{"id":"3032462c-2c9f-4acc-8e35-f6daa3d1f222","arxiv_id":"2608.08695","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 shows biangular tight frames make two projection constants coincide, derives a new lower bound from simplex edge midpoints, and conjectures exact maximal projection constants in dimensions 6 and 8.","lead":"This paper studies symmetric vector arrangements, called biangular tight frames, and shows they make two standard projection constants coincide. It derives a new lower bound from the edge midpoints of a regular simplex and conjectures exact values in dimensions 6 and 8.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Conjecture 2.1 depends on unverified numerical global searches; the small gap in Figure 1 to a non-sharp upper bound does not exclude better configurations.","rationale":"The reader's verdict is CONDITIONAL, and I agree. The load-bearing assumption is indeed the global optimality of previously reported numerical searches. I checked the internal proofs: Lemma 2.1 is a valid rearrangement argument; Example 2.1's verification that the edge-midpoint family is a Parseval frame is acceptable via the Gram-trace identity; the compressed proof of Theorem 3.2 rearranges correctly to give (8). Thus the mathematical core is not where the risk lies. The risk is epistemic: the sharp values in Conjecture 2.1 are the headline of the paper, and their only support is unexaminable numerical evidence plus a suggestive but non-rigorous gap to the upper bound. This does not require changing the verdict: since the manuscript is honest about the conjectural status, CONDITIONAL remains the right posture. If the numerical evidence were made available and independently certified, the conjecture would be placed on much firmer ground.","tokens_in":9130,"tokens_out":11549,"duration_ms":117481,"concrete_test":"Run an independent certified global optimization for λ_R(6) and λ_R(8) using the Seidel-matrix formulation (9), varying N from m up to at least 2m(m+1), with a branch-and-bound or interval-based method that returns rigorous upper bounds. If any configuration exceeds 16/7 (for m=6) or 8/3 (for m=8), Conjecture 2.1 is false; if the maximum is attained exactly at the edge-midpoint values and no higher value is found, the concern is resolved. In parallel, re-run the numerical procedure referenced as [25] from many random starting frames to confirm that the reported maxima are reproducible.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The proven content of Section 2 (Lemma 2.1, Corollary 2.1) is solid, but Conjecture 2.1 asks the reader to accept that the edge-midpoint frame attains the global maximum, not merely a local one. The evidence offered is (i) the small gap between the lower bound λ_R(V_m) and the upper bound of Theorem 1.3 for m=6,8, and (ii) numerical estimates attributed to Chalmers and to [25]. Neither is decisive: the Theorem 1.3 bound is not attained by V_m, so a gap of about 0.02 in dimension 6 and 0.015 in dimension 8 only shows closeness to the known ceiling, not that no intermediate value is achievable. Moreover, the searches behind [25] are not described in the manuscript, and no code or optimality certificates are supplied. The optimization problems in (2) and (9) are non-concave over the frame or Seidel variables, so local maxima are a real hazard. If the reported numerics were only local, the equalities λ_R(6)=16/7 and λ_R(8)=8/3 could fail even though every theorem in the paper is correct.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper studies the maximal absolute projection constant λ_K(m) and candidate extremal configurations. After recalling the frame-theoretic formulations (2) and (9), the authors prove Lemma 2.1, stating that for any biangular Parseval tight frame the relative and quasimaximal constants coincide. They then construct, in Example 2.1, the frame V_m formed by the midpoints of edges of a regular m-simplex, verify that it is a biangular Parseval tight frame, and compute λ_R(V_m)=4(m-2)/(m+1), yielding Corollary 2.1. They conjecture that this lower bound is sharp for m=6 and m=8. Section 3 gives a proof of the known König–Tomczak-Jaegermann upper bound via the spherical (2,2)-design inequality, and discusses weighted spherical (2,2)-designs in R^4 and R^5 whose sign matrices coincide with those of conjectured maximizers, leading to Conjecture 3.1.","tokens_in":9393,"tokens_out":21882,"duration_ms":206441,"significance":"The paper's main proven contribution, Lemma 2.1 and the edge-midpoint frame in Example 2.1, is explicit and verifiable; the computation gives a clean lower bound for λ_R(m). The conjectures are clearly stated and are falsifiable, and the connection between biangular tight frames and quasimaximal constants is a useful structural observation. The proof of Theorem 3.2, once the omitted algebra is supplied, provides a compact route to a known bound. The numerical evidence for Conjecture 2.1 is not documented in detail, so the conjectures should be read as plausible but not strongly established. Overall this is a useful contribution to the study of extremal projection constants, though its main claims are conjectural.","major_comments":[{"comment":"The step described as 'Squaring the addends of the left-sided sum and rearranging the latter inequality' is a substantial algebraic manipulation. Let A=Σ t_i t_j |⟨u_i,u_j⟩|²/(‖u_i‖‖u_j‖), B=Σ t_i t_j |⟨u_i,u_j⟩|, and C=Σ t_i t_j ‖u_i‖‖u_j‖. The displayed inequality reads A-2φB+φ²C ≥ (c₂C-2φ²A+φ⁴C)/(1+φ)², and rearranging yields (8). Please write out these intermediate steps; in the current form the proof of Theorem 3.2 cannot be checked without reconstructing this algebra.","section":"§3, Theorem 3.2, Eq. (8)"},{"comment":"The conjecture that λ_R(6)=16/7 and λ_R(8)=8/3 rests on the small gap between λ_R(V_m) and the non-sharp upper bound of Theorem 1.3, and on numerical estimates attributed to Chalmers and to [25]. The numerical searches behind (2) and (9) are not described, and the optimization problems are non-concave, so without details about the algorithms, tolerances, or optimality certificates the reader cannot assess whether these are global maxima. Please either document the numerical evidence or explicitly label Conjecture 2.1 as heuristic.","section":"§2, Conjecture 2.1"}],"minor_comments":[{"comment":"The reduction to nonzero columns is too terse: after deleting zero columns, the coefficient vector t should be restricted and renormalized; the displayed inequality with the same t is only valid after this is explained.","section":"§3, Theorem 3.2 proof"},{"comment":"The sentence 'A = sgn(U^T U) realizes the maximum in (9)' appears inconsistent with the definition A=I_N+B in (9), since the latter has off-diagonal entries in {0,2}. Please state precisely how the sign matrix is embedded in the Seidel-matrix formulation.","section":"§3, after Theorem 3.4"},{"comment":"The notation 'U ∈ R^m' is incorrect for a sequence of vectors; also the definition of sgn(U^T U) for vanishing inner products should be stated explicitly.","section":"Conjecture 3.1"},{"comment":"The combinatorial counts in the displayed sum (how many pairs fall into each of the three categories) would be easier to verify if a short counting argument were included.","section":"Example 2.1"},{"comment":"Reference [16] has a typo: 'Small number of vectors,,' should be 'small number of vectors'.","section":"References"},{"comment":"The figure needs axis labels, a caption, and a non-color legend; the current 'yellow dots' vs 'blue ones' description is hard to use in print.","section":"Figure 1"}],"recommendation":"minor_revision","confidential_remarks":"The paper is partly a research announcement: the main conjectures (2.1 and 3.1) are not proven, and the supporting numerical evidence is not fully documented. If the journal's scope requires complete proofs for the main claims, this should be carefully weighed. I also note that reference [25] is a very recent arXiv preprint; the editor may wish to verify its availability before publication."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Quick take: the proven part of this paper is worth having. Lemma 2.1 is correct and clean: for a biangular Parseval frame, the relative and quasimaximal projection constants coincide. The edge-midpoint construction in Example 2.1 is explicit, the Parseval-frame verification is complete, and Corollary 2.1 gives a real lower bound for lambda_R(m) that I don't think is in the literature. The re-derivation of the Koenig–Tomczak-Jaegermann bound through the spherical (2,2)-design inequality is also a nice route, though it has a gap.\n\nThe soft spots are real but not fatal. First, Conjecture 2.1 — lambda_R(6)=16/7 and lambda_R(8)=8/3 — depends on numerical searches by Chalmers and by Sivashankar–Tang–Wakhare that are described only by attribution. No code, no data, no optimality certificates are supplied. The optimization in (2) and (9) is non-concave, so local maxima are a genuine hazard, and the small gap to the upper bound in Figure 1 does not exclude intermediate configurations. The conjecture may well be true, but as presented it is an informed guess, not an established claim. Second, in the proof of Theorem 3.2, the step described as “squaring the addends” hides a nontrivial rearrangement. A referee will need to see that algebra. Third, Conjecture 3.1 is sweeping: as stated, it says that every weighted spherical (2,2)-design with a Seidel sign matrix attains the global maximum lambda_R(m). That is a strong claim and probably needs qualification; I would not be surprised if there are counterexamples.\n\nThe citation pattern is fine: the one self-citation (Theorem 1.3, attributed to [11]) is used as an upper bound and is re-proved in this paper, so there's no circular dependence. The examples in dimensions 4 and 5 are suggestive, and the sign-matrix observation is a reasonable topic for future work.\n\nWho is this for? Banach space theorists working on projection constants and frame theorists interested in extremal configurations. They will get a useful lower bound and a clean observation about biangular frames. The conjectures are a legitimate research prompt, not a result.\n\nRecommendation: send it to a serious referee. It deserves referee time, with the request that the missing algebra in Theorem 3.2 be supplied and that the numerical basis for Conjecture 2.1 be made reproducible or explicitly softened.","headline":"A clean examples-and-conjectures note: the proven lower bound via edge-midpoint frames is solid, but the headline exact values in dimensions 6 and 8 rest on undocumented numerical searches and should be treated as well-motivated guesses, not results.","tokens_in":9907,"tokens_out":3208,"would_cite":true,"duration_ms":33805,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["46B20","15A42","42C15"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper conjectures that the maximal absolute projection constant equals 16/7 in dimension 6 and 8/3 in dimension 8, attained by the edge midpoints of a regular simplex, and proves a general lower bound of 4(m−2)/(m+1) for every m≥3.","keywords":["maximal absolute projection constant","relative projection constant","quasimaximal projection constant","tight frames","biangular tight frames","spherical (2,2)-designs","Seidel matrices","edge midpoints of regular simplex"],"falsifier":"Run a global-optimization search over all Parseval frames $U\\in\\mathbb R^{6\\times 15}$ and $U\\in\\mathbb R^{8\\times 36}$, starting from the explicit edge-midpoint frames; if any such frame has $\\lambda_{\\mathbb R}(U)>16/7$ or $\\lambda_{\\mathbb R}(U)>8/3$ respectively, Conjecture 2.1 is false.","tokens_in":8927,"feed_emoji":"📐","tokens_out":11998,"duration_ms":113191,"temperature":0.7,"pith_summary":"The paper tries to pin down the largest possible absolute projection constant λ_R(m) among m-dimensional real Banach spaces, a number known exactly in only a few dimensions. Its concrete proposal is that in dimensions 6 and 8 the extremal configuration is the set of midpoints of the edges of a regular simplex, giving λ_R(6)=16/7 and λ_R(8)=8/3. The supporting result is a general lower bound: for every m≥3 these edge midpoints form a biangular Parseval tight frame whose relative and quasimaximal projection constants both equal $4(m-2)/(m+1)$. The same analysis, together with examples in dimensions 4 and 5, leads the authors to conjecture that weighted spherical (2,2)-designs—and ultimately only the sign pattern of their Gram matrices—determine maximal projection constants. If true, the problem acquires a finite, combinatorial face that existing methods can attack.","feed_headline":"Edge midpoints may maximize projection constants in dimensions 6 and 8","feed_subtitle":"Edge-midpoint frames give certified lower bounds; numerical searches suggest they are exact in dimensions 6 and 8.","key_machinery":"The mechanism carrying the argument is the identity $\\lambda_{\\mathbb K}(U)=\\mu_{\\mathbb K}(U)$ for biangular tight frames. A biangular tight frame is a Parseval frame whose pairwise inner-product magnitudes take exactly two values; the equidistribution property fixes the number of large entries in each row, and a Cauchy–Schwarz style row-sum argument shows that the uniform weight vector already maximizes the quadratic form defining $\\lambda_{\\mathbb K}(U)$. The edge-midpoint configuration $V_m$ enters as an explicit biangular Parseval frame with exactly $m(m+1)/2$ vectors, and computing its row sums gives the value $4(m-2)/(m+1)$. The second mechanism is the spherical-design inequality of Theorem 3.1, which yields the König–Tomczak-Jaegermann upper bounds, together with the sign-matrix reformulation via Seidel matrices, which turns the maximization into one over finite combinatorial sign patterns.","core_discovery":"The central discovery is that for biangular tight frames the relative projection constant and the quasimaximal relative projection constant coincide, because the equidistribution property makes every row of the Gram matrix contribute the same weighted sum. Applying this to the $m(m+1)/2$ midpoints of the edges of a regular $m$-simplex, which form a biangular Parseval frame whose inner products take exactly two absolute values, gives $\\lambda_{\\mathbb R}(V_m)=4(m-2)/(m+1)$ and hence the lower bound $\\lambda_{\\mathbb R}(m)\\ge 4(m-2)/(m+1)$. For $m=6$ and $m=8$ the gap between this value and the best known upper bound is small, and the paper's numerical evidence agrees with equality; the paper therefore conjectures $\\lambda_{\\mathbb R}(6)=16/7$ and $\\lambda_{\\mathbb R}(8)=8/3$. A separate thread identifies weighted spherical (2,2)-designs with the smallest possible number of vectors as candidate extremizers in dimensions 4 and 5, where the relevant quantity appears to depend only on the sign matrix of the Gram matrix rather than on the exact inner products.","pith_inferences":["The proof of Lemma 2.1 is field-independent, so the same equality should hold for complex biangular tight frames; if a complex analogue of the edge-midpoint frame exists, it would yield analogous lower bounds for $\\lambda_{\\mathbb C}(m)$.","The pattern in dimensions 6 and 8 suggests a heuristic: when a real maximal equiangular tight frame is impossible, the extremizer may be the next most symmetric two-distance tight frame with $N=m(m+1)/2$ vectors, singling out the edge-midpoint family as a systematic search space.","If Conjecture 3.1 transfers to arbitrary $N$, then $\\lambda_{\\mathbb R}(m)$ could be computed by enumerating Seidel matrices up to a bound related to minimal spherical designs, connecting projection constants to spectral graph theory in a concrete way.","A natural higher-precision test would be a random-restart optimization over the Stiefel manifold of Parseval frames for $m=6$ and $m=8$; if many independent runs plateau at $16/7$ or $8/3$ and never exceed them, that would strengthen, though not prove, the conjecture."],"forward_implications":["If Conjecture 2.1 is correct, $\\lambda_{\\mathbb R}(6)=16/7$ and $\\lambda_{\\mathbb R}(8)=8/3$ become the first exact maximal absolute projection constants in these two dimensions.","The lower bound $\\lambda_{\\mathbb R}(m)\\ge 4(m-2)/(m+1)$ holds in every dimension $m\\ge 3$, so the edge-midpoint configuration is always a certified starting point for searching exact values.","The equality of relative and quasimaximal constants for biangular tight frames means the uniform weighting is optimal for such frames, simplifying numerical tests of extremality.","If Conjecture 3.1 is correct, then for weighted spherical (2,2)-designs with no orthogonal pair the maximal projection constant can be obtained by maximizing only over Seidel sign matrices, a finite combinatorial problem for each fixed $N$.","The examples in dimensions 4 and 5 suggest that sign matrices, rather than exact inner products, may be the object that governs extremal projection constants."],"supporting_citations":[{"why":"Supplies the variational formula for the maximal relative projection constant and the Seidel sign-matrix reformulation used throughout the paper.","marker":"[5]"},{"why":"Provides the equidistribution property of biangular tight frames that underlies Lemma 2.1.","marker":"[7]"},{"why":"Proves that the absolute projection constant is the supremum of quasimaximal relative constants and proves the König–Tomczak-Jaegermann upper bound.","marker":"[11]"},{"why":"Gives the real-case proof that the absolute constant equals the supremum over quasimaximal constants, the route the paper follows.","marker":"[2]"},{"why":"Provides the weighted spherical (2,2)-design inequality and the example of 16 vectors in $\\mathbb R^5$.","marker":"[16]"},{"why":"States the König–Tomczak-Jaegermann upper-bound estimates that set the comparison values for $\\lambda_{\\mathbb R}(m)$.","marker":"[20]"},{"why":"Supplies the Gram matrix of an 11-vector weighted spherical (2,2)-design in $\\mathbb R^4$ used in the numerical match for dimension 4.","marker":"[22]"},{"why":"Gives the numerical estimates for $\\lambda_{\\mathbb R}(6)$ and $\\lambda_{\\mathbb R}(8)$ that support Conjecture 2.1.","marker":"[25]"},{"why":"Bounds the number of vectors in weighted spherical (t,t)-designs, identifying the minimal configurations considered in the paper.","marker":"[27]"},{"why":"Provides the conjectured sign matrix for the fifth maximal projection constant, used in the dimension-5 example.","marker":"[10]"}],"fun_headline_variants":["Edge midpoints may max out projection constants in 6D, 8D","Conjecture: Simplex edge midpoints are extremal in 6 and 8","New bounds from simplex edges: may be exact in 6, 8","Tight frames from simplex edges may solve dims 6 and 8"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The conjectured equalities in dimensions 6 and 8 rest on the unproven premise that the numerical searches for $\\lambda_{\\mathbb R}(6)$ and $\\lambda_{\\mathbb R}(8)$ found a global maximum, not merely a local one; if some other frame exceeded $16/7$ or $8/3$, the conjectures would fail.","fun_headline_variants_meta":{"raw":{"variants":["Edge midpoints may max out projection constants in 6D, 8D","Conjecture: Simplex edge midpoints are extremal in 6 and 8","New bounds from simplex edges: may be exact in 6, 8","Tight frames from simplex edges may solve dims 6 and 8"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001356,"raw_usage":{"total_tokens":5530,"prompt_tokens":1001,"completion_tokens":4529,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":617,"completion_tokens_details":{"reasoning_tokens":4442}},"tokens_in":617,"tokens_out":4529,"duration_ms":30485,"temperature":1.0,"reasoning_tokens":4442,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T04:27:31.603392+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run a global-optimization search over all Parseval frames $U\\in\\mathbb R^{6\\times 15}$ and $U\\in\\mathbb R^{8\\times 36}$, starting from the explicit edge-midpoint frames; if any such frame has $\\lambda_{\\mathbb R}(U)>16/7$ or $\\lambda_{\\mathbb R}(U)>8/3$ respectively, Conjecture 2.1 is false.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the variational formula for the maximal relative projection constant and the Seidel sign-matrix reformulation used throughout the paper."},{"cited_title":"Casazza, A","cited_arxiv_id":null,"evidence_quote":"Provides the equidistribution property of biangular tight frames that underlies Lemma 2.1."},{"cited_title":"Deregowska, B","cited_arxiv_id":null,"evidence_quote":"Proves that the absolute projection constant is the supremum of quasimaximal relative constants and proves the König–Tomczak-Jaegermann upper bound."},{"cited_title":"Basso,Computation of maximal projection constants,J","cited_arxiv_id":null,"evidence_quote":"Gives the real-case proof that the absolute constant equals the supremum over quasimaximal constants, the route the paper follows."},{"cited_title":"Hughes, S","cited_arxiv_id":null,"evidence_quote":"Provides the weighted spherical (2,2)-design inequality and the example of 16 vectors in $\\mathbb R^5$."},{"cited_title":"König, N","cited_arxiv_id":null,"evidence_quote":"States the König–Tomczak-Jaegermann upper-bound estimates that set the comparison values for $\\lambda_{\\mathbb R}(m)$."},{"cited_title":"ParkInteracion energies, lattices, and designsPh.D","cited_arxiv_id":null,"evidence_quote":"Supplies the Gram matrix of an 11-vector weighted spherical (2,2)-design in $\\mathbb R^4$ used in the numerical match for dimension 4."},{"cited_title":"Graph Eigenvalues and Projection Constants","cited_arxiv_id":"2608.02429","evidence_quote":"Gives the numerical estimates for $\\lambda_{\\mathbb R}(6)$ and $\\lambda_{\\mathbb R}(8)$ that support Conjecture 2.1."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Bounds the number of vectors in weighted spherical (t,t)-designs, identifying the minimal configurations considered in the paper."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the conjectured sign matrix for the fifth maximal projection constant, used in the dimension-5 example."}],"review_version":1}