{"id":"d27e9124-a036-401e-8e79-38fe200faf21","arxiv_id":"2412.00190","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":6,"one_line_summary":"Using a dispersive formalism with experimental transition form factors and short-distance constraints, the authors obtain a_mu^HLbL = 101.9(7.9) x 10^-11, halving the previous uncertainty.","lead":"This paper presents a full dispersive calculation of the hadronic light-by-light contribution to the muon's magnetic moment, yielding 101.9(7.9) x 10^-11. The result is twice as precise as the previous phenomenological estimate, meeting the precision target of the final Fermilab measurement.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The axial-vector TFF input is fixed by U(3) extrapolation from f1 data, with only an uncalibrated 30% systematic; if U(3) breaking in the a1 doubly-virtual TFFs exceeds 30%, the central precision claim in Eq. (8) fails.","rationale":"Good-faith reading: the paper is a milestone in dispersive HLbL; it does not fake precision, and it explicitly labels the U(3) and tensor-model inputs. The load-bearing weak point is the axial-vector TFF input. The central claim '101.9(7.9) meets Fermilab precision' hinges on the error budget. The axial-vector row of Table II (12.2(2.3)) is the largest non-pseudoscalar subleading piece, and its central value is set by a U(3) extrapolation from f1/f1' data to the a1, with only an ad hoc 30% systematic. No derivation or cross-check establishes that 30% as an upper bound on U(3) breaking. The paper cites a dispersive VVA study [81] as 'close' to the U(3) assumption, but that study primarily constrains the singly-virtual a1 TFF; the doubly-virtual components that enter HLbL are not directly tested. A 50% normalization error or an unmodeled shape difference would shift the total by several units of 10^-11, comparable to the full quoted 7.9. The effective-pole and tensor-cancellation inputs also carry model dependence, but the paper assigns uncertainties larger than the effective-pole central value and includes a 100% penalty on the tensor cancellation, so those are less likely to break the precision claim. The proposed test directly quantifies whether the 30% prior is sufficient. On this basis the reader's CONDITIONAL verdict stands; no change is needed.","tokens_in":12630,"tokens_out":13060,"duration_ms":122706,"concrete_test":"Recompute the axial-vector contribution to Eq. (8) using the dispersively reconstructed a1 TFF from Ref. [81] for the singly-virtual component, without imposing the U(3) relation to f1, and vary the doubly-virtual components between the quark-model M_rho form and the asymptotic light-cone form of Ref. [75]. Compare the resulting shift in a_mu with the 3.7 x 10^-11 axial-vector systematic. If the shift exceeds 3.7 x 10^-11, the 30% U(3) prior is insufficient and the precision claim must be qualified.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim Eq. (8) states a_mu^HLbL = 101.9(7.9) x 10^-11 and asserts this meets Fermilab final-precision requirements. The most load-bearing input is the axial-vector sector: Table II gives the axial-vector contribution as 12.2(2.3) x 10^-11, the largest non-pseudoscalar subleading piece. The a1(1260) TFFs are not extracted from direct data. Instead, the Letter uses e+e- data for f1/f1' and then defines the complete axial-vector TFF set via U(3) symmetry, assigning a global 30% uncertainty for 'possible symmetry violations'. This 30% is not derived from any spread of models, fits, or dispersion relations; it is a round-number prior. Because a1 is isovector and f1/f1' are isoscalar, U(3) breaking in TFF normalizations, and especially in the doubly-virtual Q^2 dependence, has no reason to be bounded by 30%. The paper itself concedes axial-vector uncertainties must be validated with future BESIII and Belle II data. If the true U(3) violation in the dominant a1 TFFs were 40-50%, or the TFF shapes differed beyond the parameterized form, the shift in a_mu could exceed the 7.9 total error, undermining the 'meets precision requirements' claim. This is not an internal inconsistency; it is an uncalibrated modeling uncertainty at the heart of the claimed precision.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper presents a data-driven dispersive evaluation of the hadronic light-by-light (HLbL) contribution to the muon anomalous magnetic moment. The HLbL tensor is reconstructed from its discontinuities using four-point dispersion relations, with contributions from pseudoscalar poles, meson boxes, rescattering, axial-vector and tensor resonances, heavy scalars, and a matching to pQCD and OPE short-distance constraints. The final result is a_mu^HLbL = 101.9(7.9) x 10^-11 (Eq. (8)), which the authors state meets the precision requirements of the final Fermilab result. The paper relies on a companion paper (Ref. [82]) for many technical details.","tokens_in":13094,"tokens_out":6965,"duration_ms":65783,"significance":"If correct, this would be the most precise data-driven determination of a_mu^HLbL to date, reducing the uncertainty of the 2020 White Paper evaluation by more than a factor of two and providing a sharp benchmark for comparisons with lattice QCD. The paper is notable for its detailed error budget, separating experimental, matching, systematic, and effective-pole uncertainties, and for demonstrating stability of the integral under variation of the matching scale Q0 and the OPE parameter r (Fig. 5). The result is also falsifiable in the sense that future BESIII and Belle II measurements of axial-vector transition form factors, as well as improved lattice calculations, will directly test the assumptions that underlie the axial-vector and tensor contributions.","major_comments":[{"comment":"The axial-vector sector is the most load-bearing input for the claimed precision, yet the a1(1260) transition form factors are not extracted from direct data. The Letter fixes the f1 and f1' TFF normalizations and mixing angle from e+e- data and then defines the complete set of axial-vector TFFs using U(3) symmetry, assigning a global 30% uncertainty 'to account for possible symmetry violations.' This 30% is not derived from a spread of models, fits, or dispersion relations; the supporting comparison with the dispersive VVA calculation of Ref. [81] constrains only the singly-virtual a1 form factor, not the doubly-virtual TFFs that dominate the HLbL integral. Since the axial-vector contribution is 12.2(2.3) x 10^-11 (Table II), the largest non-pseudoscalar subleading piece, a U(3) violation of 40-50% in the a1 TFF normalizations or virtuality dependence would shift the result by more than the quoted 7.9 x 10^-11 total uncertainty and would invalidate the precision claim of Eq. (8). Please calibrate the 30% by explicit sensitivity studies (e.g., 50% variations or an alternative model spread) or soften the claim accordingly.","section":"Exclusive hadronic states / Table II / Eq. (7)"},{"comment":"The effective-pole model (P(2200), A(1700)) is introduced to estimate the impact of missing higher intermediate states in the matching between hadronic and pQCD/OPE regions. The Letter assigns an effective-pole uncertainty larger than the entire effective-pole contribution (3.9 x 10^-11 vs. contributions of 2.0 and 1.2 x 10^-11 in Table II), which is conservative in that specific sense. However, the effective-pole amplitude is an ad-hoc construct with masses fixed at 2.2 GeV and 1.7 GeV and a TFF scale varied only over the same range as Q0; no explicit functional form is given in the Letter. A different pole structure or asymptotic behavior could shift the central value beyond the quoted eff uncertainty. Please provide the explicit effective-pole parameterization in the Letter or in a form that can be checked against Ref. [82], and demonstrate robustness to the assumed pole masses and the number of poles.","section":"Matching to short-distance constraints / Fig. 4 / Table II"},{"comment":"The tensor contribution is estimated using a quark-model-inspired TFF with a single nonvanishing form factor, with the scale set to M_rho. The Letter adds a 100% uncertainty on the total tensor contribution 'to protect against the cancellation observed between a_mu[Pi1,2] and a_mu[Pi3-12],' but this protects against the sign cancellation rather than against individual normalization or shape errors. The two tensor components in Table II are -2.6(3) and -5.1(7) x 10^-11; a 50% error in the f2(1270) TFF alone would shift the sum by roughly 3.9 x 10^-11, exceeding the quoted total sys error. Please provide a more robust estimate of the tensor uncertainty, for example by varying the TFF scale and form independently or by including a dispersive D-wave treatment as the paper itself identifies as future work.","section":"Tensor contributions / Table II / Eq. (7)"}],"minor_comments":[{"comment":"The phrase 'complete dispersive evaluation' overstates the treatment of axial-vector and tensor states, which are evaluated in narrow-resonance and quark-model approximations rather than fully dispersively; consider qualifying the claim as 'complete within the stated hadronic approximations' or moving the stronger claim to the conclusions.","section":"Title and abstract"},{"comment":"The caption refers to the effective pole P(2200) without defining its amplitude; please either define it in the text or refer explicitly to the equation in Ref. [82] where it is introduced.","section":"Figure 4 caption"},{"comment":"The 'Eff.' rows list central values without uncertainties, while Eq. (7) quotes an effective-pole uncertainty; please clarify whether the quoted eff uncertainty is meant to cover these entries or is an additional assignment.","section":"Table II"},{"comment":"The sentence 'all contributions but a single TFF vanish' is ambiguous; please specify which of the tensor TFFs is kept and in which basis the vanishing occurs.","section":"Exclusive hadronic states"},{"comment":"The K± box entry is quoted as -0.5(0); please provide the actual uncertainty (even if negligible) to avoid the appearance of an exactly vanishing error.","section":"Table I"}],"recommendation":"major_revision","confidential_remarks":"The paper is a strong candidate for publication if the axial-vector uncertainty can be defended or recalibrated. The uncalibrated 30% U(3) prior is the main risk to the precision claim, and the effective-pole and tensor modeling uncertainties are secondary but related concerns. The companion paper (Ref. [82]) likely contains the needed details; the editor may wish to ensure that it is available and consistent with the Letter. The heavy reliance on the authors' prior work is appropriate for a synthesis Letter, but the novelty and incremental advance beyond the White Paper should be made explicit in the response."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Bottom line up front: this is the first dispersive HLbL evaluation that includes the axial-vector TFFs, subleading short-distance constraints, and the VVA correlator, and it reduces the uncertainty from 19 to 7.9 x 10^-11. The result is consistent with the 2020 white paper but sharper, and it is not a fit to a_mu — the inputs come from external data and independent short-distance constraints. That alone makes it a serious piece of work.\n\nWhat it does well: the error budget in Eq. (7) is unusually transparent, with separate exp, match, sys, and eff entries. The stability under Q0 and r (Fig. 5) is reassuring, and the authors are explicit that the effective-pole uncertainty is larger than the effective-pole contribution itself. The inclusion of tensor and heavy-scalar estimates, even in a simplified model, gives a sense of scale. The comparison with lattice results in Fig. 6 is fair.\n\nThe soft spot is the axial-vector sector, as your stress-test notes. The a1 TFFs are fixed from f1/f1' data via U(3), with a global 30% uncertainty that is a round-number prior, not a derived spread. That is a legitimate weakness. But the paper's own mitigation deserves weight: the dispersive VVA analysis of the singly-virtual a1 TFF agrees with the U(3)-based assumption, and a 30% error on the axial-vector contribution (12.2) is 3.7 units, comparable to the total error of 7.9. A violation bigger than 30% would need to be demonstrated, not just asserted; the stress-test's '40-50%' scenario is plausible but not evidenced. So I would call this a controlled assumption that needs future data, not a load-bearing flaw that invalidates the central value.\n\nThe other soft spot is the 'complete' claim. It rests on the companion paper (2412.00178) for derivation details, and no code or data release is mentioned. For a Letter that is normal, but a referee should ask for the companion paper and ideally a reproducibility package.\n\nBottom line: this deserves a serious referee. The method is data-driven, the uncertainties are itemized, and the central number is credible. The U(3) question is the one to push on, and the authors themselves invite that. I would take it.","headline":"First complete dispersive HLbL evaluation with a realistic error budget; the U(3) axial-vector assumption is the main thing to probe, but it is handled as a transparent systematic.","tokens_in":13585,"tokens_out":2711,"would_cite":true,"duration_ms":23431,"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":"A complete dispersive evaluation determines the hadronic light-by-light contribution to the muon's magnetic moment as $a_\\mu^\\text{HLbL} = 101.9(7.9)\\times 10^{-11}$.","keywords":["muon g-2","hadronic light-by-light","dispersive approach","transition form factors","short-distance constraints","operator product expansion","axial-vector mesons"],"falsifier":"New data on the $f_1$ or $f_1'$ transition form factor that deviate from the U(3)-symmetric parameterization by more than the assigned 30% would move the axial-vector contribution beyond the quoted systematic error.","tokens_in":12429,"feed_emoji":"🧲","tokens_out":12995,"duration_ms":94389,"temperature":0.7,"pith_summary":"The paper presents the first complete dispersive evaluation of the hadronic light-by-light (HLbL) contribution to the muon's anomalous magnetic moment. The central result is $a_\\mu^\\text{HLbL} = 101.9(7.9)\\times 10^{-11}$, which improves on the previous data-driven estimate by more than a factor of two and reaches the precision required for the final result of the muon g-2 experiment. The evaluation reconstructs the HLbL tensor from its discontinuities, expressing it in terms of hadronic matrix elements measured in experiment, and matches the low-energy hadronic description to short-distance constraints from perturbative QCD and the operator product expansion. The piece of the calculation that remains most assumption-dependent is the axial-vector meson contribution, fixed through U(3) symmetry with a 30% assigned uncertainty.","feed_headline":"Hadronic light-by-light pinned to 101.9(7.9) in muon g-2","feed_subtitle":"Uncertainty drops from 19 to 7.9 in units of 10^-11, reaching the precision the final muon g-2 result requires.","key_machinery":"The central mechanism is a Bardeen–Tung–Tarrach decomposition of the four-photon HLbL tensor, which reduces the amplitude to 54 scalar coefficient functions (12 independent after the equations of motion). The master formula for $a_\\mu$ integrates these functions against known kernels over the three photon virtualities. The functions are reconstructed dispersively from their discontinuities, which are saturated by a finite set of hadronic intermediate states: pseudoscalar poles ($\\pi^0,\\eta,\\eta'$) with transition form factors from data, two-meson boxes and rescattering, and narrow-resonance approximations for axial-vector, scalar, and tensor mesons. In the deep-Euclidean region the same functions are matched to the perturbative quark loop with $\\alpha_s$ corrections, and in the asymmetric region to the operator product expansion controlled by the vector-vector-axial-vector correlator $w_{L,T}$. A newly introduced effective-pole parametrization estimates the error from intermediate states that lie between the hadronic and asymptotic descriptions.","core_discovery":"The authors establish a data-driven, dispersive value for the hadronic light-by-light contribution to the muon's $g-2$, $a_\\mu^\\text{HLbL} = 101.9(7.9)\\times 10^{-11}$, by reconstructing the HLbL tensor from its unitarity cuts and matching the sum of exclusive hadronic states to short-distance constraints. This is the first such evaluation that includes axial-vector resonances ($a_1$, $f_1$, $f_1'$), scalar and tensor resonances, the full vector-vector-axial-vector correlator for the OPE region, and a dedicated estimate of the matching uncertainty via an effective-pole ansatz. The result agrees with the 2020 phenomenological consensus, $92(19)\\times 10^{-11}$, but reduces the uncertainty by more than half, meeting the precision target set for the final result of the muon g-2 experiment.","pith_inferences":["A natural next step would be a direct dispersive calculation of the $a_1$ transition form factor from $\\tau$-decay data, which could replace the U(3)-symmetry assumption with an experimental determination.","If the effective-pole matching error is as small as quoted, then the dominant remaining model dependence in the subleading contributions sits in the tensor-meson treatment; a triangle-kinematics dispersive calculation of the $f_2$ contribution would provide a sharp cross-check.","The small tension with two of the lattice QCD evaluations, if it persists, may point to a systematic effect in one of the methods; the present result offers a data-driven reference point for locating the difference."],"forward_implications":["The uncertainty of the HLbL contribution drops from $19\\times 10^{-11}$ to $7.9\\times 10^{-11}$, so the hadronic light-by-light piece no longer dominates the theory error for the muon $g-2$ prediction.","The quoted precision meets the requirement set by the final result of the muon g-2 experiment, making the comparison between the Standard-Model prediction and experiment sensitive mainly to the hadronic vacuum polarization contribution.","The new value agrees with the 2020 phenomenological estimate but with an uncertainty reduced by more than a factor of two, strengthening the data-driven benchmark against which lattice QCD calculations are compared.","The axial-vector resonance contribution is now the largest subleading term with an assumption-based error, so new measurements of the $f_1$ and $f_1'$ transition form factors directly test the main systematic uncertainty."],"supporting_citations":[{"why":"supplies the four-point dispersive framework and the master integral for $a_\\mu^\\text{HLbL}$.","marker":"[21]"},{"why":"dispersive analyses of the pion transition form factor that determine the dominant $\\pi^0$-pole contribution.","marker":"[22,23]"},{"why":"fit the axial-vector meson transition form factors and $f_1$--$f_1'$ mixing from $e^+e^-$ and radiative-decay data.","marker":"[73,74]"},{"why":"subleading corrections to the short-distance constraints, including the OPE cancellation used in the matching.","marker":"[77–80]"},{"why":"provides the dispersive reconstruction of the vector-vector-axial-vector correlator used as OPE input.","marker":"[81]"},{"why":"is the companion paper with the full details of the evaluation summarized in this Letter.","marker":"[82]"},{"why":"is the previous phenomenological estimate whose uncertainty the present result reduces by more than a factor of two.","marker":"[6]"}],"fun_headline_variants":["Dispersive HLbL at 101.9(7.9)","Muon g-2 HLbL cut to 7.9","Hadronic light-by-light uncertainty halved","Data-driven HLbL hits 101.9(7.9)"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The calculation assumes that the axial-vector meson transition form factors follow U(3) symmetry with no more than 30% deviation; if the true symmetry breaking is larger, the central value shifts outside the quoted uncertainty.","fun_headline_variants_meta":{"raw":{"variants":["Dispersive HLbL at 101.9(7.9)","Muon g-2 HLbL cut to 7.9","Hadronic light-by-light uncertainty halved","Data-driven HLbL hits 101.9(7.9)"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000882,"raw_usage":{"total_tokens":3787,"prompt_tokens":898,"completion_tokens":2889,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":514,"completion_tokens_details":{"reasoning_tokens":2815}},"tokens_in":514,"tokens_out":2889,"duration_ms":19131,"temperature":1.0,"reasoning_tokens":2815,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T05:38:18.802534+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"New data on the $f_1$ or $f_1'$ transition form factor that deviate from the U(3)-symmetric parameterization by more than the assigned 30% would move the axial-vector contribution beyond the quoted systematic error.","supporting_citations":[],"review_version":1}