{"id":"91362d06-decf-475b-a831-c449d0cd457b","arxiv_id":"2608.10148","paper_version":1,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"low","formal_verification":"none","parameter_count":0,"one_line_summary":"No significant excess is observed in a search for top-philic resonances in four-top-quark events with two leptons, and 95% confidence-level upper limits are set for mediator masses from 500 GeV to 4 TeV.","lead":"The CMS experiment searched for a new heavy particle that would appear together with four top quarks in proton-proton collisions. It found no clear signal and set new upper limits on how often such a particle could be produced.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The background estimate's load-bearing assumption is that t-tagging correlation corrections measured in Z+jets-dominated CR2J transfer to tt-dominated SRs; the dedicated 20% flavour uncertainty is the only cover, and a direct closure test in a b-enriched region is missing.","rationale":"The paper is a complete, well-documented search. The strongest claim, no significant excess, is supported by the data: the 2.1 sigma local excess is not significant, especially before any look-elsewhere correction. The exclusion limits are conditional on the benchmark signal definition (s-channel only, no interference), which the paper states explicitly. The most load-bearing assumption is the transfer of the t-tag correlation correction from CR2J to the SRs. This is the same assumption the reader highlighted. I agree it is the weakest point, but the paper does not ignore it: it validates the shape in Fig. 4, applies the normalisation corrections, and assigns a dedicated 20% uncertainty for the flavour-composition difference. My concern is whether 20% is sufficient, given the large CR corrections (0.66 for 13.6 TeV mu+mu-) and the 22% mistag rate that amplifies correlation errors. This is testable with a closure test in a tt-enriched two-VR-jet region. Since the concern is about the quantitative reach of exclusion limits rather than the validity of the null result, and since the paper already discloses the limitation, I do not change the verdict.","tokens_in":1058,"tokens_out":1013,"duration_ms":91823,"concrete_test":"Perform a closure test in the tt-enriched control region of Section 4.5: extend the selection to require at least two VR jets and split events by 0 versus 2 t-tagged VR jets, in data and simulation, for each data-taking period. Measure the 0-to-2 tag transfer ratio (the quantity used in the SR background estimate) in this b-rich, tt-dominated phase space and compare it with the CR2J-derived correction factors applied to the SRs. If the ratio differs from the SR transfer factor by more than the assigned 20% flavour-normalisation uncertainty, the background normalisation in the SRs is not covered by the current systematic. Alternatively, rerun the combine limit fit with the flavour-normalisation uncertainty inflated to 30% and 40% and check whether the 50%-width vector exclusion at 850 GeV and the low-mass local significances survive.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central null result is robust, but the quantitative exclusion reach rests on the background normalisation in Section 5. The SR prediction is obtained from 0-tag VR jet events using simulation-derived transfer factors, then corrected by constants fit in the Z+jets-dominated CR2J (Fig. 4). These corrections are large (e.g., 0.66±0.15 for 13.6 TeV mu+mu-) and their transfer to the SRs assumes the residual mismodelling of the correlation between the two top-tagging decisions is the same in CR2J and in the SRs. The SR jet-flavour composition differs substantially: Table 2 gives 18-21% b-quark-matched VR jets in SR1b2T/SR2b2T, whereas CR2J is dominated by Z+jets with light-flavour jets. The dedicated systematic in Section 6 is estimated by removing b-matched jets from the transfer-factor computation in the same simulation; it does not directly test whether the data/MC correlation mismatch in a b-enriched, tt-dominated sample is the same as in CR2J. Because the 2-tag selection has a high mistag rate (22%), the 0-tag to 2-tag transfer factor amplifies any per-jet correlation error. A 20% normalisation shift is comparable to the 84 vs 61.5 observed/expected pre-fit excess, so the low-mass local excesses (2.1-2.4 sigma) and the vector 50%-width exclusion at 850 GeV are sensitive to this assumption. This is a modelling caveat, not an internal inconsistency; the paper discloses it and assigns an uncertainty, so the null result stands, but the exact mass reach is conditional.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper reports a search for a heavy resonance produced in association with a top-quark pair and decaying to a top-quark pair, in final states with two opposite-sign leptons and jets, using 138 fb^-1 of 13 TeV and 35 fb^-1 of 13.6 TeV CMS data. The resonance mass is reconstructed from the two leading variable-radius jets, and a boosted decision tree is used to identify hadronically decaying top quarks. The dominant background from misidentified top-tagged jets is estimated from data by transferring zero-tagged events in transfer regions to two-tagged signal regions using simulation-derived transfer factors, with normalisation corrections obtained from Z-boson control regions. A simultaneous profile-likelihood fit to the dijet invariant mass finds no significant excess (background-only p-value 0.21), and 95% CL upper limits are set on ttZ', ttphi, and tta production cross sections for mediator masses from 500 GeV to 4 TeV and relative widths from 4% to 50%. For the vector benchmark with 50% relative width, observed exclusion reaches 850 GeV, compared with 1000 GeV expected. Local significances of 2.1-2.4 standard deviations are observed for the 500 GeV signal hypotheses and are attributed to statistical fluctuations.","tokens_in":46297,"tokens_out":9864,"duration_ms":103689,"significance":"If the result holds, this is the first dedicated two-lepton four-top-resonance search and it extends sensitivity below 1 TeV using variable-radius jets; it also provides a new reinterpretation of the results in terms of axion-like particles. The paper's strengths include a data-driven background estimate with dedicated transfer and control regions, a profile-likelihood fitting procedure with CLs limits, a documented treatment of systematic uncertainties including a dedicated 20% flavour-composition uncertainty, and a public HEPData record. The central null result is robust. The main caveat, explicitly acknowledged in Section 5, is that the normalisation corrections measured in the Z+jets-dominated CR2J region are transferred to the b-enriched signal regions under an assumption about the similarity of top-tagging correlation mismodelling; the paper states this assumption and assigns a dedicated 20% uncertainty, so I do not regard this as undermining the central claim that no significant excess is observed. The exact mass reach of the exclusions is conditional on this modelling assumption, but that is a standard feature of searches of this type.","major_comments":[],"minor_comments":[{"comment":"The region names SR1b2T and SR2b2T are initially easy to misread; although Table 1 defines n_b and n_t, a one-line clarification in the text that 'b2T' denotes the b-jet multiplicity plus the requirement of at least two t-tagged VR jets would improve readability.","section":"Section 4.4 / Table 1"},{"comment":"The correction factors for the e±mu∓ channel are set to the geometric mean of the e+e- and mu+mu- factors, but no validation of this approximation is shown; a sentence describing the systematic uncertainty assigned to this choice, or a closure check in the e±mu∓ control region, would make the procedure more complete.","section":"Section 5"},{"comment":"The goodness-of-fit of the constant functions in Figure 4 is reported only as chi2/ndf ≲ 1; quoting the individual chi2/ndf values would be more quantitative and would help readers assess the quality of the normalisation corrections.","section":"Section 5 / Figure 4"},{"comment":"The legend entry 'Theory 95% CL upper limits' in Figure 6 appears to conflate the theoretical cross-section curve with the 95% CL limit curves; please relabel the theory line and the limit curves separately to avoid confusion.","section":"Figure 6"}],"recommendation":"accept","confidential_remarks":"The paper is a well-executed CMS search and is appropriate for EPJC. The only substantive caveat is the transfer of the CR2J normalisation corrections to the b-enriched signal regions, but the authors disclose the assumption and assign a dedicated 20% uncertainty, so I do not consider it blocking. The local excesses below 3 sigma are handled appropriately and do not affect the central null claim. I support acceptance."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"First things first: this is a solid, workmanlike search paper, and the central claim holds up. The genuinely new bits are the first dedicated two-lepton channel for four-top-quark resonances, the use of HOTVR variable-radius jets with a BDT top tagger in this signature, and the inclusion of 13.6 TeV data. The ALP reinterpretation is a nice extra. I agree with the reader's ACCEPT: no circular background fit, no p-hacking, and the null result is not in question.\n\nThe background estimation is the part worth scrutinising. They transfer 0-tag VR jet events to 2-tag events using simulation-derived transfer factors, then correct the normalisation with constants fit in Z+jets-dominated CR2J. As the stress-test note says, those corrections are sizeable (0.66±0.15 for 13.6 TeV mu+mu-) and the jet flavour composition in the SRs differs substantially (18-21% b-matched VR jets vs light-flavour-dominated CR2J). The dedicated 20% normalisation uncertainty covers exactly this difference, and the paper says so explicitly. But it is a single conservative blanket, and there is no direct closure test in a b-enriched tt-dominated region. Given the 2-tag selection has a 22% mistag rate, any per-jet correlation mismatch gets amplified in the transfer. This matters for the low-mass local excesses (2.1-2.4 sigma) and for whether the 50%-width vector exclusion sits at 850 GeV or somewhere else. I would not call this a flaw in the null result—the uncertainty is disclosed and propagated—but it is a modelling caveat that a referee should push on. The paper would be stronger with one validation region that is tt-dominated and b-enriched.\n\nOther soft spots are minor and disclosed. The signal definition includes only s-channel production and neglects interference with SM four-top production; the paper says the exclusions apply to that benchmark, so that is fine. The observed limits are weaker than expected because of the 84 vs 61.5 pre-fit excess, but the background-only p-value is 0.21, so nothing to chase. The citation pattern is appropriate: prior one-lepton searches, tt resonance searches, and the ALP literature are all cited.\n\nWho gets value: CMS/ATLAS exotica groups, pheno people mapping top-philic mediator parameter space, and anyone building four-top combinations. It deserves a serious referee. My recommendation: send to peer review, and ask the referee to focus on the background estimation closure and the size of the flavour-mismatch uncertainty.","headline":"A competent, honest null result that opens the two-lepton channel for four-top resonance searches; the central claim holds, with a disclosed but real caveat on the background normalisation that sets the exact mass reach.","tokens_in":46846,"tokens_out":1271,"would_cite":true,"duration_ms":16494,"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":"Four-top search finds no resonance and sets exclusion limits on a top-philic vector boson.","keywords":["four top quarks","top-philic resonance","vector mediator","boosted top tagging","variable-radius jets","dijet invariant mass","95% confidence limits","axion-like particle"],"falsifier":"With the next $100\\,\\text{fb}^{-1}$ of 13.6 TeV data, count events in SR2b2T near $m_{JJ}\\approx500$ GeV: a real 500 GeV resonance would show a growing peak, while a fluctuation would track the transfer-factor background. The transfer assumption itself can be checked by measuring the two-tag correlation in a same-flavour dilepton region outside the Z window and verifying that the control-region correction factor does not change by more than its uncertainty.","tokens_in":45757,"feed_emoji":"⚛️","tokens_out":8337,"duration_ms":71806,"temperature":0.7,"pith_summary":"This paper reports a search for a heavy boson that decays to a top-quark pair while being produced together with another top-quark pair, giving events with four top quarks. Using $173\\,\\text{fb}^{-1}$ of proton-proton collisions at 13 and 13.6 TeV, the CMS Collaboration finds no significant excess over the standard-model background. It sets 95% confidence-level upper limits on the production cross section of vector, scalar, and pseudoscalar mediators with masses between 500 GeV and 4 TeV and relative widths of 4-50%. For the benchmark top-philic vector mediator, resonances with 50% relative width are excluded up to 850 GeV, with an expected exclusion of 1000 GeV. The result is the first dedicated search for these models in the two-lepton channel and the first four-top resonance search to include 13.6 TeV data.","feed_headline":"Four-top search finds no resonance; vector boson excluded to 850 GeV","feed_subtitle":"First dedicated two-lepton search places 95% CL limits on top-philic Z' bosons.","key_machinery":"The mass of the putative resonance is reconstructed as $m_{JJ}$, the invariant mass of the two leading variable-radius jets formed with the HOTVR algorithm, whose distance parameter shrinks as $R_{\\text{eff}}=600\\,\\text{GeV}/p_T$ to follow collimated top-quark decays. A boosted decision tree tags whether each VR jet comes from a hadronically decaying top quark, and the background of misidentified top-tagged jets is estimated from data by transferring events with zero t-tagged VR jets into the two-tagged signal regions using simulation-derived transfer factors, with the overall normalisation corrected in Z-boson control regions. A simultaneous profile-likelihood fit of the $m_{JJ}$ templates in two signal regions and three dilepton channels yields the 95% confidence-level limits via the $CL_s$ criterion.","core_discovery":"On its own terms, the paper establishes a null result: the dijet invariant mass spectrum in the two-lepton, two-top-tagged signal regions is consistent with the standard-model background, with a background-only $p$-value of 0.21. The largest local deviation is a 2.1 standard deviation excess for a 500 GeV vector mediator with 4% width, with similar deviations of 2.0 and 2.4 standard deviations for scalar and pseudoscalar hypotheses. The paper therefore reports upper limits rather than evidence: for the $\\Gamma_{Z'}/m_{Z'}=50\\%$ benchmark, $Z'$ bosons are excluded up to 850 GeV, while the observed limits for 10 and 20% widths are weaker than expected and exclude no mediators. Scalar and pseudoscalar mediators are not excluded. These exclusions apply to the resonant $s$-channel contribution only; non-$s$-channel diagrams and interference with standard-model four-top production are not included, as the paper explicitly states.","pith_inferences":["If the mild 2.1σ excess near 500 GeV is real, additional 13.6 TeV data should produce a growing narrow peak in $m_{JJ}$; if it is a fluctuation, the spectrum should continue to follow the transfer-factor background.","Because the shown exclusions omit interference with standard-model four-top production, a full top-philic model could be more or less constrained than the benchmark limits suggest, particularly for broad mediators.","The same two-lepton plus VR-jet topology could be combined with single-lepton four-top searches, and applied to other resonances such as heavy Higgs bosons, to improve overall coverage of top-philic scenarios.","A direct measurement of the correlation between the two top-tagging decisions in a dedicated control region would test the background-transfer assumption and could shrink the roughly 20% normalisation uncertainty that currently dominates the background systematics."],"forward_implications":["The benchmark top-philic vector mediator with 50% relative width is excluded below 850 GeV, extending the mass reach of earlier four-top searches.","Scalar and pseudoscalar mediators are not excluded, so those interpretations remain viable targets for future searches.","Variable-radius jets extend sensitivity below 1 TeV mediator mass, a region where fixed-radius top tagging was inefficient.","The analysis provides the first four-top resonance limits that include 13.6 TeV data, combining 138 fb${}^{-1}$ at 13 TeV with 35 fb${}^{-1}$ at 13.6 TeV.","The ALP reinterpretation sets new upper limits on $c_t/f_A$ in the mass range around 500 GeV to 4 TeV, complementing tt-resonance and missing-transverse-momentum searches."],"supporting_citations":[{"why":"Documents the observation of four-top production in the multilepton final state, the process this search builds on.","marker":"[1]"},{"why":"Reports the CMS observation of four-top production that motivates four-top final states as a BSM probe.","marker":"[2]"},{"why":"Defines the top-philic vector mediator model and Lagrangian used as the benchmark signal.","marker":"[9]"},{"why":"Shows boosted top tagging for top-philic resonances, informing the reconstruction strategy.","marker":"[10]"},{"why":"Provides tt-resonance pseudoscalar limits used for comparison in the ALP reinterpretation.","marker":"[22]"},{"why":"Supplies low-mass ALP and dark-matter limits against which the new ALP coupling limits are compared.","marker":"[24]"},{"why":"Prior ATLAS search for top-philic heavy resonances in four-top events that this search extends to a new final state.","marker":"[25]"},{"why":"Companion CMS search for a top-philic Z' in a single-lepton final state, the previous four-top approach.","marker":"[26]"},{"why":"Introduces the HOTVR variable-radius jet algorithm used to reconstruct boosted top quarks.","marker":"[27]"},{"why":"Describes the machine-learning tagger for hadronically decaying heavy particles used to build the top-tagging BDT.","marker":"[28]"}],"fun_headline_variants":["Four-top search yields no resonance, constrains top-philic Z' bosons","CMS four-top search: no new boson, Z' excluded to 850 GeV","No excess in four-top search, but Z' mass limit set at 850 GeV","Top-philic boson search: no signal, but Z' excluded to 850 GeV","First dedicated two-lepton four-top search: no resonance, limits set"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The background estimate assumes that the residual mismodelling of the correlation between the two top-tagging probabilities, measured in Z-boson control regions and applied as a normalisation correction, is the same in the signal regions; if that correlation differs there, the background yield could shift by about 20% and change the mass reach.","fun_headline_variants_meta":{"raw":{"variants":["Four-top search yields no resonance, constrains top-philic Z' bosons","CMS four-top search: no new boson, Z' excluded to 850 GeV","No excess in four-top search, but Z' mass limit set at 850 GeV","Top-philic boson search: no signal, but Z' excluded to 850 GeV","First dedicated two-lepton four-top search: no resonance, limits set"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000886,"raw_usage":{"total_tokens":3862,"prompt_tokens":1018,"completion_tokens":2844,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":634,"completion_tokens_details":{"reasoning_tokens":2736}},"tokens_in":634,"tokens_out":2844,"duration_ms":20116,"temperature":1.0,"reasoning_tokens":2736,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T04:11:21.526008+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"With the next $100\\,\\text{fb}^{-1}$ of 13.6 TeV data, count events in SR2b2T near $m_{JJ}\\approx500$ GeV: a real 500 GeV resonance would show a growing peak, while a fluctuation would track the transfer-factor background. The transfer assumption itself can be checked by measuring the two-tag correlation in a same-flavour dilepton region outside the Z window and verifying that the control-region correction factor does not change by more than its uncertainty.","supporting_citations":[{"cited_title":"Search for dark matter produced in association with one or two top quarks in proton-proton collisions at $\\sqrt{s}$ = 13 TeV","cited_arxiv_id":"2505.05300","evidence_quote":"Supplies low-mass ALP and dark-matter limits against which the new ALP coupling limits are compared."},{"cited_title":"Search for a top-philic Z' boson decaying into a $\\mathrm{t\\bar{t}}$ pair in a final state with jets and an electron or muon in proton-proton collisions at $\\sqrt{s}$ = 13.6 TeV","cited_arxiv_id":"2608.01251","evidence_quote":"Companion CMS search for a top-philic Z' in a single-lepton final state, the previous four-top approach."},{"cited_title":"A new tagger for hadronically decaying heavy particles at the LHC","cited_arxiv_id":"1606.04961","evidence_quote":"Introduces the HOTVR variable-radius jet algorithm used to reconstruct boosted top quarks."}],"review_version":1}