{"id":"3da779de-8626-4b54-b680-1f39e83a0139","arxiv_id":"2607.11016","paper_version":1,"verdict":"ACCEPT","confidence":"HIGH","novelty_score":6.5,"correctness_risk":"low","formal_verification":"none","parameter_count":3,"one_line_summary":"No significant excess is observed in the four-prong large-radius jet + MET signature; 95% CL upper limits are set on the signal strength versus mediator mass or χ₂χ₁Y₀ coupling.","lead":"CMS searched 138 fb⁻¹ of 13 TeV LHC data for nonprompt dark-matter pairs that produce a single four-prong large-radius jet plus missing momentum, using a graph-neural-network jet tagger. No excess was seen; the first limits in this boosted displaced topology constrain the mediator mass and dark-sector coupling.","discovery_kind":"new_application","skeptic_critique":{"model":"grok-4.5","headline":"No significant objection identified","rationale":"The manuscript is a standard, transparent CMS search: LO signal generation, GNN jet tagger with secondary-vertex inputs, data-driven multijet/W/Z estimation via orthogonal CRs and transfer factors, simultaneous binned likelihood fit of pTmiss (or hadronic recoil), full experimental systematics, and asymptotic CLs limits. No excess is observed; limits are set on the two one-dimensional scans of Table 1. The large LJP-derived tagger SFs (50–70%) are the dominant signal systematic and are correctly flagged by the reader, yet they act only to broaden the expected limit bands; they cannot manufacture a null result when the background model is data-driven and the post-fit CRs agree with data. Because the central claim is the absence of excess plus the first limits in this topology, the analysis holds under the stated uncertainties. Verdict remains ACCEPT; no adjustment required.","tokens_in":37699,"tokens_out":471,"duration_ms":6606,"concrete_test":"Re-derive the observed 95% CL limits of Fig. 4 after fixing the GNN scale-factor uncertainty to zero (or to a flat 10% envelope) while keeping all other nuisances; if the observed limits move by less than the width of the current expected band, the large SF uncertainty is not load-bearing for the null result.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The reader's weakest assumption (50–70% GNN tagger scale-factor uncertainties from LJP reweighting for displaced four-prong jets) is real and dominates the signal systematic (Table 2, §6). However, it is not load-bearing against the central claim. The paper reports a null result with data-driven QCD/W/Z backgrounds constrained in situ via transfer factors and simultaneous CR+SR fits (Fig. 3, §5, §7). Large, conservatively assigned signal-efficiency uncertainties weaken the expected limits but do not create a false null or invalidate the CLs upper limits on µ. The first-search claim and the observed limits remain intact under the stated systematics. No internal inconsistency or unaccounted bias that would reverse the null result is evident.","agreement_with_reader":"agree"},"referee_report":{"model":"grok-4.5","summary":"This CMS paper presents a search for nonprompt dark-matter production in the Lorentz-boosted topology using 138 fb^{-1} of 13 TeV pp data. A heavy vector/axial-vector mediator Y1 produces a pair of long-lived dark fermions χ2 that each decay to a stable DM particle χ1 plus a light scalar Y0 (m_Y0 = 1 GeV) decaying to quarks; the four quarks are reconstructed as a single four-prong AK8 jet accompanied by large pTmiss and an ISR jet. Events are selected with a GNN jet-substructure tagger that incorporates secondary-vertex information, and the three dominant backgrounds (QCD multijet, W+jets, Z\to\nu\nu+jets) are estimated from data via transfer factors in dedicated control regions that are fitted simultaneously with the signal region. No significant excess is observed; 95% CL upper limits are set on the signal strength as functions of the Y1 mass or the coupling g_χ2χ1Y0 that controls the χ2 lifetime. The analysis is presented as the first search for this signature in the boosted regime.","tokens_in":37932,"tokens_out":909,"duration_ms":7585,"significance":"The result is a solid, first-of-its-kind experimental constraint on a previously unexplored boosted displaced DM topology. The analysis follows standard CMS Run-2 practice: data-driven transfer-factor background estimates constrained in situ by simultaneous CR+SR fits, a full suite of experimental systematics, and CLs limits with the asymptotic approximation. The GNN tagger that exploits both multiprong substructure and secondary vertices, together with the first extension of Lund-jet-plane reweighting to displaced four-prong jets, constitutes a useful technical advance even though the associated 50–70% efficiency uncertainties dominate the signal systematic. The null result and the reported limits remain robust under the stated systematics; the paper therefore supplies a concrete, reproducible exclusion of the tested benchmark points and a clear path for future improvements.","major_comments":[],"minor_comments":[{"comment":"Section 6 and Table 2: the 50–70% GNN scale-factor uncertainties are large and dominate the signal systematic. A short additional sentence clarifying how residual lifetime dependence was checked (or why a more differential uncertainty is not assigned) would help readers assess the conservatism of the assignment.","section":"Section 6 / Table 2"},{"comment":"Figure 2 caption and text: the data/simulation discrepancy in the GNN score is correctly attributed to imperfect QCD modelling and is irrelevant because the QCD yield is taken from data; a one-line reminder that the tagger is used only for region definition (not for absolute rate) would make this even clearer.","section":"Figure 2"},{"comment":"Table 1: the theoretical cross sections are quoted without PDF or scale uncertainties. Even a brief statement that these are LO MADGRAPH values and that the limits are reported on μ = σ/σ_theory would remove any ambiguity.","section":"Table 1"},{"comment":"Section 4: the precise definition of the secondary-vertex features fed to the interaction network is only sketched. A short list of the SV observables (or a reference to the training paper) would improve reproducibility.","section":"Section 4"},{"comment":"Figure 3: the pre-fit signal overlays in the SR are useful, but the vertical scale of the pull panels makes small coherent residuals hard to judge; a linear pull range of ±2 or ±3 would be more informative.","section":"Figure 3"}],"recommendation":"accept","confidential_remarks":"The manuscript is a clean, well-executed first search that fits comfortably in JHEP. The large tagger uncertainties weaken the expected reach but do not threaten the validity of the null result or the reported limits; I see no reason to request major revision. The paper can be accepted essentially as is, with only the usual minor polishing."},"author_rebuttal":null,"desk_editor":{"model":"grok-4.5","letter":"This is the first CMS search for a pair of nonprompt dark-matter candidates in the Lorentz-boosted four-prong large-radius-jet + MET topology. They take a simplified model with a heavy vector/axial-vector mediator Y1 producing long-lived χ2 that decay to stable χ1 + light Y0 → qq, reconstruct the four quarks as one AK8 jet, and tag it with an interaction-network GNN that folds in secondary-vertex information. Full Run-2 dataset (138 fb−1), data-driven transfer-factor estimates for QCD, W+jets and Z\to\nu\nu from dedicated CRs, simultaneous binned likelihood fit of pTmiss, standard CLs limits. No excess; they set 95 % CL limits on signal strength versus mediator mass or the g_χ2χ1Y0 coupling that controls lifetime.\n\nWhat they do well is the experimental craft. Backgrounds are constrained in situ from data; post-fit agreement in the CRs looks clean (Fig. 3). The GNN is a genuine technical step beyond ParticleNet-style taggers for displaced multiprong jets, and they document its AUC range and the LJP-based scale factors. The first-search claim is accurate; the model and topology were not previously covered in this boosted regime.\n\nThe soft spot is real but not fatal: the tagger efficiency scale factors carry 50–70 % uncertainties (Table 2) because subjet matching in the Lund-jet-plane reweighting is hard for highly collimated, displaced four-prong jets. That dominates the signal systematic and weakens the expected limits, yet it is assigned conservatively and does not invent a null result. The observed limits remain valid statements under the stated systematics. Benchmarks are sparse (fixed mY0 = 1 GeV, a few mass/coupling points), which is fine for a first look but leaves the five-dimensional space only sparsely sampled.\n\nThis paper is for people working on dark-sector phenomenology at the LHC, jet-tagging methods for displaced objects, and anyone planning Run-3 or HL-LHC extensions of nonprompt DM searches. The methods and GNN are reusable. Math, data handling and citation pattern are standard and solid. I would send it to peer review without hesitation; it is a clean experimental result that fills a genuine gap.","headline":"Solid first CMS search for boosted nonprompt DM pairs; null result with data-driven backgrounds and a reusable GNN, limited mainly by large tagger SFs.","tokens_in":38487,"tokens_out":585,"would_cite":true,"duration_ms":7333,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":["13.85.Rm","14.80.-j","95.35.+d"],"model":"grok-4.5","headline":"CMS finds no excess of boosted nonprompt dark matter in 138 fb−¹ of 13 TeV data and sets first limits on a four-prong large-radius-jet plus missing-momentum signature.","keywords":["dark matter","long-lived particles","large-radius jets","jet substructure","graph neural network","missing transverse momentum","CMS","LHC"],"falsifier":"A statistically significant excess of events in the high-missing-transverse-momentum tail of the signal region relative to the background prediction obtained from the simultaneous control-region fit, after the graph-neural-network tagger requirement has been applied.","tokens_in":38574,"feed_emoji":"⚲","tokens_out":1000,"duration_ms":7844,"temperature":0.7,"pith_summary":"This paper reports the first search for a pair of long-lived dark-sector particles that each decay into a stable dark-matter particle and a light boson that decays to quarks, all produced via a heavy mediator in association with an energetic initial-state-radiation jet. Because of the large Lorentz boost, the four quarks from the two light bosons are reconstructed as a single large-radius jet whose four-prong, potentially displaced substructure is identified by a graph-neural-network tagger. Using the full 2016–2018 CMS data set of 138 fb−¹ at 13 TeV, the analysis examines the missing-transverse-momentum spectrum after estimating the dominant multijet, W+jets and Z+jets backgrounds from dedicated control regions. No significant excess above the Standard Model expectation is observed. The result therefore places the first 95 % confidence-level upper limits on the signal strength as functions of the mediator mass or the coupling that controls the lifetime of the long-lived particles, thereby opening a previously unexplored experimental window on nonprompt dark matter in the boosted topology.","feed_headline":"No excess of boosted nonprompt dark matter at CMS","feed_subtitle":"First limits set on four-prong large-radius jets plus missing momentum with 138 fb−¹ of 13 TeV data","key_machinery":"A graph-neural-network jet tagger trained on particle-flow constituents and secondary-vertex information, which selects large-radius jets whose multiprong, potentially displaced substructure matches the four-quark decay of a boosted pair of light dark mediators; the tagger score defines the signal region and the orthogonal control regions used for data-driven background estimation.","core_discovery":"No significant excess over the Standard Model background is observed in the missing-transverse-momentum spectrum of events that contain a four-prong large-radius jet tagged by a graph neural network. Consequently the analysis sets the first 95 % CL upper limits on the production of a pair of nonprompt dark-matter candidates in the Lorentz-boosted topology, expressed as functions of either the heavy mediator mass or the coupling that sets the lifetime of the intermediate dark-sector particles.","pith_inferences":["Because the tagger performance degrades for the most displaced and heaviest signals, dedicated lifetime-binned or secondary-vertex-aware taggers could recover sensitivity in the long-lifetime regime that this analysis leaves comparatively unconstrained.","The same boosted four-prong topology could be re-examined with Run-3 data and improved secondary-vertex inputs to test whether residual data–simulation discrepancies in the Lund-plane reweighting shrink, thereby tightening the dominant systematic.","If a future excess appears in the same final state, the control-region transfer-factor method already provides a ready-made background model for a discovery claim."],"forward_implications":["The parameter space of vector/axial-vector mediators that produce long-lived dark-sector fermions decaying to stable dark matter plus a light quark-pair boson is now directly constrained for the first time in the boosted four-prong topology.","Future analyses can reuse the same GNN-plus-control-region strategy for other multiprong displaced signatures once higher-luminosity data become available.","The limits already exclude the benchmark points with the largest theoretical cross sections for the couplings and masses listed in the paper.","The result demonstrates that a single large-radius jet plus missing momentum is a viable experimental handle for nonprompt dark matter even when the intermediate particles have millimetre-to-metre lifetimes."],"fun_headline_variants":["CMS finds no excess in boosted nonprompt dark matter search","First CMS limits on nonprompt DM via four-prong large-radius jets","No signal seen for boosted nonprompt DM pairs at CMS","CMS sets first limits on nonprompt dark matter in boosted topology","GNN-tagged four-prong jets show no nonprompt DM excess at CMS"],"cache_read_input_tokens":32896,"weakest_assumption_plain":"The large (50–70 %) scale-factor uncertainties that correct the tagger efficiency from simulation to data for highly collimated, highly displaced four-prong jets correctly capture residual reconstruction differences; if those corrections are wrong the extracted signal limits shift substantially.","fun_headline_variants_meta":{"raw":{"variants":["CMS finds no excess in boosted nonprompt dark matter search","First CMS limits on nonprompt DM via four-prong large-radius jets","No signal seen for boosted nonprompt DM pairs at CMS","CMS sets first limits on nonprompt dark matter in boosted topology","GNN-tagged four-prong jets show no nonprompt DM excess at CMS"]},"model":"grok-4.5","effort":"low","cost_usd":0.00634,"raw_usage":{"total_tokens":1690,"prompt_tokens":854,"num_sources_used":0,"completion_tokens":93,"cost_in_usd_ticks":63400000,"prompt_tokens_details":{"text_tokens":854,"audio_tokens":0,"image_tokens":0,"cached_tokens":256},"completion_tokens_details":{"audio_tokens":0,"reasoning_tokens":743,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"tokens_in":854,"tokens_out":93,"duration_ms":5520,"temperature":1.0,"reasoning_tokens":743,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-07-14T07:37:33.490519+00:00","model_set":{"reader":"grok-4.5"},"falsifier":"A statistically significant excess of events in the high-missing-transverse-momentum tail of the signal region relative to the background prediction obtained from the simultaneous control-region fit, after the graph-neural-network tagger requirement has been applied.","supporting_citations":[],"review_version":1}