{"id":"724ee396-888c-4e96-bffa-21ae3b83a5ab","arxiv_id":"2412.05989","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":8,"one_line_summary":"Using the latest Belle and Belle II data, CLN, BGL, and HQET fits give consistent exclusive |Vcb| values near 39.7 times 10^-3, confirming the persistent 2 to 3 sigma gap with the inclusive value.","lead":"This paper re-extracts the CKM matrix element Vcb from B to D* decays using the newest Belle, Belle II, lattice QCD, and light-cone sum rule data, testing three standard form-factor parameterizations. It finds all three agree on Vcb around 39.5 to 39.9 times 10^-3, still about 2 to 3 sigma below the inclusive determination, and uses a Bayesian method to argue that missing higher-order terms in the heavy quark expansion cannot close that gap.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The Bayesian truncation analysis behind the 'higher-order effects cannot resolve the Vcb puzzle' claim can return a near-zero uncertainty by making the large NLO-to-NNLO shift the reference scale and then dropping it; this needs a direct test.","rationale":"The paper's central numerical result—consistent exclusive |Vcb| values around 39.5-39.9×10^-3 that are 2-3σ below inclusive determinations—is a plausible reading of the current data, and the cross-parameterization consistency is supported by the fits in Tables I-III. The genuinely new claim, that a Bayesian analysis shows higher-order HQET effects cannot resolve the puzzle, is not supported by the presented convergence evidence. The LO, NLO, and NNLO sequence for |Vcb| is 36.57, 36.69, and 39.48, i.e., a 7.6% jump at the last step, and the NLO fit is statistically poor; this is not a series from which one can infer a negligible tail. The X_ref definition in Eq. (20) is the mechanism: by taking a maximum of the LO value and the scaled LO→NLO and NLO→NNLO shifts, it can set the largest correction equal to 1, and the 'cm=1' rule then removes that term from the convergence diagnostic. For the |Vcb| series, the NLO→NNLO term is approximately 85×10^-3 and dominates X_ref, so c2=1 is discarded; with only c0≈0.43 and c1≈0.008 left, the posterior is artificially narrow. Even if the authors applied the procedure to differential rates rather than to |Vcb| directly, the burden is on them to show that the derived |Vcb| truncation uncertainty is not an artifact of this scale choice. The abstract's LFU-violation statement is also stronger than the 1.8-1.9σ tensions support, but that is secondary. The reader's weakest assumption identifies the same issue; my check would settle whether the Bayesian claim survives. I therefore endorse the CONDITIONAL verdict without moving it.","tokens_in":27012,"tokens_out":10078,"duration_ms":96694,"concrete_test":"Re-run the Bayesian procedure with |Vcb| itself as the observable: compute the three candidates in Eq. (20) from Table IV (36.57, 36.69, 39.48). If X_ref is set by |X_NLO-X_NNLO|/Q^2 ≈ 85.4 and c2=1 is excluded from c_k^2, repeat the posterior calculation with c2 retained (or with X_ref=|X_NNLO|). If the 68% truncation interval on |Vcb| then exceeds ≈0.5×10^-3, the conclusion that higher-order effects are negligible and cannot close the 2-3σ gap is not supported.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim that missing higher-order HQET terms cannot resolve the Vcb puzzle rests on a Bayesian truncation error that is effectively zero. This result is driven by the scale prescription in Eq. (20), not by evidence that the series has converged. The three candidates in Eq. (20) are |X_LO|, |X_LO-X_NLO|/Q, and |X_NLO-X_NNLO|/Q^2, and the coefficient equal to 1 is excluded from the posterior as the 'cm=1' term. For |Vcb|, Table IV gives 36.57, 36.69, and 39.48 at LO, NLO, and NNLO: the third candidate is |39.48-36.69|/0.1807^2 ≈ 85.4 and dominates X_ref, so c2=1 and is dropped. The remaining coefficients c0≈0.43 and c1≈0.008 contain almost no convergence information, yielding Δ≈0. More generally, the NLO fit is poor (χ^2/d.o.f. = 324.7/171) and the NNLO correction moves |Vcb| by 7.6%; a model that returns a negligible tail from such a sequence is not a reliable basis for excluding higher-order effects.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper revisits the extraction of |Vcb| from exclusive B -> D* l nu decays using the recently published Belle 2023 and Belle II differential distributions together with lattice QCD and light-cone sum rule form factors. It analyzes three form-factor parameterizations (CLN, BGL, HQET) and finds consistent exclusive values, |Vcb|_CLN = (39.69 +/- 0.56) x 10^-3, |Vcb|_BGL = (39.90 +/- 0.55) x 10^-3, and |Vcb|_HQET = (39.48 +/- 0.59) x 10^-3, which deviate from recent inclusive determinations by about 2-3 sigma. The paper then employs a Bayesian truncation-error framework to estimate the effect of missing higher-order HQET terms and concludes that this effect is negligible, so the Vcb puzzle persists. Finally, using form-factor parameters fitted only to LQCD and LCSR data, it predicts R_D*, F_L^D*, and P_tau^D* and claims that lepton-flavour universality is violated in b -> c tau nu transitions.","tokens_in":27383,"tokens_out":9079,"duration_ms":86492,"significance":"If the conclusions are robust, the parameterization-independence of |Vcb| across CLN, BGL, and HQET is a valuable update to the Vcb puzzle literature, and the use of LQCD/LCSR-only inputs for the R_D*, F_L^D*, and P_tau^D* predictions avoids the circularity of fitting to the very observables being predicted. The central numerical fits are standard and the dataset is comprehensive. However, the paper's most distinctive claim - that higher-order HQET corrections cannot resolve the Vcb puzzle - rests on a Bayesian truncation estimate that has a structural weakness; and the final lepton-flavour-universality conclusion is drawn from 1.8-1.9 sigma tensions, which is stronger than the evidence warrants.","major_comments":[{"comment":"The near-zero Bayesian truncation uncertainty for |Vcb| is driven by the definition of X_ref in Eq. (20), not by evidence of convergence. With Q = max{0.0716, 0.0522, 0.1807} = 0.1807, the three candidates in Eq. (20) are |X_LO| = 36.57, |X_LO - X_NLO|/Q = 0.12/0.1807 approx 0.66, and |X_NLO - X_NNLO|/Q^2 = 2.79/0.1807^2 approx 85.4 (in the units of Table IV). The third term dominates, so X_ref = 85.4, and c2 = (X_NNLO - X_NLO)/(X_ref Q^2) = 1 by construction. The posterior in Eq. (24) then excludes c2 as the 'cm = 1' term, leaving only c0 approx 0.43 and c1 approx 0.008. These two coefficients contain essentially no information about whether the NNLO shift was anomalously large or part of a normal fluctuation, so the resulting Delta near zero is an artifact of the normalization. The observed NLO-to-NNLO shift of 2.79 x 10^-3 (7.6%) is absorbed into X_ref instead of being treated as a correction that should inform the truncation error. I ask the authors to provide a direct test: repeat the estimate with X_ref = |X_LO| or X_ref = |X_NNLO|, or with c2 included in the posterior, and report the resulting truncation error on |Vcb|. This test is essential because the paper's conclusion that higher-order effects cannot eliminate the deviation relies on this near-zero Bayesian error.","section":"Sec. III B, Eq. (20), Table IV"},{"comment":"The HQET series is not visibly converged at NNLO in this analysis. The NLO fit has chi^2/d.o.f. = 324.68/171, whereas the NNLO fit has chi^2/d.o.f. = 222.80/167, and the NNLO correction moves |Vcb| by 7.6%. A model that returns a negligible tail from such a sequence is not a reliable basis for the claim that higher-order contributions are 'completely negligible'. At minimum, the authors should provide a convergence diagnostic - for example, the values of c0, c1, and c2 under alternative normalizations, or a stability check of the Bayesian error against the choice of expansion parameter Q in Eq. (21). Without such a check, the statement that higher-order HQET effects cannot solve the Vcb puzzle is not supported.","section":"Table IV, Sec. III B"},{"comment":"The conclusion that 'lepton-flavour universality violations still exist in the b -> c tau nu transitions' is stronger than the numerical results warrant. The predicted R_D* = 0.262 +/- 0.003 differs from the HFLAV value by about 1.8 sigma, and the predicted F_L^D* differs from the Belle result by about 1.9 sigma, while P_tau^D* is consistent. Tensions at the 1.8-1.9 sigma level do not establish the existence of lepton-flavour-universality violation. The abstract and summary should be reworded to say that the predictions show mild tensions, unless a combined significance that properly accounts for correlations is provided.","section":"Abstract, Sec. III C, Sec. IV"},{"comment":"The re-fit procedure that 'incorporates their Bayesian uncertainties' into the data covariance is self-referential in an important way. The Bayesian uncertainties are computed from the same LO, NLO, and NNLO fits whose central values are then used to generate the final Bayesian fit, and because X_ref in Eq. (20) is chosen so that the largest known increment defines cm = 1 and is then excluded, the added uncertainties are near zero by construction. The final parameter uncertainty in the 'Bayesian results' column of Table IV therefore does not provide an independent propagation of truncation uncertainty. A cleaner approach would be to add a nuisance parameter with a scale set by X_ref * Delta_k to the |Vcb| extraction, or to show explicitly how the posterior for |Vcb| changes when the truncation error is included as a separate source.","section":"Sec. III B, final paragraph"}],"minor_comments":[{"comment":"Please clarify the units and dimensionality of X_ref: since X in Eq. (18) is the observable, X_ref carries the same units, and the coefficients cn are dimensionless. The text should state this explicitly, especially because Eq. (20) combines X_LO and differences divided by powers of Q.","section":"Sec. II C, Eq. (20)"},{"comment":"The statement that h = 10 is chosen 'based on comprehensive numerical tests' is not supported by any results in the paper. A short table or figure showing convergence of the Bayesian uncertainty as a function of h would make this choice reproducible.","section":"Sec. II C, after Eq. (24)"},{"comment":"The treatment of correlations within and between the LQCD data sets (FNAL/MILC, HPQCD, JLQCD) and between the Belle and Belle II measurements is not described. If only diagonal uncertainties were used for the 153 or 179 fitted data points, the reported parameter uncertainties could be underestimated. Please state the covariance treatment explicitly.","section":"Sec. III A, datasets"},{"comment":"There are several typographical issues: 'Beysian' in Sec. I should be 'Bayesian'; 'LSCR' in the text preceding Table VIII should be 'LCSR'; and 'the the preferred' appears in Sec. III C. These should be corrected in a revision.","section":"Various"}],"recommendation":"major_revision","confidential_remarks":"The numerical fits and the parameterization-independence result are solid and well within the scope of the journal. The main obstacle to acceptance is the Bayesian truncation analysis: the near-zero error on |Vcb| appears to be an artifact of Eq. (20), and this issue is load-bearing for the 'higher-order effects cannot solve the puzzle' claim. If the authors can add a direct robustness test that shows the truncation estimate is not sensitive to the X_ref prescription, the paper would be suitable; otherwise the Bayesian conclusion should be withdrawn or heavily qualified. The LFU conclusion should also be softened."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The headline result is solid: using the latest Belle and Belle II differential distributions plus the newest LQCD and LCSR form factors, the three parameterizations give consistent exclusive values around (39.5–39.9) × 10^-3, and the 2–3σ gap from inclusive determinations persists. That directly contradicts older claims of parameterization dependence and is worth knowing. The data compilation is careful, the fits are standard, and the SM predictions for R_D*, F_L^D*, and P_tau^D* are genuine predictions — form-factor parameters come from LQCD and LCSR only, not from the experimental observables being predicted.\n\nThe soft spot is the Bayesian truncation analysis in Sec. II C and its use in Sec. III B. The claim that higher-order HQET effects cannot resolve the Vcb puzzle rests on a Bayesian uncertainty for |Vcb| that is essentially zero. That zero is not evidence of convergence. For |Vcb|, the LO, NLO, and NNLO values are 36.57, 36.69, and 39.48 (Table IV). The third candidate in Eq. (20), |X_NLO − X_NNLO|/Q^2, dominates X_ref, so c2 becomes 1 and is excluded from the posterior as the “cm = 1” term. The remaining coefficients, c0 ≈ 0.43 and c1 ≈ 0.008, contain almost no convergence information, and Δ comes out tiny. The same mechanism is visible in the worked example: Xref selects the smallest-difference term, the largest coefficient is dropped, and the reported error is small. This is a scale choice, not a test of higher-order contributions. The NLO fit is also poor (χ²/dof ≈ 1.9), so the sequence is not obviously convergent; a 7.6% shift between NLO and NNLO should not yield a near-zero truncation error.\n\nA secondary issue: the abstract says lepton-flavour universality violations “still exist,” but the tensions with HFLAV and Belle are only 1.8–1.9σ. That is an overstatement.\n\nI would send this to a serious referee. The central numerical result and the parameterization consistency are valuable and deserve publication. The Bayesian section needs either a direct validation of Eq. (20) against the observed coefficient pattern or removal of the claim that higher-order effects cannot resolve the puzzle. With that fixed, this is a useful contribution to the Vcb literature.","headline":"Consistent |Vcb| from CLN, BGL, and HQET with the newest data is a solid result, but the Bayesian truncation analysis that claims higher-order HQET terms cannot solve the puzzle is an artifact of the scale prescription in Eq. (20), not evidence of convergence.","tokens_in":27974,"tokens_out":2165,"would_cite":true,"duration_ms":20381,"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":"Using the newest Belle and Belle II differential data together with lattice QCD and light-cone sum-rule form factors, the paper finds that the CLN, BGL, and HQET parameterizations give mutually consistent values of |Vcb| around 39.5–39.9…","keywords":["Vcb puzzle","B to D* semileptonic decays","CKM matrix element","HQET form factors","CLN parameterization","BGL parameterization","Bayesian truncation uncertainty","lepton flavour universality"],"falsifier":"Find the next-order shift: a future fit that includes $O(1/m_c^3)$ or NNNLO HQET corrections (or lattice determinations of the next subleading Isgur–Wise functions) should move $|V_{cb}|$ by an amount comparable to the observed NLO-to-NNLO shift of about $2.8\\times10^{-3}$ (36.69 to 39.48) if the near-zero Bayesian error is wrong; if it moves by only a few $0.1\\times10^{-3}$, the paper's conclusion is supported. Alternatively, re-run the Bayesian estimate with $X_{\\rm ref}$ defined as $|X_{\\rm LO}|$ or as the maximum of the three unnormalized scales; if the resulting truncation uncertainty grows to the size of the inclusive-exclusive gap (about $2\\times10^{-3}$), the near-zero result is an artifact of the scale choice. A new Belle II measurement that raises the exclusive central value by roughly $2\\times10^{-3}$ would also dissolve the puzzle.","tokens_in":26742,"feed_emoji":"⚛️","tokens_out":13274,"duration_ms":115459,"temperature":0.7,"pith_summary":"The paper tries to establish whether the long-standing discrepancy between exclusive and inclusive determinations of the CKM element $|V_{cb}|$ stems from the way $B\\to D^*$ form factors are parameterized, from neglected higher-order corrections, or from something more interesting. Fitting the latest Belle and Belle II differential data together with recent lattice QCD and light-cone sum-rule form factors, the three standard parameterizations (CLN, BGL, and HQET) return mutually consistent values, $|V_{cb}|=(39.69\\pm0.56)\\times10^{-3}$, $(39.90\\pm0.55)\\times10^{-3}$, and $(39.48\\pm0.59)\\times10^{-3}$, respectively. These sit about 2–3 standard deviations below the inclusive determinations, so the puzzle is reconfirmed rather than explained away. A Bayesian estimate of the terms missing from the truncated heavy-quark expansion gives an almost vanishing truncation error for $|V_{cb}|$, indicating that higher-order HQET contributions cannot close the gap. The paper also predicts the observables $R_{D^*}$, $F_L^{D^*}$, and $P_\\tau^{D^*}$ from form-factor inputs alone and reads the resulting tensions with current measurements as evidence that lepton-flavour universality violation in $b\\to c\\tau\\nu$ persists.","feed_headline":"Exclusive Vcb stays 2-3 sigma below inclusive in new fits","feed_subtitle":"All three parameterizations agree near 39.6 x 10^-3; Bayesian errors rule out higher-order fixes.","key_machinery":"The argument is carried by three form-factor parameterizations and a Bayesian truncation-error model. CLN is a compact dispersion-relation and heavy-quark expansion using four parameters; BGL is a model-independent $z$-expansion with Blaschke factors and weak-unitarity constraints; HQET(2/1/0) expands the leading, subleading, and sub-subleading Isgur–Wise functions in powers of $z$, with 2, 1, and 0 $z$-terms respectively. The Bayesian machinery treats the truncated expansion of any observable as $X=X_{\\rm ref}\\sum_n c_n Q^n$, with $Q=\\max\\{\\alpha_s/\\pi, \\bar\\Lambda/(2m_b), \\bar\\Lambda/(2m_c)\\}$, defines $X_{\\rm ref}$ through Eq. (20) as the largest of the LO, NLO, and NNLO scales, and constructs a posterior for the dimensionless remainder $\\Delta_k$. From the known coefficients $c_1$, $c_2$ this yields the truncation uncertainty $X_{\\rm ref}\\Delta_k$, which for $|V_{cb}|$ comes out close to zero and is then folded into a refit of all parameters.","core_discovery":"The central claim is that, with the newest data set, the $V_{cb}$ puzzle is real and independent of form-factor parameterization. Fitting the Belle 2023 and Belle II differential distributions together with the latest non-zero-recoil lattice QCD results and the two recent LCSR computations, the CLN, BGL, and HQET(2/1/0) parameterizations agree on $|V_{cb}|$ at the level of $39.5$–$39.9$ in units of $10^{-3}$, with uncertainties near $\\pm0.55$–$0.59$. The same fits leave a gap of about 2–3$\\sigma$ relative to the inclusive values $(41.69\\pm0.63)\\times10^{-3}$ and $(41.97\\pm0.48)\\times10^{-3}$. Applying a Bayesian truncation-error analysis to the HQET expansion, the paper concludes that neglected terms beyond $O(1/m_c^2)$ cannot account for the deviation, so the discrepancy has to be resolved by improved data or by new physics. Using only lattice and sum-rule inputs, it then predicts the Standard Model values of $R_{D^*}$, $F_L^{D^*}$, and $P_\\tau^{D^*}$, finding tensions with the current HFLAV average for $R_{D^*}$ and with the Belle measurement of $F_L^{D^*}$, which it interprets as a persistent hint of lepton-flavour universality violation.","pith_inferences":["The near-zero Bayesian truncation error may be an artifact of the $X_{\\rm ref}$ choice: the large NLO-to-NNLO jump in $|V_{cb}|$ (36.69 to 39.48) is divided by $Q^2\\simeq0.033$ before being compared to the other scales, so a robustness check with alternative $X_{\\rm ref}$ definitions is warranted before concluding that higher-order effects are truly negligible.","If the $R_{D^*}$ and $F_L^{D^*}$ tensions survive future high-luminosity measurements, the $b\\to c\\tau\\nu$ transition would become the strongest single hint of new physics, and the paper's form-factor-only predictions give a concrete target ($R_{D^*}\\sim0.262$) to test.","A direct test of the Bayesian assumption would be to estimate the next coefficient $c_3$ from a model NNNLO HQET fit or from lattice calculations of the next subleading Isgur–Wise functions; a large $c_3$ would revise the 'near-zero' truncation error.","The paper's decision not to rescale the NLO form factors to the lattice value at zero recoil explains part of the difference from earlier NLO HQET results; a systematic comparison of rescaling choices could help separate genuine higher-order effects from fitting conventions."],"forward_implications":["Parameterization dependence is ruled out as the source of the puzzle: CLN, BGL, and HQET fits to the same combined data agree, so future work should target experimental systematics, form-factor inputs, or new physics.","The 2–3$\\sigma$ gap between exclusive and inclusive $|V_{cb}|$ persists under the newest Belle and Belle II data, so the puzzle is not a relic of older measurements.","Missing higher-order HQET terms cannot close the gap, meaning the exclusive determination's uncertainty is dominated by data and form-factor errors rather than by truncation of the heavy-quark expansion.","The Standard Model predictions for $R_{D^*}$, $F_L^{D^*}$, and $P_\\tau^{D^*}$ derived only from lattice and sum-rule inputs provide benchmarks for future measurements; the predicted $R_{D^*}=0.262\\pm0.003$ sits about $1.8\\sigma$ below the current experimental average.","Combining $B\\to D$ and $B\\to D^*$ constraints in the HQET framework, flagged by the paper as future work, is the natural next step to sharpen the exclusive $|V_{cb}|$ value."],"supporting_citations":[{"why":"Supplies the new hadronic-tagged Belle differential B->D* l nu distributions that anchor the exclusive fits.","marker":"[29]"},{"why":"Supplies the Belle II differential distributions used alongside Belle in every fit scenario.","marker":"[30]"},{"why":"Provides the FNAL/MILC non-zero-recoil lattice form factors hV, hA1, hA2, hA3 included in the 'Data+LQCD' fits.","marker":"[31]"},{"why":"Provides the HPQCD non-zero-recoil lattice form factors used to constrain the fits.","marker":"[32]"},{"why":"Provides the JLQCD lattice form factors g, f, and F1 used in the non-zero-recoil constraints.","marker":"[33]"},{"why":"Provides the earlier LCSR B->D* form factors used as input.","marker":"[22]"},{"why":"Provides the newer LCSR results for V, A1, A2 that are added in the 'Data+LQCD+LCSR' scenario.","marker":"[34]"},{"why":"Defines one of the inclusive |Vcb| determinations that the exclusive results are compared against.","marker":"[55]"},{"why":"Defines the other modern inclusive |Vcb| determination that sets the benchmark for the puzzle.","marker":"[56]"},{"why":"Supplies the Bayesian framework used to estimate the HQET truncation uncertainty.","marker":"[42]"}],"fun_headline_variants":["V_cb puzzle persists: all parameterizations agree on 2-3 sigma gap","Bayesian analysis: higher-order QCD terms can't resolve V_cb tension","Exclusive V_cb remains ~40e-3; inclusive higher by 2-3 sigma","B→D* decays: LFU violation hints persist in new combined fits","Lattice and sum rules predict SM tensions, supporting new physics"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the Bayesian truncation-error model correctly measures the size of the missing higher-order heavy-quark expansion terms, even though its definition of the overall scale $X_{\\rm ref}$ and its use of only the first two known coefficients can return a near-zero truncation error when the NLO-to-NNLO shift in $|V_{cb}|$ itself is about 7.6% (from 36.69 to 39.48 in units of $10^{-3}$), so if the true higher-order corrections are comparable to that observed shift, the claim that they cannot resolve the $V_{cb}$ puzzle would fail.","fun_headline_variants_meta":{"raw":{"variants":["V_cb puzzle persists: all parameterizations agree on 2-3 sigma gap","Bayesian analysis: higher-order QCD terms can't resolve V_cb tension","Exclusive V_cb remains ~40e-3; inclusive higher by 2-3 sigma","B→D* decays: LFU violation hints persist in new combined fits","Lattice and sum rules predict SM tensions, supporting new physics"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000403,"raw_usage":{"total_tokens":2196,"prompt_tokens":1138,"completion_tokens":1058,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":754,"completion_tokens_details":{"reasoning_tokens":952}},"tokens_in":754,"tokens_out":1058,"duration_ms":9998,"temperature":1.0,"reasoning_tokens":952,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T20:07:19.743884+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Find the next-order shift: a future fit that includes $O(1/m_c^3)$ or NNNLO HQET corrections (or lattice determinations of the next subleading Isgur–Wise functions) should move $|V_{cb}|$ by an amount comparable to the observed NLO-to-NNLO shift of about $2.8\\times10^{-3}$ (36.69 to 39.48) if the near-zero Bayesian error is wrong; if it moves by only a few $0.1\\times10^{-3}$, the paper's conclusion is supported. Alternatively, re-run the Bayesian estimate with $X_{\\rm ref}$ defined as $|X_{\\rm LO}|$ or as the maximum of the three unnormalized scales; if the resulting truncation uncertainty grows to the size of the inclusive-exclusive gap (about $2\\times10^{-3}$), the near-zero result is an artifact of the scale choice. A new Belle II measurement that raises the exclusive central value by roughly $2\\times10^{-3}$ would also dissolve the puzzle.","supporting_citations":[],"review_version":1}