{"id":"aff4c974-3f0d-46ec-87fe-27792e02e7e1","arxiv_id":"2412.03244","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"The five- and six-loop clustering logarithm coefficients for the kt jet algorithm are computed for the first time, and they are small enough to leave the four-loop result unchanged.","lead":"A new calculation pushes the perturbative series for 'clustering logarithms' in jet mass distributions to six loops, two orders beyond the previous frontier. The new five- and six-loop terms are very small, so practitioners can trust lower-order predictions and the series appears to converge quickly.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"New F5/F6 coefficients rest on an unverified master formula, while the paper's own factorization pattern omits required reducible terms and the MC benchmark is insensitive to the new orders.","rationale":"The reader's weakest assumption is that the master formula from Ref. [1] is correct; I agree that this is the primary unverified input for F5 and F6. My stress test adds two sharper observations beyond that: (i) the all-orders MC comparison is not a test of the new coefficients because the five- and six-loop curves are indistinguishable from the four-loop curve, so the quoted agreement with Monte Carlo does not validate F5 or F6; and (ii) the paper's own factorization equations are internally inconsistent — Eq. (40) omits (1/2) Σ1 (Σ2)^2 and Eq. (43) omits Σ1 Σ2 Σ3 — even though the corresponding cluster partitions appear in the integrand decomposition of Eq. (39). These omissions do not prove F5 and F6 are numerically wrong, since those coefficients are defined through the connected cluster integrals, but they reveal that the claimed exponentiation pattern is not checked in the text. The appropriate verdict remains CONDITIONAL / UNCHANGED: the paper is a plausible extension with lower-order results that reproduce earlier work, but the decisive five- and six-loop numbers are not independently verified and the presentation contains concrete internal slips that should be corrected before full acceptance. I therefore do not move the verdict; the conditional status already reflects this risk.","tokens_in":13585,"tokens_out":16242,"duration_ms":142138,"concrete_test":"Independently evaluate F5 and F6 by implementing the master formula of Ref. [1], or by a fixed-order Monte Carlo for primary abelian emissions with the kt algorithm under the same strong-energy-ordering and single-jet-mass setup, and compare F5(R) and F6(R) with Fig. 1 for R = 0.7 and R = 1.0. In the same run, verify that the full five-loop result decomposes as in Eq. (40) once the omitted (1/2) Σ1 (Σ2)^2 term is included; if the identities fail, the exponentiation pattern is mis-stated.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central coefficients F5(R) ≈ −0.013 and F6(R) ≈ +0.010 are taken directly from the author's master formula [1] and from Mathematica notebook integrals that are not displayed in the paper. No independent derivation or cross-check of these integrands is given, so a mistake in applying the master formula would propagate directly into the headline results. The external all-orders comparison cannot rescue this: as the paper itself notes in Sec. V, the five- and six-loop curves coincide with the four-loop curve, so the Monte Carlo comparison is essentially insensitive to the exact values of F5 and F6. There is also an internal inconsistency in the claimed exponentiation pattern. The five-loop factorization, Eq. (40), omits the reducible term (1/2) Σ1 (Σ2)^2 that is required by the cluster expansion whose integrand structure appears in Eq. (39) and that is part of the full exp(Σ1) C(t) expansion. Likewise, Eq. (43) omits the Σ1 Σ2 Σ3 term required at six loops. These omissions do not by themselves change the connected-cluster integrals defining F5 and F6, but they show that the paper's factorization statements are not self-consistently verified and cast doubt on the claimed pattern of exponentiation. The decisive input for the new orders therefore remains unvalidated.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript computes the leading clustering logarithms (abelian non-global logarithms) for the kt jet algorithm through six loops in the eikonal, strong-energy-ordered approximation, using a master formula for the analytic structure of kt clustering introduced in the author's companion paper Ref. [1]. After recapping the two- through four-loop results for the single-jet-mass observable, it presents new five- and six-loop coefficients F5(R) ≈ −0.013 and F6(R) ≈ +0.010, claims a factorization pattern reminiscent of an exponential, and proposes the all-orders form C(t) = exp[Σ_{n≥2} (−1)^n/n! F_n(R) (2 C_F t)^n]. The resummed and truncated forms are compared with the Monte Carlo program of Dasgupta and Salam, with percent-level agreement reported. The paper concludes that the CL series converges rapidly and that four-loop accuracy is sufficient phenomenologically.","tokens_in":13902,"tokens_out":7899,"duration_ms":75576,"significance":"If the new five- and six-loop results are correct, the leading clustering logarithms for the kt algorithm are indeed rapidly convergent, and the existing four-loop calculations are enough for phenomenology. The two- through four-loop recap reproduces previously published coefficients, and the comparison with Monte Carlo is a useful convergence check. The principal significance of the paper, however, lies in the first-ever five- and six-loop coefficients. Because those coefficients are not displayed in the manuscript and are relegated to an external Mathematica file, the central new content is not verifiable from the paper itself. The paper also does not rederive or independently test the master formula of Ref. [1] on which the higher-loop integrands rest. The exponentiation claim is weakened by an apparent incompleteness in the reducible terms of the factorized cross-section.","major_comments":[{"comment":"The five- and six-loop irreducible clustering functions Ξ^kt_12345 and Ξ^kt_1...6 are not presented; the text states only that they are provided in an accompanying Mathematica notebook file 'Xi5-Xi6.nb', which is not part of the manuscript. Consequently the numerical coefficients F5(R) and F6(R) quoted near Eqs. (42) and (45) and plotted in Fig. 1 cannot be checked from the paper. Since these coefficients are the sole new quantitative results, the manuscript should either include the notebook as supplementary material or display the integrands explicitly with sufficient detail (including the numerical integration method, grid sizes, and estimated uncertainties) to allow reproduction.","section":"Secs. III and IV, Eqs. (38)–(45)"},{"comment":"The claimed factorization pattern is not consistent with the cluster expansion of the lower-order clustering functions. At five loops, Eq. (39) contains the reducible contribution Ξ^kt_ij Ξ^kt_kl Θ^in_m, which upon integration and symmetrization produces a term proportional to Σ1 (Σ2)^2. This term is missing from Eq. (40). At six loops, Eq. (43) similarly omits the Σ1 Σ2 Σ3 term that is part of the exponential expansion of the connected-cluster series. The paper should either correct the reducible terms in Eqs. (40) and (43) or explain why those terms are excluded from the stated factorization. As written, the claimed pattern of exponentiation is not self-consistently verified.","section":"Sec. III, Eq. (40) and Sec. IV, Eq. (43)"},{"comment":"The all-orders exponential C(t) is defined so that its expansion reproduces the fixed-order coefficients F_n; it is therefore a repackaging of the computed F_n rather than a predictive resummation obtained from an evolution equation. The paper's own Monte Carlo comparison, Fig. 2 and the associated text, states that the three-, four-, five-, and six-loop curves coincide, so the comparison is insensitive to the precise values of F5 and F6. This means that the new coefficients are not independently validated by the numerical comparison. To strengthen the central claim, the author should either derive the exponent from an evolution or recurrence, or provide a direct MC-based extraction of F5 and F6 (for example, from logarithmic moments of the CL form factor).","section":"Sec. V, Eq. (46)"},{"comment":"The five- and six-loop integrands are taken verbatim from the master formula of Ref. [1] by the same author, and this master formula is not rederived or independently tested in the present manuscript. A mistake in that formula would propagate directly into the new F5 and F6. The lower-loop recap in Sec. II provides some validation for two through four loops, but it does not test the master formula's predictions at five and six loops. Please include a self-contained derivation or at least a nontrivial consistency check at the new orders (for example, a symmetry or limiting-behavior test of the five- and six-loop integrands).","section":"Sec. III, Eq. (38) and Sec. IV"}],"minor_comments":[{"comment":"The product of Θ-factors reads '¯Ω15 ¯Ω25 ¯Ω45 ¯Ω45'; presumably the last factor should be '¯Ω35' (and ¯Ω45 should appear once). Please correct this typographical error.","section":"Eq. (38), last line"},{"comment":"The sentence giving the fitting parameters contains a stray period: 'for R = 0.7. and 0.83, 3.16 and 0.31, respectively'. It should read 'for R = 0.7, and 0.83, 3.16 and 0.31, respectively, for R = 1.0'.","section":"Sec. V, Eq. (48)"},{"comment":"The sentence 'although the signs of the CLs contributions ... alternate, they are in fact all positive' is confusing; the signs of the coefficients F_n alternate, whereas the physical contributions, after including the prefactors, are all positive. Please rephrase to avoid the apparent contradiction.","section":"Sec. V, paragraph after Eq. (47)"},{"comment":"The fitted Monte Carlo parameters a, b, and c are quoted without uncertainties or a description of the fit procedure, and the MC statistical errors are not shown in Fig. 2. Since the paper claims 'percent-level accuracy', the precision of the comparison would be clearer if the MC uncertainties and fit residuals were reported.","section":"Sec. V, Fig. 2 and Eq. (48)"}],"recommendation":"major_revision","confidential_remarks":"The new five- and six-loop results depend on the master formula of an unpublished companion paper by the same author (arXiv:2409.14029). The editor may wish to consider requesting a review of that companion paper or a self-contained derivation in this manuscript, since the present paper cannot be evaluated independently of it. The apparent incompleteness of the reducible terms in Eqs. (40) and (43) is the main technical issue that needs to be resolved before publication."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Dear Colleague,\n\nThe take-away: this is a genuine but modest extension of the author's own four-loop calculation, and it cannot be fully verified from the manuscript alone. The new five- and six-loop clustering-log coefficients F5 ≈ -0.013 and F6 ≈ +0.010 are new; prior fixed-order work stopped at four loops. The systematic use of the master formula [1] and the clean recap of orders two through four are the paper's real strengths.\n\nThe soft spots are real. First, the decisive input—the irreducible five- and six-loop clustering functions Ξ_12345 and Ξ_1...6—is relegated to an external Mathematica file that is not included. Without those integrands, the headline numbers cannot be checked. Second, the master formula from [1] is assumed correct and is not rederived or cross-checked; an error there would propagate directly into F5 and F6. Third, the Monte Carlo comparison is insensitive to the new orders: the paper's Fig. 2 shows the five- and six-loop curves coincide with the four-loop curve, so the benchmark cannot validate these coefficients.\n\nMore seriously, the claimed factorization pattern appears to have errors. Eq. (40) for the five-loop jet mass fraction omits the reducible term (1/2) Σ1 (Σ2)^2, which is required by the cluster expansion structure the paper itself invokes. Eq. (43) omits the Σ1 Σ2 Σ3 term at six loops. These omissions do not change the connected-cluster integrals that define F5 and F6, but they contradict the statement that expanding the exponential (46) reproduces the fixed-order distributions. Also, Eq. (38) contains an obvious typo—the last line has ¯Ω45 ¯Ω45 where ¯Ω35 ¯Ω45 is clearly intended.\n\nThat said, the central idea—that these logs exponentiate and that the fixed-order series converges rapidly—is plausible and supported by the lower-order data. The numerical values may well be correct. This is a paper a specialist in non-global logarithms will want to see, but it needs revision: the notebook must be provided, the factorization formulas should be corrected or the omission explained, and the typo fixed. The MC comparison should also be presented with the caveat that it does not discriminate the new coefficients.\n\nWho is this for? People actively working on jet clustering and non-global observables. It is not a breakthrough, but it is a legitimate fixed-order extension. I would accept it for peer review, with the expectation of major revision.\n\nCandidly yours,\n[Your name]","headline":"New five- and six-loop clustering-log coefficients, but the decisive integrals are hidden in a notebook and the factorization pattern has gaps.","tokens_in":14374,"tokens_out":3855,"would_cite":false,"duration_ms":32518,"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":"This paper computes the leading clustering logarithms for the $k_t$ jet algorithm through six loops and shows their resummed exponential reproduces the all-orders Monte Carlo result at the percent level.","keywords":["clustering logarithms","non-global logarithms","kt jet algorithm","resummation","eikonal approximation","single-jet mass","abelian non-global logarithms","jet substructure"],"falsifier":"Compute the five-loop coefficient directly by a brute-force phase-space integration of the $k_t$-clustered eikonal amplitude for the single-jet mass observable, without using the master formula, and compare with Eq. (42): a value for $F_5(R)$ that differs from approximately $-0.013$ by more than the numerical integration uncertainty would invalidate the central claim. A second, model-independent check is to run the all-orders Monte Carlo of Ref. [2] at larger values of $t$ (say $t=0.3$ through $0.5$): if the measured clustering form factor departs from the exponential (46) by more than a few percent there, the claimed all-orders accuracy fails where the missing coefficients would matter.","tokens_in":13366,"feed_emoji":"⚛️","tokens_out":8102,"duration_ms":69641,"temperature":0.7,"pith_summary":"Clustering logarithms are large QCD corrections that enter jet-mass distributions only because the $k_t$ algorithm clusters soft gluons; previous calculations reached four loops. This paper carries the fixed-order series to five and six loops, finding $F_5(R)\\simeq-0.013$ and $F_6(R)\\simeq0.010$, both nearly constant in the jet radius. It then exponentiates the coefficients and compares the result with an all-orders Monte Carlo simulation, obtaining agreement at the percent level. If the calculation is right, the clustering-logarithm series is under control at four loops, and the missing higher orders cannot affect precision jet phenomenology.","feed_headline":"Six-loop clustering logarithms match Monte Carlo to a percent","feed_subtitle":"New five- and six-loop coefficients are tiny; four loops already capture the full jet-mass correction.","key_machinery":"The machinery is the master formula for the analytic structure of $k_t$ clustering from Ref. [1], which writes the all-orders measurement operator as products of step functions: $\\Theta_i^{\\mathrm{in/out}}$ for jet-region membership and $\\Omega_{ij}$ for pairwise clustering. The paper uses this formula to build the five- and six-loop clustering functions $\\Xi^{kt}_{1\\cdots5}$ and $\\Xi^{kt}_{1\\cdots6}$; at every order the clustering function splits into reducible products of lower-order functions plus one irreducible term $\\Xi^{kt}_{1\\cdots n}$, whose phase-space integral defines the new coefficient $F_n(R)$. The eikonal approximation and strong energy ordering reduce the squared amplitudes to products of single-gluon antenna functions, making the multidimensional integrals tractable.","core_discovery":"The paper's claim is that the leading clustering (abelian non-global) logarithms for the $k_t$ jet algorithm obey a pattern of exponentiation in which each loop order contributes one new irreducible coefficient, $F_n(R)$, and that the first six of these coefficients are $F_2(R)\\simeq0.183$, $F_3(R)\\simeq-0.052$, $F_4(R)\\simeq0.022$, $F_5(R)\\simeq-0.013$, and $F_6(R)\\simeq0.010$. Substituted into the all-orders exponential $C(t)=\\exp\\big[\\sum_{n\\ge2}(-1)^n F_n(R)\\,(2C_F t)^n/n!\\big]$, these coefficients reproduce the all-orders Monte Carlo clustering form factor at the percent level for jet radii $R=0.7$ and $R=1.0$ over a wide range of $t$. The paper reads this as evidence that the clustering-logarithm series converges rapidly and that the four-loop result already captures the phenomenologically relevant contribution.","pith_inferences":["Because the master formula is claimed to be observable-independent, the same $F_n(R)$ coefficients should appear for other non-global observables; testing this on, say, the gap-between-jets observable would extend the paper's single-observable check.","The near-constancy of $F_n(R)$ as $R\\to0$ points to a boundary/edge effect; if that is universal, small-$R$ clustering corrections could be predicted from jet geometry rather than from a new integration at each order.","The calculations treat only abelian primary emissions; whether non-abelian correlated emissions generate comparable clustering logarithms at leading-log level is left open, and a positive answer would change the interpretation of these coefficients as 'leading.'","Since percent-level agreement is reached already at four loops, a practical next step would be to embed the four-loop exponential as a default clustering correction in jet-substructure codes, a step the paper notes but does not take."],"forward_implications":["The $n$-loop coefficients follow the alternating pattern $0.183$, $-0.052$, $0.022$, $-0.013$, $0.010$, so the CL series converges rapidly and the $1/n!$ suppression is effective.","The resummed exponential $C(t)$ tracks the Monte Carlo form factor to about $0.5\\%$ or better at $t=0.15$ and $t=0.25$ for $R=1.0$, and in the exponential form the five- and six-loop curves are effectively indistinguishable from the four-loop curve.","For the single-jet mass observable, four-loop information is sufficient: five- and six-loop terms change the distribution by less than the already-small $\\sim5\\%$ CL contribution.","The same exponentiation pattern applies to generic non-global observables in $e^+e^-$, lepton-hadron, and hadron-hadron collisions, because the master formula is observable-independent.","The factorization of the jet mass fraction at each order into products of lower-order terms plus a new irreducible term confirms the CL distribution is an exponential of the irreducible coefficients, not a power series with unrelated numbers."],"supporting_citations":[{"why":"Supplies the master formula for the analytic structure of kt clustering used verbatim to write the five- and six-loop integrands.","marker":"[1]"},{"why":"Provides the all-orders Monte Carlo clustering form factor against which the resummed exponential and its series expansion are compared.","marker":"[2]"},{"why":"Earlier calculation of clustering logarithms for the single-jet mass observable up to four loops, the baseline this paper extends.","marker":"[7]"},{"why":"First identification of clustering logarithms in the kt jet algorithm, defining the class of corrections under study.","marker":"[13]"},{"why":"Extended clustering-logarithm calculations to four loops for the gap-between-jets observable, the prior higher-loop benchmark.","marker":"[22]"},{"why":"Earlier application of the master formula to a dijet mass observable, establishing the computational setup adopted here.","marker":"[25]"},{"why":"Numerical integration library used to evaluate the multidimensional coefficient integrals F2(R) through F6(R).","marker":"[28]"}],"fun_headline_variants":["Clustering logs resummed to six loops hit Monte Carlo","Six-loop QCD logs: percent-level match with MC","Exponentiation proven: clustering logs to six loops","Tiny high-order terms: four loops suffice","Rapid convergence: six-loop clustering logs"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The calculation inherits the master formula for the analytic structure of $k_t$ clustering from Ref. [1] and does not rederive or independently test it; if that formula mis-describes how clustering acts on the phase-space measurement operator at five or six gluons, the new coefficients $F_5$ and $F_6$ are wrong.","fun_headline_variants_meta":{"raw":{"variants":["Clustering logs resummed to six loops hit Monte Carlo","Six-loop QCD logs: percent-level match with MC","Exponentiation proven: clustering logs to six loops","Tiny high-order terms: four loops suffice","Rapid convergence: six-loop clustering logs"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000592,"raw_usage":{"total_tokens":2721,"prompt_tokens":838,"completion_tokens":1883,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":454,"completion_tokens_details":{"reasoning_tokens":1807}},"tokens_in":454,"tokens_out":1883,"duration_ms":11777,"temperature":1.0,"reasoning_tokens":1807,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T22:37:02.387263+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute the five-loop coefficient directly by a brute-force phase-space integration of the $k_t$-clustered eikonal amplitude for the single-jet mass observable, without using the master formula, and compare with Eq. (42): a value for $F_5(R)$ that differs from approximately $-0.013$ by more than the numerical integration uncertainty would invalidate the central claim. A second, model-independent check is to run the all-orders Monte Carlo of Ref. [2] at larger values of $t$ (say $t=0.3$ through $0.5$): if the measured clustering form factor departs from the exponential (46) by more than a few percent there, the claimed all-orders accuracy fails where the missing coefficients would matter.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the master formula for the analytic structure of kt clustering used verbatim to write the five- and six-loop integrands."},{"cited_title":"R is the jet-radius parameter and B denotes the beam direction","cited_arxiv_id":null,"evidence_quote":"Provides the all-orders Monte Carlo clustering form factor against which the resummed exponential and its series expansion are compared."},{"cited_title":"Azimuthal decorrelation between a jet and a Z boson at hadron colliders","cited_arxiv_id":"2207.10147","evidence_quote":"Earlier application of the master formula to a dijet mass observable, establishing the computational setup adopted here."}],"review_version":1}