{"id":"b4c0a11b-505e-473d-8a28-c5ecb81311e9","arxiv_id":"2412.15044","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Bourgain's slicing problem is resolved: every volume-one convex body in R^n has a hyperplane section of (n-1)-volume at least a universal constant c > 0.","lead":"Two mathematicians have finished the proof of Bourgain's slicing problem, a famous open question about high-dimensional convex shapes: every such shape has a flat slice through its middle whose area is always at least a fixed positive fraction. The proof combines recent progress on random heat flows with classic tools from convex geometry and information theory.","discovery_kind":"first_principles","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Lemma 3.4's stated consequence appears to need ε² rather than ε; the pivotal estimate (65) is therefore not justified as written.","rationale":"I read the paper in good faith: the overall architecture is coherent, the M-ellipsoid projection argument in Lemma 2.5 is plausible, the covariance-process bootstrap in Section 4 is elegant, and the final entropy closing would be valid if the quoted estimates hold. The reader's concern about Guan's Lemma 2.3 is legitimate and remains the main external risk: equation (28) depends on a bound imported from a one-week-old preprint, and the title's claim is conditional on it. However, the single most load-bearing internal concern is different. Lemma 3.4, as stated and used at equation (65), appears to contain a wrong power of ε. The matrix inverse in the integrand is of order ε² under the assumption Γ≤ε^{-1}Id, not ε. A scalar two-point example satisfying EΓ≤Id and Γ≤ε^{-1}Id shows the claimed ε-factor inequality is false for small ε. Since Section 4 sets ε=ξ and ξ can be smaller than 1/16 depending on the universal constant in Guan's bound, the derivation of (65) is not sound as written. This is not an attack on the theorem or the authors: the proof can very likely be fixed by replacing ε with ε², which still yields a universal constant bound because ξ is a universal constant. But the preprint's written proof is not yet correct at this step, so the conclusion should remain conditional pending this correction and independent verification of Guan's bound.","tokens_in":14547,"tokens_out":21681,"duration_ms":200247,"concrete_test":"Set ε=0.01, A=100Id and B=δId with δ small. Evaluate the scalar analogue of the trace in Lemma 3.4: (A−B)² / ((A²+B²)/2 + (A+B)/2) = 10000/5050 ≈ 1.98 for each off-diagonal pair, while ε|A−B|²/8 = 12.5. Averaging over the distribution P(Γ=100Id)=0.01, P(Γ=δId)=0.99 for independent copies gives left side ≈0.039 and right side ≈0.248, violating the claimed implication. If this computation is correct, Lemma 3.4's 'Consequently' statement must be corrected (e.g., to ε²), and Section 4's constants must be rerun.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The proof's key entropy lower bound comes from Lemma 3.4, quoted from Eldan–Mikulincer. The first display is a trace inequality, and the 'Consequently' step claims that Γ_r ≤ ε^{-1} Id almost surely on r∈(ξ,1) implies δ_KL(μ) ≥ ε ∫_ξ^1 E|Γ_r^(1) − Γ_r^(2)|^2 / (8(1−r)) dr. The pointwise matrix bound behind this claim is false: for Γ^(1)=ε^{-1}Id and Γ^(2)=0, the inverse matrix in the first display equals (2ε²/(1+ε))Id, which is smaller than εId whenever ε<1. Equivalently, the denominator is of order ε^{-2}, not ε^{-1}. A concrete two-point example with Γ=ε^{-1}Id with probability ε and Γ=0 otherwise satisfies EΓ=Id, compatible with Lemma 3.3(ii), yet the expected matrix trace is about 4ε while ε E|Γ_1−Γ_2|^2/8 is about 1/4; for ε<1/16 the claimed inequality fails. In Section 4, ε is taken to be ξ, and if Guan's constant forces ξ<1/16, the line leading to (65) is invalid. The argument can likely be repaired by replacing ε with ε², which changes only the universal constants, but the manuscript as printed contains a false estimate at the core of the entropy closing step.","agreement_with_reader":"disagree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper claims to resolve Bourgain's slicing problem by proving sup_n L_n < ∞, subject to Guan's bound ETr[A_t^2] ≤ Cn for the stochastic localization covariance process. The proof selects an isotropic log-concave measure in a lower-dimensional subspace that nearly attains the maximal isotropic constant and whose expected covariance has a uniform lower bound, then closes the argument using the Eldan--Mikulincer stability estimate for the Shannon--Stam inequality together with an entropy bound of Ball--Nguyen. The internal chain from Guan's bound to the theorem is coherent, and I have verified the entropy-closing estimates, including the epsilon factor in Lemma 3.4.","tokens_in":1191,"tokens_out":1979,"duration_ms":146348,"significance":"If Guan's bound is correct, this is a landmark result: it provides the final step toward the affine-invariant hyperplane conjecture. The paper's own contribution is the new reduction from the covariance-process bound to a uniform isotropic constant, combining Milman's M-ellipsoid theory with the recent stability estimates. The derivations in Sections 2--4 are clear, and the proof is genuinely non-circular: Guan's bound is strictly weaker than the target statement. The principal caveat is the complete reliance on an unreviewed external preprint for the key estimate, which currently prevents the manuscript from being a self-contained proof of the slicing problem. I also examined the concern about the epsilon factor in Lemma 3.4; it does not land, because the matrix bound M ≤ 2ε^{-1}I yields M^{-1} ≥ (ε/2)I and hence exactly the claimed ε in the consequent, not ε².","major_comments":[{"comment":"The central estimate ETr[A_t^2] ≤ Cn is imported verbatim from Guan's preprint arXiv:2412.09075 and is not proved or independently verified in this manuscript. This estimate is load-bearing: it enters at equation (28), where it yields the linear decay d/dt ETr[A_t] ≥ −Cn, and it drives Proposition 2.4 and the final entropy closing argument in Section 4. The footnote pointing to informal notes at a tinyurl is not a substitute for a proof. The manuscript should either include a complete proof of Lemma 2.3 (with appropriate attribution) or explicitly state in the abstract and introduction that the main theorem is conditional on this external result, pending independent verification. As written, the abstract's unconditional assertion is not supported by the body of the paper.","section":"Section 2, Lemma 2.3"},{"comment":"The paper's title and abstract state an unconditional resolution of Bourgain's slicing problem, yet the proof depends entirely on an unreviewed preprint posted six days earlier. This is a substantive accuracy issue: the reader is not told that the theorem is conditional unless they notice the footnote. The authors should make the conditional status explicit in the abstract and introduction, or include a proof of the external bound, before the paper can be accepted.","section":"Abstract and Introduction"}],"minor_comments":[{"comment":"The notation '/greaterorsimilar' and '/lessorsimilar' (e.g., in Proposition 2.4 and Lemma 2.5) is nonstandard; the paper already defines ≲ and should use it consistently for both upper and lower bounds.","section":"Section 2 and throughout"},{"comment":"The footnote referencing an informal note via a tinyurl is not a stable scholarly reference. If the notes are needed for context, they should be made into a permanent appendix or replaced by a direct reference to Guan's preprint.","section":"Section 2, Lemma 2.3 footnote"},{"comment":"The consequence of Lemma 3.4 is correct, but the authors could make the derivation more transparent by explicitly stating that from Γ_r ≤ ε^{-1}I one has M := sqrt(((Γ^(1))^2+(Γ^(2))^2)/2) + (Γ^(1)+Γ^(2))/2 ≤ 2ε^{-1}I, so M^{-1} ≥ (ε/2)I; this intermediate step is what turns the displayed trace inequality into the ε-factor consequence.","section":"Section 3, Lemma 3.4"}],"recommendation":"major_revision","confidential_remarks":"The manuscript's validity is entirely conditional on Guan's preprint arXiv:2412.09075, which is not yet refereed. The editors may want to solicit an independent review of that preprint, or to require the authors to reproduce the proof of Lemma 2.3 in an appendix, before accepting the paper. If Guan's result is correct, the paper is of exceptional significance; the current presentation is appropriate only as a conditional announcement. The title may be premature until the external bound is fully verified and publicly available in a peer-reviewed form."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Klartag and Lehec prove sup_n L_n < ∞, resolving Bourgain's slicing problem, contingent on Guan's covariance bound (Lemma 2.3, quoted from a preprint posted six days earlier). The genuinely new idea is the use of Milman's M-ellipsoids: from Guan's bound they derive the existence of a low-dimensional projection with large covariance, and then Eldan–Mikulincer's Shannon–Stam stability converts that into an entropy lower bound, giving the universal upper bound on L_n. The internal chain is coherent and the strategy is fresh. This is a real advance, not a repackaging of known results.\n\nThe paper is honest about its external dependency: Lemma 2.3 is not proved here, though the authors provide notes on Guan's proof. That is the main soft spot. The resolution is conditional on a recently posted, not yet independently verified preprint. This is not circularity—Guan's bound is strictly weaker than the target and derived elsewhere—but it does mean the final claim should be read as \"if Guan is right, then Bourgain's slicing problem is solved.\" The other external inputs (M-ellipsoids, Ball–Nguyen, Fradelizi) are established and used correctly.\n\nThe stress-test note about Lemma 3.4 does not hold up. It claims the consequence needs ε² instead of ε. That is a miscalculation: for Γ1=ε^{-1}I and Γ2=0, the inverse matrix in the first display is (1/(1/√2+1/2)) ε I ≈ 0.828 ε I, not order ε². The note's two-point example gets the trace wrong. There is a small real blemish: the \"Consequently\" in Lemma 3.4 states the constant ε, but the first display only supports a universal multiple cε (for example ε/2). Since the proof only uses universal constants, this is a harmless typo-level fix, not a structural flaw.\n\nThis paper is for convex geometers and anyone tracking high-dimensional volume/covariance problems. It deserves a serious referee. The referee should check Guan's preprint carefully and note the small constant correction in Lemma 3.4. If those check out, this is a major result. Send it to peer review.","headline":"Likely-correct resolution of Bourgain's slicing problem, conditional on Guan's bound; the stress-test concern about Lemma 3.4 is a miscalculation and does not survive reading.","tokens_in":15368,"tokens_out":10898,"would_cite":true,"duration_ms":69128,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["52A20","60D05","46B06"],"pacs":[],"model":"deepseek-v4-flash","headline":"For every convex body of volume one, in every dimension, there exists a hyperplane whose intersection with the body has volume larger than a universal constant; equivalently, the isotropic constant is uniformly bounded across all…","keywords":["slicing problem","isotropic constant","convex body","hyperplane section","stochastic localization","Shannon-Stam inequality","M-ellipsoid","log-concave measure"],"falsifier":"Compute the quantity $\\mathbb{E}\\operatorname{Tr}[A_t^2]$ for the uniform measure on the simplex in high dimensions; if it ever exceeds $C n$ for a fixed universal $C$ and all $t$, the imported bound fails and the present proof collapses. More directly, any sequence of volume-one convex bodies whose maximal hyperplane-section volumes tend to zero would refute the theorem itself.","tokens_in":14232,"feed_emoji":"📐","tokens_out":10815,"duration_ms":76367,"temperature":0.7,"pith_summary":"The paper claims to settle the slicing problem, a long-standing question in convex geometry, in the affirmative. Its main theorem says that every convex body of volume one has a hyperplane section whose volume is larger than a universal constant that does not depend on the dimension. The equivalent formulation proved is a uniform bound on the isotropic constant of all convex bodies, a quantity that measures how far a body is from being Gaussian-like. The proof combines stochastic localization, a recently imported covariance bound, and a stability version of the Shannon-Stam inequality to force the entropy of any extremal body to be at least a constant times the dimension.","feed_headline":"Every convex body has a hyperplane slice of constant size","feed_subtitle":"The guarantee holds in every dimension: a volume-one body always has a hyperplane section of size at least a universal constant.","key_machinery":"The central object is the covariance process $(A_t)_{t \\ge 0}$ generated by stochastic localization: starting from an isotropic log-concave measure, $A_t$ is the covariance of the tilted density $p_{t,\\theta_t}$. Its evolution obeys $\\frac{d}{dt} \\mathbb{E} A_t = -\\mathbb{E} A_t^2$, and the imported bound $\\mathbb{E} \\operatorname{Tr}[A_t^2] \\le C n$ makes the expected trace decay at most linearly. With the substitution $r = t/(t+1)$ and $\\Gamma_r = (1+t)A_t$, the process is related by $\\mathbb{E}|v_r|^2 = J(\\nu_r \\|\\gamma_r)$ to the Fisher information of the running law against a Gaussian, so integrating $r$ from 0 to 1 recovers the relative entropy $D(\\mu \\|\\gamma_1)$. A stability estimate for the Shannon-Stam inequality turns the variance of $\\Gamma_r$ into a lower bound on the same relative entropy, and an entropy jump bound $\\delta_{KL}(\\mu) \\le 2m$ closes the argument.","core_discovery":"The paper proves Theorem 1.1: for any convex body $K \\subset \\mathbb{R}^n$ of volume one, there exists a hyperplane $H$ such that $\\operatorname{Vol}_{n-1}(K \\cap H) > c$, where $c > 0$ is universal. This is shown to follow from Theorem 1.2, $\\sup_{n \\ge 1} L_n < \\infty$, where $L_n$ is the largest isotropic constant among $n$-dimensional convex bodies. The proof runs by contradiction at the extremal body: M-ellipsoid theory supplies a projection of the body onto a subspace of dimension $n/3$ whose isotropic constant is comparable to $L_n$, and stochastic localization generates a covariance process $A_t$ whose expected trace decays at most linearly thanks to the imported bound $\\mathbb{E} \\operatorname{Tr}[A_t^2] \\le C n$. After a change of variables into the F\\\"ollmer drift, the same process controls the Fisher information along the heat flow, which integrates to the relative entropy with respect to the Gaussian law. A stability estimate for the Shannon-Stam inequality, combined with an entropy jump bound, then shows that the entropy of the projected measure is at least $-C m$; this forces $L_n \\le C$ and completes the proof.","pith_inferences":["The theorem's correctness currently rests on an imported covariance bound that the paper does not prove; a self-contained verification of that bound would fully settle the argument.","The proof identifies the entropy jump bound as the only place where the dimension enters linearly, so refining the stability constants in the Shannon-Stam inequality would directly reduce the universal constant.","The mechanism may extend beyond log-concave measures: any family of measures with a covariance-process bound analogous to $\\mathbb{E}\\operatorname{Tr}[A_t^2] \\le C n$ would inherit a slicing-type statement.","A natural next target is the strong slicing conjecture; the information-theoretic formulation points to equality cases in the entropy jump bound rather than to geometric symmetrization."],"forward_implications":["The slicing problem is settled: every convex body of volume one has a hyperplane section of volume at least a universal constant, in every dimension.","The isotropic constant $L_n$ is uniformly bounded across all dimensions, so every log-concave probability measure has isotropic constant within a universal multiplicative range of the Gaussian value.","The theorem strengthens M-ellipsoid theory: uniform lower bounds on hyperplane sections now hold without any dimensional factor.","The proof yields, in principle, an explicit -- though astronomically large -- universal constant for the slicing bound.","Conditionally, if the extremal body for $L_n$ is the simplex, the argument connects to Mahler's conjecture on volume products."],"supporting_citations":[{"why":"Supplies the bound $\\mathbb{E}\\operatorname{Tr}[A_t^2] \\le C n$ imported as Lemma 2.3, the essential new estimate that drives the linear decay of the expected trace.","marker":"[15]"},{"why":"Provides the stability estimate for the Shannon-Stam inequality used as Lemma 3.4 to convert variance of the covariance process into relative entropy.","marker":"[13]"},{"why":"Gives the entropy jump bound $\\delta_{KL}(\\mu) \\le 2n$ used as Lemma 3.5 to close the bootstrap.","marker":"[2]"},{"why":"Establishes the symmetrization relations $L_m \\le C L_n$ and the subspace-section estimate used in Proposition 2.4.","marker":"[7]"},{"why":"Founds the M-ellipsoid covering theory used in Lemma 2.5 to find a projection that preserves the isotropic constant.","marker":"[30]"},{"why":"Supplies the de Bruijn-type representation of relative entropy used in Corollary 3.2.","marker":"[28]"}],"fun_headline_variants":["Bourgain's slicing problem settled: constant slices for all","Universal hyperplane slice constant proven for convex bodies","Every convex body has a slice of universal size","Constant-size hyperplane sections exist in all dimensions"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The proof relies on the imported bound that, for an isotropic log-concave measure, the expected squared trace of the stochastic-localization covariance is at most a universal constant times the dimension; this lemma is not proved in the paper and, if it fails or needs extra hypotheses, the main theorem is unsupported.","fun_headline_variants_meta":{"raw":{"variants":["Bourgain's slicing problem settled: constant slices for all","Universal hyperplane slice constant proven for convex bodies","Every convex body has a slice of universal size","Constant-size hyperplane sections exist in all dimensions"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000817,"raw_usage":{"total_tokens":3570,"prompt_tokens":925,"completion_tokens":2645,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":541,"completion_tokens_details":{"reasoning_tokens":2583}},"tokens_in":541,"tokens_out":2645,"duration_ms":13936,"temperature":1.0,"reasoning_tokens":2583,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T11:41:32.727897+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute the quantity $\\mathbb{E}\\operatorname{Tr}[A_t^2]$ for the uniform measure on the simplex in high dimensions; if it ever exceeds $C n$ for a fixed universal $C$ and all $t$, the imported bound fails and the present proof collapses. More directly, any sequence of volume-one convex bodies whose maximal hyperplane-section volumes tend to zero would refute the theorem itself.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the stability estimate for the Shannon-Stam inequality used as Lemma 3.4 to convert variance of the covariance process into relative entropy."},{"cited_title":"H., Entropy jumps for isotropic log-concave random vectors and spectral gap","cited_arxiv_id":null,"evidence_quote":"Gives the entropy jump bound $\\delta_{KL}(\\mu) \\le 2n$ used as Lemma 3.5 to close the bootstrap."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Establishes the symmetrization relations $L_m \\le C L_n$ and the subspace-section estimate used in Proposition 2.4."},{"cited_title":"D., An inverse form of the Brunn-Minkowski inequality, with app lications to the local theory of normed spaces","cited_arxiv_id":null,"evidence_quote":"Founds the M-ellipsoid covering theory used in Lemma 2.5 to find a projection that preserves the isotropic constant."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the de Bruijn-type representation of relative entropy used in Corollary 3.2."}],"review_version":1}