{"id":"d0d18365-8bd4-44c2-a417-f6dacc23fec8","arxiv_id":"2411.12079","paper_version":1,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"On the complete graph K_n, the sharp Poincare-Wirtinger constant and the full set of extremizers are determined for p in [1, 3+δ1_n) ∪ (3+δ2_n, ∞).","lead":"This paper finds the exact best constant in a discrete Poincare inequality on complete graphs for almost every exponent p, and it lists all functions that make the inequality an equality. Because the complete graph is the densest possible network, these constants and extremizers serve as natural benchmarks for Poincare-type estimates on other graphs.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Lemma 3(iii) is false: for n=3, p=4 the inequality is an identity and every f is a maximizer, contradicting the claimed Dirac-only equality classification.","rationale":"The reader's accepted verdict relied on Lemma 3, but the specific weakness identified was the Descartes sign analysis and the compressed maximizer transfer. The actual counterexample is different: Lemma 3's equality classification is false for n=3, p=4. The sharp constants may well be correct, and the displayed theorem only says equality is 'attained' by Dirac deltas rather than 'only by', but the paper's abstract and introduction explicitly promise a complete characterization of all maximizers, and Lemma 3(iii) asserts exactly the false Dirac-only classification. Since the central claim as stated is false, the verdict cannot remain ACCEPT. A revision that adds the exceptional case n=3, p=4 (where the inequality is an identity and all functions are extremizers) and corrects Lemma 3(iii) could restore the result; as written, the paper should be rejected.","tokens_in":11738,"tokens_out":44944,"duration_ms":423021,"concrete_test":"Set n=3, p=4 and f=(1,1/2,-1). Compute the left side of (2.5): (1-1/6)^3+(1+1/6)^3 = (5/6)^3+(7/6)^3 = 13/6. Compute the right side: (1/9)[16+(1/2)^3+(3/2)^3] = (1/9)(16+1/8+27/8) = 13/6. Equality holds for this non-Dirac function, so Lemma 3(iii) is false. A stronger check: verify the identity Σ_{i=1}^3 |f_i-m|^4 = (1/9)Σ_{1≤i<j≤3}|f_i-f_j|^4 for arbitrary f, e.g. f=(1,2,4), confirming that every f maximizes the inequality when n=3 and p=4.","verdict_should_be":"REJECT","load_bearing_attack":"The paper's central claim includes a complete characterization of maximizers, and Lemma 3(iii) asserts that for p∈[4,∞) the only maximizers of (2.5) are Dirac deltas. This is false for n=3, p=4. Take f=(1,x,-1) with 0≤x≤1, so m=x/3. Then the two sides of Lemma 3's inequality (2.5) are (1-x/3)^3+(1+x/3)^3 and (1/9)[16+(1-x)^3+(1+x)^3], and both expand to 2+2x^2/3. Thus equality holds for every x, including non-Dirac functions such as f=(1,1/2,-1). Equivalently, for n=3 and p=4 one has the identity Σ|f_i-m|^4 = (1/9)Σ_{i<j}|f_i-f_j|^4 for all f, so every function is an extremizer. The source of the error is the non-strict monotonicity step: in the proof of the p≥4 case, T is proved to be decreasing with T(1)=0, but for n=3, p=4, T is identically zero, not strictly decreasing. The sharp constant 1/9 in Theorem 1(vi) is correct, but the claimed uniqueness of Dirac deltas as maximizers is not. This invalidates the stated equality-case characterization in the abstract, the introduction, and Lemma 3.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the sharp Poincaré-Wirtinger inequality on the complete graph K_n, namely the optimal constant C_{n,p} in \\|f-m_f\\|_p ≤ C_{n,p} Var_p(f), for p in the ranges stated in Theorem 1. The proof proceeds by first establishing existence of extremizers, deriving a first-order condition (2.3) involving only the maximum and minimum values of f, and then proving an auxiliary sharp inequality, Lemma 3, for the extreme values. The auxiliary inequality is used to identify the optimal constant, after which the extremizers are transferred back to the original problem. The paper gives explicit constants for p=1, 1<p<2, p=2, 2<p≤3+δ_n^1, the regime (3,4)∩A_n, and p≥4, and it claims a complete characterization of the corresponding maximizers.","tokens_in":12014,"tokens_out":15751,"duration_ms":150039,"significance":"If the main results were correct as stated, the paper would provide a complete, explicit solution for the sharp Poincaré-Wirtinger constant on complete graphs, with a clean phase transition in the structure of extremizers. The lower-bound test functions are natural, the p=1 and p=2 cases are elementary and correct, and the proof is self-contained, using standard tools such as compactness, Karamata's inequality, convexity, and Descartes' rule of signs. There are no fitted constants. However, the claimed equality-case characterization is not correct as stated: for n=3 and p=4 every function is an extremizer, contradicting the asserted uniqueness of Dirac-delta-type maximizers in Lemma 3(iii) and the abstract's promise of a complete characterization. This is a load-bearing issue for the equality part of the main theorem, although the sharp constants themselves appear to survive.","major_comments":[{"comment":"Lemma 3(iii) is false for n=3, p=4. Take f=(1,x,-1) with 0≤x≤1, so m=x/3. For p=4 the two sides of (2.5) are (1-x/3)^3+(1+x/3)^3 and (1/9)[16+(1-x)^3+(1+x)^3], and both sides expand to 2+2x^2/3. Hence equality holds in (2.5) for every such f, including f=(1,1/2,-1), which is not of the form asserted in Lemma 3(iii). The same phenomenon occurs in the original inequality: for n=3 and p=4 one has the identity Σ_i |f_i-m|^4 = (1/9)Σ_{i<j}|f_i-f_j|^4 for every f, so every f is an extremizer of (2.1), not only Dirac-delta functions. This invalidates the complete characterization claimed in the abstract, in Theorem 1(vi), and in Lemma 3(iii). The sharp constant C=1/9 is still correct, but the equality-case statement must be amended, for example by adding the exceptional case n=3, p=4 where equality holds for all functions.","section":"§2, Lemma 3(iii) and Theorem 1(vi)"},{"comment":"The proof asserts that 'the maximizers of (2.1) are the same as for (2.5)' solely because the constants coincide. The forward direction is valid: a maximizer of (2.1) satisfies the Euler-Lagrange equation (2.3), which is exactly the equality case of (2.5). The reverse direction, however, requires an argument that every equality case of Lemma 3 attains the value C^* in the original quotient (2.1); this is not supplied. Since Lemma 3(iii) is false as stated, the transfer statement is currently false in the exceptional case. Even after correcting the exception, the reverse implication needs a separate proof rather than the current one-sentence assertion.","section":"§2, paragraph following Eq. (2.6)"}],"minor_comments":[{"comment":"The statement of item (iv) is organized awkwardly: the displayed inequality is introduced with the condition 'If 2<p≤3 and n≥3', while the extension to the interval (3,3+δ_n^1) is described only after the equality-case paragraph. This should be restructured so that the actual hypotheses for the displayed inequality are unambiguous.","section":"Theorem 1(iv)"},{"comment":"The heading reads 'Cases: 4 < p and p ∈ (3,4) ∩ A_n', but the text immediately says 'We study first the case p ∈ [4, ∞]'. The heading should include p=4 to match the proof and the statement of Lemma 3(iii).","section":"§2, proof of Lemma 3, p≥4 heading"},{"comment":"In the displayed expression '22l+2−p' the superscript is lost; this should read '2^{2l+2-p}'. As written, the power is unclear and the subsequent sign discussion is hard to follow.","section":"§2, Eq. (2.9)"},{"comment":"In the proof of Proposition 5, the majorization statement is presented as 'the (k+1)-tuple (1+ky,...,1+ky,y) majorizes (k+y,ky,...,ky) or (ky,...,ky,k+y)'. Since the ordering of the tuple depends on whether y is above or below 1, the proof should specify which majorization applies for which range of y.","section":"§2, Proposition 5"}],"recommendation":"major_revision","confidential_remarks":"The main constants of the paper appear to be correct, and the p=1 and p=2 cases are certainly correct. The equality-case classification, however, is genuinely wrong for n=3, p=4, and the transfer from Lemma 3 to the original maximizer problem is not proved as written. These are localized issues: the sharp constants can likely be kept, but the equality statements and the proof of the transfer need substantial correction. If the authors fix the exceptional case and supply the missing converse argument, the paper may be suitable for publication."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Two things to know. The sharp constants for the Poincaré-Wirtinger inequality on complete graphs are real: the paper computes C_{n,p} for essentially all p, and those constants are almost certainly correct. The second thing is that the equality-case classification, a headline claim, is false as stated for n=3, p=4, where the inequality is an identity for every function; the Dirac-delta-only maximizer list is wrong.\n\nThe new material is substantial. Lemma 3, the auxiliary sharp inequality between the extreme-vertex terms and the total variation, is the load-bearing piece and a genuine contribution. The lower-bound test functions match the constants, the p=1 and p=2 cases are elementary and correct, and the use of Karamata, convexity, and Descartes' rule of signs in the delicate ranges is mostly coherent. For a graph-inequality audience, the sharp constants themselves are a useful benchmark.\n\nThe soft spot is the strictness analysis. In the p≥4 case the proof shows the auxiliary function T is decreasing with T(1)=0, which yields the inequality but does not identify the maximizers unless the decrease is strict. At n=3, p=4, T is identically zero: for f=(1,x,-1), both sides of Lemma 3's inequality expand to 2+2x^2/3, so every function is a maximizer. The authors' Lemma 3(iii) asserts maximizers must be Dirac deltas, which is not true in that case. Since Theorem 1 inherits the classification, the 'characterize all maximizers' claim overreaches. The constants survive; the identity case needs to be carved out and the strictness argument tightened.\n\nThe Descartes/binomial-series part for 3<p≤3+δ1_n is terse and I couldn't fully certify it, but I did not find a concrete error there. The citation pattern is normal, no self-citation inflation. The paper is worth a serious referee, but as it stands it needs a correction, not just tweaks. My recommendation: send it to review with a clear request to fix the equality-case characterization.","headline":"Sharp constants on complete graphs are likely correct, but the equality-case classification is false at n=3, p=4, making the claim overbroad; fixable with a stricter monotonicity argument.","tokens_in":12549,"tokens_out":6748,"would_cite":true,"duration_ms":57548,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["26A45","39A12","46E39","05C12"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper determines the exact optimal constant and all equality cases for the Poincaré–Wirtinger inequality on the complete graph, for every exponent p in the stated ranges.","keywords":["Poincaré inequalities","complete graphs","p-variation","sharp constants","extremizers","Wirtinger inequality","discrete inequalities"],"falsifier":"For n=5 and a fine grid of p in (3, 3+$δ^{1}$_5), compute the maximum of \\|f-m_f\\|_p^p / \\mathrm{Var}_p(f)^p over the one-parameter family f(a_1)=1, f(a_5)=-1, f(a_2)=f(a_3)=f(a_4)=x, x∈[0,1]. The theorem predicts this maximum is exactly 1/($2^{{p-1}}$+3) and is attained at x=0. If any x≠0 yields a strictly larger value, the sharp constant for that p is wrong. A second check: evaluate the power-series inequality (2.14) at x=1/2 for n=5 and p=3+$δ^{1}$_5; if the infinite sum is negative, the auxiliary lemma fails.","tokens_in":11524,"feed_emoji":"📐","tokens_out":10686,"duration_ms":97134,"temperature":0.7,"pith_summary":"The paper solves the extremal problem of comparing a function's deviation from its mean with its total p-variation on the complete graph K_n. For each p in [1, 3+$δ^{1}$_n) ∪ (3+$δ^{2}$_n, ∞), it pins down the sharp multiplicative constant and, more, identifies every function that attains equality. The answer is piecewise: balanced two-valued functions for 1<p<2, a one-high-one-low shape with the interior value exactly midway for 2<p≤3+$δ^{1}$_n, and a single spike for the large-p range. If correct, this closes the problem on K_n and provides a clean model for how extremizers change as the exponent crosses 2 and 3.","feed_headline":"Exact constants settle Poincaré inequality on complete graphs","feed_subtitle":"The optimal constant and every extremal function are now known for all p, in three regimes.","key_machinery":"The load-bearing object is an auxiliary sharp inequality (Lemma 3) comparing (max f - m)^{p-1} + (m - min f)^{p-1} with the sum over vertices of (max f - f(a_j))^{p-1} + (f(a_j) - min f)^{p-1}. The argument shows that any extremizer of the original quotient satisfies the Euler-Lagrange-type equation (2.3), and that equation is exactly the equality case of this auxiliary inequality; classifying the maximizers of the auxiliary inequality therefore classifies the extremizers of the main problem. The proof of the auxiliary inequality uses symmetrization, convexity and concavity via majorization arguments, and, in the delicate interval 3<p≤3+$δ^{1}$_n, a binomial-series expansion whose coefficients are controlled by a sign-counting rule for polynomial roots.","core_discovery":"On the complete graph with n≥3 vertices, for every exponent p in the stated union of intervals the paper proves that the optimal constant C_{n,p} in \\|f-m_f\\|_p ≤ C_{n,p} \\mathrm{Var}_p(f) is exactly: 2/n for p=1; (⌊n/2⌋^{p-1}+⌈n/2⌉^{p-1})/n^p for 1<p<2; 1/n (an identity) for p=2; 1/($2^{{p-1}}$+n-2) for 2<p≤3+$δ_n^{1}$; and (1+(n-1)^{p-1})/n^p for p≥4 and for p∈(3,4) satisfying the explicit condition (n-2)/n ≥ (1+(n-1)^{p-1})/$n^{{p-1}}$. It also lists the equality cases in each regime: two-valued functions with balanced level sizes for 1<p<2; functions taking values a+c, a, ..., a, a-c (with the interior value exactly the midpoint) for 2<p≤3+$δ_n^{1}$; and single-spike functions for the large-p regime. The proof obtains these constants by showing that every extremizer satisfies a pointwise critical equation that is precisely the equality case of an auxiliary sharp inequality on the extreme deviations.","pith_inferences":["The same auxiliary-inequality strategy could be tried on other vertex-transitive graphs; if it works, one would expect the extremizers to organize into the same three families, with crossover exponents set by the graph's diameter or spectral gap rather than by n alone.","The small window (3+δ^1_n, 3+δ^2_n] that the theorem leaves untreated is a natural place to search for a fourth family of maximizers, or for a continuum of extremizers; the paper's own formulas make the gap visible and give a concrete boundary to test.","Because the constants are explicit and the equality cases are finite-dimensional, the result yields a ready-made finite set of test functions for numerical algorithms that estimate Poincaré constants on graphs."],"forward_implications":["Any use of the Poincaré–Wirtinger inequality on K_n can now quote the sharp constant instead of a bound; no improvement is possible within the stated p-ranges.","The equality-case list is complete: for each regime, the only functions attaining the constant are the specified two-valued, three-valued, or spike functions up to translation and scaling.","For p=2 the inequality is an identity with constant 1/n, so K_n is a graph on which every function is an extremizer.","The threshold structure shows a genuine phase transition at p=3: below it the extremizer is a midpoint three-valued function, above it (in the covered range) the extremizer is a single spike, with a small open window between 3+δ^1_n and 3+δ^2_n where the sharp constant is not claimed.","The p=1 endpoint has constant 2/n and its extremizers are exactly two-valued functions with any split, while for 1<p<2 the split must be as balanced as possible."],"supporting_citations":[],"fun_headline_variants":["Exact constants and all extremizers for complete graph inequality","Sharp inequality on complete graphs: optimal constants known for all p","Complete graph Poincare: exact constants and all sharp maximizers","Exact optimal constants for sharp complete-graph inequality, all p","On complete graphs, exact Poincare constants and all maximizers"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"Everything for 3<p≤3+$δ^{1}$_n rests on the auxiliary inequality (2.5) being exactly sharp, and its proof in that window depends on a sign analysis of a truncated binomial series; if that sign analysis has a gap, the claimed sharp constant in that interval collapses even if the surrounding cases stand.","fun_headline_variants_meta":{"raw":{"variants":["Exact constants and all extremizers for complete graph inequality","Sharp inequality on complete graphs: optimal constants known for all p","Complete graph Poincare: exact constants and all sharp maximizers","Exact optimal constants for sharp complete-graph inequality, all p","On complete graphs, exact Poincare constants and all maximizers"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001139,"raw_usage":{"total_tokens":4784,"prompt_tokens":1059,"completion_tokens":3725,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":675,"completion_tokens_details":{"reasoning_tokens":3653}},"tokens_in":675,"tokens_out":3725,"duration_ms":26734,"temperature":1.0,"reasoning_tokens":3653,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T17:59:29.484338+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"For n=5 and a fine grid of p in (3, 3+$δ^{1}$_5), compute the maximum of \\|f-m_f\\|_p^p / \\mathrm{Var}_p(f)^p over the one-parameter family f(a_1)=1, f(a_5)=-1, f(a_2)=f(a_3)=f(a_4)=x, x∈[0,1]. The theorem predicts this maximum is exactly 1/($2^{{p-1}}$+3) and is attained at x=0. If any x≠0 yields a strictly larger value, the sharp constant for that p is wrong. A second check: evaluate the power-series inequality (2.14) at x=1/2 for n=5 and p=3+$δ^{1}$_5; if the infinite sum is negative, the auxiliary lemma fails.","supporting_citations":[],"review_version":1}