{"id":"e29ec77c-9d70-4258-a7e6-ae091cf61fbf","arxiv_id":"2412.12419","paper_version":3,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"The cyclic polytope attains the maximum number of vertices in a hyperplane slice, and the paper determines the missing vertex counts, or gaps, for slices of hypercubes up to dimension seven.","lead":"This paper asks how many vertices a flat slice of a convex polytope can have, and which numbers of vertices are impossible. It proves exact maximums and computes the full list of possible counts for hypercubes up to dimension seven.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Proof of Theorem 1.1 derives the global upper bound from the Upper Bound Theorem; UBT controls total edge count, not the number of edges a hyperplane can cut, so the central upper bound is not established as written.","rationale":"The reader's weakest assumption is about Theorem 1.5 and the hypercube gap computation; that is a real secondary concern. However, the most load-bearing issue is earlier and more central: the proof of Theorem 1.1, which is the paper's headline result, uses the Upper Bound Theorem in a way that does not control the quantity ν(P). The UBT bounds global face counts, whereas ν(P) is bounded by the number of edges in a single cut of the 1-skeleton. For d > 3, the UBT's edge bound C(n,2) is roughly twice the claimed slice maximum ⌊n/2⌋⌈n/2⌉, so the proof as written does not establish the upper bound for arbitrary polytopes. The gap is repairable: a vertex-free slice defines a cut, and the trivial |S|(n−|S|) bound gives the d > 3 statement, with the d = 3 case requiring a planar/dual-cycle argument. Because the theorem is very likely true but the submitted proof omits the crucial step, the reader's conditional verdict remains appropriate; I would not reject the paper outright, but the proof of Theorem 1.1 must be corrected. The reader did note a 'repairable logical gap' in Theorem 1.1, but their identified weakest link was Theorem 1.5, so my agreement is only partial.","tokens_in":20876,"tokens_out":17206,"duration_ms":156778,"concrete_test":"Repair the upper-bound proof by replacing the UBT step with a cut argument: for any hyperplane H avoiding vertices, let S = V(P) ∩ H⁻; then the slice has one vertex per edge of δ(S), so cv ≤ |S|(n−|S|) ≤ ⌊n/2⌋⌈n/2⌉. Then verify whether the same cut perspective, together with the planar dual-cycle bound for d = 3, yields exactly 2(n−2). If this derivation succeeds, Theorem 1.1 is true but Section 2 must be rewritten; if it fails, the paper's central upper bound is genuinely unproved.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Section 2, in the proof labeled 'Proof of Theorem 1.1' (Theorem 2.4), argues that because the Upper Bound Theorem makes the cyclic polytope maximize the number of edges, 'to bound ν(P), it suffices to bound ν(Cd(n)).' This is a non sequitur. UBT gives f1(P) ≤ f1(Cd(n)), but a hyperplane slice counts edges that cross a particular cut, not all edges. For d > 3, f1(Cd(n)) = C(n,2), whereas the claimed slice bound is ⌊n/2⌋⌈n/2⌉, which is strictly smaller; for d = 3, f1(Cd(n)) = 3n−6, also larger than 2(n−2). The paper computes ν(Cd(n)) and gives a spanning-tree upper bound for C3(n), but it never proves that an arbitrary P cannot have a slice with more vertices than the cyclic polytope. A missing argument is needed: a vertex-free slice corresponds to a cut δ(S) in the 1-skeleton, so its size is at most |S|(n−|S|) ≤ ⌊n/2⌋⌈n/2⌉; for d = 3 a separate planar/dual-cycle bound is needed to get 2(n−2). Without such an argument, the central theorem's upper bound is unsupported.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the function cv_P(H), the number of vertices in a hyperplane section of a convex polytope. Its central claim (Theorem 1.1) is that among d-polytopes with n vertices the cyclic polytope maximizes ν(P), with value 2(n−2) for d=3 and ⌊n/2⌋⌈n/2⌉ for d>3. The paper also introduces a slicing poset (Theorem 1.2), constructs infinitely many 3-polytopes with gaps in their vertex slice sequence (Theorem 1.3), and analyzes the vertex slice sequence of hypercubes up to dimension 7, including two general gap theorems (Theorems 1.4 and 1.5). The methods combine Gale evenness, the Upper Bound Theorem, graph connectivity, poset width, and computational experiments with provided code.","tokens_in":21198,"tokens_out":28665,"duration_ms":249637,"significance":"If the main theorem is correct, it gives a clean upper bound theorem for the number of vertices in hyperplane sections of polytopes, directly addressing part of Open Problem 7.7 in [16]. The explicit VSS description for cyclic polytopes, the poset framework for enumerating slices, and the new hypercube gap data are concrete and useful contributions. The paper is constructive, offers code for the hypercube computations, and does not rely on fitted parameters or circular reasoning; previous algorithmic work is used only as a comparison. However, two load-bearing proofs are incomplete as written: the reduction in Theorem 1.1 is a non sequitur, and the proof of Theorem 1.5 contains incorrect degree counts. The main claims may well be true, but the current version does not establish them.","major_comments":[{"comment":"The step to bound ν(P) is not justified. The proof says that because the Upper Bound Theorem makes the cyclic polytope maximize the number of edges, it suffices to bound ν(Cd(n)). This is a non sequitur: UBT controls the total number of edges f1(P) ≤ f1(Cd(n)), whereas a slice counts only the edges crossed by one hyperplane. For d>3, f1(Cd(n)) = C(n,2), which is much larger than the claimed bound ⌊n/2⌋⌈n/2⌉, and for d=3, f1(Cd(n)) = 3n−6 is larger than 2(n−2). An additional argument is needed: for d>3, a vertex-free slice is a cut δ(S) in the 1-skeleton, so it has size at most |S|(n−|S|) ≤ ⌊n/2⌋⌈n/2⌉; for d=3 a separate planar dual or cycle bound is needed to obtain 2(n−2). Without such an argument, the upper-bound half of Theorem 1.1 is unsupported.","section":"§2, Theorem 2.4 / Proof of Theorem 1.1"},{"comment":"The characterization of maximum antichains in Lemma 5.7 is stated and proved with incorrect level indices. With Ei = {e=(v1,v2) : ||v1||+1=||v2||=i}, the path count through an edge at level i is (d−i)!(i−1)!, whose minimum for even d occurs at i = d/2 and i = d/2+1; the lemma's E⌈d/2⌉ ∪ E⌊d/2⌋ collapses to E_{d/2} for even d, and the proof's 'i∈{d/2,d/2−1}' is inconsistent with that minimum. More seriously, the degree-count observation used in the proof of Theorem 1.5 is false as written: for e={u,v}∈Ei with u at level i−1, u is incident to d−i+1 edges in Ei and to i−1 edges in E_{i−1}, not to d−i and i. Consequently the assertion that endpoints of central-level edges have more than two incident edges in each of the consecutive levels is not true as stated (for d=6, a level-2 endpoint of an E3 edge has two edges in E2). Since this is exactly the step meant to exclude a drop of exactly one vertex when translating back to a vertex-containing slice, the proof of Theorem 1.5 is incomplete.","section":"§5.1, Lemma 5.7 and proof of Theorem 1.5"},{"comment":"The exactness of Table 1 for Q6 and Q7 is certified only by the statement that all other values are realized by slices computed with the GitHub code. The paper does not specify the finite search space: which directions were used, how the authors know that these directions produce all possible values of cv_Qd, or how the enumeration was certified. Since Corollary 1.5.1 is a claim of exact VSS, the computational proof should include this information, for example a list of representative directions or an argument that the chosen set realizes every attainable slice type.","section":"§5.2, Corollary 1.5.1"}],"minor_comments":[{"comment":"The statement should specify that an edge is counted only when H meets its relative interior. If edges are counted by mere intersection, then a hyperplane through a vertex makes all incident edges 'intersected' at that vertex, and the resulting set is not an antichain. The proof itself uses the relative-interior convention, so the statement should be aligned with it.","section":"§4, Theorem 1.2"},{"comment":"The set of central levels for even d should be E_{d/2} ∪ E_{d/2+1} in the statement, and the proof's minimum calculation should refer to the same two levels; the current E⌈d/2⌉ ∪ E⌊d/2⌋ is a typo that obscures the argument.","section":"§5.1, Lemma 5.7"},{"comment":"In the case |H−∩V(Qd)| = 3, 'N(v1)∩N(v1) = v4' should read 'N(v1)∩N(v3) = v4'.","section":"§5.1, proof of Lemma 5.5"},{"comment":"For d>3, the proof asserts that every triple (a,b,i) with a+b=n−i and i≤d is realizable by a suitable hyperplane, but it does not give the construction. A brief Gale-evenness argument showing how to place i points on H and split the remaining a and b vertices would make the VSS description of the cyclic polytope fully rigorous.","section":"§2, Theorem 2.3"},{"comment":"Please include the version or commit of the GitHub repository and a short description of the computational enumeration so that the claim that all listed values are attained can be reproduced independently.","section":"§5.2, Corollary 1.5.1"}],"recommendation":"major_revision","confidential_remarks":"I believe the central claims are likely correct and repairable, so the appropriate decision is major revision rather than rejection. The Theorem 1.1 gap can likely be fixed with the standard cut bound for d>3 and a planar-dual argument for d=3; the Theorem 1.5 proof needs corrected degree counts and a clarified statement of central levels. The computational certification for Q6 and Q7 should also be made explicit. I saw no circularity or attribution issues: the arguments use standard external results, and the authors' previous work appears only as a comparison."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Bottom line: this is a paper worth taking seriously. The main new results—the cyclic polytope slice vertex bound, the VSS formula for cyclic polytopes, the slicing poset, and the first complete gap lists for hypercubes up to dimension seven—are real. Theorem 1.1 as stated is true, and the proof gap flagged in the stress test is genuine but repairable. The UBT does not control the number of edges cut by a single hyperplane, so the sentence \"it suffices to bound ν(Cd(n))\" doesn't follow. But the upper bound itself is easy: a vertex-free slice partitions the vertex set into two sides, and the slice vertices are exactly the edges crossing that cut. For d>3, that is at most floor(n/2)ceil(n/2), full stop. For d=3, Proposition 4.1 gives that the two induced subgraphs are connected, so the number of uncut edges is at least n−2; the planar graph bound f1≤3n−6 then gives crossing ≤2(n−2). So the theorem survives; the proof just needs to be rewritten. The cyclic polytope attaining the bound is also shown correctly.\n\nThe poset framework in Section 4 is a clean generalization of O'Neil's idea, and Theorem 1.2 seems correct. The hypercube gap results are the most interesting part. Theorem 1.4 and the low-dimensional table are supported by the lemmas plus code, though the code at the GitHub link is unversioned and the verification for Q6/Q7 has no certificate. That is acceptable as experimental data, but the paper should say so more explicitly. The weak spot is Theorem 1.5: Lemma 5.7 (uniqueness of the maximum antichain) is plausible and the path-counting argument is reasonable, but the final geometric step—that moving the maximum slice back to a vertex cannot drop the count by exactly one—is sketched and relies on a degree observation at central levels. That could use a full proof. Lemma 5.6 also hides a lot in case analysis and figures; a skeptical referee should check the figures.\n\nThe 3-polytope gap construction (Theorem 1.3) is modest but solid, building on stacked polytopes and connectivity.\n\nNo circularity concerns; the derivations use standard external results. Citation pattern is fine.\n\nWho this is for: people working on polytope sections, hypercube cuts, and slice combinatorics. It deserves a serious referee. I would send it to review with the expectation of a major revision: fix the Theorem 1.1 proof, expand Theorem 1.5, and clean up the computational claims.","headline":"The main theorem is true and the paper has real new content, but the proof of Theorem 1.1 has a logical gap that needs a repair; the hypercube gap theorems are plausible but partly sketched.","tokens_in":21740,"tokens_out":4230,"would_cite":true,"duration_ms":37737,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["52B05","52B12","52B11"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper proves that the cyclic polytope maximizes the number of vertices a hyperplane can cut from any convex polytope: $2(n-2)$ vertices in dimension 3 and $\\lfloor n/2\\rfloor\\lceil n/2\\rceil$ in higher dimensions.","keywords":["hyperplane sections","convex polytopes","cyclic polytopes","upper bound theorem","vertex slice sequence","hypercube","slicing poset","gaps in slice counts"],"falsifier":"For any even dimension $d>4$, find an explicit hyperplane $H$ such that $H\\cap Q_d$ has exactly $(d/2)\\binom{d}{d/2}-1$ vertices; the simplest target is $d=6$, where the paper's table declares 59 a gap, so a single slice with 59 vertices would refute Theorem 1.5.","tokens_in":20706,"feed_emoji":"✂️","tokens_out":11717,"duration_ms":89025,"temperature":0.7,"pith_summary":"The paper proves a sharp upper bound theorem: for any $d$-dimensional convex polytope with $n$ vertices, every hyperplane slice has at most $2(n-2)$ vertices when $d=3$, and at most $\\lfloor n/2\\rfloor\\lceil n/2\\rceil$ vertices when $d>3$; the cyclic polytope $C_d(n)$ attains both bounds. It also describes the full set of possible slice vertex counts (the vertex slice sequence) for cyclic polytopes, and proves that hypercubes in even dimensions $d>4$ always miss the value $(d/2)\\binom{d}{d/2}-1$. Since slicing by a hyperplane is a basic operation in optimization, geometry, and probability, identifying the extremal polytope and the exact missing counts answers a natural analogue of the classical Upper Bound Theorem.","feed_headline":"Cyclic polytopes set the record for vertices in a hyperplane slice","feed_subtitle":"For any d-polytope with n vertices, the maximum slice size is 2(n−2) in 3D and about n²/4 beyond.","key_machinery":"The argument is carried by the cyclic polytope $C_d(n)$ (the convex hull of $n$ points on the moment curve) together with a perturbation lemma, Lemma 2.1, that lets one replace any slicing hyperplane by one avoiding all vertices without decreasing the number of slice vertices. Gale's Evenness Condition then gives a complete description of how a hyperplane separates the vertices of $C_d(n)$: it leaves $a$ and $b$ vertices on the two sides and $i$ on the hyperplane, and the slice has exactly $ab + i$ vertices. For enumeration and hypercube results, the slicing poset $(P_u,\\le)$ encodes all vertices and non-orthogonal edges in a direction $u$, and every slice is a maximal antichain in this poset; for the hypercube this poset is the extended hypercube poset, whose width is $\\lceil d/2\\rceil\\binom{d}{\\lfloor d/2\\rfloor}$.","core_discovery":"The central claim is Theorem 1.1: among all $d$-polytopes with $n$ vertices, the maximum number $\\nu(P)$ of vertices in a hyperplane section is attained by the cyclic polytope, with $\\nu(C_d(n)) = 2(n-2)$ for $d=3$ and $\\nu(C_d(n)) = \\lfloor n/2\\rfloor\\lceil n/2\\rceil$ for $d>3$. The proof shows, via Lemma 2.1, that a maximizing slice can be assumed to avoid all vertices of $P$, and then uses the Upper Bound Theorem and Gale's Evenness Condition to count the edges crossed by such a slice. The paper also determines the complete vertex slice sequence of $C_d(n)$: in dimension 3 it is $[2(n-2)]$, and in higher dimensions it consists exactly of values $ab + i$ with $a+b = n-i$ and $0\\le i\\le d$. For hypercubes, Theorem 1.4 locates the first gaps of the vertex slice sequence for every $d\\ge4$, and Theorem 1.5 shows that for every even $d>4$ the number $(d/2)\\binom{d}{d/2}-1$ is a gap in $VSS(Q_d)$.","pith_inferences":["Going beyond the paper: the same extremal question can be turned around, asking how small $\\nu(P)$ can be among $d$-polytopes with $n$ vertices; the paper fixes only the top end of this range.","Going beyond the paper: because the slicing poset captures every slice as a maximal antichain, the same machinery could bound the number of $k$-dimensional faces of a slice, not just vertices, moving toward the broader open problem cited in [16].","Going beyond the paper: the even-dimensional hypercube gap is parity-forced (vertex-free slices of $Q_d$ have even vertex counts), suggesting that other centrally symmetric polytopes may show analogous near-maximum gaps; a computational scan of cross-polytopes would test this."],"forward_implications":["For any 3-polytope with $n$ vertices, no hyperplane slice can have more than $2(n-2)$ vertices, and cyclic polytopes realize every count from 1 to $2(n-2)$.","In dimensions $d>3$, the maximum slice size grows like $n^2/4$, so a single hyperplane cannot cut exponentially many vertices even when the polytope itself has exponentially many.","The complete description of $VSS(C_d(n))$ shows that high-dimensional cyclic polytopes have gaps exactly when the required number $i$ of vertices lying on the slicing hyperplane exceeds $d$.","For hypercubes, the first missing vertex counts are known for all $d\\ge4$, and in even dimensions the last value below the maximum is always missing, so $VSS(Q_d)$ always has a gap at the top.","The poset-antichain theorem provides a finite combinatorial certificate for every slice of a given polytope and direction, enabling complete enumeration of combinatorially distinct sections."],"supporting_citations":[{"why":"supplies the first general algorithm for computing $\\nu(P)$, whose methods the paper's poset approach complements","marker":"[6]"},{"why":"introduced the hypercube poset and proved the width result for central-level edges used in the hypercube gap theorems","marker":"[25]"},{"why":"the Upper Bound Theorem showing the cyclic polytope maximizes every face count, including edges, the key input to Theorem 1.1","marker":"[22]"},{"why":"Gale's Evenness Condition, which describes the facets of cyclic polytopes and drives the slice analysis","marker":"[11]"},{"why":"prior complete characterization of slices of $Q_4$ and $Q_5$, used as base data for the low-dimensional VSS table","marker":"[9]"},{"why":"poses the open problem on maximizing $i$-dimensional faces of slices that Theorem 1.1 partially resolves","marker":"[16]"}],"fun_headline_variants":["Cyclic polytopes maximize hyperplane slice vertices","Hyperplane slice vertex max: cyclic polytopes rule","Slicing polytopes: max vertices from cyclic case","Vertex slice sequences: gaps found in hypercubes","Theorem: cyclic polytopes cap slice vertex count"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The most fragile step in the proof of Theorem 1.5 is the unproved geometric assertion that translating the unique maximizing slice of an even-dimensional hypercube back until it first passes through a vertex cannot reduce the vertex count by exactly one, because each central-level vertex has more than two incident edges in each neighboring level.","fun_headline_variants_meta":{"raw":{"variants":["Cyclic polytopes maximize hyperplane slice vertices","Hyperplane slice vertex max: cyclic polytopes rule","Slicing polytopes: max vertices from cyclic case","Vertex slice sequences: gaps found in hypercubes","Theorem: cyclic polytopes cap slice vertex count"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000206,"raw_usage":{"total_tokens":1395,"prompt_tokens":945,"completion_tokens":450,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":561,"completion_tokens_details":{"reasoning_tokens":371}},"tokens_in":561,"tokens_out":450,"duration_ms":3963,"temperature":1.0,"reasoning_tokens":371,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T14:08:07.973784+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"For any even dimension $d>4$, find an explicit hyperplane $H$ such that $H\\cap Q_d$ has exactly $(d/2)\\binom{d}{d/2}-1$ vertices; the simplest target is $d=6$, where the paper's table declares 59 a gap, so a single slice with 59 vertices would refute Theorem 1.5.","supporting_citations":[{"cited_title":"The best ways to slice a polytope","cited_arxiv_id":null,"evidence_quote":"supplies the first general algorithm for computing $\\nu(P)$, whose methods the paper's poset approach complements"},{"cited_title":"Hyperplane cuts of an n-cube","cited_arxiv_id":null,"evidence_quote":"introduced the hypercube poset and proved the width result for central-level edges used in the hypercube gap theorems"},{"cited_title":"The maximum numbers of faces of a convex polytope","cited_arxiv_id":null,"evidence_quote":"the Upper Bound Theorem showing the cyclic polytope maximizes every face count, including edges, the key input to Theorem 1.1"},{"cited_title":"Neighborly and cyclic polytopes","cited_arxiv_id":null,"evidence_quote":"Gale's Evenness Condition, which describes the facets of cyclic polytopes and drives the slice analysis"},{"cited_title":"Sections of hyper-cube in five dimensions","cited_arxiv_id":null,"evidence_quote":"prior complete characterization of slices of $Q_4$ and $Q_5$, used as base data for the low-dimensional VSS table"},{"cited_title":"Combinatorics of sections of polyt opes and Coxeter groups in Lobachevsky spaces","cited_arxiv_id":null,"evidence_quote":"poses the open problem on maximizing $i$-dimensional faces of slices that Theorem 1.1 partially resolves"}],"review_version":1}