{"id":"94e890b2-40be-46d5-b1e3-c683bf9fcafd","arxiv_id":"2608.09338","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":1.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"A comprehensive review of BESIII's amplitude-analysis methodology for charmed meson decays, presenting no new measurement, derivation, or formalism.","lead":"This paper reviews how the BESIII collaboration extracts the properties of short-lived intermediate particles from measurements of charm meson decays. It is a practical methods guide covering likelihood fits, Monte Carlo normalization, machine-learning background modeling, and resonance line shapes, useful to practitioners, but it contains no new measurement or derivation.","discovery_kind":"review","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Eq. (11) inverts the signal-MC importance weight: the odds ratio should be divided by |M_gen|^2, not multiplied, so the background normalization in Eq. (12) is biased for nonuniform generation amplitudes.","rationale":"The reader correctly identified the XGBoost background modeling as a load-bearing weak spot and flagged Eq. (7) and the resolution remark as issues. However, the most concrete and easily settled error is in Eq. (11): the importance-sampling weight for the background normalization is algebraically inverted. This is not a calibration or approximation concern; it is a mathematical misweighting that biases the background shape whenever the signal MC generation amplitude is nonuniform. The reader's density-ratio assumption concern is closely related but distinct: even with a perfectly calibrated classifier, Eq. (11) as written gives the wrong answer. The resolution claim in Sec. II A is also questionable, but it is more of a conceptual approximation than a formula-level inconsistency. Since the paper is a methods review intended as a reference, a reader implementing Eqs. (11)–(12) would obtain incorrect fit results, so the paper should not be accepted without correction. The overall verdict remains conditional: the main structure is sound, but this equation and the other flagged statements must be fixed or clarified before the document is used as a methodological reference.","tokens_in":12032,"tokens_out":16240,"duration_ms":167123,"concrete_test":"Implement a one-dimensional toy: set R=1 and ϵ=1, choose a nonuniform true background B(x) and a nonconstant generation amplitude M_gen(x) (e.g., x). Train a perfect classifier so the odds ratio equals B(x) up to a constant. Generate N_gen events from |M_gen(x)|^2, pass them through a trivial 'reconstruction', and evaluate the right-hand side of Eq. (11) both as printed (×|M_gen|^2) and with the corrected weight (÷|M_gen|^2). Compare each to the analytic ∫B(x)dx. If the printed version matches ∫B(x)|M_gen(x)|^4 dx while the corrected version matches ∫B(x)dx, Eq. (11) is demonstrably wrong and should be corrected or restricted to PHSP MC.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Section II B, Eq. (11) claims that the background normalization integral, ∫ϵ(p)B_ϵ(p)R(p)dp = ∫B(p)dp, is estimated as (1/N_gen)∑_{rec}[P_BKG/P_PHSP]|M_gen|^2. By importance sampling, the reconstructed signal-MC sample is distributed as ϵ(p)|M_gen(p)|^2R(p), so the weight that removes the generation density is the odds ratio divided by |M_gen(p)|^2, not multiplied by it. With the printed weight, the sum converges to (constant) × ∫B(p)|M_gen(p)|^4 dp rather than to ∫B(p)dp. Eq. (12) then uses this biased denominator to normalize the background shape, so the term entering Eq. (6) is suppressed or enhanced in regions where the generation amplitude is large. This is an internal algebraic error independent of XGBoost calibration; it only vanishes in the special case |M_gen|=1 (PHSP MC), which is not what the text prescribes when it says the sum runs over a 'generated signal MC sample' and stresses using the same MC footprint as the signal term. Because Eqs. (11)–(12) are the mechanism by which the multidimensional background enters the likelihood, a reader copying them will obtain fit results corrupted by the choice of signal MC model.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper is a methodological review of the amplitude-analysis framework used by the BESIII Collaboration for multibody charmed-meson decays. It develops the signal probability density function and the unbinned likelihood, describes Monte Carlo normalization with PHSP and signal MC samples, discusses efficiency and resolution treatment, introduces an XGBoost-based multidimensional background reweighting procedure, and explains projection and fit-fraction evaluation. It then summarizes the resonance parametrizations used in the collaboration's hadronic and semileptonic analyses, including relativistic Breit-Wigner, Gounaris-Sakurai, Flatté, coupled-channel, K-matrix, and LASS forms.","tokens_in":12199,"tokens_out":12787,"duration_ms":135743,"significance":"If the formulas are correct, the paper would serve as a useful practical reference for amplitude analyses at BESIII, consolidating the standard isobar-model and K-matrix machinery with concrete normalization and background-modeling recipes. It is honest about the limitations of Gaussian uncertainties for fit fractions near physical boundaries and clearly distinguishes reconstructed MC from generator-level PHSP MC, which is valuable for practitioners. The paper presents no new data or measurements, so its significance rests entirely on the accuracy and reproducibility of the methodological equations; one of the load-bearing equations, Eq. (11), is erroneous as written, and the key density-ratio assumption of Section II B is not validated.","major_comments":[{"comment":"The importance-sampling estimator for the background normalization integral is written with the wrong power of |M_gen|^2. If the reconstructed signal-MC sample is distributed as ε(p)|M_gen(p)|^2 R(p), then the estimator for ∫ ε(p)B_ε(p)R(p)dp = ∫ B(p)dp should use the weight [P_BKG/P_PHSP]/|M_gen|^2, not [P_BKG/P_PHSP]·|M_gen|^2. As printed, the sum converges to a weighted integral proportional to ∫ B(p)|M_gen(p)|^2 dp, so the normalized background shape in Eq. (12) and hence the background term in Eq. (6) inherit a spurious dependence on the generator amplitude M_gen. This is an internal algebraic error that does not vanish unless |M_gen|=1, which is inconsistent with the text's instruction to use the same signal-MC footprint as the signal term in Eq. (10). Please correct Eqs. (11) and (12) and re-derive the normalization, or explicitly clarify the density of the MC sample being summed.","section":"II B, Eqs. (11)–(12)"},{"comment":"The central identification P_BKG(p)/P_PHSP(p) = B_ε(p) = B(p)/[ε(p)R(p)] is presented as a direct consequence of density-ratio estimation, but the paper gives no validation that the trained XGBoost classifier reliably recovers this ratio over the full phase space. A miscalibrated classifier, an input feature set that misses discriminating correlations, or a PHSP MC sample that does not faithfully map ε(p)R(p) would bias the background shape and propagate into every fitted amplitude. The review should state explicitly that calibration curves, closure tests, or pseudo-experiments are required for this step and should point to the BESIII analyses (e.g., Refs. [7], [29], [34]) where such validation is performed.","section":"II B"},{"comment":"Equation (7) is not unambiguously typeset: the fraction structure among N_data, wN_bkg, and N_data + w^2 N_bkg is unclear in the provided text, and no derivation or reference is given for the prefactor that is claimed to ensure correct statistical uncertainties. Since this equation is offered as an alternative background-subtracted likelihood, the authors should rewrite the formula unambiguously and either derive the prefactor from the statistical weights or cite the specific BESIII analysis where this form is used.","section":"II, Eq. (7)"}],"minor_comments":[{"comment":"Please clarify whether |M_gen| in the denominator is evaluated at the generator-level momentum or at the reconstructed momentum p_rec. As written, using p_rec in both numerator and denominator is not the standard importance-sampling weight unless resolution effects are deliberately included through a separate empirical procedure; this should be stated explicitly.","section":"II A, Eq. (10)"},{"comment":"The description of the XGBoost procedure would be more reproducible if the input feature set and training hyperparameters were listed, or if the authors stated that these are channel-dependent and referred to the original BESIII analyses for details.","section":"II B"},{"comment":"The text should make explicit that A_i, P_j, K, and ρ are matrices/vectors in the same five-channel basis (ππ, KKbar, 4π, ηη, ηη′), and that the indices i, j denote these channels throughout.","section":"III C 6, Eq. (27)"},{"comment":"The bare P used for the Cauchy principal value should be typeset as ℘ or “PV” to avoid confusion with a kinematic variable.","section":"III C 5, Eq. (26)"},{"comment":"Refs. [38]–[40] are arXiv preprints; if they have appeared in journals by the time of publication, the published versions should be cited.","section":"References"}],"recommendation":"major_revision","confidential_remarks":"The paper is heavily self-referential: most of Refs. [7], [9], [17]–[20], [28]–[40] are BESIII papers or papers by the present authors. For a methodology review this is not itself disqualifying, but the authors should be asked to anchor the new formulas—especially Eqs. (7), (11), and (12)—to specific published analyses so that the reader can verify the claimed collaboration practice. The load-bearing error in Eq. (11) is fixable in revision, and the overall scope is appropriate for a journal-style review, so I do not see this as a rejection; however, the validation discussion in Section II B needs to be substantially strengthened."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: this is a review of BESIII's amplitude-analysis machinery, not a new measurement or method. It is mostly competently assembled and would be a useful entry point for practitioners, but it has one genuine algebraic error that needs fixing before anyone should copy the recipes. Eq. (11) inverts the importance-sampling weight for the background normalization. The signal MC sample is distributed as eps(p)|M_gen(p)|^2 R(p), so to estimate the background integral ∫ B(p) dp the odds ratio P_BKG/P_PHSP, which estimates B_eps(p) = B(p)/(eps R), must be divided by |M_gen|^2, not multiplied by it. As printed, the sum converges to something like ∫ B(p)|M_gen(p)|^4 dp, which biases the background shape wherever the generation amplitude is large. This is load-bearing: Eq. (12) uses that normalization to normalize the background term entering Eq. (6). The error is independent of XGBoost calibration and only goes away in the special case |M_gen|=1, which is not what the text prescribes when it says to sum over the signal MC sample.\n\nThe paper does a number of things well. The likelihood construction, MC normalization, fit-fraction definitions, and the summary of resonance parametrizations are standard material and are mostly stated correctly. The discussion of fit-fraction uncertainties, including the warning about non-Gaussian distributions near boundaries, is honest. The heavy self-citation cluster is to be expected in a collaboration methods review and is not circular, since the cited results are published measurements.\n\nThe soft spots beyond Eq. (11): Eq. (7), the background-subtracted log-likelihood, is asserted without derivation and the second sum appears to be missing the weight w defined in that same equation, so the formula is at least mis-scaled. And the claim that evaluating |M|^2/|M_gen|^2 at reconstructed kinematics automatically folds in detector resolution is an approximation; the 'm-shaped weighting curve' is not a general property of resolution smearing. These are details a reader would copy, so they need to be corrected or qualified.\n\nWho is this for? A student or postdoc wanting a single place to see BESIII's amplitude-analysis conventions. It deserves a serious referee because it could be a useful reference, but the referee should insist on fixing Eq. (11) before publication. My verdict: send to peer review with major revisions.","headline":"A useful but flawed methods review: the background-normalization formula in Eq. (11) inverts an importance-sampling weight and must be corrected before anyone uses it as a reference.","tokens_in":12921,"tokens_out":4474,"would_cite":false,"duration_ms":43862,"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 claims to give the complete BESIII amplitude-analysis methodology, from measured charm-decay events to intermediate-resonance magnitudes, phases, fit fractions, and branching fractions.","keywords":["amplitude analysis","charmed meson decays","BESIII","unbinned maximum likelihood","isobar model","XGBoost background modeling","resonance propagators","fit fractions"],"falsifier":"Generate pseudo-experiments from a known amplitude model plus a known background density, apply the XGBoost-based likelihood of Eq. (6), and check whether the fitted magnitudes and phases recover the injected values within statistical uncertainty. A systematic miss concentrated in any phase-space region would show that the density-ratio assumption behind $B_\\epsilon$ fails.","tokens_in":11651,"feed_emoji":"⚛️","tokens_out":12342,"duration_ms":110376,"temperature":0.7,"pith_summary":"Amplitude analysis is the bridge between what a detector records — stable final-state particles — and the short-lived intermediate resonances whose couplings theory can predict. This paper lays out the complete machinery BESIII uses for charmed-meson decays: a signal density built from a coherent sum of intermediate amplitudes, normalized by Monte Carlo integration that folds in detection efficiency and resolution, and a likelihood extended with a background shape learned by an XGBoost classifier. If the account is correct, Eqs. (1)–(18) form the operational route from BESIII's $20.3\\ \\mathrm{fb}^{-1}$ dataset at $\\sqrt{s}=3.773$ GeV to fitted resonance magnitudes, phases, fit fractions, and branching fractions. The paper's value is that it makes that route explicit enough to be checked, reproduced, or modified.","feed_headline":"A single method maps 20.3 fb-1 of charm data to hidden resonances","feed_subtitle":"The paper's formulas connect the data to the resonance parameters behind each published BESIII decay.","key_machinery":"The load-bearing object is the unbinned maximum-likelihood signal density $f_S(p)=\\epsilon(p)|M(p)|^2R(p)/\\int\\epsilon(p)|M(p)|^2R(p)\\,dp$, together with the extended log-likelihood of Eq. (6). Three mechanisms carry the argument: Monte Carlo integration over PHSP, reconstructed, and signal-MC samples turns normalization, efficiency, and resolution into sums over simulated events; the XGBoost classifier's odds ratio supplies the corrected background density $B_\\epsilon(p)$; and the isobar amplitude $A_n=P_n S_n F_n^r F_n^D$ connects resonance lineshapes to data through spin factors, Blatt-Weisskopf barriers, and propagators such as Breit-Wigner, Gounaris-Sakurai, Flatt\\'e, $K$-matrix, and LASS forms.","core_discovery":"The central claim is that Eqs. (1)–(18) faithfully describe the amplitude-analysis methodology used by the BESIII Collaboration. The signal probability density is $f_S(p)=\\epsilon(p)|M(p)|^2 R(p)/\\int \\epsilon(p)|M(p)|^2 R(p)\\,dp$, and the total amplitude is $M(p)=\\sum_n c_n A_n(p)$, a coherent sum over intermediate processes. The likelihood is extended with an efficiency- and phase-space-corrected background density $B_\\epsilon(p)$ obtained from an XGBoost odds ratio $P_{\\rm BKG}(p)/P_{\\rm PHSP}(p)$. Monte Carlo integration over reconstructed samples folds in detector efficiency and resolution without analytic modeling; fit fractions are defined at generator level to isolate dynamics from acceptance; and the semileptonic decay rate is factorized into hadronic and leptonic currents in five kinematic variables. The authors present this formula chain as the map from measured charm-decay events to the intermediate-resonance parameters BESIII publishes.","pith_inferences":["A validation the authors leave out: calibration curves or pseudo-experiment closure tests for the XGBoost odds ratio; a miscalibrated classifier would shift every fitted phase and magnitude in the same direction.","The same density-ratio normalization trick could be transplanted to unbinned amplitude fits at other charm or beauty facilities, where background compositions differ but the mathematics of Eq. (6) does not.","The fit-fraction uncertainty prescription (Gaussian width from covariance-matrix sampling) is explicitly flagged by the paper as fragile near boundaries; a natural extension is to report full posterior shapes instead of single Gaussian widths."],"forward_implications":["Every published BESIII branching fraction or fit fraction from these channels can be traced to the likelihood, amplitude, and background definitions in Eqs. (1)–(18).","Generator-level PHSP Monte Carlo gives fit fractions that are independent of detector acceptance, so comparisons between experiments are meaningful only when the same propagator and production-vector conventions are used.","Reconstructed signal-MC normalization removes the need for multidimensional resolution convolution: detector smearing is folded in by evaluating $|M(p^{\\rm rec})|^2/|M_{\\rm gen}(p^{\\rm rec})|^2$.","The XGBoost background model lets arbitrarily complex multidimensional background shapes enter the fit, provided the odds ratio is a faithful density ratio.","Simultaneous fits across multiple $D$ and $D_s$ channels can be performed on the same likelihood footing, which is the paper's route to probing $K^0_S$–$K^0_L$ asymmetries and $U$-spin breaking."],"supporting_citations":[{"why":"Supplies the kinematic phase-space counting, resonance quantum numbers, masses, widths, and Blatt-Weisskopf barrier definitions used throughout the amplitude formalism.","marker":"[1]"},{"why":"Defines the 20.3 fb^-1 dataset at sqrt(s)=3.773 GeV that the described analysis framework is built to fit.","marker":"[3]"},{"why":"Provides the detector simulation through which Monte Carlo samples are reconstructed, making the efficiency and resolution folding in Eqs. (8)-(10) operational.","marker":"[4]"},{"why":"Supplies the XGBoost binary classifier whose odds ratio is used as the efficiency- and phase-space-corrected background density in Eqs. (11)-(12).","marker":"[5, 6]"},{"why":"The D_s^+ -> K0S K0L pi+ analysis used as the illustrative example of the XGBoost background-shape projection.","marker":"[7]"},{"why":"Supplies the covariant tensor isobar amplitude formalism, including spin-projection factors and Blatt-Weisskopf barriers.","marker":"[8]"},{"why":"Supplies the semileptonic D -> M1 M2 l nu parametrization with five kinematic variables and the intensity decomposition used in Eq. (18).","marker":"[9]"},{"why":"Supplies the Gounaris-Sakurai lineshape used for broad vector resonances such as rho(770).","marker":"[12]"},{"why":"Supplies the Flatte propagator with channel couplings used for f0(980).","marker":"[16]"},{"why":"Supplies the K-matrix and LASS parametrizations that enforce unitarity for the pi pi and K pi S-waves.","marker":"[24, 25]"}],"fun_headline_variants":["Charm decay formula chain maps data to resonances","BESIII amplitude analysis: data to hidden resonances","One unified method decodes charm decay data at BESIII","From 20 fb-1 charm data to resonance parameters via formulas","Amplitude analysis: the key to BESIII's charm decay results"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The whole background treatment depends on the XGBoost odds ratio being the true density ratio of background to phase-space events, and the paper does not show a calibration or closure test of that equality.","fun_headline_variants_meta":{"raw":{"variants":["Charm decay formula chain maps data to resonances","BESIII amplitude analysis: data to hidden resonances","One unified method decodes charm decay data at BESIII","From 20 fb-1 charm data to resonance parameters via formulas","Amplitude analysis: the key to BESIII's charm decay results"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000161,"raw_usage":{"total_tokens":1208,"prompt_tokens":887,"completion_tokens":321,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":503,"completion_tokens_details":{"reasoning_tokens":234}},"tokens_in":503,"tokens_out":321,"duration_ms":4050,"temperature":1.0,"reasoning_tokens":234,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T19:07:59.541043+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Generate pseudo-experiments from a known amplitude model plus a known background density, apply the XGBoost-based likelihood of Eq. (6), and check whether the fitted magnitudes and phases recover the injected values within statistical uncertainty. A systematic miss concentrated in any phase-space region would show that the density-ratio assumption behind $B_\\epsilon$ fails.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the kinematic phase-space counting, resonance quantum numbers, masses, widths, and Blatt-Weisskopf barrier definitions used throughout the amplitude formalism."},{"cited_title":"Their detailed parametrizations and numerical values are taken from Ref","cited_arxiv_id":null,"evidence_quote":"Defines the 20.3 fb^-1 dataset at sqrt(s)=3.773 GeV that the described analysis framework is built to fit."},{"cited_title":"Agostinelliet al.(GEANT4 Collaboration),Nucl","cited_arxiv_id":null,"evidence_quote":"The D_s^+ -> K0S K0L pi+ analysis used as the illustrative example of the XGBoost background-shape projection."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the semileptonic D -> M1 M2 l nu parametrization with five kinematic variables and the intensity decomposition used in Eq. (18)."},{"cited_title":"Zhang, B","cited_arxiv_id":null,"evidence_quote":"Supplies the Gounaris-Sakurai lineshape used for broad vector resonances such as rho(770)."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the Flatte propagator with channel couplings used for f0(980)."}],"review_version":1}