{"id":"5a4fe274-3efc-451e-8bb9-1142b60ff65c","arxiv_id":"2607.20777","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"Near the boundary of large complex structure, the first worldsheet instanton can displace flux vacua significantly while higher instantons stay negligible, and such vacua are common in a bounded two-modulus scan.","lead":"This paper finds type IIB string flux vacua near the boundary of the large complex structure region where the first instanton correction, not just the polynomial part, controls the vacuum position. The same effect appears in a non-negligible fraction of the scanned vacua, which matters for building controlled string models away from asymptotic limits.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The d'=100 single-degree check in Eq. (3.4) is an unvalidated proxy for the full instanton tail; App. B's own counterexample shows low-degree convergence is not sufficient, so the 'higher instantons negligible' claim needs a direct tail test for each reported vacuum.","rationale":"The reader's weakest_assumption already identifies the d'=100 truncation check as load-bearing, and Appendix B provides a concrete failure mode where low-degree convergence ratios are misleading. This is the most fundamental threat to the paper's central claim because it applies not just to the statistical extrapolation but to the very identification of the reported vacua as 'controlled by perturbative + first instanton only.' The statistical representativeness issue is secondary: even if the ensemble is biased, the explicit examples would remain valid; but if the tail check is insufficient, the examples themselves could be mischaracterized. My proposed test directly uses the degree-300 GV data already present in the paper to compute the full tail and to re-solve with deeper truncations. I recommend keeping the reader's CONDITIONAL verdict: the concern is real and addressable, but there is independent evidence (the d=10-to-d=20 stability in Tabs. 1–2, the exponential fit, and the stringent d'=100 threshold) that the explicit examples are likely safe. The conditionality should remain until the direct tail test is performed and reported.","tokens_in":20861,"tokens_out":7190,"duration_ms":60029,"concrete_test":"For each vacuum in Tabs. 1–7 and for a random sample of the statistical ensemble, compute the tail sum S_tail = Σ_{d=11}^{300} |ΔF^(d)_inst|/|F^(10)| using the degree-300 GV invariants from App. A. Also solve the F-term equations at d_max=100 and d_max=300 and compare the resulting moduli and τ. If any selected vacuum has S_tail > 10^-10 or a moduli shift > 10^-6 between the d_max=100 and d_max=300 solutions, the d'=100 criterion is insufficient and the 'higher instantons negligible' claim fails for that vacuum. If all pass, the concern is resolved.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim—that the vacuum is determined by the perturbative term plus the first instanton, with all higher instantons negligible—rests on the selection criterion (3.4): |ΔF^(100)_inst|/|F^(d)| ≤ 10^-10, plus the individual degree-1 check (3.5). This is a single-degree proxy for the full tail; it does not bound the sum S_tail = Σ_{d=11}^∞ |ΔF^(d)_inst|/|F^(10)|, and the paper never reports the actual tail sum for the reported vacua. Appendix B explicitly demonstrates that low-degree convergence diagnostics can be deceptive: the flux configuration (B.2) has ε_ratio^(10) ≤ 10^-6 and yet the instanton series loses control at d≈50. The authors use d'=100 to catch such cases, but there is no argument that d'=100 is a sufficient diagnostic for every vacuum in the ensemble, especially those with Im z_i near the convergence boundary (Im z ≈ 0.862 from the exponential fit of App. B). If the tail beyond d=100 is not negligible for some selected vacuum, the characterization of the minimum and the statistical 'quite common' claim would be quantitatively wrong. The explicit low-shift examples in Tabs. 1 and 2 are likely safe, but the ensemble statistics inherit this unvalidated proxy.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies supersymmetric flux vacua in the large complex structure (LCS) patch of a two-modulus Calabi-Yau orientifold, including worldsheet instanton corrections to the prepotential. It constructs explicit examples in which the first instanton correction produces a large displacement of the perturbative minimum while higher instantons are claimed to be negligible, and it presents a statistical scan suggesting that such vacua are 'quite common'. It also reports examples with instanton-generated multiplicity, monodromy shifts, destabilization of perturbative minima near the LCS boundary, and instanton-induced minima. The main technical tools are a degree-10 truncated prepotential with degree-300 Gopakumar-Vafa data and convergence diagnostics based on high-degree ratios.","tokens_in":21290,"tokens_out":7335,"duration_ms":64747,"significance":"If the central claim holds, the paper provides a controlled exploration away from the deep LCS asymptotics, a regime usually avoided. The explicit construction with high-order GV invariants, the deformation-tracking method, and the honest Appendix B counterexample are valuable methodological contributions. The explicit examples in Tabs. 1-2 are convincing, with independent d=10 to d=20 convergence checks. However, the statistical analysis supports the 'quite common' claim only under a convergence proxy that is not fully validated, and the scan is heavily bounded; the abstract and conclusions overstate the result as it stands.","major_comments":[{"comment":"The selection criterion (3.4) bounds only |ΔF^(100)_inst|/|F^(d)|; it does not bound the tail sum S_tail = Σ_{d=11}^∞ |ΔF^(d)_inst|/|F^(10)|. Appendix B's counterexample (B.2) shows that ε_ratio^(10) ≤ 10^-6 can coexist with loss of control at d≈50, and no argument is given that d'=100 is a sufficient diagnostic for every selected vacuum, especially near the Im z ≈ 0.862 boundary from the exponential fit. The explicit Tabs. 1-2 have independent d=10 to d=20 convergence checks, but the ensemble claim in Sec. 4 inherits the unvalidated proxy. Please report actual tail estimates, e.g. ε_reference (B.1), for all selected vacua and exclude vacua in the non-convergent region of Fig. 7.","section":"Sec. 3.1 (Eq. 3.4), App. B"},{"comment":"The statement that the phenomenon is 'statistically quite common' is supported only by a scan with N_flux ≤ 10, Im(z_i) ∈ [0.5,5], further restricted to min{Im(z_i)}<1. The paper does not test how the 9.3% fraction depends on N_max or on the search region U, nor does it compare with any prior expectation. The abstract and conclusions present the claim without these qualifications. Please either weaken the wording to 'common within this bounded ensemble' or add robustness checks showing that the fraction is stable under reasonable variations of the scan parameters.","section":"Sec. 4 and Abstract"}],"minor_comments":[{"comment":"The text defines d=0,d_max for the condition but the first condition is written with d=0; please clarify whether F^(0) denotes F_pert and define the notation explicitly.","section":"Sec. 3.1, Eq. (3.4)"},{"comment":"The stated validity condition δφ^T M δφ + b·δφ = 0 is not the standard quadratic-stationarity condition (which would be M δφ = -b); please correct or clarify this equation.","section":"Sec. 3.2.1, footnote 5"},{"comment":"The caption should explicitly state that the 9.3% and 1.2% fractions correspond to thresholds ||Δz||/||z_P|| ≥ 0.1 and ≥ 0.3; presently this is only given in the text.","section":"Sec. 4, Fig. 4 caption"},{"comment":"The GV invariants grow extremely rapidly with degree; a brief note on the expected exponential growth and on how the invariants were cross-checked would aid reproducibility.","section":"Appendix A, Eq. (A.6)"},{"comment":"The Acknowledgements thank CYTools and JAXVacua, but no statement is made about availability of the scan data or code. A data/code availability statement would strengthen the paper.","section":"General"}],"recommendation":"major_revision","confidential_remarks":"This is a careful computational paper with interesting explicit examples. The main risk is the statistical overreach and the unvalidated tail proxy for the ensemble claim. The explicit Tabs. 1-2 and the App. B convergence caution are valuable. I would be willing to accept after the statistical claim is qualified or the tail estimates are provided for the full ensemble."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Two things to know. The explicit two-modulus examples are the real content, and they hold up: the deformation-from-perturbative-vacua procedure plus the tiny d=1 to d=10 shifts make the claim 'first instanton plus perturbative determines the minimum' credible for those configurations. The statistical claim, 'quite common', is weaker than the paper implies — the scan is small, bounded, and lacks even a report of how many vacua were found, so 9.3% is a number in search of an error bar.\n\nWhat's good: the paper computes GV invariants to degree 300 and uses a d=100 control check; it includes a rare piece of honesty in App. B, showing a flux configuration where low-degree convergence diagnostics fail. The monodromy and multiplicity phenomena (Sec. 3.2.2–3.2.4) are genuinely new and well-illustrated by the continuous deformation tracking. The linear-response diagnostic v=-M^{-1} b, and the correlation of |v| with actual displacement in Fig. 6, is a useful tool that appears to be a posteriori rather than engineered.\n\nSoft spots: (1) The tail-bound criterion (3.4) checks the single degree-100 term, not the sum of the tail. App. B shows a configuration where low-degree ratios are small but the series breaks down at d~50, so the d'=100 check is doing real work; but the paper never exhibits the actual tail sum for the reported vacua. This is not fatal for Tabs. 1–2, where the degree-10 vs degree-20 shifts are below 10^-8, but it is exactly the kind of gap that matters for the ensemble statistics. (2) The 'quite common' claim is based on a scan with N_flux<=10 and Im(z) in [0.5,5], with the subset min Im(z)<1; the total number of vacua is not given, so the reader cannot judge sampling uncertainty. A bootstrap over the ensemble would be easy and should be added. (3) No code or data is shipped, which hampers independent verification of the scan.\n\nBottom line: this is a serious paper with a real new phenomenon, and those working on explicit moduli stabilization and landscape statistics will get value from it. For the explicit examples, I'd trust the conclusion. For the frequency claim, the evidence is suggestive, not 'common' as established. A referee should ask for a tail-sum check and a fuller statistical report, but this deserves to go to review, not desk rejection.","headline":"Concrete two-modulus vacua where the first instanton dominates — but the frequency claim rides on a proxy tail check and an underspecified scan.","tokens_in":21738,"tokens_out":5557,"would_cite":true,"duration_ms":41703,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":null,"created_at":"2026-08-01T09:25:08.804635+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":null,"supporting_citations":[],"review_version":1}