{"id":"4597b85c-e110-47f0-a289-437f7bdacac8","arxiv_id":"2611.17071","paper_version":1,"verdict":"ACCEPT","confidence":"HIGH","novelty_score":9.0,"correctness_risk":"low","formal_verification":"none","parameter_count":0,"one_line_summary":"A sharp two-sided characterization of Gaussian small ball probabilities is established, involving metric entropy and local widths, with applications to intrinsic volumes and random sections.","lead":"This paper provides a sharp two-sided characterization of Gaussian small ball probabilities for symmetric convex sets, expressed in terms of metric entropy and local Gaussian widths. It also establishes analogous characterizations for intrinsic volumes and the radii of random sections, unifying these geometric quantities through the lens of convexity.","discovery_kind":"first_principles","skeptic_critique":{"model":"deepseek/deepseek-v4-flash","headline":"The lower bound in Lemma 2.4 uses a covering-number duality step that may require the polar body K^o to have nonempty interior, which holds for symmetric convex bodies with nonempty interior; but the proof of Theorem 1.5 applies to all symmetric convex bodies, including those that are degenerate…","rationale":"The reader identifies convexity as the weakest assumption, but does not flag the interior condition for the duality step. The paper's own sentences (e.g., b24, b133, b275) often assume compactness or nonempty interior, but Theorem 1.5 is stated without such qualification. The proof of Lemma 2.4 uses Theorem 2.1, which is known to require nonempty interior of the polar (or equivalently, that K has nonempty interior). This is a concrete technical gap that, if unresolved, would mean the theorem is not fully proved for all symmetric convex bodies. The fix is straightforward (restrict to bodies with nonempty interior, or add a brief embedding argument), so the verdict should be CONDITIONAL rather than REJECT. The reader's verdict ACCEPT is too strong given this oversight.","tokens_in":62850,"tokens_out":650,"duration_ms":10602,"concrete_test":"Check whether the duality theorem (Theorem 2.1) is cited as requiring K to have nonempty interior. If it does, then verify that every step in Lemma 2.4 that invokes Theorem 2.1 is applied only to K with nonempty interior. If the paper intends to cover degenerate K, provide a separate argument (e.g., by embedding into the affine hull) or explicitly restrict the theorem to bodies with nonempty interior.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The central result Theorem 1.5 is stated for every symmetric convex body K in R^n, without requiring nonempty interior. However, the proof of the lower bound (Lemma 2.4) relies on the duality of covering numbers (Theorem 2.1, from Artstein–Milman–Szarek 2004). That duality theorem is known to hold for symmetric convex bodies with nonempty interior; when K is contained in a proper subspace, its polar K^o is unbounded and has empty interior, and the duality relation (2.2) may fail or require a different formulation. The paper does not discuss this case, nor does it provide a separate argument for degenerate K. Since the statement of Theorem 1.5 claims universality for all symmetric convex bodies, this gap is load-bearing: if the lower bound cannot be justified for degenerate K, the theorem is not fully proved as stated. The authors could restrict to bodies with nonempty interior without losing much generality (the small-ball probability for a degenerate K is trivial), but the paper does not make this restriction explicit.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek/deepseek-v4-flash","summary":"The paper claims to provide a sharp two-sided characterization of Gaussian small ball probabilities for symmetric convex bodies, involving both metric entropy and local widths. It also derives analogous characterizations for intrinsic volumes and the radii of random sections. The main result is Theorem 1.5, which gives upper and lower bounds on -log γ_n(u K^o) in terms of r_*(K, u) and er(K, u). The paper is well-structured and builds on known results.","tokens_in":63183,"tokens_out":2391,"duration_ms":20723,"significance":"If correct, the results close a long-standing gap in the understanding of Gaussian small ball probabilities, providing a characterization that is both sharp and general. The connections to intrinsic volumes and random sections are also significant. The paper is rigorous and provides detailed proofs.","major_comments":[{"comment":"The proof of the lower bound (Lemma 2.4) relies on the duality of covering numbers (Theorem 2.1) which is known to hold for symmetric convex bodies with nonempty interior. However, Theorem 1.5 is stated for every symmetric convex body K in R^n, without requiring nonempty interior. The paper does not discuss the case where K is degenerate (e.g., contained in a proper subspace). The duality theorem may fail or require a different formulation for such K. Since the central claim is for all symmetric convex bodies, this gap is load-bearing. The authors should either provide a separate argument for degenerate K or explicitly restrict the statement to bodies with nonempty interior (which would still cover the main cases of interest).","section":"Theorem 1.5, Lemma 2.4, Section 2.3","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The definition of er(K, u) in (1.15) involves log N(K, r), which is defined for compact sets. For a symmetric convex body K that is not full-dimensional, the covering numbers are still defined, but the duality step in Lemma 2.4 uses the polar body K^o, which may be unbounded or have empty interior. The paper should clarify that the covering numbers and polar are taken in the ambient space R^n, and that the duality theorem applies only when K has nonempty interior. This is related to the previous comment.","section":"Section 1.2, Theorem 1.5","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The proof of Lemma 2.4 uses the inclusion (K^o + r^{-1} B_2) ⊃ (K ∩ r B_2)^o. This is a standard polarity relation, but it requires that K ∩ r B_2 is a symmetric convex body with nonempty interior (or at least that its polar is defined appropriately). If K is degenerate, K ∩ r B_2 may also be degenerate, and the polarity relation may not hold in the same form. The paper should address this.","section":"Section 2.3, Lemma 2.4","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"Theorem 1.10 characterizes the inradius of Gaussian projections G_j K. The proof of the upper bound (Section 5.2) uses Lemma 5.2, which assumes that T is a closed set. However, the application to K ∩ r S^{n-1} (the intersection of K with the sphere of radius r) may not be closed if K is not full-dimensional? Actually K ∩ r S^{n-1} is closed because K is compact and the sphere is closed. That seems fine. But the proof also uses Lemma 5.3, which requires K to be a symmetric convex body. That is satisfied. The main issue is again the reliance on the duality argument in the lower bound for the inradius? Actually the lower bound uses Lemma 5.1, which is standard and does not require nonempty interior. So this may be less problematic.","section":"Section 1.5, Theorem 1.10","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The proof of the lower bound in Theorem 1.9 uses the reverse Blaschke-Santaló inequality (Lemma 3.10), which holds for symmetric convex bodies with nonempty interior. The proof also uses the representation (1.32) for intrinsic volumes, which holds for any convex body (including degenerate ones) if we define V_j(K)=0 for j > dim(K). However, the application of Lemma 3.10 to G_j K requires that G_j K has nonempty interior almost surely, which holds if the linear span of K has dimension at least j. The paper should discuss the case where dim(K) < j, in which case θ_j(K)=0 and the inequality is trivial. This is a minor oversight but should be clarified.","section":"Section 3.8, Proof of Theorem 1.9","authors_acknowledge":null,"authors_acknowledge_p":null}],"minor_comments":[{"comment":"The abstract states 'no general sharp characterization of the lower tail is available'. This is a bit strong given the existing results of Kuelbs-Li and Talagrand, but the paper later qualifies this. Consider softening slightly.","section":"Abstract","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The condition (1.12) is derived from Theorem 1.3 under additional regularity assumptions. The paper should make clear that this is not a direct consequence of Theorem 1.3 alone, but requires the duality of covering numbers and the assumption that log N(K, r) behaves regularly.","section":"Section 1.2, equation (1.12)","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The duality theorem is cited from Artstein-Milman-Szarek 2004. The paper should note that this theorem holds for symmetric convex bodies with nonempty interior, which is the standard setting. The authors should explicitly state this condition when citing it.","section":"Section 2.1, Theorem 2.1","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The condition u ∈ (0, w(K)/(1+ε)) is used to ensure that r_*(K, (1+ε)u) is defined. This is fine, but the statement could be simplified by assuming u < w(K) and then using a slightly different argument. Not a major issue.","section":"Section 2.2, Lemma 2.2","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The proof uses the inequality (K^o + r^{-1} B_2) ⊃ (K ∩ r B_2)^o. This is a known polarity relation, but a brief justification or reference would be helpful.","section":"Section 2.3, Lemma 2.4","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The proof of strict monotonicity uses Fact 3.1, which relies on the fact that K ∩ r B_2 is strictly contained in K ∩ r' B_2. This is true if K has nonempty interior, but if K is degenerate, the strict inequality may fail. The lemma should be stated for symmetric convex bodies with nonempty interior, or the proof should be adjusted.","section":"Section 3.1, Lemma 3.1","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The lemma uses the constant c from (3.4), which is from [Mou25]. The paper should ensure that the constant is universal and does not depend on K. This is fine.","section":"Section 3.2, Lemma 3.4","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The proof uses [Mou25, Lemma A.1], which assumes that the sequence is ultra-log-concave and has finite support? The paper should check that the lemma applies to the infinite sequence (v_j) with v_j=0 for j>n. This is likely fine, but should be mentioned.","section":"Section 3.5, Lemma 3.8","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The example uses the Blaschke-Santaló inequality and its reverse. The paper should cite the reverse inequality from Bourgain-Milman correctly. It does so.","section":"Section 3.7, Example 3.2","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The lemma is cited from [Mou25, Corollary 4.1]. The paper should ensure that the definition of γ_* matches the one in that reference. It seems consistent.","section":"Section 4.1, Lemma 4.1","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The proof uses [AAGM15, Theorem 9.2.4] which is a consequence of Gordon's inequality. The paper should note that this holds for any symmetric convex body K, not necessarily with nonempty interior. That is fine.","section":"Section 5.1, Lemma 5.1","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The lemma uses [AAGM15, Theorem 9.3.3] which is a min-max inequality. The paper should ensure that the conditions are satisfied for T = K ∩ r S^{n-1}. This set is closed and bounded, so it is compact. That is fine.","section":"Section 5.2, Lemma 5.2","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The example of the ellipsoid with half-lengths 1/√i is used to show that the Kuelbs-Li condition may have no solution. The paper should note that this ellipsoid is a symmetric convex body with nonempty interior, so it is within the scope of Theorem 1.5. That is fine.","section":"Section 7.1","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The example of the ℓ_1 ball is used to show that the Kuelbs-Li condition may have many solutions. The paper should note that this body has nonempty interior. That is fine.","section":"Section 7.2","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The paper uses the notation γ_n for Gaussian measure and γ_2 for the generic chaining functional. This is potentially confusing, but the paper clarifies it in a footnote. It is acceptable.","section":"General","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The paper states in the acknowledgements that Theorem 1.9 was obtained using GPT-6. This is unusual but the authors take responsibility. It does not affect the mathematical validity, but some readers may find it noteworthy. The paper should ensure that the proof is fully verified and that the AI contribution is appropriately disclosed. The current disclosure seems adequate.","section":"General","authors_acknowledge":null,"authors_acknowledge_p":null}],"recommendation":"major_revision"},"author_rebuttal":null,"desk_editor":{"model":"deepseek/deepseek-v4-flash","letter":"Hey,\n\nThis paper is the real deal. It gives the first sharp two-sided characterization of Gaussian small ball probabilities for symmetric convex bodies, in terms of local Gaussian widths and covering numbers. That's been open for decades, and the author nails it. The proof is clean, the constants are universal, and the paper also draws nice connections to intrinsic volumes, the Wills functional, and random sections. It's a significant piece of work.\n\nWhat's new: Theorem 1.5 is the headline. It shows that -log γ_n(u K^o) is equivalent (up to universal constants) to u^2 / min(r_*(K, u)^2, er(K, u)^2). This is genuinely new and improves on the one-sided or regularity-dependent results of Kuelbs-Li and Talagrand. The paper also gives a chaining description of lower quantiles (Corollary 1.4), a comparison inequality (Corollary 1.5), and a precise equivalence between intrinsic volumes and negative moments (Theorem 1.9). The writing is clear and the proofs are rigorous.\n\nSoft spots: The stress-test note flags a real issue. The lower bound (Lemma 2.4) uses the Artstein-Milman-Szarek duality theorem for covering numbers, which is stated for symmetric convex bodies with nonempty interior. Theorem 1.5 claims to hold for all symmetric convex bodies, including those contained in a proper subspace (where K^o is unbounded and has empty interior). The paper doesn't address this case. However, this is a minor gap: for degenerate K, the small ball probability is trivial (either 0 or 1), and the result could be restricted to bodies with nonempty interior without losing anything of substance. The author should add a remark or a short argument. It's not a load-bearing flaw, but it should be fixed.\n\nOtherwise, the paper is solid. The reliance on cited results is appropriate, the constants are explicit, and the examples are helpful. The AI disclosure for Theorem 1.9 is transparent and the proof has been reorganized by the author.\n\nWho this is for: Anyone working in Gaussian processes, convex geometry, or high-dimensional probability. It's a major result that will be cited.\n\nRecommendation: Accept with minor revision. The degenerate case needs a brief clarification, but the paper is ready for a serious journal.","headline":"Sharp two-sided characterization of Gaussian small ball probabilities, closing a long-standing gap. The paper is well-written, rigorous, and the main result is genuinely new. The only real concern is a minor gap in handling degenerate convex bodies, but it's easily fixable and doesn't undermine the core contribution.","tokens_in":63662,"tokens_out":616,"duration_ms":10778},"pith_extraction":null,"created_at":"2026-11-13T10:16:51.026790+00:00","model_set":{"reader":"deepseek/deepseek-v4-flash"},"pipeline_version":"pith-pipeline@v1.4.1-daily-deepseek-or-medium","input_kind":null,"input_chars":null,"input_total_chars":null,"passes_missing":null,"published_standing":null,"published_confidence":null,"objection_audit":null,"run_record":null,"basis":{"kind":"not_recorded","source":null,"chars_given":null,"chars_total":null},"receipt":null,"falsifier":null,"supporting_citations":[],"lane_accuracy":{"status":"not yet measured","lane":{"model":"deepseek/deepseek-v4-flash","input_kind":null},"url":"https://pith.science/accuracy"},"review_version":1}