{"id":"3e532e2f-e107-4c60-a1f9-7557107e96ee","arxiv_id":"2507.21689","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"The paper introduces spectral generalized Turán numbers, proves a general transfer theorem from counting stability to spectral extremality, and derives a spectral Erdős Pentagon Theorem and an entropy formula.","lead":"Turán problems usually ask how many edges or pattern copies can fit in a graph without a forbidden pattern; this paper adds a spectral twist by maximizing a weighted, vertex-weighted sum of pattern copies. The main result transfers known stability theorems for ordinary counting into spectral extremal theorems, and the paper applies it to a spectral version of the Erdős Pentagon Theorem and to an entropy formula.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Lemma 3.3's proof uses the reversed inequality (1−ε)^{(α−1)/α} ≤ (1−ε), so the small-coordinate step that starts the vertex-deletion in Theorem 1.5 is not established as written.","rationale":"The reader's conditional verdict is, in my view, the right overall judgment, but the most load-bearing defect is not the one singled out in the reader's weakest_assumption. The false inequality in Lemma 3.3 sits at the engine of Theorem 1.5: without a small coordinate, the recursive removal cannot begin, and the entire transfer argument collapses. The reader's two noted gaps (the Fact 2.5 constant in the final lower bound and the entropy support issue) are real but more clearly cosmetic. I do not believe the central theorem is false: the Lemma 3.3 comparison can be repaired by using a δ' in Fact 2.5 much smaller than δπhat, and the intended contradiction then holds because πhat/(q−1)<1. But as written, the proof has a genuine gap in a parameter range (α close to 1) that the statements explicitly cover. Therefore the verdict should remain conditional: the paper is a promising and likely correct contribution, but it needs a corrected Lemma 3.3 and minor fixes elsewhere before the proof is fully rigorous.","tokens_in":18556,"tokens_out":38063,"duration_ms":387044,"concrete_test":"Recompute the final comparison in Lemma 3.3 with the correct upper factor (1−ε)^{(α−1)/α}πhat and lower factor (πhat−δ')(1−δ)^{α−1}, where δ=πhat ε/((q−1)α), and determine the maximal δ' for which the lower bound exceeds the upper bound for all α>1. Then check numerically the pentagon case (q=5, πhat=10/5^5, α=1.0005, ε=0.5): if the inequality fails even as δ'→0, attempt to construct a K3-free graph with δQ(H)≤(1−ε)qπhat n^{q−1} but λ_{α,C5}(H)≥μ_n; such a graph would falsify Lemma 3.3 as stated.","verdict_should_be":"UNCHANGED","load_bearing_attack":"In Lemma 3.3 (Section 3.1), the proof upper-bounds ∂_{i*}P/q by (1−ε)^{(α−1)/α} πhat n^{(q−1)(α−1)/α} and then asserts this is at most (1−ε)πhat n^{(q−1)(α−1)/α}. For α>1 and ε∈(0,1), (α−1)/α<1, so (1−ε)^{(α−1)/α} > (1−ε); the inequality is reversed. The intended contradiction needs the lower bound (1−αδ)πhat n^{...}, δ=πhat ε/((q−1)α), to exceed the actual upper bound. This requires (1−πhat ε/(q−1)) > (1−ε)^{(α−1)/α}, which in the small-ε limit is πhat/(q−1) < (α−1)/α. No such condition appears in Theorem 1.5. In the pentagon application πhat=10/5^5≈0.0032, q=5, so πhat/(q−1)≈0.0008, while for α=1.0005, (α−1)/α≈0.00050; the comparison fails. Since Lemma 3.3 is the step that produces the first small-weight coordinate for the iterative deletion, the proof of Theorem 1.5 is incomplete in this regime. The gap appears repairable by invoking Fact 2.5 with a much smaller error term instead of δπhat, but this correction is not in the manuscript.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper introduces the spectral generalized Turán problem: for an r-graph Q and a family F, it studies the maximum of the (α,Q)-spectral radius λ_{α,Q}(H) over F-free r-graphs H on n vertices. The main result (Theorem 1.5) asserts that under three hypotheses on F and a hereditary family H—smoothness of inj(n,Q,F), Q-degree stability of F with respect to H, and spectral balance of H—every F-free r-graph H on n vertices satisfies λ_{α,Q}(H) ≤ λ_{α,Q}(n,H), with equality only when H∈H; consequently spex_α(n,Q,F)=λ_{α,Q}(n,H). Theorem 1.2 derives the spectral Erdős Pentagon Theorem from this framework. Section 5 defines the (α,Q)-entropic density and proves η_{α,Q}(H)=λ_{α,Q}(H), extending a result of Chao and Hans.","tokens_in":18834,"tokens_out":14303,"duration_ms":167267,"significance":"The framework is natural and the main theorem, if correct, provides a useful transfer principle: it reduces spectral generalized Turán problems to three extremal/structural quantities, in the spirit of Keevash–Lenz–Mubayi. The pentagon application is a genuine new result and combines the Erdős Pentagon Theorem with the Kang–Nikiforov–Yuan spectral bound in a non-obvious way. The entropy identity is elegant and appears essentially correct. The paper is not accompanied by code or machine-checked proofs, but the argument is classical analytic/combinatorial in style. The proof gaps identified below are substantial, though each appears local and repairable.","major_comments":[{"comment":"The display after Claim 3.4 uses the inequality (1−ε)^((α−1)/α) ≤ (1−ε). For α>1 and ε∈(0,1), this inequality is reversed: since the exponent (α−1)/α is smaller than 1, (1−ε)^((α−1)/α) is larger than (1−ε). Consequently the claimed upper bound in the contradiction is not established. The intended contradiction requires (1−αδ) > (1−ε)^((α−1)/α) with δ = πhat ε/((q−1)α), which in the small-ε limit becomes πhat/(q−1) < (α−1)/α. No such condition appears in Theorem 1.5, and it fails in the pentagon application for α close to 1. Since Lemma 3.3 produces the first small coordinate for the iterative deletion, the proof of Theorem 1.5 is incomplete as written. The lemma appears repairable by choosing a different tolerance, but the repair must be supplied.","section":"Section 3.1, Lemma 3.3"},{"comment":"Fact 2.5 is invoked with δ′ replaced by 1/2 in order to assert μ_n ≥ πhat n^{q−q/α}. Fact 2.5 gives only μ_n ≥ (πhat−1/2)n^{q−q/α}, which is useless and even negative when πhat is small; in the pentagon application πhat≈10/5^5≈0.0032. The correct choice would be δ′=πhat/2 or another positive constant smaller than πhat, after which the displayed product lower bound still yields a contradiction provided condition (19) is adjusted accordingly. As written, this is a second load-bearing gap in the final step of the iteration.","section":"Section 3.2, proof of Theorem 1.5, final displayed lower bound"}],"minor_comments":[{"comment":"The first displayed inequality claims λ_{α,Q}(H) ≥ |H| n^{−q/α} for arbitrary Q; this is false, since |H| is the number of edges. The correct inequality is λ_{α,Q}(H) ≥ inj(Q,H)n^{−q/α}. For Q=K_r^r the two statements coincide up to the factor r!, which is the case actually used in Lemma 4.1, but the stated general fact should be corrected.","section":"Fact 1.3"},{"comment":"The notation λ_α(G) is used in the statement of Theorem 1.2 and in the surrounding discussion, but the quantity being bounded is the (α,C5)-spectral radius; it should be written λ_{α,C5}(G), or the abbreviation should be defined explicitly.","section":"Theorem 1.2 and Section 1.1"},{"comment":"The claim says \"Let (X_1,...,X_n) be the random embedding of Q in H\", but the tuple should have length q, not n. In the same proof, terms of the form y_j log(x_j^α/y_j) with y_j=0 and x_j=0 require a convention; the paper only sets 0·log 0=0, which does not directly cover the ratio 0/0.","section":"Section 5, Claim 5.2"},{"comment":"In the proof of Fact 2.5, the transition δ/n + δ′/2 ≤ δ′ is not immediate as written; it requires choosing n large enough so that δ/n ≤ δ′/2. This is a minor omission, but it should be stated explicitly.","section":"Fact 2.5"}],"recommendation":"major_revision","confidential_remarks":null},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The paper opens the spectral generalized Turán direction and proves a transfer theorem that converts smoothness, degree stability, and spectral balance into a spectral extremal result, generalizing Keevash–Lenz–Mubayi. That is genuinely new and useful. The pentagon application is clean and the entropy equivalence is a nice extension of Chao–Hans. The setup is careful and the writing is mostly clear.\n\nThe proofs have a few real soft spots. The most serious is in Lemma 3.3: the proof claims (1−ε)^((α−1)/α) ≤ (1−ε), which is reversed for α>1. This means the contradiction is not established as written. The gap looks repairable by choosing δ smaller than πhat ε/((q−1)α), specifically satisfying 1−αδ > (1−ε)^((α−1)/α), but the manuscript does not do this. The stress-test note is correct.\n\nThere is also an issue near the end of Theorem 1.5. The text applies Fact 2.5 with δ′=1/2 and then asserts μ_n ≥ πhat n^{q−q/α}. Fact 2.5 only gives μ_n ≥ (πhat − δ′)n^{...}, so you cannot just drop the factor. This is likely a minor fix—use a small δ′ and keep the (1−δ′) factor—but as written it is wrong.\n\nThe entropy proof in Proposition 5.1 has the stationarity equation on the support only; the text writes it as if it holds for all classes. The subsequent inequality still goes through if you define β over the support and note that β ≤ P_{Q,H}(x), so this is a minor presentational flaw, not a logical one.\n\nNo circularity, no data issues. The main ideas are sound and the theorems are plausible. The gaps are in the execution, not the architecture. This is the kind of paper that belongs in the refereeing pipeline—it will need a revision, but it deserves a serious referee.\n\nThe paper is for extremal combinatorics readers who work on spectral Turán problems or generalized Turán problems. I would not cite it in the next year unless I were working directly on spectral extremal problems, but I would bring it to a reading group.","headline":"New framework for spectral generalized Turán problems with a transfer theorem, but the proof has a few real gaps, including a reversed inequality in Lemma 3.3.","tokens_in":19454,"tokens_out":7009,"would_cite":false,"duration_ms":67657,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["05C35","05C65","05C50"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper introduces the spectral generalized Turán problem—maximizing the $(\\alpha,Q)$-spectral radius among $F$-free hypergraphs—and proves a general stability-to-spectrum theorem whose first application is the spectral Erdős Pentagon…","keywords":["spectral Turán problem","generalized Turán problem","spectral generalized Turán problem","(α,Q)-spectral radius","Erdős Pentagon Theorem","degree stability","entropic density","hypergraphs"],"falsifier":"Compute, for $\\alpha=2$ and increasing $n$, the maximum $(\\alpha,C_5)$-spectral radius among triangle-free graphs on $n$ vertices and compare it with the maximum over $C_5$-colorable graphs; an infinite sequence of triangle-free graphs whose value strictly exceeds every $C_5$-colorable graph on the same $n$ would refute the spectral Erdős Pentagon Theorem. A cheaper check on the machinery is to test Claim 4.4 numerically: the difference $|\\lambda_{\\alpha,C_5}(n,\\mathcal C_5)-|\\mathrm{Aut}(C_5)||T^5_{5,n}|/n^{5/\\alpha}|$ must grow at most like $n^{4-5/\\alpha}$; faster growth would invalidate the balance verification.","tokens_in":18253,"feed_emoji":"⭐","tokens_out":18356,"duration_ms":175494,"temperature":0.7,"pith_summary":"This paper sets up a new combined extremal problem: instead of counting copies of a fixed $r$-graph $Q$ in an $F$-free hypergraph, or taking the classical spectral radius, it maximizes the $(\\alpha,Q)$-spectral radius—the maximum of the Lagrangian $Q$-polynomial over vectors of unit $\\ell^\\alpha$ norm—among $F$-free $r$-graphs on $n$ vertices. The main theorem (Theorem 1.5) shows that if the forbidden family is 'smooth' and 'degree stable' with respect to a hereditary family $\\mathcal H$ of $F$-free $r$-graphs, and $\\mathcal H$ is 'balanced in spectral', then for all large $n$ every $F$-free $r$-graph has $(\\alpha,Q)$-spectral radius at most the maximum inside $\\mathcal H$, with equality only there. This determines the spectral generalized Turán number and extends the classical spectral Turán theorem of [KLM14] from edges to arbitrary $Q$. A short application is the spectral Erdős Pentagon Theorem: large triangle-free graphs maximize their $C_5$-spectral radius only on $C_5$-colorable graphs. The paper also proves that the $(\\alpha,Q)$-spectral radius equals an entropic density defined by random embeddings of $Q$, extending a recent identity of [CY24].","feed_headline":"Spectral Erdős Pentagon Theorem proven for large graphs","feed_subtitle":"The C5-spectral radius of large triangle-free graphs is maximized by C5-colorable graphs","key_machinery":"The central object is the $(\\alpha,Q)$-spectral radius $\\lambda_{\\alpha,Q}(H)$: the maximum, over vectors $x$ with $\\sum_i|x_i|^\\alpha=1$, of the Lagrangian $Q$-polynomial $P_{Q,H}(x)=\\sum_{\\phi\\in\\mathrm{Inj}(Q,H)}\\prod_{i\\in\\phi(V(Q))}x_i$. The argument is carried by a vertex-removal iteration. The Lagrange multiplier rule (Lemma 3.1) equates partial derivatives of $P_{Q,H}$ at an optimal vector with $q\\lambda_{\\alpha,Q}(H)x_i^{\\alpha-1}$; from this, Lemma 3.3 shows that a graph with large spectral radius but small minimum $Q$-degree must have an optimal vector with a very small coordinate, and Lemma 3.5 shows that deleting that vertex barely reduces the spectral radius. Iterating, the spectral radius is driven down until it contradicts the asymptotic value forced by the spectral balance condition. In the pentagon application, the proof translates $C_5$-copies in a $C_5$-colorable graph into edges of a 5-partite 5-graph and bounds its spectral radius with the $m$-partite bound of [KNY15] together with an estimate for the balanced complete 5-partite 5-graph.","core_discovery":"The central claim is Theorem 1.5. Let $\\alpha>1$ and $\\varepsilon>0$ be real numbers, $Q$ an $r$-graph on $q$ vertices, $F$ a family of $r$-graphs with positive embedding density $\\hat\\pi(Q,F)$, and $\\mathcal H$ a hereditary family of $F$-free $r$-graphs. Suppose $F$ is $(\\delta,M,Q)$-smooth, $F$ is $(\\varepsilon,M,Q)$-degree stable with respect to $\\mathcal H$, and $\\mathcal H$ is $(\\alpha,\\delta,M,Q,F)$-balanced in spectral. Then every $F$-free $r$-graph $H$ on $n\\ge N$ vertices satisfies $\\lambda_{\\alpha,Q}(H)\\le\\lambda_{\\alpha,Q}(n,\\mathcal H)$, with equality only if $H\\in\\mathcal H$; hence $\\mathrm{spex}_\\alpha(n,Q,F)=\\lambda_{\\alpha,Q}(n,\\mathcal H)$ for all large $n$. Smoothness controls the growth of the injection count $\\mathrm{inj}(n,Q,F)$, degree stability says the family of near-extremal graphs in $Q$-degree is inside $\\mathcal H$, and spectral balance pins the maximum $(\\alpha,Q)$-spectral radius inside $\\mathcal H$ to the asymptotic embedding count $\\mathrm{inj}(n,Q,F)/n^{q/\\alpha}$ up to an error of order $n^{q-q/\\alpha-1}$. The paper verifies these hypotheses for $Q=C_5$, $F=\\{K_3\\}$, and $\\mathcal H$ the $C_5$-colorable graphs, using the combinatorial Erdős Pentagon Theorem and an $m$-partite spectral bound, and obtains the spectral Erdős Pentagon Theorem. It also proves the identity $\\eta_{\\alpha,Q}(H)=\\lambda_{\\alpha,Q}(H)$ for the newly defined $(\\alpha,Q)$-entropic density.","pith_inferences":["The spectral balance condition is the genuine bottleneck: it ties the spectral maximum inside $\\mathcal H$ to a purely combinatorial embedding count, so for most natural problems verifying it may be as hard as the spectral problem itself; the pentagon case succeeds only because the full combinatorial Erdős Pentagon Theorem supplies the extremal family. This is an editorial reading of where the dif","The entropy identity suggests a testable strategy for open generalized Turán problems: maximize $\\eta_{\\alpha,Q}$ over $F$-free hypergraphs through entropy inequalities, then read off the spectral extremal value; this could bypass flag-algebra computations in cases where balanced families are known.","Since the combinatorial pentagon theorem holds for every $n$, a natural extension is to ask whether the spectral pentagon theorem also holds for all $n$ and for $\\alpha=1$, and whether similar transfers work for other odd cycles using the degree-stability results of [CHHL24]."],"forward_implications":["For any family $F$ and pattern $Q$ satisfying smoothness, $Q$-degree stability, and spectral balance, the spectral generalized Turán number $\\mathrm{spex}_\\alpha(n,Q,F)$ is simply the maximum $(\\alpha,Q)$-spectral radius inside the stable family $\\mathcal H$, for all large $n$.","Under the spectral Erdős Pentagon Theorem, for $\\alpha>1$ and large $n$, every triangle-free graph $G$ has $\\lambda_{\\alpha,C_5}(G)\\le\\max_{H\\in\\mathcal C_5,\\,v(H)=n}\\lambda_{\\alpha,C_5}(H)$, with equality only for $C_5$-colorable $G$.","Because $\\eta_{\\alpha,Q}(H)=\\lambda_{\\alpha,Q}(H)$, spectral extremal problems for a fixed pattern $Q$ can be reformulated as entropy maximization over random embeddings of $Q$ in $F$-free hypergraphs.","The framework turns any degree-stability theorem for a generalized Turán problem into a spectral statement, provided the spectral balance condition is verified, thereby extending the reach of the classical reduction of [KLM14]."],"supporting_citations":[{"why":"Supplies the smoothness, degree-stability, and spectral-balance framework that Theorem 1.5 generalizes from ordinary spectral Turán problems to arbitrary patterns $Q$.","marker":"[KLM14]"},{"why":"Provides one of the two large-$n$ proofs of the combinatorial Erdős Pentagon Theorem, identifying $C_5$-colorable graphs as the extremal family for the $C_5$-count in triangle-free graphs.","marker":"[Grz12]"},{"why":"Provides the independent large-$n$ proof of the same pentagon theorem, supporting the identification of the extremal family used in the spectral application.","marker":"[HHK+13]"},{"why":"Extends the pentagon theorem to every $n$, giving the exact value $|\\mathrm{Aut}(C_5)||T^5_{5,n}|$ used in Claim 4.4 to verify spectral balance.","marker":"[LP18]"},{"why":"Supplies the $C_5$-degree stability of $K_3$ with respect to the $C_5$-colorable graphs (Theorem 2.3), which is the degree-stability hypothesis in the pentagon application, and the Lagrangian $Q$-polynomial definition.","marker":"[CL24]"},{"why":"Provides the $m$-partite spectral bound used to estimate the spectral radius of the 5-partite 5-graph associated to $C_5$-colorable graphs, a key step in verifying spectral balance.","marker":"[KNY15]"},{"why":"Establishes the entropic-density identity in the edge case (Proposition 5.4), which Proposition 5.1 extends to arbitrary patterns $Q$.","marker":"[CY24]"}],"fun_headline_variants":["Spectral generalized Turán theorem proven with new conditions","Entropic density equals spectral radius in generalized Turán","C5-spectral radius of large triangle-free graphs determined","General spectral Turán bound for hereditary r-graph families","Erdős Pentagon Theorem now proven via spectral methods"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The spectral balance condition (Definition 1.4(iii))—that the maximum $(\\alpha,Q)$-spectral radius inside the hereditary extremal family $\\mathcal H$ matches $\\mathrm{inj}(n,Q,F)/n^{q/\\alpha}$ up to an error of order $n^{q-q/\\alpha-1}$—is the load-bearing premise; if it fails, Theorem 1.5 does not apply, and in the pentagon application it is established by combining the full combinatorial Erdős Pentagon Theorem with the $m$-partite spectral bound, so it is a substantial hypothesis rather than a formality.","fun_headline_variants_meta":{"raw":{"variants":["Spectral generalized Turán theorem proven with new conditions","Entropic density equals spectral radius in generalized Turán","C5-spectral radius of large triangle-free graphs determined","General spectral Turán bound for hereditary r-graph families","Erdős Pentagon Theorem now proven via spectral methods"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001312,"raw_usage":{"total_tokens":5402,"prompt_tokens":1059,"completion_tokens":4343,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":675,"completion_tokens_details":{"reasoning_tokens":4265}},"tokens_in":675,"tokens_out":4343,"duration_ms":43311,"temperature":1.0,"reasoning_tokens":4265,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T12:32:19.022869+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute, for $\\alpha=2$ and increasing $n$, the maximum $(\\alpha,C_5)$-spectral radius among triangle-free graphs on $n$ vertices and compare it with the maximum over $C_5$-colorable graphs; an infinite sequence of triangle-free graphs whose value strictly exceeds every $C_5$-colorable graph on the same $n$ would refute the spectral Erdős Pentagon Theorem. A cheaper check on the machinery is to test Claim 4.4 numerically: the difference $|\\lambda_{\\alpha,C_5}(n,\\mathcal C_5)-|\\mathrm{Aut}(C_5)||T^5_{5,n}|/n^{5/\\alpha}|$ must grow at most like $n^{4-5/\\alpha}$; faster growth would invalidate the balance verification.","supporting_citations":[],"review_version":1}