{"id":"1c016e9c-b4f7-4ef6-9f05-8f4b11856a86","arxiv_id":"2506.21676","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"A combined fit to four LHC Run 2 searches prefers a wino-bino supersymmetric spectrum with m(wino) ~ 320 GeV and mass splitting ~ 20 GeV, compatible with bino dark matter.","lead":"A new combination of four LHC searches favors a supersymmetric 'wino-bino' spectrum with a wino-like particle around 320 GeV and a bino dark matter candidate about 20 GeV lighter. It ties together mild excesses in soft-lepton and monojet data and points to a region Run 3 can test.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The headline best-fit point (320,20) with B10=7.3 depends on a post-hoc modified simplified likelihood for CMS-SUS-18-004; under the full likelihood the CMS soft-lepton B10 for that point is 0.85, so the joint-preference claim is not yet established.","rationale":"The reader's weakest assumption is exactly where I find the decisive soft spot. The paper's most striking claims—B10=7.3 at (320,20) and the statement that all four analyses favour the wino-bino region—depend on the modified CMS simplified likelihood. With the full likelihood, Table 10 gives CMS soft-lepton B10=0.85 for that point, so the 'joint' preference is carried by the ATLAS and monojet channels. The bin-merging was motivated by the observed-limit discrepancy and is not independently validated; the authors' own suggestion of a COMBINE model is the appropriate resolution. I considered whether the non-Bayesian B10 statistic or the DM-band mu-dependence could be more load-bearing, but those are acknowledged limitations and do not change the LHC-only claim as much as the modified likelihood does. The recasts themselves are valuable and the nearby (330,22) point remains preferred even with the full likelihood, so this is a reason for a conditional verdict, not a rejection. A single COMBINE-based recomputation would settle the question. Therefore I keep the reader's CONDITIONAL verdict.","tokens_in":26012,"tokens_out":6170,"duration_ms":68822,"concrete_test":"Obtain the official CMS COMBINE model for CMS-SUS-18-004 (or, if unavailable, a dense set of official benchmark cutflows covering m(chi2~0)=280–340 GeV and Delta m=10–20 GeV) and recompute the combined B10 map using that model with all original signal and control regions and no ad hoc bin merging. If the (320,20) point then has CMS soft-lepton B10<1 or the combined B10 falls below the full-likelihood value of 4.9, the headline preference is an artifact of the modified likelihood.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central quantitative claim—preferred region around m(chi2~0)=320 GeV, Delta m=20 GeV, combined B10=7.3—rests on the modified simplified likelihood introduced in Sec. 2.3 for CMS-SUS-18-004. With the full covariance-matrix simplified likelihood, the recast does not reproduce the published observed exclusion limit for Delta m in [5,20] GeV, and the authors acknowledge 'there could be a subtle error in our recast' (Sec. 2.3). To repair this, they merge the Ultra and High m_ll in [4,10] GeV bins and the 3-lepton m_ll in [1,4] and [4,10] GeV bins, a choice made after seeing the discrepancy. In Table 10, the point (320,20) has CMS soft-lepton B10=0.85 with the full likelihood, but B10=1.26 with the modified likelihood; the combined B10 rises from 4.90 to 7.3. Thus the claim that all four analyses jointly favour this point, and the 'most likely' mass extraction, depend on an ad hoc and unvalidated statistical model. The paper itself proposes a COMBINE model as the resolution; until then, the magnitude and precise location of the claimed excess are not robust.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This manuscript presents new public recasts of the CMS soft-lepton search CMS-SUS-18-004 (in HackAnalysis) and the ATLAS multijet search ATLAS-SUSY-2018-22 (in MadAnalysis 5), validates them against official cutflows and exclusion contours, and combines them with earlier recasts of ATLAS soft-lepton and CMS monojet analyses. For the wino-bino(+) simplified model the authors find a region around m(chi_2^0) ~ 320-340 GeV and Delta m ~ 20 GeV in which the data are better described by the model plus background than by background alone, with combined profile-likelihood ratios B10 ~ 4.5-7.3. They further show that the same region lies near the band where a nearly pure bino LSP has the observed relic abundance, and they discuss how the simplified model can be embedded in the MSSM. The paper also constrains a frustrated dark matter model with the new ATLAS multijet recast.","tokens_in":26336,"tokens_out":7808,"duration_ms":89490,"significance":"If the claimed preference is robust, the paper would provide a coherent, testable explanation of several mild LHC excesses and a dark-matter candidate, and it would supply the community with two validated recasting tools, with inputs deposited on Zenodo. The validations against official cutflows and exclusion curves are generally good, and the relic-density calculation is performed independently of the LHC fit. However, the headline numerical result depends on a post-hoc modification of the CMS simplified likelihood, and the full-likelihood result is weaker; the paper's own text acknowledges a possible recast error. The significance is therefore conditional on resolving this statistical-model issue.","major_comments":[{"comment":"The central claim that the wino mass is 'most likely' around 320 GeV with Delta m ~ 20 GeV and combined B10 = 7.3 rests on the modified simplified likelihood introduced in Section 2.3, in which two pairs of CMS-SUS-18-004 bins are merged. This modification was made after the full simplified likelihood failed to reproduce the published CMS observed limit for Delta m in [5,20] GeV, and the authors state that 'there could be a subtle error in our recast.' With the full likelihood, Table 10 gives CMS soft-lepton B10 = 0.85 and combined B10 = 4.90 for (320,20), so the exact best-fit point and the magnitude 7.3 are not supported by the primary statistical model. I request either an independent validation of the modified likelihood (for example, reproduction of the official observed limit for additional benchmark points, or use of a COMBINE model), or a presentation that makes the full-likelihood results primary and treats the modified likelihood only as a cross-check.","section":"2.3, Table 10, Figure 5"},{"comment":"The dark-matter compatibility, called the 'most striking result,' depends on the position of the relic-density band, but the paper reports in Section 4.4 that the band shifts by 6-7 GeV in Delta m when mu is raised from 500 GeV to 2 TeV, and that the simplified-model samples are generated without true mixing matrices ('in some sense, a fake model'). Because the LHC-preferred region is narrow in Delta m, please state the exact (mu, At) values used for the band in Figure 6, quantify the uncertainty of the band location, and show explicitly whether the full preferred region remains within the band over the range of parameters allowed by direct detection. Without this, the compatibility is not yet quantitatively established.","section":"4.4, Figure 6"}],"minor_comments":[{"comment":"The bullet for 'Higgsino180/10' lists (m_chi2^0, Delta m) = (160,3) GeV; this should be (180,10) GeV.","section":"2.2"},{"comment":"Although the text disclaims the terminology, referring to a profile likelihood ratio as a 'Bayes factor' and quoting B10 = 7.3 without a prior will be confusing to readers; consider renaming it or adding a prominent caveat in the abstract and conclusions.","section":"4.1, Eq. (4.1)"},{"comment":"Several MB-GGd bins show differences larger than 50% (e.g., SR4 340010cuts -187.77% and SR4 400022cuts 73.22%); since the exclusion contour agrees well, please comment on whether these outliers affect the CFDM limits in Figure 3.","section":"Table 8"},{"comment":"The p-values listed are local and not corrected for the grid scan; a sentence noting this would prevent readers from interpreting them as global significances.","section":"4.3, Table 10"},{"comment":"The text says 'the data favours a region around m_chi2^0 ~ 330 GeV and Delta m ~ 20 GeV' while the conclusions say 'around 320 GeV'; please harmonize these statements and tie them explicitly to the full-likelihood results.","section":"4.2"},{"comment":"Please clarify the statistical independence of the two ATLAS soft-lepton analyses combined in one column of Table 10, since the combined B10 is a product of their individual factors.","section":"4.3, Table 10"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is a follow-up in a series by the same group, and the central numerical claim relies on their own recasting chain. The paper is transparent about the unresolved discrepancy, which is a point in its favor, but the discrepancy is in the very analysis that drives the headline. I see this as fixable by reprioritizing the full-likelihood results and validating the modified likelihood, hence major revision rather than reject."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"First thing you should know: this paper provides the first public recast of CMS-SUS-18-004 and combines four Run 2 analyses in the wino-bino(+) model, with careful documentation and a Zenodo release. The central claim is a preferred region near m(chi2~0) ~ 320-330 GeV, Delta m ~ 20 GeV, overlapping the bino dark-matter relic band. That is new and worth engaging.\n\nWhat the paper does well: the recasting work is the core contribution. The CMS soft-lepton recast had no public predecessor, and the ATLAS multijet recast is carefully validated against official cutflows and exclusion contours. The authors are transparent about where their recast fails to reproduce the observed CMS limit for moderate Delta m, and they discuss a possible overspill between bins. The DM relic-density band is computed independently with MSSM scans using SPheno and micrOMEGAs; it is not fitted to the LHC data. The discussion of how the simplified model maps onto a realistic MSSM, with the dependence on mu and the higgsino admixture, is honest and useful.\n\nThe soft spot is that the headlining point depends on a post-hoc statistical fix. To recover the observed CMS limit, they merge two pairs of bins in the simplified likelihood. That is a legitimate robustness check, but the (320,20) best-fit point with combined B10=7.3 appears only with the modified likelihood. Under the full likelihood, the CMS soft-lepton B10 at that point is 0.85 and the combined B10 is 4.90. There is a nearby point, (330,22), with all individual factors above 1 even with the full likelihood, so the region is likely real, but its precise location and the size of the preference are not stable. Also, B10 is called a Bayes factor but it is actually a profiled likelihood ratio; the authors acknowledge this, but readers will take 7.3 at face value.\n\nI would send this to peer review. The recasts are valuable to the community regardless of whether the joint-excess interpretation survives; the paper is honest about its limitations and names the concrete follow-up needed (a COMBINE model from CMS). A referee should focus on the bin-merging statistics and the independence of the four likelihoods before the preferred region is treated as settled.","headline":"Useful new recasts and a coherent wino-bino excess story, but the headline B10 depends on a post-hoc bin-merging fix for CMS-SUS-18-004, so treat the (320,20) preferred point as tentative.","tokens_in":26893,"tokens_out":2709,"would_cite":true,"duration_ms":29174,"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":"Soft-lepton and monojet excesses at the LHC are jointly consistent with a wino-bino supersymmetric spectrum in which a wino-like neutralino pair sits about 20 GeV above a nearly pure bino dark matter candidate.","keywords":["supersymmetry","electroweakinos","wino-bino model","compressed spectra","LHC excesses","soft leptons","monojet search","dark matter relic density"],"falsifier":"If the CMS collaboration releases a full statistical model and it shows that the point $(m_{\\tilde\\chi_2^0},\\Delta m)=(320,20)$ GeV is excluded, or if Run 3 data show the soft-lepton and monojet excesses fading to background, the claimed preference would be refuted.","tokens_in":25784,"feed_emoji":"🌌","tokens_out":5776,"duration_ms":58222,"temperature":0.7,"pith_summary":"This paper argues that several modest excesses seen in LHC Run 2 data are not independent accidents but overlapping views of the same compressed supersymmetric spectrum. By recasting the CMS soft-lepton analysis, which previously had no public recast, and combining it with the ATLAS soft-lepton searches and the CMS monojet search, the authors find that a wino-bino(+) parameter region around a second-neutralino mass of roughly 330 GeV and a mass split of about 20 GeV fits the combined data better than the Standard Model. The paper reports combined Bayes factors up to 7.3 for the alternative likelihood. The same region passes close to the band where a nearly pure bino freezes out with the observed dark matter relic abundance, so the excesses, if real, would point to the mass and properties of the dark matter particle.","feed_headline":"Four LHC searches point to a 320-GeV dark matter candidate","feed_subtitle":"Soft-lepton and monojet excesses together favour a wino-bino spectrum whose bino has the observed relic abundance","key_machinery":"The argument rests on a full recast of the CMS-SUS-18-004 soft-lepton search built from the provided simplified likelihood, covariance matrices and cutflows; on a Bayes factor $B_{10}$, defined as the ratio of profiled signal-plus-background to background-only likelihoods, which directly shows where the model beats the Standard Model; and on an alternative simplified likelihood that merges two pairs of $m_{\\ell\\ell}$ bins, correcting an observed-limit discrepancy. This modified likelihood is the one that produces the headline $B_{10}=7.3$ at $(320,20)$ GeV, while the unmodified likelihood gives a combined $B_{10}=4.9$ at the same point.","core_discovery":"In the wino-bino(+) simplified model, an MSSM limit with decoupled higgsinos where only two parameters matter, the heavier neutralino mass $m_{\\tilde\\chi_2^0}$ and the splitting $\\Delta m$, the paper claims the data favour $m_{\\tilde\\chi_2^0} \\sim 320\\text{--}330$ GeV and $\\Delta m \\sim 20$ GeV over background-only. At those values all four analyses have favouring Bayes factors $B_{10} > 1$. The preferred point overlaps the band in which bino-wino coannihilation during freeze-out gives $\\Omega h^2 \\simeq 0.12$, making the bino LSP a viable thermal dark matter candidate. The authors present this as the first time combined compressed-electroweakino and monojet data point to a specific dark matter mass and its properties.","pith_inferences":["If the excesses are real, the small mass splitting implies the chargino may travel a detectable distance before decaying, so displaced-track or disappearing-chargino searches are an independent corroboration the paper does not perform.","A full statistical model from CMS, with all control regions and correlations, would settle whether the bin-merging likelihood is the right description; the paper itself asks for exactly this input.","The dark matter band shifts by about 6-7 GeV in $\\Delta m$ as the higgsino decoupling scale $\\mu$ is raised, so mapping the simplified wino-bino scenario onto a complete MSSM moves the predicted target; this mapping sensitivity is an implicit caveat.","The same recast machinery could be applied to other compressed electroweakino templates that may fit the combined data even better, a possibility the paper notes but does not pursue."],"forward_implications":["Run 3 searches in soft-lepton plus missing-energy and monojet channels should accumulate more events near $m_{\\tilde\\chi_2^0} \\sim 320$ GeV and $\\Delta m \\sim 20$ GeV if the signal is real.","The multijet recast of ATLAS-SUSY-2018-22 gives only weak constraints on the wino-bino region, so it does not undermine the excess interpretation; it excludes bino LSP masses below roughly 170 GeV.","A nearly pure bino with mass around $m_{\\tilde\\chi_1^0} \\sim 300$ GeV coannihilating with winos can simultaneously explain the LHC excesses and the Planck relic abundance, assuming standard freeze-out cosmology.","The compatibility of the CMS excess region with the ATLAS soft-lepton and monojet regions is demonstrated by overlapping $B_{10} > 1$ contours in the $m_{\\tilde\\chi_2^0}$ versus $\\Delta m$ plane."],"supporting_citations":[{"why":"Supplies the CMS soft-lepton search data, cutflows, covariance matrix, and simplified likelihood on which the new recast is built and validated.","marker":"[4]"},{"why":"Provides one of the ATLAS soft-lepton compressed-spectra analyses whose excess enters the combination.","marker":"[2]"},{"why":"Provides the other ATLAS soft-lepton search and the benchmark wino-bino parameter cards used for event generation.","marker":"[3]"},{"why":"Supplies the CMS monojet search data whose excess is combined with the soft-lepton ones.","marker":"[6]"},{"why":"Predecessor work establishing that monojet and soft-lepton excesses can arise from the same compressed models.","marker":"[7]"},{"why":"Previous combined analysis of these excesses that the present work extends by adding the CMS soft-lepton channel.","marker":"[8]"},{"why":"Defines the simplified-likelihood bin aggregation procedure used to construct the alternative CMS likelihood.","marker":"[23]"},{"why":"Introduces the Bayes-factor-surface approach used to quantify preference over the Standard Model.","marker":"[56]"},{"why":"Provides the Planck relic abundance value $\\Omega h^2 \\approx 0.12$ used to draw the coannihilation band.","marker":"[14]"},{"why":"Computes bino-wino coannihilation relic densities and underlies the dark matter band plotted against the LHC fit.","marker":"[68]"}],"fun_headline_variants":["Four LHC searches converge on a 320-GeV wino-bino dark matter","Wino-bino model unifies soft-lepton and monojet excesses","Soft-lepton + monojet excesses explained by 320 GeV wino-bino dark matter","Four LHC excesses favor 320 GeV wino-bino dark matter candidate","320 GeV wino-bino dark matter unifies LHC excesses"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the modified CMS simplified likelihood, created by merging the Ultra and High $m_{\\ell\\ell}$ bins and two three-lepton low-mass bins, correctly captures the experimental correlations that the unmodified simplified likelihood fails to reproduce; the paper itself notes that the unmodified likelihood does not match the CMS observed limit and acknowledges a subtle error in the recast cannot be excluded.","fun_headline_variants_meta":{"raw":{"variants":["Four LHC searches converge on a 320-GeV wino-bino dark matter","Wino-bino model unifies soft-lepton and monojet excesses","Soft-lepton + monojet excesses explained by 320 GeV wino-bino dark matter","Four LHC excesses favor 320 GeV wino-bino dark matter candidate","320 GeV wino-bino dark matter unifies LHC excesses"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000941,"raw_usage":{"total_tokens":4004,"prompt_tokens":908,"completion_tokens":3096,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":524,"completion_tokens_details":{"reasoning_tokens":2987}},"tokens_in":524,"tokens_out":3096,"duration_ms":20861,"temperature":1.0,"reasoning_tokens":2987,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T22:21:21.374331+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"If the CMS collaboration releases a full statistical model and it shows that the point $(m_{\\tilde\\chi_2^0},\\Delta m)=(320,20)$ GeV is excluded, or if Run 3 data show the soft-lepton and monojet excesses fading to background, the claimed preference would be refuted.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Defines the simplified-likelihood bin aggregation procedure used to construct the alternative CMS likelihood."}],"review_version":1}