables, control region definitions, systematic uncertainty list, and fit outputs align, and the quoted limits follow from the described profile likelihood procedure. The excess is reported with local significances and is not over-interpreted. The benchmark signal model, couplings, and leading-order cross sections are external to the analysis (LHC Dark Matter Working Group recommendations), and the background prediction is anchored to data control regions. The main caveat is the per-bin transfer-factor shape extrapolation for the dominant tt background in the high-pTmiss tails used in the hadronic and single-lepton channels (Section 6, Ref.","major_comments":[],"minor_comments":[{"comment":"The text states that the neural networks used a \"Recified Linear Unit\" activation function; this should read \"Rectified Linear Unit\".","section":"4.3.1"},{"comment":"The sentence \"Categorical cross-entropy was used for the activation function\" is imprecise; categorical cross-entropy is a loss function, not an activation function, and the sentence should be rephrased accordingly.","section":"4.3.1"},{"comment":"In Figure 5, the legend entries \"Expected AH\", \"Expected SL\", and \"Expected DL\" are not explained in the caption; the caption should clarify that these denote the channel-specific expected limits from the all-hadronic, single-lepton, and dileptonic channels, respectively.","section":"7 (Fig. 5)"},{"comment":"The phrase \"Expected exclusion limits ... are set\" is imprecise: expected limits are projected, while observed limits are set; consider rephrasing to distinguish the expected exclusion reach from the observed exclusions.","section":"Abstract and Section 7"},{"comment":"The paper reports local significances for the excess but does not quote a global significance including the look-elsewhere effect; adding this information would help the reader calibrate the significance of the observed excess.","section":"7"}],"recommendation":"accept","confidential_remarks":"This is a well-executed and clearly documented experimental search. The transfer-factor assumption is the most delicate point but is standard, internally flagged, and backed by control-region fits; a dedicated closure test would strengthen confidence but is not indispensable for publication. The claims of being the first simultaneous optimization across the three lepton channels and the first ALP interpretation in this final state appear appropriate. No concerns about citation practices or fit to journal scope."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Quick take: this is a well-executed, incremental CMS Run 2 search for dark matter produced with one or two top quarks. It deserves a serious referee and, on the evidence here, acceptance.\n\nWhat is actually new: the paper is the first to optimize t+DM and tt+DM simultaneously across 0, 1, and 2 lepton final states. The new dilepton channel and the lower jet-multiplicity selections for single-top production give roughly 40% better sensitivity than the previous CMS combination, and the results include the first limits on |ct|/fA for ALP mediators in this signature. The model-independent visible cross section limits, from 1 pb down to 0.02 pb, are the kind of output other model builders can reuse. The paper is clearly written; the selection tables, CR definitions, systematic lists, and fit outputs line up, and the data in HEPData is a plus.\n\nThe soft spots are proportionate. The load-bearing assumption is the per-bin background transfer-factor method in the all-hadronic and single-lepton channels. Unconstrained multiplicative parameters connect CRs and SRs in each pTmiss bin, so the shape and composition transfer from CR to SR is carried by MC simulation. The stress-test note identifies this correctly as the fragile piece, especially in the high-pTmiss tails where the AH 2b excess lives. But I do not think this rises to a flaw. The method is standard in CMS DM searches, the paper explicitly names these parameters as a leading uncertainty, and the CRs are kinematically close to the SRs apart from the inverted selections. The post-fit plots do not show an obvious breakdown. A referee should ask for a validation plot in the extreme tail, but the concern is a caveat, not a killer.\n\nOther soft spots are minor. Signal cross sections are LO, the benchmark point is gq = gchi = 1 with mchi = 1 GeV, and the ALP interpretation assumes negligible ALP couplings to gluons and electroweak bosons. All of this is stated plainly. The 1.9 sigma excess is not over-interpreted; the local significances are reported, and the paper explains why the excess is consistent across mediator masses.\n\nI largely agree with the reader's scores. I would put novelty at 7 rather than 6, since the simultaneous optimization and the ALP interpretation are real additions, but significance 6 and soundness 8 are about right. The central claim, that no significant deviation is observed, holds. This paper is for the LHC DM working group, phenomenologists who want reinterpretable limits, and the collaborations planning the next round of top+DM searches. Send it to referees.","headline":"A well-executed, incremental CMS Run 2 search for DM plus top quarks that handles its 1.9 sigma excess honestly; the transfer-factor shape caveat is real but standard and not shown to fail, so it deserves refereeing and likely acceptance.","tokens_in":49699,"tokens_out":3151,"would_cite":true,"duration_ms":29886,"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":"A search for dark matter produced with top quarks in 13 TeV proton-proton collisions finds no significant signal; a 1.9-standard-deviation excess at the 150 GeV pseudoscalar hypothesis weakens the expected exclusion limits from 410/380…","keywords":["dark matter","top quark","missing transverse momentum","scalar mediator","pseudoscalar mediator","axion-like particles","simplified model","hadron collider"],"falsifier":"Reanalyze the same $138\\,\\mathrm{fb}^{-1}$ data set with an independent data-driven estimate of the top-quark-pair background in the high-$p_{\\mathrm{T}}^{\\text{miss}}$ tail, for example using a sideband defined by inverting the $m_T^b$ or $m_{T2}^W$ requirement and extrapolating by simulation-free ratios; if the signal-like excess at the 150 GeV pseudoscalar hypothesis disappears or falls below 1 standard deviation, the null-result claim is supported, while a growth beyond 3 standard deviations would refute it.","tokens_in":48535,"feed_emoji":"🌌","tokens_out":7821,"duration_ms":74360,"temperature":0.7,"pith_summary":"The paper searches for dark matter produced together with a single top quark or a top-quark pair, using large missing transverse momentum as the experimental handle. It aims to establish whether a benchmark simplified model with a scalar or pseudoscalar mediator, unit couplings, and a 1 GeV dark matter particle is excluded by the $138\\,\\mathrm{fb}^{-1}$ data set, and whether any excess over standard model backgrounds is real. The central result is a null signal: expected 95% CL exclusions of 410 and 380 GeV for scalar and pseudoscalar mediators are weakened by a signal-like excess, with the largest local significance 1.9 standard deviations at the 150 GeV pseudoscalar hypothesis, leaving observed exclusions of 310 and 320 GeV. A curious reader should care because this probes collider-accessible dark matter coupled to top quarks, a channel that is decisive for simplified dark matter models.","feed_headline":"Dark matter search finds no signal, only a 1.9-sigma hint","feed_subtitle":"A 13 TeV search excludes scalar and pseudoscalar mediators below 310/320 GeV; the mild excess is a likely fluctuation.","key_machinery":"The central object is the simplified spin-0 mediator model, defined by Lagrangians for a scalar $\\phi$ and a pseudoscalar $a$ that couple through Yukawa-like interactions to standard model quarks, preferentially the top quark, and to a Dirac dark matter fermion $\\chi$. The analysis machinery combines orthogonal signal regions split by lepton multiplicity (0, 1, 2), b-tagged jet count, and forward-jet presence, with discriminating observables that include the missing transverse momentum, the transverse masses $m_T$, $m_{T2}^W$, and $m_T^b$, a modified topness variable, and, in the dilepton channel, a neural network fed by kinematic reconstruction of the top quarks. The load-bearing statistical mechanism is a simultaneous fit in which per-bin unconstrained 'transfer factor' parameters adjust the dominant backgrounds, especially $t\\bar{t}$, as functions of $p_{\\mathrm{T}}^{\\text{miss}}$ while linking control regions to signal regions; this is what converts control-region event counts into signal-region background predictions.","core_discovery":"After analyzing events with zero, one, or two leptons and at least one b-tagged jet, the paper states that no significant deviations from the standard model predictions are observed. Under the simplified model with mediator couplings to fermions and dark matter both set to unity and a dark matter mass of 1 GeV, mediator masses are excluded at 95% confidence level below 310 GeV for the scalar and 320 GeV for the pseudoscalar, whereas 410 and 380 GeV were expected. The shortfall is attributed to a signal-like excess that is visible mainly in the all-hadronic two-b-tag region and in three dileptonic regions, is consistent across signal regions, and has local significances between 1.4 and 1.9 standard deviations. The paper also derives model-independent 95% CL upper limits on the visible cross section of top-associated dark matter production, from 1 pb down to 0.02 pb, and, for the first time, limits on the coupling of axion-like particles to top quarks in this invisible signature.","pith_inferences":["The per-bin transfer-factor assumption is the most promising point to attack: a closure test using an alternative top-quark-pair shape model, or a background estimate from same-sign leptons, could reveal whether the mild excess is a symptom of shape mismatch rather than new physics.","If the excess is genuine, its consistency with many mediator masses suggests a broad, low-mass production mode; a dedicated scan of mediator masses below 50 GeV and higher couplings could be the fastest way to confirm it with this same data set.","The model-independent limits could be recast, without new experimental analysis, onto 2HDM+a or other top-philic dark matter models, since the paper only interprets the benchmark spin-0 simplified model."],"forward_implications":["If the central claim is right, the benchmark scalar and pseudoscalar mediator model with unit couplings and a 1 GeV dark matter particle is excluded for mediator masses below 310 and 320 GeV, closing a simple thermal weakly interacting massive particle window.","The 1.9-standard-deviation excess, treated in the paper as a background fluctuation, implies that more data at the same center-of-mass energy should either make the excess grow into a signal or wash it out; the next data-taking period provides a direct test.","The model-independent visible cross section limits, from 1 pb to 0.02 pb, bound any new physics process producing top quarks plus invisible particles, regardless of the mediator's spin, as long as the kinematics fall in the searched phase space.","The first axion-like-particle limits in this final state constrain the top-quark coupling to ALPs for mediator masses between 50 and 500 GeV, complementing constraints from $t\\bar{t}$ resonance searches."],"supporting_citations":[{"why":"the previous search in the same final state that this analysis extends and improves","marker":"[16]"},{"why":"the combination from the other collaboration that provides the boundary it must match or surpass","marker":"[17]"},{"why":"the earlier combined search whose sensitivity this paper improves by about 40%","marker":"[21]"},{"why":"the source of the transfer-factor background estimation method used in the simultaneous fit","marker":"[79]"},{"why":"the benchmark choice of unit mediator couplings and 1 GeV dark matter mass","marker":"[50]"},{"why":"the scalar simplified model Lagrangian that defines the signal interpretation","marker":"[4]"},{"why":"the definition of $m_{T2}$ used to suppress dileptonic top-quark-pair background","marker":"[59]"},{"why":"the modified frequentist CLs method used to set the 95% confidence limits","marker":"[80]"}],"fun_headline_variants":["Dark matter search: no signal, only a 1.9σ hint","CMS sees no dark matter, only a 1.9σ excess","Top-associated DM? No, but a 1.9σ tease remains","DM mediators excluded below 310 GeV with a hint","No dark matter signal, just a 1.9σ bump in data"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The result depends on the assumption that the per-bin transfer factors, calibrated by Monte Carlo simulation, correctly describe how the standard model background shape and composition, in particular top-quark pair events, transfers from control regions into the far high-missing-transverse-momentum tail; if that shape transfer is wrong, the exclusion boundaries and the 1.9-standard-deviation excess both shift.","fun_headline_variants_meta":{"raw":{"variants":["Dark matter search: no signal, only a 1.9σ hint","CMS sees no dark matter, only a 1.9σ excess","Top-associated DM? No, but a 1.9σ tease remains","DM mediators excluded below 310 GeV with a hint","No dark matter signal, just a 1.9σ bump in data"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000799,"raw_usage":{"total_tokens":3591,"prompt_tokens":1102,"completion_tokens":2489,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":718,"completion_tokens_details":{"reasoning_tokens":2394}},"tokens_in":718,"tokens_out":2489,"duration_ms":19334,"temperature":1.0,"reasoning_tokens":2394,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T23:08:20.196376+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Reanalyze the same $138\\,\\mathrm{fb}^{-1}$ data set with an independent data-driven estimate of the top-quark-pair background in the high-$p_{\\mathrm{T}}^{\\text{miss}}$ tail, for example using a sideband defined by inverting the $m_T^b$ or $m_{T2}^W$ requirement and extrapolating by simulation-free ratios; if the signal-like excess at the 150 GeV pseudoscalar hypothesis disappears or falls below 1 standard deviation, the null-result claim is supported, while a growth beyond 3 standard deviations would refute it.","supporting_citations":[{"cited_title":"Stop the Top Background of the Stop Search","cited_arxiv_id":"1203.4813","evidence_quote":"the definition of $m_{T2}$ used to suppress dileptonic top-quark-pair background"},{"cited_title":"Confidence level computation for combining searches with small statistics","cited_arxiv_id":null,"evidence_quote":"the modified frequentist CLs method used to set the 95% confidence limits"}],"review_version":1}