{"id":"b90e9b15-a985-43e7-9b57-7cd52e9c8ce8","arxiv_id":"2608.18506","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":6,"one_line_summary":"A Higgs-portal axion-like particle with mass near 13.6 GeV, decaying into a photon pair and a gluon pair, can account for the CMS low-mass diphoton excess and predicts Higgs decays into two 2-prong wide jets.","lead":"A 13.6 GeV particle that decays into two photons or two gluons could be produced in pairs from Higgs boson decays, offering an explanation for a CMS diphoton excess. The same model predicts a new Higgs decay into two wide jets with a two-prong inner structure that the LHC experiments have not yet searched for.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Rate comparison collapses if the DELPHES A×ε=12.7% overestimates CMS/ATLAS photon selection; a factor-2 efficiency drop pushes required Br(h0→AhAh) above the 12% bound.","rationale":"Good-faith reading: the paper is a candidate explanation, not a discovery claim; the model is self-consistent and the hand-derived cross sections and branching-fraction ratios (Eqs. 2.8, 2.16, 2.17, 3.9) are not the weak point. The point that makes or breaks the paper is whether the h0→AhAh→(γγ)(gg) rate is actually at the observed CMS/ATLAS level. The theoretical inputs to Eq. (3.9) are parameter-free relations (α, αs) plus the Higgs branching-fraction formula; the weak point is the conversion of a generated sample into an experimental signal yield. Section 3.2.1 replaces the CMS multivariate selection with a single pT cut and disables photon isolation, and Table 1 provides no systematic uncertainty on A×ε. A factor-of-two efficiency error is not extreme for fast simulation of nonstandard, collimated low-mass diphotons: the CMS NN score is dominated by isolation and shower-shape variables that DELPHES cannot reproduce, and the signal photons sit near the isolation-cone boundary at ΔRγγ~0.3-0.4. Because Fig. 5 and the entire phenomenological conclusion are linear in A×ε, a drop to ~6% moves the required Br(h0→AhAh) past the |κV|≤1 bound. The local-vs-global significance issue is real but secondary: a 3.5σ local excess can still be used as a candidate benchmark, provided the rate comparison is credible, and it cannot be made credible until the detector emulation is validated. Hence the reader's CONDITIONAL verdict is the right one, and no verdict change is needed.","tokens_in":14420,"tokens_out":9890,"duration_ms":106694,"concrete_test":"One check: rerun the exact DELPHES setup of §3.2.1 with default photon isolation (physical PTRatioMax for the CMS and ATLAS cards, track isolation, and photon-ID efficiency maps) instead of the disabled isolation, keeping all other cuts and the same event sample; recompute the final row of Table 1. If the CMS A×ε falls below ~0.064, or equivalently the required Br(h0→AhAh) for the CMS 1σ band exceeds 0.12, the Fig. 5 overlap is an artifact of the emulation and the central claim fails at the rate-comparison level. A stronger version of the same check is to run a Rivet/MadAnalysis5 recast of the published CMS [10] and ATLAS [11] analyses, or to use the collaborations' public efficiency maps, on the same UFO/PYTHIA8 events.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central quantitative step is §3.2.2/Fig. 5, where the claimed consistency with the CMS 3.5σ excess and the compatibility with ATLAS are read off from a predicted vertical band B(h0→AhAh→γγgg) = 2 B(h0→AhAh) B(Ah→γγ) B(Ah→gg), scaled by an acceptance-efficiency of 12.7% (CMS) and 11.9% (ATLAS) from Table 1. Every point in Fig. 5 inherits these numbers linearly. The simulation in §3.2.1 deliberately (i) emulates the CMS neural-network selection by a flat pT(γγ)>50 GeV cut, (ii) sets PTRatioMax large so calorimeter isolation rejects nothing, and (iii) approximates photon identification by a ΔRγγ cut only. No uncertainty is quoted on A×ε. The CMS NN was trained on shower-shape, isolation, and vertex variables; a flat pT cut cannot reproduce its turn-on, and disabling isolation removes a known rejection source for close photon pairs in the 13.6 GeV signal, where ΔRγγ distributions peak near the 0.3 isolation-cone boundary. If the true CMS+ATLAS efficiency is lower by as little as a factor of two (e.g., 12.7%→6.4%), the Br(h0→AhAh) required to match the 1σ CMS band increases from ~10% to ≳20%, above the 12% ATLAS |κV|≤1 bound quoted in §3.2.2, and the central explanation breaks. Since no public analysis artifacts or closure tests validate the emulation, the rate agreement is not yet established.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper constructs an effective field theory of an axion-like particle A_h coupled to the Standard Model Higgs boson through a Higgs-portal term and to new heavy colored scalars. The ALP acquires a mass from explicit breaking of a global U(1)_phi symmetry, and its loop-induced decays are predominantly to gluons with a small branching fraction to photons. The main phenomenological claim is that for M_A around 13.6 GeV, the process pp -> h0 -> A_h A_h -> (gamma gamma)(g g) yields a boosted diphoton resonance consistent with the CMS 3.5 sigma excess at 13.6 GeV while remaining compatible with the ATLAS search, provided Br(h0 -> A_h A_h) is around 10%. The paper also proposes the fully hadronic channel h0 -> A_h A_h -> 4g as a new signature of two wide jets with 2-prong substructure forming a 125 GeV resonance. The field-theory part is presented with explicit formulas for widths and the ALP mass, and the collider analysis is based on a DELPHES simulation with simplified implementations of the CMS and ATLAS photon selections.","tokens_in":14922,"tokens_out":2412,"duration_ms":24674,"significance":"If the rate comparison is correct, the paper offers a concrete and falsifiable interpretation of the CMS 13.6 GeV diphoton excess as arising from Higgs decays to ALPs, and it identifies a genuinely new search channel (Higgs to two wide jets) that has not been experimentally explored. The model-building part is economical and internally consistent: the width formulas (2.8), (2.16), and (2.17) follow standard loop computations, and the parametric relation (2.19) is a reasonable way to locate M_A at the 10 GeV scale. The central quantitative step, however, rests on a simplified detector simulation whose acceptance-efficiency is not validated against the public CMS and ATLAS analyses; this is where the main risk lies. The paper does not provide machine-checked proofs or released analysis code, but the analytic widths are simple enough to be independently checked.","major_comments":[{"comment":"The acceptance-efficiency values A*epsilon = 12.7% (CMS) and 11.9% (ATLAS) are the backbone of the rate comparison in Figure 5, yet the simulation deliberately disables photon isolation by setting PTRatioMax large and emulates the CMS neural-network selection with a flat pT(gamma gamma) > 50 GeV cut. The paper itself characterizes these choices as \"an approximation\" rather than a detailed implementation. No uncertainty is quoted on A*epsilon, and no closure test against the public CMS or ATLAS selection efficiencies is provided. If the true efficiencies are lower by a factor of two, the required Br(h0 -> A_h A_h) to match the CMS 1-sigma band rises from ~10% to ~20%, which exceeds the 12% ATLAS bound under |kappa_V| <= 1 quoted in Sec. 3.2.2 and would invalidate the central explanation. This is a load-bearing point, not a presentation issue.","section":"Sec. 3.2.1, Table 1"},{"comment":"The overlap of the CMS and ATLAS 1-sigma bands in Figure 5 is claimed to show that the two experiments are consistent with the model, but the vertical position of every band inherits the simulated A*epsilon linearly. Because the ATLAS bound Br(h0 -> undetected) < 12% is only a short distance above the central value Br(h0 -> A_h A_h) ~ 10% required to match the CMS excess, the apparent consistency is fragile. The paper should propagate a realistic systematic uncertainty on A*epsilon (including isolation, photon reconstruction, and trigger turn-on) and show how the allowed region in Figure 5 changes; without that, the rate agreement is not established.","section":"Sec. 3.2.2, Figure 5"},{"comment":"The model contains several free parameters (mu_Hphi/v_phi, Lambda_H, m_R/v_H, Lambda_tilde_u/M_tilde_u, mu_tilde_uphi/mu_Hphi, plus the colored-scalar masses and mixing parameters), and Eq. (2.19) is a parameter relation, not a prediction. This is acceptable for a phenomenological model, but the paper should state more explicitly which combinations are constrained by naturalness and which are simply chosen to place M_A at 13.6 GeV and Br(h0 -> A_h A_h) near 10%. In particular, the choice mu_Hphi/v_phi ~ 10^-2 that yields Br(h0 -> A_h A_h) ~ 10% is close to the experimental upper limit, so the paper should discuss the sensitivity of the conclusion to variations of this parameter within its natural range.","section":"Sec. 2.2, Eq. (2.19)"}],"minor_comments":[{"comment":"The phrase \"an ALPA h lighter\" in the abstract and the introduction reads as a typo; it should be \"an ALP A_h lighter.\"","section":"Abstract and Sec. 1"},{"comment":"The estimate Delta R ~ 0.32 for M_A = 10 GeV is stated as an order-of-magnitude guide, but the following sentence about the CMS and ATLAS separation cuts could benefit from a reference to the actual photon-reconstruction separation limits in the two experiments, since the 0.3 cut is close to the peak of the Delta R distribution.","section":"Sec. 3.1, Eq. (3.3)"},{"comment":"The right-hand panel of Figure 6 shows a broad M_gamma gamma Jgg peak near 125 GeV, but the caption does not specify the selection efficiency after requiring the 11 GeV < m_gamma gamma < 16 GeV window and Delta R(Jgg, gamma) > 0.8; reporting this efficiency would help the reader judge the usefulness of the Higgs-mass reconstruction for background suppression.","section":"Sec. 3.2.3, Fig. 6"},{"comment":"The statement that the ratio (2.17) is independent of the spin of the particles in the loop is not justified in the text; for scalar and fermion loops the functional dependence on the mass ratio can differ, even though the leading charge/color factors may be similar. A brief comment on the loop-function cancellation would avoid confusion.","section":"Sec. 2.2, Eq. (2.17)"},{"comment":"The paper states that varying the pT(gamma gamma) threshold between 40 and 60 GeV changes the efficiency only slightly, but no quantitative result is shown. This claim is relevant to the robustness of the central A*epsilon value, so a small table or plot would be more convincing.","section":"Sec. 3.2.1"}],"recommendation":"major_revision"},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The paper does something genuinely new: it interprets the CMS 13.6 GeV diphoton excess as h0→AA→(γγ)(gg) with A an ALP from a Higgs-portal complex scalar, and points out that the same model gives h0→AA→4g as two wide 2-prong jets. The field theory is standard but cleanly presented: the ALP mass from explicit and spontaneous U(1) breaking, loop decays via TeV-scale colored scalars, and the γγ/gg branching ratio around 0.15%. I see no load-bearing error in the analytic part. The authors are also straight about what is approximate and what is tuned.\n\nThe strength is the physics case for the mixed γγ+gg channel and the explicit suggestion to search for Higgs decays into wide jets. That is a well-motivated, possibly important signature, independent of the current excess.\n\nThe soft spot is exactly where the reader puts it. The claimed consistency with CMS and ATLAS in Fig. 5 depends on an acceptance-efficiency of 12.7% from a DELPHES emulation that (i) replaces the CMS neural-network selection by a pT(γγ)>50 GeV cut, (ii) sets PTRatioMax so isolation rejects nothing, and (iii) approximates photon ID by a ΔRγγ cut. No uncertainty is quoted, and no closure tests or public artifacts are provided. Because the needed Br(h→AA) sits just under the ATLAS |κV|≤1 bound of 12%, a factor-two drop in efficiency would push the central explanation above the bound. That is a real problem, not a minor quibble. I also think the paper underplays the local vs global significance of the 3.5σ excess; in the 10–70 GeV range, the look-elsewhere effect is substantial, and the mass is chosen after the fact.\n\nThat said, the paper is not sloppy. The authors are transparent about their approximations. The wide-jet signature is a concrete, new target that could be useful even if the CMS excess evaporates.\n\nBottom line: send it to referees. It deserves careful review, and the efficiency question is exactly what a referee should push on. If the simulation cannot be made more realistic or conservative, the excess interpretation should be softened to a possible explanation; the wide-jet proposal stands on its own.","headline":"A plausible ALP explanation of the CMS 13.6 GeV excess with a new wide-jet Higgs decay, undercut by an unvalidated detector simulation that could break the rate claim.","tokens_in":15429,"tokens_out":4097,"duration_ms":41766},"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"An axion-like particle of mass near 13.6 GeV, produced in pairs from Higgs decay and decaying one to two photons and one to two gluons, can explain the reported diphoton excess while staying consistent with the other search.","keywords":["axion-like particle","Higgs portal","diphoton resonance","13.6 GeV excess","wide jets","pseudo-Nambu-Goldstone boson","nonstandard Higgs decay","gluon fusion Higgs production"],"falsifier":"Using the real selection algorithms on simulated $h^0\\to A_h A_h\\to(\\gamma\\gamma)(gg)$ events would settle it: if the measured acceptance times efficiency is below about 6%, the branching fraction required to reproduce the 13.6 GeV excess exceeds the 12% undetected-Higgs bound under $|\\kappa_V|\\le1$. Alternatively, a dedicated reconstruction of the $\\gamma\\gamma$-plus-wide-jet invariant mass in the current data that excludes a 125 GeV peak at the predicted rate would rule out the explanation.","tokens_in":14226,"feed_emoji":"⚛️","tokens_out":13713,"duration_ms":130472,"temperature":0.7,"pith_summary":"This paper proposes that a 3.5σ excess in low-mass diphoton data at 13.6 GeV is not a single new particle produced alone, but an axion-like particle (ALP) that is produced in pairs from the decay of the 125 GeV Higgs boson. The model gives the ALP a mass of order 10 GeV through explicit breaking of a global symmetry, and loop effects from new colored, electrically charged scalars make it decay almost always to two gluons, with a much smaller but observable branch to two photons. When one ALP decays to two photons and the other to two gluons, the photon pair is automatically boosted, which is why the signal passes the tight kinematics of existing low-mass diphoton searches. With a Higgs branching fraction into ALP pairs of about 10%, the predicted rate matches the 13.6 GeV excess and stays compatible with the milder excess and limits from the other search, while remaining under the 12% bound on undetected Higgs decays. Independently of that excess, the same cascade gives a Higgs decay into four gluons, which would appear as two wide jets with two-prong substructure forming a 125 GeV resonance, a signature no current search has looked for.","feed_headline":"A new axion-like particle explains the 13.6 GeV diphoton excess","feed_subtitle":"One ALP decays to photons, the other to gluons; the same cascade predicts a new wide-jet Higgs decay.","key_machinery":"The load-bearing object is the axion-like particle $A_h$, a pseudo-Nambu-Goldstone boson whose mass near 10 GeV arises from three contributions to $M_A^2$: a tree-level Higgs-portal term, a quadratically divergent one-loop term regulated at an effective cut-off near 1 TeV, and a logarithmically divergent term from the new colored scalars. What carries the argument is the pair decay $h^0\\to A_h A_h$ followed by the one-loop decays $A_h\\to gg$ and $A_h\\to\\gamma\\gamma$, whose relative rate is fixed by the loop-particle charges through $\\Gamma(A_h\\to\\gamma\\gamma)/\\Gamma(A_h\\to gg)=8\\alpha^2/(9\\alpha_s^2)\\simeq1.5\\times10^{-3}$. Because each $A_h$ is boosted with $\\gamma_{A_h}\\simeq6.3\\,(10\\,\\mathrm{GeV}/M_A)$, the two photons from one ALP remain separated enough to pass low-mass diphoton triggers, while the two gluons from the other merge into a single wide jet with 2-prong substructure; this kinematic split is the mechanism that links the diphoton excess to the new wide-jet signature.","core_discovery":"On the paper's own terms, the central claim is that a light ALP $A_h$ with the Higgs-portal coupling of Eq. (2.6) and the loop-induced couplings of Eq. (2.15) is a viable explanation of the observed 13.6 GeV diphoton excess: for $M_A\\approx13.6$ GeV and $Br(h^0\\to A_h A_h)\\approx10\\%$, the $h^0\\to A_h A_h\\to(\\gamma\\gamma)(gg)$ cascade yields a boosted diphoton resonance at $M_A$ with an acceptance times efficiency of 12.7% under one experiment's selection and 11.9% under the other's, putting the predicted rate inside the 1σ bands of both searches. The same model predicts a hadronic cascade $h^0\\to A_h A_h\\to 4g$ with branching fraction of order 10%, in which the two gluons from each ALP merge into a wide jet with 2-prong substructure and the jet pair reconstructs a 125 GeV Higgs; this is a new, unsearched nonstandard Higgs decay.","pith_inferences":["If the excess is confirmed as $h^0\\to A_h A_h\\to(\\gamma\\gamma)(gg)$, the companion all-hadronic decay $h^0\\to A_h A_h\\to 4g$ has a much larger rate, so a future wide-jet search that finds nothing at 125 GeV would pressure the same parameter space even if the diphoton signal is reproduced.","Because the ratio $Br(A_h\\to\\gamma\\gamma)/Br(A_h\\to gg)=8\\alpha^2/(9\\alpha_s^2)$ is independent of the loop particles' spin and weak representation, the same ALP could be searched for in other production modes, and the 13.6 GeV excess may imply additional signals at masses and rates the paper does not compute.","The model's consistency argument assumes the $|\\kappa_V|\\le1$ constraint on the Higgs couplings to W and Z; if future global fits allow larger values, the required $A_h$ branching could be accommodated more easily, making the explanation harder to falsify."],"forward_implications":["The 13.6 GeV excess, if real, is evidence for the nonstandard Higgs decay $h^0\\to A_h A_h$ rather than for a singly produced ALP, so the apparent tension between the two collider diphoton searches is resolved by the presence of the hadronic branch.","Combining the diphoton with the wide jet from the second ALP should produce a second resonance near 125 GeV, giving a nearly background-free confirmation channel for the explanation.","The model predicts $h^0\\to A_h A_h\\to 4g$ at the percent level, visible as two wide 2-prong jets; a dedicated search for a 125 GeV resonance in this topology could discover it or exclude the mechanism.","Improving the measurement of undetected Higgs branching fractions beyond the current 12% bound will directly probe the required $Br(h^0\\to A_h A_h)\\sim10\\%$, so near-future Higgs coupling measurements can confirm or rule out this parameter region.","Run 3 and High-Luminosity collider data multiply the number of Higgs bosons by a factor of a few, making both the photonic and fully hadronic $A_h$ cascade searches statistically sensitive at the relevant branching fractions."],"supporting_citations":[{"why":"The low-mass diphoton search that reports the 3.5σ excess at 13.6 GeV that the paper sets out to explain.","marker":"[10]"},{"why":"The other experiment's boosted low-mass diphoton search whose milder excess and limits constrain the proposed signal.","marker":"[11]"},{"why":"Establishes the $h^0\\to A_hA_h$ decay and the loop-induced ALP couplings to gluons and photons used as the model's skeleton.","marker":"[4]"},{"why":"Provides the $|\\kappa_V|\\le1$ assumption that converts measured Higgs rates into an upper bound on undetected decay modes.","marker":"[29]"},{"why":"Supplies the 12% upper limit on undetected Higgs branching fractions used to constrain $Br(h^0\\to A_hA_h)$.","marker":"[30]"},{"why":"Supplies the combined Higgs measurements whose looser limits provide the comparison point.","marker":"[31]"},{"why":"Provides the gluon-fusion Higgs production cross section used to derive expected signal rates.","marker":"[21]"},{"why":"Provides the Standard Model Higgs width and production inputs for the rate calculation.","marker":"[19]"}],"fun_headline_variants":["ALP explains 13.6 GeV diphoton excess","Axion-like particle links Higgs to diphoton excess","ALP decay solves 13.6 GeV anomaly, predicts wide jets","Light ALP from Higgs decay: diphoton and dijet signatures","ALP cascade: 13.6 GeV diphoton, 125 GeV wide dijet"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The argument's quantitative anchor is the assumption that the simplified emulation of the two collider event selections, which replaces the multivariate selection by the single requirement that the photon-pair transverse momentum exceed 50 GeV and turns off photon isolation, correctly predicts that about 12.7% of signal events pass; if the true acceptance is even a factor of two lower, the required Higgs branching fraction into ALPs exceeds the 12% upper bound on undetected Higgs decays, and the explanation fails.","fun_headline_variants_meta":{"raw":{"variants":["ALP explains 13.6 GeV diphoton excess","Axion-like particle links Higgs to diphoton excess","ALP decay solves 13.6 GeV anomaly, predicts wide jets","Light ALP from Higgs decay: diphoton and dijet signatures","ALP cascade: 13.6 GeV diphoton, 125 GeV wide dijet"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000933,"raw_usage":{"total_tokens":4033,"prompt_tokens":1023,"completion_tokens":3010,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":639,"completion_tokens_details":{"reasoning_tokens":2924}},"tokens_in":639,"tokens_out":3010,"duration_ms":22376,"temperature":1.0,"reasoning_tokens":2924,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-27T19:56:47.408462+00:00","model_set":{"reader":"deepseek-v4-flash"},"pipeline_version":"pith-pipeline@v1.4.0-alltime-deepseek-medium","input_kind":null,"input_chars":null,"input_total_chars":null,"passes_missing":null,"published_standing":null,"published_confidence":null,"objection_audit":null,"run_record":null,"basis":{"kind":"not_recorded","source":null,"chars_given":null,"chars_total":null},"receipt":null,"falsifier":"Using the real selection algorithms on simulated $h^0\\to A_h A_h\\to(\\gamma\\gamma)(gg)$ events would settle it: if the measured acceptance times efficiency is below about 6%, the branching fraction required to reproduce the 13.6 GeV excess exceeds the 12% undetected-Higgs bound under $|\\kappa_V|\\le1$. Alternatively, a dedicated reconstruction of the $\\gamma\\gamma$-plus-wide-jet invariant mass in the current data that excludes a 125 GeV peak at the predicted rate would rule out the explanation.","supporting_citations":[{"cited_title":"Search for a narrow resonance with a mass between 10 and 70 GeV decaying to a pair of photons in proton-proton collisions at √s= 13 TeV,","cited_arxiv_id":"2603.03250","evidence_quote":"The low-mass diphoton search that reports the 3.5σ excess at 13.6 GeV that the paper sets out to explain."},{"cited_title":"Higgs boson decays to CP odd scalars at the Tevatron and beyond,","cited_arxiv_id":null,"evidence_quote":"Establishes the $h^0\\to A_hA_h$ decay and the loop-induced ALP couplings to gluons and photons used as the model's skeleton."},{"cited_title":"Coupling spans of the Higgs-like boson","cited_arxiv_id":"1210.3342","evidence_quote":"Provides the $|\\kappa_V|\\le1$ assumption that converts measured Higgs rates into an upper bound on undetected decay modes."},{"cited_title":"Review of particle physics,","cited_arxiv_id":null,"evidence_quote":"Provides the Standard Model Higgs width and production inputs for the rate calculation."}],"lane_accuracy":{"status":"not yet measured","lane":{"model":"deepseek-v4-flash","input_kind":null},"url":"https://pith.science/accuracy"},"review_version":1}