{"id":"8b65cd9a-9cbd-424b-a64b-2786179dc462","arxiv_id":"2608.12094","paper_version":1,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"The paper proves the Cohn-Kumar 196560 auxiliary-function conjecture by showing that all functions g_C = a_2 - C b_2 with C above a finite threshold satisfy the required sign conditions and Poisson-summation identity.","lead":"This paper constructs an explicit family of 24-dimensional functions that satisfies the 2009 Cohn-Kumar auxiliary-function conjecture for the Leech lattice, proving a previously open existence problem. The proof combines Fourier interpolation with exact cusp expansions and yields every Leech-shell coefficient as a bonus.","discovery_kind":"first_principles","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The infinity control rests on a sign-sensitive, unverified coefficient extraction; a single error in Tables 1–2 would break the boundedness proof, so full acceptance should wait for an independent recomputation.","rationale":"The manuscript gives a coherent proof: the sign reduction is sound given Proposition 2.2, the quotient extensions across nodes are handled correctly, and the final Poisson-summation step is exact. I checked several of the displayed series manipulations (theta expansions, the identity Δ j = E4^3, the inverse j expansion, and the aggregate sums in Tables 1–2) and found them consistent. The load-bearing risk I see is not the external sign anchor from [4], which is established, but the internal finite coefficient extraction in Sections 4–5. All of the infinity control (boundedness of R and Rhat, common limit ρ, first correction with its sign) flows from (4.34)–(4.41) and Tables 1–2. Those are long, sign-dependent formal-power-series identities, and the paper explicitly disclaims computer verification while disclosing LLM assistance in the extraction. That does not make the proof wrong, but it makes independent verification the appropriate condition for full acceptance. If the proposed CAS check reproduces the tables, the concern is resolved and the reader's ACCEPT is justified; otherwise the cusp expansions and the finiteness of C* must be re-examined.","tokens_in":15459,"tokens_out":26772,"duration_ms":247239,"concrete_test":"Run a computer-algebra recomputation of the [X^4] coefficients of K+ and K- from (4.11)–(4.14) using the truncated series (4.18)–(4.33), and compare with (4.38)–(4.41). Then recompute the sums in Tables 1 and 2, verifying the zero cancellations through degree 4 and the sixth-degree entries [x^6]A = 104448960π^2 and [x^6]B = 152409600π^2. If these match, re-derive (5.14)–(5.19) and confirm c0 = 458752/π and d1 = 10080π(713 - 840 log 2) > 0.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The proof of finiteness of C* (Proposition 6.2) and of C* > ρ (Proposition 6.3) is carried entirely by the exact coefficient extraction in Sections 4–5: the S-cusp expansions (5.14)–(5.19) depend on (4.34)–(4.41) and on the row-by-row cancellations in Tables 1–2. These identities are finite but long and sign-sensitive: they involve branch-dependent logarithms (4.29)–(4.30), the sign flips caused by S2 = -z^{-2}V(w) in Section 5.1, and the parity substitution x ↦ -x used for the hatted coefficients. The manuscript states that no computer-assisted verification was used and discloses that portions of the symbolic extraction were developed with an OpenAI language model. A single incorrect scalar, a missed term at the truncation boundary, or a sign error in (4.38)–(4.41) or Tables 1–2 would change the leading constants c0, d1, or ρ; boundedness of R and Rhat would then not follow, and the existence of the ray [C*,∞) would collapse. This step is internal and not corroborated by an external theorem, unlike the sign anchor from [4], and no machine-checked certificate is supplied.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper constructs radial Schwartz functions on R^24 that prove the 2009 Cohn-Kumar auxiliary-function conjecture for the Leech lattice. Working with the radial Fourier interpolation basis of [3], the authors form the one-parameter family g_C = a_2 - C b_2. They identify b_2 with -16380 times the sphere-packing magic function of [4], which supplies strict signs for b_2 and its Fourier transform. The main technical work is an exact analysis of the quotients a_2/b_2 and hat-a_2/hat-b_2 on their respective half-lines: after proving that all apparent singularities are removable, the paper computes the common limit rho=(43+240 log 2)/15 through an S-cusp expansion of the n=2 interpolation coefficients and obtains an explicit first exponential correction. It follows that the admissible parameters form a nonempty closed ray [C*, infinity), and every member satisfies the required sign conditions and the ratio identity (hat-g(0)-g(0))/g(2)=196560 by Poisson summation over the Leech lattice. A corollary recovers every nontrivial Leech-shell coefficient from the interpolation basis.","tokens_in":15655,"tokens_out":21403,"duration_ms":212003,"significance":"This is a substantial result: it settles a conjecture that has been open since 2009 and gives a clean, parameter-free construction from Fourier interpolation rather than a numerical search. The proof has several strong features: the target value 196560 enters only through Poisson summation and is not an input; the sign anchor is the independent sphere-packing function from [4]; the asymptotic analysis is carried out by explicit finite identities; and the paper honestly discloses which inputs come from [3] and [4]. I spot-checked representative entries in the coefficient tables and found them consistent with the displayed formulas. The absence of a machine-checked certificate is not, in my view, a defect, because the coefficient extraction is presented as finite identities in formal power series rather than as an opaque computer calculation.","major_comments":[],"minor_comments":[{"comment":"The blank entries in Tables 1 and 2 are easy to misinterpret in a text rendering; please ensure that the final typeset version has explicit column alignment (for example, by repeating the power of x in each header or by using a vertical layout) so that each coefficient can be checked unambiguously.","section":"Tables 1-2"},{"comment":"The branch choices for L and L_S are described as those used in [3], but since the S-cusp calculation is sign-sensitive, it would help to state the branch convention in a self-contained sentence before these expansions are used.","section":"Section 4.3, Eqs. (4.29)-(4.30)"},{"comment":"The statement that the leading constant in the Bessel asymptotics 'simplifies exactly to 1' would benefit from a one-line derivation, since the powers of a and the r^{-23/2} factor are not immediately obvious from the preceding display.","section":"Section 6, Eq. (6.7)"},{"comment":"Given the disclosure that portions of the symbolic extraction were developed with an OpenAI language model, I suggest adding a sentence clarifying whether the displayed identities (4.38)-(4.41) and Tables 1-2 were subsequently re-derived and checked independently by the authors; this would directly address reproducibility concerns.","section":"Section 8"}],"recommendation":"minor_revision","confidential_remarks":"This is a strong paper and well within the journal's scope. The only caveat I have is the LLM-assisted coefficient extraction, but the authors have disclosed it transparently and the identities are finite and checkable. I would be satisfied with minor revisions; I would not require a machine certificate, although the editor may reasonably ask the authors for one as an additional safeguard."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short take: it's the real thing—it proves the 196560 auxiliary-function conjecture—but the load-bearing finite computation is exactly where I'd want a referee to spend time.\n\nWhat's new: the affine ansatz g_C = a_2 - C b_2, the reduction of the conjecture to boundedness of the two quotients, and the exact S-cusp expansion showing both quotients tend to ρ = (43 + 240 log 2)/15 with opposite-side exponential corrections. That last bit is nice: it gives C* > ρ explicitly. The corollary recovering every Leech-shell coefficient from interpolation is a clean bonus.\n\nWhat's good: the sign reduction is clean and uses only the sign anchor from [4] plus removability of quotient singularities. The cusp-to-Laplace transfer in Lemma 6.1 is a solid piece of analysis: the Bessel function ratio handles the exponential decay properly and the Gaussian tail bounds are correct. The modular-form machinery is used with the right amount of care, and the authors are explicit about which parts come from [3] and [4]. There's no fitting: the ratio 196560 comes out of Poisson summation, not from tuning coefficients.\n\nThe soft spot is the one the stress-test flags: Sections 4–5 rest on a sign-sensitive coefficient extraction. Tables 1–2 and identities (4.38)–(4.41) are finite but long, with branch-dependent logs and a parity substitution. A single scalar error would change the leading constants and break the boundedness proof. The authors disclose that part of the extraction was developed with an AI language model and that no machine check was run. That is not a sin, but it makes independent verification important. The good news: everything needed is in the paper, so a referee (or the authors) can check the tables with a computer algebra system in an afternoon. The valuation argument at the end of Section 4 correctly explains why the truncation is exact, so this is a human-verifiable proof, not an appeal to a black box.\n\nI don't see a circularity problem: the external inputs are independent published theorems, and the new contribution is the quotient analysis. The theorem itself is a real open-problem resolution, significant for the Fourier-interpolation and sphere-packing communities. The paper deserves a serious referee, and the referee should be asked to verify Tables 1–2 and the signs in (5.7)–(5.8). I'd accept subject to that verification.","headline":"Proves the Cohn–Kumar 196560 conjecture via a clean quotient argument; the main open risk is the explicit but unverified coefficient extraction in Tables 1–2.","tokens_in":16206,"tokens_out":2727,"would_cite":true,"duration_ms":27438,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["42A38","11F27","11H31","52C17"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper proves the 2009 auxiliary-function conjecture for the Leech lattice by constructing an explicit one-parameter family of radial Schwartz functions, the admissible members of which form exactly a closed ray.","keywords":["Fourier interpolation","Leech lattice","radial Schwartz function","auxiliary function","Poisson summation","sphere packing","modular forms","theta series"],"falsifier":"Evaluate the paper's own explicit Laplace-integral formulas for b_2 and hat b_2 at a single non-node point, say r=3 for b_2 and r=1/2 for hat b_2; if b_2(3) were not positive, or hat b_2(1/2) not negative, the sign reduction would fail and the ray [C*, infinity) would not be admissible.","tokens_in":15217,"feed_emoji":"📐","tokens_out":6442,"duration_ms":62856,"temperature":0.7,"pith_summary":"This paper proves a 2009 conjecture about the Leech lattice: there exists a radial Schwartz function g on $R^{24}$ whose direct side is nonpositive outside radius $\\sqrt$(6), whose Fourier transform is nonnegative everywhere, which is positive at radius 2, and whose Poisson-summing ratio (hat g(0)-g(0))/g(2) equals 196560, the number of shortest Leech vectors. The proof constructs g as g_C = a_2 - C b_2, a one-parameter family built from the first two radial Fourier interpolation basis functions, and shows the admissible parameters are exactly a closed ray [C*, infinity). This matters because such an auxiliary function is the missing input that makes the linear-programming optimality argument for the Leech lattice fully explicit, with no numerical search. The same basis also recovers every nontrivial Leech-shell coefficient.","feed_headline":"Construction finishes the 196560 Leech-lattice conjecture","feed_subtitle":"A closed ray of Schwartz functions yields the 196560 ratio through Poisson summation.","key_machinery":"The central object is the radial Fourier interpolation basis in dimension 24: Schwartz functions a_n, b_n, tilde a_n, tilde b_n indexed by n >= 2 that interpolate the value and first radial derivative of a function and of its Fourier transform at the radii sqrt(2n). Only the first value-derivative pair a_2, b_2 is used, with b_2 identified as a fixed multiple of the dimension-24 sphere-packing magic function, which supplies the strict signs b_2(r)>0 for r>2 away from nodes and hat b_2(r)<0 for r>=0 away from nodes. The argument is carried by the removable quotients R(r)=a_2/b_2 and hat R(r)=hat a_2/hat b_2: their boundedness converts the sign conditions into C >= sup, and their asymptotic limits are computed by exact coefficient extraction from the interpolation kernel, an S-cusp expansion at z=it, and a Laplace-Bessel transfer to large radii.","core_discovery":"The paper's central claim is that the 2009 auxiliary-function conjecture for the Leech lattice is true, with a complete description of the admissible functions in its natural affine family. For every real C at least C*, where C* is the maximum of the suprema of the smooth quotient functions R(r)=a_2(r)/b_2(r) on r >= $\\sqrt$(6) and hat R(r)=hat a_2(r)/hat b_2(r) on r >= 0, the function g_C = a_2 - C b_2 is a radial Schwartz function on $R^{24}$ satisfying g_C(r) <= 0 for r >= $\\sqrt$(6), hat g_C(r) >= 0 for r >= 0, g_C(2)=1, and (hat g_C(0)-g_C(0))/g_C(2)=196560. Poisson summation over the self-dual Leech lattice gives exactly this identity, because the interpolation node data force g_C to vanish on every nonzero Leech shell except the shortest one and force hat g_C to vanish on every nonzero Leech shell. The paper also shows the admissible set in the affine line is exactly the closed ray [C*, infinity), with C* > rho = (43+240 log 2)/15.","pith_inferences":["Because C* is a maximum of two explicit smooth quotients with known limits and corrections, one could compute rigorous two-sided bounds on C* by evaluating those quotients on a finite grid plus a tail estimate; the paper does not attempt this.","The quotient strategy should transfer to other dimensions with an interpolation basis and a signed anchor function, such as the analogous setting in dimension 8, yielding an exact ray of auxiliary functions for the corresponding ratio conjecture; the paper only treats dimension 24.","The exact coefficient cancellations at the cusp suggest a general modular identity governing the n=2 coefficients; testing whether similar finite identities hold for n>2 could make shell-coefficient recovery uniform."],"forward_implications":["The complete admissible set in the affine family is [C*, infinity), so validity is not a single lucky function but a whole ray of valid auxiliary functions.","Every member g_C with C >= C* proves the 196560 identity through Poisson summation, with no hidden numerical verification.","The explicit asymptotics R(r), hat R(r) -> rho with opposite exponential corrections imply C* > rho and pin down the direction from which the quotients approach their limit.","The same interpolation basis yields exact Leech theta coefficients N_m = (65520/691)(sigma_11(m) - tau(m))."],"supporting_citations":[{"why":"Poses the auxiliary-function conjecture and supplies the Poisson lower-bound framework that makes the 196560 ratio the target.","marker":"[1]"},{"why":"Supplies the radial Fourier interpolation theorem, the basis functions a_n, b_n and their tilded counterparts, and the integral formulas used in the Laplace transfer.","marker":"[3]"},{"why":"Supplies the sphere-packing magic function f_sp, its strict sign properties and integral representations; b_2 is identified with a nonzero multiple of f_sp.","marker":"[4]"},{"why":"Provides the standard facts about the Leech lattice: self-duality, the absence of vectors of squared length 2, and the count 196560 on the shortest shell.","marker":"[7]"}],"fun_headline_variants":["Explicit Schwartz functions close Leech lattice 196560 conjecture","Construction yields all admissible functions for Leech lattice 196560","Leech lattice conjecture proved via interpolation basis and Poisson sum","196560 achieved: closed ray of admissible Schwartz functions"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"Everything rests on the sign anchor—b_2 > 0 on r > 2 except at nodes and hat b_2 < 0 on r >= 0 except at nodes—which is inherited from the previously constructed sphere-packing magic function and not reproved here.","fun_headline_variants_meta":{"raw":{"variants":["Explicit Schwartz functions close Leech lattice 196560 conjecture","Construction yields all admissible functions for Leech lattice 196560","Leech lattice conjecture proved via interpolation basis and Poisson sum","196560 achieved: closed ray of admissible Schwartz functions"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000182,"raw_usage":{"total_tokens":1374,"prompt_tokens":1075,"completion_tokens":299,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":691,"completion_tokens_details":{"reasoning_tokens":231}},"tokens_in":691,"tokens_out":299,"duration_ms":3441,"temperature":1.0,"reasoning_tokens":231,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-16T00:17:54.610530+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Evaluate the paper's own explicit Laplace-integral formulas for b_2 and hat b_2 at a single non-node point, say r=3 for b_2 and r=1/2 for hat b_2; if b_2(3) were not positive, or hat b_2(1/2) not negative, the sign reduction would fail and the ray [C*, infinity) would not be admissible.","supporting_citations":[{"cited_title":"Cohn and A","cited_arxiv_id":null,"evidence_quote":"Poses the auxiliary-function conjecture and supplies the Poisson lower-bound framework that makes the 196560 ratio the target."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the radial Fourier interpolation theorem, the basis functions a_n, b_n and their tilded counterparts, and the integral formulas used in the Laplace transfer."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the sphere-packing magic function f_sp, its strict sign properties and integral representations; b_2 is identified with a nonzero multiple of f_sp."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the standard facts about the Leech lattice: self-duality, the absence of vectors of squared length 2, and the count 196560 on the shortest shell."}],"review_version":1}