{"id":"d1dde4c5-40e8-4eb6-ae01-c37dae3bb1a9","arxiv_id":"1908.09358","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"high","formal_verification":"none","parameter_count":0,"one_line_summary":"Every hyperplane section of the unit cube at distance at most 1/2 from the origin has (n-1)-dimensional volume greater than 1/17, with a complex analogue for polydiscs.","lead":"This paper proves that slices of a large cube by flat cuts staying within distance one half of the center always have volume at least a fixed positive constant that depends only on the cut's codimension, not on the cube's dimension. For hyperplane cuts of real cubes the constant is 1/17, and a similar statement holds for complex polydiscs.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Proposition 5.4's printed tail bound is vacuous for every t>1, invalidating the Section 6 computation as written, though the derivation shows a correctable exponent typo.","rationale":"The central claim is the explicit dimension-free lower bound for non-central hyperplane sections. The proof structure is coherent: layer-cake/Fourier formulas in Propositions 3.1–3.2, concentration and tail estimates in Sections 4–5, and a short final integration in Section 6. The decisive step is the tail bound of Proposition 5.4; if it is false, the Section 6 integrals cannot be evaluated and nothing in the paper proves the constants 1/17 and 1/27. The printed bound is not merely weak but vacuous, and the integral evaluation that follows is an identity that cannot hold for the printed integrand. However, the derivation immediately above Proposition 5.4 supplies the intended correction, and the numerical constants are consistent with the corrected tail. This is a localized, repairable gap rather than a structural flaw, so the verdict remains conditional rather than reject. I also note smaller typographical slips elsewhere, e.g., the supporting hyperplane in Lemma 2.4 should be attached to (1/2)v, and the displayed inner product in the proof of Lemma 4.2 appears to miss the factor a_n; these do not change the main diagnosis but underscore that a correction pass is needed. I therefore recommend no change to the reader's conditional verdict.","tokens_in":21038,"tokens_out":14288,"duration_ms":128050,"concrete_test":"Re-derive Proposition 5.4 by substituting c = (k/2)(1 − 1/t^2) into g_k(c) = (1 − 2c/k)^{−k/2} e^{−ct^2}; confirm the resulting tail is t^k exp(k/2 − k t^2/2), not t^k exp(k/2 − k/(2t^2)). Then recompute the Section 6 bounds with the corrected tail: set t0 via t0^3 exp(3/2 − 3 t0^2/2) = p0 for k=3 and t1 via t1^4 exp(2 − 2 t1^2) = p1 for k=4, and check p0/t0 − (1/3)exp(3/2 − 3 t0^2/2) > 1/17 and p1/t1^2 − (1/2)exp(2 − 2 t1^2) > 1/27. If these inequalities hold, the theorem survives with the corrected typo.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Proposition 5.4 states P(|Σ a_j U_j| ≥ t) ≤ t^k exp(k/2 − k/(2t^2)) for t>1. Since k/(2t^2) < k/2 for t>1, the right-hand side is strictly larger than 1, so the inequality is vacuous. Section 6 uses this bound as a genuine tail estimate: the real-case computation treats ∫_{t0}^∞ v exp(3/2 − 3/(2v^2)) dv as though it equaled (1/3)exp(3/2 − 3/(2t0^2)), but the printed integrand grows without bound as v→∞. The same problem appears in the complex case. The proof of Theorem 1.2 therefore does not go through as printed. The error is localized and almost certainly typographical: minimizing f_k(c)e^{−ct^2} at c = (k/2)(1 − 1/t^2) gives t^k exp(k/2 − k t^2/2), the exponent needed for convergence and for the displayed numerical constants. Still, the central claim currently rests on a printed inequality that cannot support it.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies lower bounds for the volume of non-central sections of the unit-volume cube Q_n and of the complex polydisc. The first main result, Theorem 1.1, asserts that for every codimension d there is a constant ε(d) > 0 such that every affine (n−d)-plane at distance at most 1/2 from the origin cuts a section of volume at least ε(d), uniformly in the ambient dimension n and the direction of the section. The second main result, Theorem 1.2, gives explicit uniform bounds for hyperplane sections at distance exactly 1/2: A_R(a,1) > 1/17 in the real case and A_C(a,1) > 1/27 in the complex case, for every unit direction a. The proofs combine a probabilistic representation of section volumes as integrals over independent spherical vectors, exponential tail estimates obtained via Orlicz-space duality, and known results of Vaaler, Ball, and Veraar. The paper also contains a new method for bounding probabilities of the form P(|Σ a_j U_j| ≥ 1) for spherical vectors U_j, which may be of independent interest.","tokens_in":1485,"tokens_out":1544,"duration_ms":47342,"significance":"If the central claims hold, the paper provides the first explicit dimension-free lower bounds for non-central hyperplane sections of the cube and polydisc at the maximal distance where sections can be nonempty, complementing the classical central-section results of Vaaler and Ball. The constants are derived rather than numerically fitted, and the proof introduces a technically interesting Orlicz-space duality argument for spherical sums that goes beyond the standard Hanson-Wright framework. The lower bounds are not optimal, as the discussion in Remark 6.1 shows, but the qualitative uniformity in dimension and direction is the main contribution.","major_comments":[{"comment":"The tail bound in Proposition 5.4 is printed with the wrong exponent: the statement P(|Σ a_j U_j| ≥ t) ≤ t^k exp(k/2 − k/(2t^2)) is vacuous for every t > 1, since the exponent k/2 − k/(2t^2) is positive and the right-hand side exceeds 1. The displayed minimization leading to the bound actually yields t^k exp(k/2 − k t^2/2); the factor t^{-2} in the exponent is a typographical error. This is not a cosmetic issue because Section 6 relies on the bound as a genuine tail estimate: the integral ∫_{t0}^∞ v exp(3/2 − 3/(2v^2)) dv, used in the real case, diverges, and the claimed equality with (1/3)exp(3/2 − 3/(2t0^2)) is false. The numerical constants 0.06011 and 0.03789, and the equations defining t0 and t1, correspond to the corrected exponent −(k/2)t^2, not to the printed inequality. The argument can likely be repaired by a local correction, but as written the proof of Theorem 1.2 does not go through.","section":"§5, Proposition 5.4; §6, real and complex cases"},{"comment":"The derivation of the explicit lower bounds uses a relationship between the threshold t0 (or t1) and the probability p0 (or p1) that is stated with the incorrect tail expression. The displayed equations t0^3 exp(3/2 − 3/(2t0^2)) = p0 and t1^4 exp(2 − 2t1^2) = p1 are inconsistent with the subsequent integration: the numerical values t0 ≈ 1.9182 and t1 ≈ 1.7657 solve the equations with the exponent −(3/2)t^2 and −2t^2, respectively. The printed versions make the constants p0 and p1 incompatible with the stated t0 and t1, so the numerical verification of the inequalities must be redone with the corrected tail bound.","section":"§6, real and complex computations"}],"minor_comments":[{"comment":"The phrase 'probabiliy' in the reference [He] (Hensley, 'Slicing the cube in Rn and probabiliy') appears to be a misspelling of 'probability'; the reference should be checked against the original publication.","section":"§1, Introduction"},{"comment":"In the proof of Lemma 4.2, the inequality ∏_{j=2}^n (1+y_j) ≥ 1 + ∑_{j=2}^n y_j is stated for y_j ∈ (−1,1) 'having the same sign'; in the application all y_j are nonnegative, so the condition is satisfied, but this should be made explicit to avoid ambiguity.","section":"§4, Lemma 4.2"},{"comment":"The counting argument for the fourth moment E(S^4) is compressed: the sentence about '6 times' and the bound 24(∑ a_i^2 a_j^2 a_l^2 a_m^2) ≤ 4(∑ a_i^2 a_j^2)^2 is hard to follow. A short derivation of the displayed inequality would improve readability.","section":"§5, proof of Proposition 5.1, part (b)"},{"comment":"After correcting the tail exponent, the displayed intermediate expression 'p0/t0 (1 − 1/(3t0^3))' should read 'p0/t0 (1 − 1/(3t0^2))' in the real case; the printed form appears to be a typographical error.","section":"§6, real case"}],"recommendation":"major_revision","confidential_remarks":"The central result is plausible and the overall strategy is sound, but the printed proof contains a consequential error in a key tail estimate that invalidates the Section 6 computation as written. Because the error is localized and the corrected inequality is essentially already present in the derivation, I do not recommend rejection; a careful revision that fixes the exponent and the numerical details should suffice. The referee's stress-test concern is fully supported by the manuscript text."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: this is a good paper with a fixable typo in the middle. The main results — dimension-free lower bounds for non-central sections of the cube, with explicit 1/17 and 1/27 for hyperplanes — are new and, as far as I can tell, true. The Orlicz-space duality argument for lower-bounding P(|Σ a_j U_j| ≥ 1) is a real methodological novelty; it improves on Veraar's bound and looks reusable. I would not desk-reject this. But the version I read does not support its central claim as printed, and a referee should say so plainly.\n\nThe problem is Proposition 5.4. The printed tail bound is P(|Σ a_j U_j| ≥ t) ≤ t^k exp(k/2 − k/(2t^2)), which for t>1 is always >1, so it is vacuous. The proof, however, minimizes the Laplace bound and arrives at the right minimizer c = k/2(1 − 1/t^2), which gives exponent −k t^2/2. Section 6 then uses exactly that correct exponent: the integrand v exp(3/2 − 3v^2/2) converges, and the numeric constants 0.06011 and 0.03789 match that corrected bound. So this is a typo in the exponent, not a broken idea. But as printed, the integral in Section 6 is ∫ v exp(3/2 − 3/(2v^2)) dv, which diverges; the displayed evaluation is nonsense. Same for the complex case. That is a load-bearing error and it has to be fixed before this paper is certifiable.\n\nThere is also a smaller issue in Lemma 2.4: the claimed τ ≤ 1/2 looks inconsistent with the symmetry of D, since (1/2)v lies in D and a supporting hyperplane in direction v should be at distance at least 1/2. Likely the intended supporting point is something like (1/2)v or the inequality is reversed. This affects only Section 2, not Theorem 1.2, but it needs correction too.\n\nEverything else holds up. Vaaler, Ball, and the earlier König–Koldobsky references are used honestly; the constants come from the derivation, not from fitting. The remark on near-optimality based on the diagonal direction is fair. The numerical maximizations in Proposition 5.1 are routine; I would want the code or a reproducibility note, but nothing smells circular.\n\nBottom line: this is for convex geometers, functional analysts, and anyone interested in cube sections or spherical tail bounds. Send it to a referee. It deserves careful review, and the authors should get a clear request to fix Proposition 5.4 and Lemma 2.4, then re-check the constants. Once those typos are corrected, the paper should be acceptable.","headline":"Good paper, true-looking results, but a typo in Proposition 5.4 breaks the printed proof of Theorem 1.2; it is fixable and worth refereeing.","tokens_in":21778,"tokens_out":5014,"would_cite":true,"duration_ms":46430,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["52A38","52A40","52A20"],"pacs":[],"model":"deepseek-v4-flash","headline":"Every off-center hyperplane slice of the unit cube has volume at least 1/17, in every dimension.","keywords":["non-central sections","unit cube","hyperplane sections","dimension-free lower bounds","polydisc","slicing problem","Orlicz space duality","tail estimates"],"falsifier":"Check Proposition 5.4 numerically for $k=3$, $t=1.1$: the printed bound is $1.1^3\\exp(3/2-3/(2\\cdot1.1^2))\\approx1.73$, which is greater than 1 and therefore cannot be a probability tail; the integration in Section 6 needs a genuine exponential tail near $t=1$, so replacing the printed bound by any valid tail $\\le C_k e^{-c_k t^2}$ would either restore the proof with adjusted constants or reveal a genuine obstruction.","tokens_in":20755,"feed_emoji":"🧊","tokens_out":18392,"duration_ms":160601,"temperature":0.7,"pith_summary":"This paper proves that off-center slices of the unit cube cannot become arbitrarily thin. For every unit vector $a\\in\\mathbb{R}^n$, the hyperplane $\\{x:\\langle x,a\\rangle=1/2\\}$ cuts the cube $Q_n=[-1/2,1/2]^n$ in a section of volume $A_{\\mathbb R}(a,1)>1/17$, independent of the dimension $n$ and of the direction $a$; for complex polydiscs the analogous constant is $1/27$. More generally, for every codimension $d$ there is a positive $\\varepsilon(d)$ such that any affine section of codimension $d$ at distance at most $1/2$ from the origin has volume at least $\\varepsilon(d)$. The distance $1/2$ is maximal, since beyond it one can choose parallel sections that miss the cube entirely. These bounds turn the cube into one of the few bodies for which all sections in the guaranteed non-empty range are controlled from below.","feed_headline":"Off-center cube slices stay above 1/17 in any dimension","feed_subtitle":"No matter the dimension, no off-center hyperplane cut can get arbitrarily thin; polydisc slices stay above 1/27.","key_machinery":"The load-bearing machinery is a probabilistic representation of section volumes. With independent random vectors $U_j$ uniformly distributed on $S^2$ in the real case and on $S^3$ in the complex case, the volume of the hyperplane section is $A_{\\mathbb R}(a,t)=\\int_{|\\sum a_jU_j|\\ge t} dP/|\\sum a_jU_j|$ and $A_{\\mathbb C}(a,t)=\\int_{|\\sum a_jU_j|\\ge t} dP/|\\sum a_jU_j|^2$. The identity $|\\sum a_jU_j|^2=1+2S$, with $S=\\sum_{i<j}a_ia_j\\langle U_i,U_j\\rangle$, turns the threshold event into $S\\ge0$; the paper controls $P(S\\ge0)$ from below via a Laplace-transform estimate and duality in Orlicz spaces, and controls the upper tail of $|\\sum a_jU_j|$ via a Khintchine-type inequality for rotationally invariant vectors. For general codimension $d$, the section volume is interpreted as the density of a projected uniform random point, and the classical lower bound for central sections together with the known upper bound for central sections complete the reduction.","core_discovery":"The central claim is that dimension does not enter any lower bound for non-central sections of the cube. Specifically, Theorem 1.2 asserts $A_{\\mathbb R}(a,1)>1/17$ for every unit $a\\in\\mathbb{R}^n$ and $A_{\\mathbb C}(a,1)>1/27$ for the polydisc in $\\mathbb{C}^n$, while Theorem 1.1 asserts that for each $d$, all $(n-d)$-dimensional affine sections at distance at most $1/2$ have volume $\\ge\\varepsilon(d)>0$. The proof derives integral representations for these volumes over products of unit spheres and bounds the resulting probabilities from below and above. The constants are not claimed to be optimal: Remark 6.1 shows they cannot be improved by more than factors $\\simeq5.1$ (real) and $\\simeq7.1$ (complex), since the diagonal-direction limits are $\\sqrt6/(\\pi e^3)\\simeq0.3084$ and $2/e^2\\simeq0.2707$.","pith_inferences":["The sphere-integral method should extend to other rotationally symmetric bodies whose section volumes admit Bessel-function or Fourier representations, giving dimension-free lower bounds for non-central sections of such bodies in their guaranteed non-empty distance range.","The true infimum of $A_{\\mathbb R}(a,1)$ is probably larger than $1/17$; the diagonal-direction limit $\\sqrt6/(\\pi e^3)$ and the exact small-dimension values suggest that the extremal direction may vary with $n$.","Since the printed upper-tail estimate of Proposition 5.4 is vacuous for $t$ close to $1$, a necessary repair is a valid exponential tail bound with constants depending only on $k$; if such a bound holds, the qualitative theorem survives with possibly different constants."],"forward_implications":["For every dimension $n$ and every unit direction $a$, the slice $Q_n\\cap\\{x:\\langle x,a\\rangle=1/2\\}$ has $(n-1)$-volume strictly larger than $1/17$, so high-dimensional cubes have uniformly thick off-center hyperplane sections.","Because the bound is independent of $n$, the result controls all non-central hyperplane sections simultaneously as $n$ grows, complementing the known upper bounds by the same dimension-free character.","The distance $1/2$ is the entire non-empty range: for any subspace $E$ and any $v\\in E^\\perp$ with $|v|\\le1/2$, the parallel section has volume at least $\\varepsilon(d)$, while for $|v|>1/2$ some choices give empty sections.","For every codimension $d$ a constant $\\varepsilon(d)$ exists, so the phenomenon is not special to hyperplanes.","In the complex case, the same argument gives $A_{\\mathbb C}(a,1)>1/27$ for polydisc hyperplane sections, a dimension-free lower bound in $\\mathbb{C}^n$."],"supporting_citations":[{"why":"Supplies the central-section lower bound of 1 used in the reduction of general codimension-d sections.","marker":"[Va]"},{"why":"Gives the Fourier-based volume formula for real hyperplane sections and the maximal-section upper bound motivating the problem.","marker":"[B]"},{"why":"Provides the upper bound $2^{d/2}$ for central sections used in the geometric lemmas of Section 2.","marker":"[B1]"},{"why":"Establishes the complex polydisc slicing formula used in Proposition 3.1 and the upper bound for complex sections.","marker":"[OP]"},{"why":"Supplies the lower probability estimate for centered random variables that Proposition 5.1 sharpens.","marker":"[V]"},{"why":"Provides the best-constant inequality for rotationally invariant vectors used in the upper-tail estimate of Proposition 5.4.","marker":"[KKw]"},{"why":"Supplies the quadratic-form concentration estimate used in the Laplace-transform bound of Lemma 4.2.","marker":"[R V]"}],"fun_headline_variants":["Cube hyperplane slices always keep 1/17 area, any dimension","Off-center cube cuts never go below 1/17","Cube slice lower bounds: 1/17 real, 1/27 complex","Dimension-free lower bound for cube sections: 1/17","For any n, off-center cube hyperplane slice at least 1/17"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The numerical content of Theorem 1.2 depends on the upper-tail estimate in Proposition 5.4; as printed, that estimate gives a bound larger than 1 for all $t>1$, so it cannot support the integration that produces $1/17$ and $1/27$.","fun_headline_variants_meta":{"raw":{"variants":["Cube hyperplane slices always keep 1/17 area, any dimension","Off-center cube cuts never go below 1/17","Cube slice lower bounds: 1/17 real, 1/27 complex","Dimension-free lower bound for cube sections: 1/17","For any n, off-center cube hyperplane slice at least 1/17"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001427,"raw_usage":{"total_tokens":5739,"prompt_tokens":912,"completion_tokens":4827,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":528,"completion_tokens_details":{"reasoning_tokens":4732}},"tokens_in":528,"tokens_out":4827,"duration_ms":33205,"temperature":1.0,"reasoning_tokens":4732,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T11:25:20.412002+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Check Proposition 5.4 numerically for $k=3$, $t=1.1$: the printed bound is $1.1^3\\exp(3/2-3/(2\\cdot1.1^2))\\approx1.73$, which is greater than 1 and therefore cannot be a probability tail; the integration in Section 6 needs a genuine exponential tail near $t=1$, so replacing the printed bound by any valid tail $\\le C_k e^{-c_k t^2}$ would either restore the proof with adjusted constants or reveal a genuine obstruction.","supporting_citations":[],"review_version":1}