{"id":"8c9e4439-3997-463c-ac3c-4b71055b0445","arxiv_id":"2412.00178","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":12,"one_line_summary":"A dispersive evaluation gives a_mu^HLbL subleading = 33.2(7.2) x 10^-11 and total = 101.9(7.9) x 10^-11.","lead":"This paper evaluates the subleading hadronic light-by-light contributions to the muon anomalous magnetic moment using dispersion relations, obtaining a subleading value of 33.2(7.2) x 10^-11 and a total HLbL value of 101.9(7.9) x 10^-11. The result matters because it sharpens the Standard Model prediction ahead of the final Fermilab muon g-2 result.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The effective-pole estimate of omitted hadronic states is the load-bearing weak point: the central value and uncertainty rest on an ansatz whose low-energy representation is not independently validated, and the paper's own asymmetric-matching factors show the model dependence.","rationale":"The reader's verdict already identifies the effective-pole construction as the weakest assumption, and my reading agrees: this is exactly where the central claim 33.2(7.2) x 10^-11 is least secure. The paper is honest and internally consistent; the matching to SDCs, the VVA constraints, and the scale variation are all treated carefully. However, the effective poles are the only component of the subleading evaluation that tries to account for all omitted states, and their construction is purely asymptotic. The large spread between symmetric and asymmetric matching conditions, and the sign change of bar-Pi_3-12 between formulations, show that the result depends on choices that are not fixed by first principles. A concrete continuum test is needed to determine whether the quoted 3.9 x 10^-11 effective-pole uncertainty covers the representability problem. Since the reader already requested conditional acceptance and my concern does not change that assessment, the verdict remains UNCHANGED.","tokens_in":22225,"tokens_out":6682,"duration_ms":65151,"concrete_test":"Recompute the effective-pole contribution to a_mu[bar-Pi_1-12] using a continuum or multi-pole spectral function (for example, five narrow poles spaced between 1.5 and 2.5 GeV) that is constrained to reproduce exactly the same symmetric and asymmetric asymptotic coefficients as in Eqs. (C.3)-(C.5), and evaluate it both in the four-point optimized basis of Ref. [92] and with the triangle-kinematics expression of Eq. (C.1). If the resulting central contribution differs from the quoted effective-pole entries in Table 4 by more than the 3.9 x 10^-11 effective-pole error, or if the four-point and triangle-kinematics results differ by that amount, then the single-pole ansatz is not a reliable representation of missing states and the total uncertainty in Eq. (4.2) is understated.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The most load-bearing assumption is the effective-pole construction in Sec. 3 and App. C, which contributes 3.2 x 10^-11 to the central subleading value and 3.9 x 10^-11 to its uncertainty. The single pseudoscalar and single axial-vector poles have quark-model TFFs, and their couplings are fixed by requiring the sum of explicit hadronic states plus poles to reproduce the pQCD short-distance constraints in the symmetric asymptotic limit. This fixes only the asymptotic coefficient, but a_mu is dominated by photon virtualities around 1 GeV, where the asymptotic condition has little leverage. The model dependence is visible inside the paper itself: asymmetric matching changes the required couplings by factors of about 1.5 for the pseudoscalar and 2.8 for the axial-vector pole, and the contribution from bar-Pi_3-12 is reported to change sign between the four-point and triangle-kinematics formulations. A single-pole ansatz is therefore not demonstrated to represent the low-energy effect of the omitted continuum, and the assigned 3.9 x 10^-11 error only covers variations within this ansatz, not failure of the ansatz itself.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper presents an evaluation of the subleading hadronic light-by-light (HLbL) contributions to the muon anomalous magnetic moment within the dispersive approach. The authors combine axial-vector transition form factors from Ref. [90], scalar and tensor narrow-resonance estimates, perturbative QCD above a matching scale Q0, and OPE constraints in the mixed region using the VVA correlator analysis of Ref. [91]. Missing higher states are modeled by effective pseudoscalar and axial-vector poles whose couplings are fixed by matching to short-distance constraints. The main result is a_mu^HLbL|_subleading = 33.2(7.2) x 10^-11, and when combined with previously evaluated dispersive contributions and the charm loop, a_mu^HLbL|_total = 101.9(7.9) x 10^-11 (Eq. 4.2). The paper gives a detailed error decomposition into experimental, matching-scale, systematic, and effective-pole uncertainties, and studies the dependence on Q0 and r in Figs. 5 and 6.","tokens_in":22678,"tokens_out":4491,"duration_ms":45187,"significance":"If the result stands, this is the most complete dispersive evaluation of HLbL to date and the first to integrate axial-vector states, tensor mesons, and short-distance constraints in a single framework with a sub-8 x 10^-11 total uncertainty. The paper is transparent: the error budget in Eq. (4.1) is explicit, the matching scale and OPE parameter r are varied, and the authors state the limitations of the tensor TFF approximation and the effective-pole construction. The use of the dispersive VVA analysis of Ref. [91] is a genuine improvement over earlier models. However, the central value and its uncertainty rest on two model-dependent elements, the effective-pole estimate and the simplified tensor TFFs, whose low-energy validation is incomplete. These elements are acknowledged in the text but are not yet demonstrated to be robust enough to support the claimed precision.","major_comments":[{"comment":"The effective-pole construction is the load-bearing element of the final error budget: it contributes 2.0 x 10^-11 to the central subleading value and 3.9 x 10^-11 to its uncertainty in Eq. (4.1). The couplings are fixed by matching the sum of explicit states plus poles to the pQCD short-distance constraints in the symmetric asymptotic limit, but the a_mu integral is dominated by photon virtualities around 1 GeV, where the asymptotic condition has little leverage. The paper itself shows that asymmetric matching changes the required couplings by factors of about 1.5 (pseudoscalar) and 2.8 (axial-vector) (App. C, Eqs. C.3-C.7), and it notes that the contribution from Pi-bar_3-12 can change sign between four-point and triangle kinematics (Sec. 2 and App. C). This indicates that the single-pole ansatz is not demonstrated to represent the low-energy effect of the omitted continuum. The assigned 3.9 x 10^-11 error only covers variations within this ansatz, not failure of the ansatz itself. I request a concrete validation or a more conservative treatment: for example, comparing the effective-pole low-energy contribution with an explicit dispersive estimate of the f2(1270) via pi-pi D-wave rescattering, or testing stability under several pole masses, TFF shapes, and matching directions simultaneously.","section":"Sec. 3 and App. C"},{"comment":"The OPE matching sets the ambiguity coefficients c_5^(2), c_6^(2), c_7^(5), and c_9^(3) to zero. The text correctly states that this strict identification holds only for asymptotically large q3^2 and that finite-q3^2 contributions from higher-dimensional operators are expected (App. B, discussion after Eq. B.3). However, the OPE region contributes 10.9 x 10^-11 to the central value in Table 4, which is a substantial part of the subleading result, and no explicit uncertainty is assigned to the c_i^(n)=0 assumption in the final error budget. The matching uncertainty in Eq. (4.1) is defined as the variation from Q0 and r only. I ask the authors to quantify the sensitivity of the OPE-region contribution to the neglected ambiguity terms, for example by using the VVA dispersive analysis of Ref. [91] or by assigning a systematic uncertainty based on the size of the residual mismatches discussed in App. B.","section":"App. B, Eq. (B.3)"},{"comment":"The tensor-meson contributions are estimated with the simplified TFFs F_2-5^T = 0 and a dipole form with Lambda_T = M_rho. The paper acknowledges this is a drastic approximation and adds a 100% uncertainty on the total tensor contribution, but the central value and the sign of the tensor contribution are controlled by the strong cancellation between a_mu[Pi_1,2] and a_mu[Pi_3-12] that this form produces (Table 1). A 100% uncertainty on the sum does not necessarily cover a different shape or sign of the individual basis-function contributions if the cancellation is altered. Since the f2(1270) is a broad pi-pi resonance, the narrow-resonance approximation with a quark-model TFF should be tested against existing gamma* gamma* -> pi-pi helicity amplitudes or by varying the TFF parametrization (monopole vs dipole, different Lambda_T). This is a load-bearing point for the subleading central value and should be addressed before final publication.","section":"Sec. 2, Eq. (2.2)"}],"minor_comments":[{"comment":"The footnote to Table 4 clarifies that the errors in the main part of the table exclude the systematic and effective-pole uncertainties, but this is easy to miss; I suggest adding one sentence in Sec. 4 stating explicitly that the experimental and pQCD errors in Table 4 must be combined with the matching, systematic, and effective-pole errors from Eq. (4.1).","section":"Sec. 4 and Table 4"},{"comment":"Please define more explicitly how F_1^T(0,0) is normalized to the two-photon width and how the sign of F_1^T is chosen; currently the reader has to reconstruct this from Ref. [86] and Table 1.","section":"Sec. 2, Eq. (2.2)"},{"comment":"The parameters beta and gamma in Eq. (A.6) are introduced with only a brief motivation; please state explicitly which values are used in the numerical analysis (gamma = 1.5 is mentioned, but beta is not numerically specified in the main text).","section":"App. A"},{"comment":"The sentence in the introduction 'In this work, describing the details of Ref. [94]' is confusing; if Ref. [94] is the companion Letter, please rephrase to avoid the appearance that this paper merely describes the Letter rather than being a full account.","section":"Sec. 2"}],"recommendation":"major_revision","confidential_remarks":"The paper is technically strong and within the scope of the journal, and the authors are unusually transparent about the limitations of their estimates. My main concern is that the central value and the quoted uncertainty in Eq. (4.1) depend on the effective-pole ansatz in a way that is not validated at low energies, and the OPE ambiguity coefficients in App. B are set to zero without a quantitative uncertainty estimate. These are fixable in revision, but they currently prevent me from recommending acceptance. I do not see any concerns about overlap with the companion Letter or about citation practice."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The main news is a new number: subleading HLbL = 33.2(7.2) x 10^-11 and total = 101.9(7.9) x 10^-11, which cuts the white-paper uncertainty by more than half. That is a real step forward for the field. The matching of axial-vector, tensor, and pseudoscalar-tail contributions to short-distance constraints at the level of the scalar basis functions (Figs. 2-4) is a genuine technical contribution, and the error budget in Eq. (4.1) is the most explicit I have seen in this line of work. The paper also makes good use of external inputs: the axial-vector TFFs come from a global fit, and the comparison to the dispersive a1 TFF in Fig. 1 is a nice cross-check. The authors are honest about the simplified tensor TFF and the U(3) assumptions; the 30% systematic and the extra 100% on the tensor piece show they are not hiding the model dependence.\n\nThe soft spot is the effective-pole estimate in Sec. C. The central value includes 3.2 x 10^-11 from these poles, and the effective-pole error of 3.9 x 10^-11 is the largest single uncertainty. But the poles are constructed to satisfy the asymptotic SDCs, not pinned down by low-energy data, and the paper itself shows that asymmetric matching changes the couplings by factors of 1.5 and 2.8. That means the low-energy effect of omitted hadronic states is not independently validated; the error covers variations within the ansatz, not failure of the ansatz. I do not think this sinks the paper, because the effective-pole contribution is only about 10% of the subleading total and the final total agrees with the white paper and with lattice results. But it does mean a reader should not treat the 7.2 x 10^-11 error as a fully model-independent statement. The choice of the axial-vector fit variant without f1 -> phi gamma is also motivated by the matching, which is a bit post-hoc, but the 30% systematic covers it.\n\nWho is this for? Anyone working on muon g-2 phenomenology or HLbL dispersive methods. It is a solid, serious calculation with honestly disclosed limitations. It deserves a serious referee. I would send it to review, asking the referee to scrutinize the effective-pole error and the tensor TFF assumption, but I would expect it to be published, possibly with a clarified error discussion.","headline":"A careful, mostly convincing dispersive update of the HLbL subleading contributions; the effective-pole estimate is the soft spot, but the paper's own error accounting keeps the central claim credible.","tokens_in":23143,"tokens_out":2501,"would_cite":true,"duration_ms":22681,"reading_group":"yes","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 reports subleading hadronic light-by-light effects summing to $33.2(7.2)\\times10^{-11}$ and a full dispersive total of $a_\\mu^{\\mathrm{HLbL}}=101.9(7.9)\\times10^{-11}$.\n","keywords":["hadronic light-by-light scattering","muon anomalous magnetic moment","dispersion relations","axial-vector mesons","transition form factors","short-distance constraints","operator product expansion","effective poles"],"falsifier":"Replacing the simplified tensor transition form factors by a dispersive calculation in triangle kinematics, specifically the D-wave $\\pi\\pi$ rescattering contribution to the $f_2(1270)$, would test the total directly: if the resulting $a_\\mu$ shifts by more than the quoted effective-pole error of $3.9\\times10^{-11}$, the simplified tensor treatment is invalid.\n","tokens_in":22012,"feed_emoji":"⚛️","tokens_out":18587,"duration_ms":147598,"temperature":0.7,"pith_summary":"This paper sets out to complete the dispersive, i.e., analyticity-based, evaluation of the hadronic light-by-light contribution to the muon's anomalous magnetic moment. Its central result is that the subleading effects—axial-vector mesons, tensor mesons, heavy scalars, the matching to perturbative QCD, and the transition region between low and high photon virtualities—sum to $33.2(7.2)\\times10^{-11}$, which combined with previously evaluated contributions gives $a_\\mu^{\\mathrm{HLbL}}=101.9(7.9)\\times10^{-11}$. This matters because the muon $g-2$ experiment is expected to report a final value with roughly twice the current precision, so the theory error on this hadronic piece must shrink accordingly. The paper presents this as the most complete dispersive evaluation available, with uncertainties propagated from data, matching-scale variation, and a deliberate estimate of hadronic states not explicitly included.\n","feed_headline":"Muon g-2 light-by-light set at 101.9(7.9)","feed_subtitle":"Subleading hadronic effects sum to 33.2(7.2), sharpening the dispersive prediction ahead of the final muon g-2 result.","key_machinery":"The load-bearing object is the optimized basis of scalar functions $\\bar{\\Pi}_i$ for the hadronic light-by-light tensor, which lets axial-vector states with $J=1$ and tensor mesons be evaluated without kinematic singularities while leaving previously computed pieces unchanged. On top of this basis, the matching procedure divides the photon-virtuality space into three regions: below a scale $Q_0$ the explicit hadronic states are summed; above $Q_0$ a perturbative QCD quark loop with $\\alpha_s$ corrections is used; and where two virtualities are large while the third is small, an operator-product expansion relates the tensor to the vector–vector–axial-vector correlator through the longitudinal and transverse form factors $w_L$ and $w_T$. Effective poles—one pseudoscalar and one axial-vector—with couplings set by the asymptotic matching condition estimate the low-energy effect of hadronic states not listed explicitly.\n","core_discovery":"The paper's central claim is that the dispersive framework can now account for every significant subleading contribution to hadronic light-by-light scattering without modeling the transition to the short-distance region by hand. In the optimized scalar basis, the narrow-resonance contributions of axial-vector, tensor, and heavy scalar states are evaluated with transition form factors; the high-virtuality region is described by the perturbative quark loop; and the mixed region is constrained through the operator product expansion and the vector–vector–axial-vector correlator. The sum of all these pieces matches the short-distance constraints reasonably well, and the remaining mismatch is estimated with effective poles whose couplings are fixed by the asymptotic matching. The resulting subleading contribution is $a_\\mu^{\\mathrm{HLbL}}|_{\\mathrm{subleading}}=33.2(7.2)\\times10^{-11}$, and adding the previously evaluated dispersive contributions and the charm loop gives $a_\\mu^{\\mathrm{HLbL}}|_{\\mathrm{total}}=101.9(7.9)\\times10^{-11}$. The paper finds good agreement with the previous consensus evaluation and with one lattice determination, with a slightly lower central value than two other lattice calculations.\n","pith_inferences":["If the effective-pole estimate is essentially right, the largest remaining route to precision is experimental: better axial-vector transition form factor data would directly reduce the dominant experimental and $U(3)$-symmetry systematic errors, since the matching-scale error is already subdominant.","One could test the effective-pole construction outside the symmetric limit by matching along the asymmetric line $Q_1^2\\gg Q_2^2=Q_3^2$; the paper reports that this would require axial-vector couplings about $2.8$ times larger, so an independent determination of an excited axial-vector two-photon coupling would discriminate between the two implementations.","A full dispersive treatment of the tensor-meson channel in triangle kinematics, replacing the simplified tensor form factor assumption, could shift the central value by more than the current tensor error because the tensor contribution is a difference of two larger numbers; this is the most direct computation that would sharpen the final number."],"forward_implications":["If the result stands, the dispersive prediction for the hadronic light-by-light contribution is $101.9(7.9)\\times10^{-11}$, with an uncertainty small enough to make this part of the muon $g-2$ theory competitive with the forthcoming final measurement.","The subleading effects are not a small correction: axial-vector mesons, tensor mesons, heavy scalars, and the short-distance matching together contribute $33.2(7.2)\\times10^{-11}$, roughly one third of the total.","The result is stable under variation of the matching scale $Q_0\\in[1.2,2.0]$ GeV and the operator-product-expansion parameter $r\\in[1/8,1/2]$, so the quoted uncertainty is not dominated by the matching-scale choice.","Heavy scalars contribute almost nothing, while tensor mesons cancel strongly between the two groups of scalar basis functions, making the simplified tensor transition form factors the most pressing input to replace."],"supporting_citations":[{"why":"Provides the optimized scalar basis that lets all $J\\le1$ states be evaluated without kinematic singularities.","marker":"[92]"},{"why":"Supplies the axial-vector transition form factors for the $f_1(1285)$, $f_1'(1420)$, and $a_1(1260)$, the main subleading hadronic input.","marker":"[90]"},{"why":"Gives the dispersive analysis of the vector–vector–axial-vector correlator used to implement the operator-product-expansion constraints in the mixed region.","marker":"[91]"},{"why":"Provides the perturbative QCD quark-loop result with $\\alpha_s$ corrections used above the matching scale $Q_0$.","marker":"[96]"},{"why":"Shows that higher-order operator-product-expansion form-factor effects cancel in the $a_\\mu$ integral, so that the leading $w_L/w_T$ constraint captures the result.","marker":"[98]"},{"why":"Supplies the asymptotic short-distance constraints that fix the couplings of the effective poles in the symmetric limit.","marker":"[26]"},{"why":"Provides the pion-pole evaluation that enters the total and whose high-virtuality tails are subtracted in the perturbative and operator-product-expansion regions.","marker":"[21, 22]"},{"why":"Provides the $\\eta$ and $\\eta'$ pole evaluations included in the total and subtracted in the asymptotic regions.","marker":"[84, 85]"}],"fun_headline_variants":["Subleading HLbL sums to 33.2(7.2), total 101.9(7.9)","Dispersive HLbL: subleading 33.2(7.2), total 101.9(7.9)","Muon g-2: subleading light-by-light at 33.2(7.2)","HLbL subleading effects: 33.2(7.2) added to total"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that a single effective pseudoscalar pole and a single effective axial-vector pole, with couplings chosen so that the explicit hadronic states plus these poles reproduce the known high-energy behavior when all three photon virtualities are large and equal, capture the low-energy effect of every hadronic state not explicitly listed.\n","fun_headline_variants_meta":{"raw":{"variants":["Subleading HLbL sums to 33.2(7.2), total 101.9(7.9)","Dispersive HLbL: subleading 33.2(7.2), total 101.9(7.9)","Muon g-2: subleading light-by-light at 33.2(7.2)","HLbL subleading effects: 33.2(7.2) added to total"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000419,"raw_usage":{"total_tokens":2194,"prompt_tokens":1019,"completion_tokens":1175,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":635,"completion_tokens_details":{"reasoning_tokens":1061}},"tokens_in":635,"tokens_out":1175,"duration_ms":8016,"temperature":1.0,"reasoning_tokens":1061,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T05:39:10.944542+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Replacing the simplified tensor transition form factors by a dispersive calculation in triangle kinematics, specifically the D-wave $\\pi\\pi$ rescattering contribution to the $f_2(1270)$, would test the total directly: if the resulting $a_\\mu$ shifts by more than the quoted effective-pole error of $3.9\\times10^{-11}$, the simplified tensor treatment is invalid.","supporting_citations":[],"review_version":1}