{"id":"4b08716f-0b66-4f67-8cb4-1fac9ba233d4","arxiv_id":"2504.20564","paper_version":1,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"For G = SL_ℓ with ℓ prime, and for G = Sp4 or Sp6, the sum over semisimple conjugacy classes of modified L-functions L_S(M_{Gγ}) is a Lefschetz-type function of the base change degree m; assuming Gross's trace formula conjectures, this governs sums of multiplicities of cuspidal representations.","lead":"This paper counts cuspidal representations of split algebraic groups over function fields, assuming a conjectural trace formula, and studies how the counts change when the base field is extended. The main proved results show that for SL_ℓ, Sp4, and Sp6 the relevant L-function sums vary in a predictable Lefschetz-type way under base change.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Sp6 proof in Theorem 6.6 mis-evaluates Hτi at √−1; claimed cancellation fails for nonempty J, leaving L(Sp6,m) Lefschetz claim unproved.","rationale":"The paper's central unconditional claims are Theorems 5.4, 6.5, and 6.6. I checked the SL_ℓ and Sp4 arguments and found no issue. However, for Sp6 the proof of Theorem 6.6 contains a concrete algebraic error in the verification of divisibility by 1+x^2: it evaluates several different Hτi at the same value and cancels them incorrectly. A direct computation with J={1,2} contradicts the asserted identity. Consequently the proof that L(Sp6,m) is a polynomial in q^m and α^m—and hence of Lefschetz type—does not go through. The cuspidal-representation interpretation is honestly conditional on Gross's Conjectures 1.1 and 1.2, so that is not a fault; my concern is in the unconditional Sp6 part. The reader's verdict should be CONDITIONAL: the paper is acceptable only after the Sp6 computation is corrected or verified, or the theorem should be restricted to cases where the identity is proven.","tokens_in":26698,"tokens_out":48885,"duration_ms":416394,"concrete_test":"With a computer algebra system, define Rτi and Hτi exactly as in the proof of Theorem 6.6 (q odd) and evaluate P4(i)=(1+x^2)∑_{i=1,2,3,7,11}RτiHτi at x=i for the multiset J={1,2}. If the result is nonzero (it is 600−1140i), the claimed identity fails. Then check the full polynomial divisibility P(x,J)∈Z[x,J] for several random small J; this settles whether Theorem 6.6's Lefschetz conclusion is established.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"In the proof of Theorem 6.6 (q odd, part (d)), the author needs P4(√−1)=0 where P4=(1+x^2)P(x). The displayed computation replaces Hτ1,Hτ2,Hτ3,Hτ7 at x=√−1 by the same product ∏(1−√−1α)(1+α^2). But from the definitions in the same proof, Hτ1(√−1)=∏(1−√−1α)(1−√−1α^3)(1−√−1α^5), Hτ2(√−1)=∏(1−√−1α)^2(1−√−1α^3), and Hτ3(√−1)=∏(1+α)(1−√−1α)(1−√−1α^3); these are not equal to ∏(1−√−1α)(1+α^2) for general J. Concretely, with J={1,2}, computing the five non-vanishing terms (coefficients (1−i)/2, (1−i)/2, i, i−1, −i for τ1,τ2,τ3,τ7,τ11) gives 600−1140i ≠ 0. The q-even part has the same defect: it identifies Hτ1(√−1) with Hτ7(√−1) and Hτ3(√−1) with Hτ11(√−1). Hence divisibility by 1+x^2 is not proved; without it, P(x,J) need not lie in Z[x,J], and L(Sp6,m) may fail to be of Lefschetz type.","agreement_with_reader":"disagree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper assumes Gross's conjectural trace formula to express sums of multiplicities of cuspidal representations with prescribed local conditions as sums of Artin-Tate L-functions, and studies the behavior of these sums under the base change F_q ↦ F_{q^m}. The main unconditional results are Theorem 5.4 for SL_ℓ (ℓ prime) and Theorems 6.5 and 6.6 for Sp4 and Sp6, which assert that the corresponding sums over semisimple conjugacy classes are Lefschetz type functions of m; Proposition 4.4 treats the case where both S and T are nonempty. The connection to cuspidal representations is conditional on Gross's conjectures as stated in Theorem 4.2.","tokens_in":26962,"tokens_out":43610,"duration_ms":370246,"significance":"The SL_ℓ result is a clean, self-contained verification of Conjecture 4.6 in a non-trivial case, and the paper is transparent about the conditional nature of the trace-formula applications. If the Sp4 and Sp6 proofs were correct, they would provide further strong evidence for the expected Lefschetz-type behavior. However, as detailed in the major comments, the proofs of Theorems 6.5 and 6.6 contain a load-bearing error in the evaluation at x = -1, and the resulting expressions appear to fail for an explicit elliptic curve; hence the main new claims are not established.","major_comments":[{"comment":"The assertion that H_{τ_i}(-1) = ∏_{α∈J}(1+α)^2 for i = 1,...,5 is false for i = 1. From the definition H_{τ_1}(x) = ∏_{α∈J}(1-αx)(1-αx^3), one obtains H_{τ_1}(-1) = ∏_{α∈J}(1+α)(1+α^3), not A. Consequently P_2(-1) = ∏(1+α)(1+α^3) - ∏(1+α)^2, which is not identically zero; for J = {2} it equals 18. This invalidates the proof that P(x) ∈ Z[x], so the Lefschetz-type conclusion for Sp4 is not established.","section":"Section 6, proof of Theorem 6.5, part (b)"},{"comment":"The same type of error occurs in the Sp6 proof. One has H_{τ_1}(-1) = ∏(1+α)(1+α^3)(1+α^5), and H_{τ_2}(-1) = H_{τ_3}(-1) = ∏(1+α)^2(1+α^3), none of which equals A = ∏(1+α)^3 for general J. Writing B = ∏(1+α), C = ∏(1+α^3), D = ∏(1+α^5), the value P_3(-1) becomes (1/3)BCD - B^2C + (2/3)B^3, which is nonzero for example when J = {2}. Thus the claimed vanishing at x = -1 is not proved, and Theorem 6.6 is unsupported.","section":"Section 6, proof of Theorem 6.6, part (c)"},{"comment":"The error also affects the q-even parts of Theorems 6.5 and 6.6. For the elliptic curve E: y^2+y = x^3 over F_4, the Frobenius eigenvalues on H^1 are -2,-2; choosing S to be two F_4-rational points gives J = {-2,-2}. In the q-even part of Theorem 6.5 one computes Q_2(-1) = (B/2)(C-B) with B = 1 and C = 49, so Q_2(-1) = 24 ≠ 0. Since the proof identifies L(Sp4,m) with Q(q^m, J^m), a pole at x = -1 implies L(Sp4,m) cannot be of Lefschetz type as a function of m. This strongly indicates that the failure is not merely a proof gap but a genuine counterexample to the stated theorem if the identification is correct.","section":"Section 6, q-even cases and a concrete counterexample"}],"minor_comments":[{"comment":"The sentence 'Note that JX, JS, JT ⊂ Z' is false for JX: Frobenius eigenvalues on H^1(X) are Weil numbers and need not be rational integers. The subsequent argument treats J as indeterminates, so the incorrect claim is not used, but the sentence should be corrected.","section":"Section 3, Proposition 3.6"},{"comment":"The verification steps are mislabeled: the label '(c)' appears twice, and part (e) follows. The numbering should be corrected for readability.","section":"Section 6, proof of Theorem 6.6"},{"comment":"The evaluation at x = √-1 in part (d) is correct when the terms are grouped as τ1, τ2, τ7 for the first product and τ3, τ11 for the second; the alleged mis-evaluation in the stress-test note does not occur. The actual obstruction lies in the x = -1 evaluations discussed in the major comments.","section":"Section 6, proof of Theorem 6.6, part (d)"}],"recommendation":"reject","confidential_remarks":"I checked the stress-test concern about the √-1 evaluation and found it to be based on a misreading of the grouping; that specific point is not an error. However, a more serious issue appears in the x = -1 evaluations in Theorems 6.5 and 6.6, and the explicit elliptic curve y^2+y = x^3 over F_4 gives a concrete setting where the proof's own expression develops a pole, suggesting the theorems themselves may be false. This is why I recommend rejection rather than major revision."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Fast take: this is a careful, honest computation paper in Gross's program, and the specific stress-test worry about the Sp6 cancellation does not survive contact with the paper. The proofs are mostly long elementary manipulations, but they are real proofs.\n\nWhat is new: the paper proves that for G = SL_ℓ (ℓ prime), Sp4, Sp6, the base-change function m ↦ L(G,m) = Σ L_S(M_{Gγ}) is of Lefschetz type. That is new as far as I can tell; the framework comes from Gross and Yu, but the explicit centralizer-motive bookkeeping is the contribution. Theorem 5.5 with Proposition 5.8 (the polynomiality statement for SL_n when gcd(n, q^m−1) = 1) is a substantial algebraic result, and the S,T nonempty case is handled cleanly via Proposition 3.6.\n\nWhere it is soft: the title's cuspidal-representation claim is conditional on Gross's Conjectures 1.1 and 1.2, and the author says so plainly. If those conjectures fail, the unconditional Lefschetz-type results for the L-function sums still stand. The Sp6 proof is dense; I did not machine-check every table entry or congruence, and there are minor typos (e.g., the duplicated (c) label in Theorem 6.6's proof). But I did check the specific cancellation at x = √−1 flagged in the stress-test. The paper's grouping is correct: Hτ1, Hτ2, Hτ7 all evaluate to ∏(1−iα)(1+α²), while Hτ3 and Hτ11 evaluate to ∏(1+α)(1+α²); the coefficients for the first group sum to zero and the coefficients for the second group are i and −i. So P4(i) = 0 does hold. The stress-test simply misread which H's land in which product.\n\nBottom line: this deserves a serious referee. The claims are honestly delimited, the computations are substantial, and the conditional interpretation is clearly labeled. The citation pattern is standard for the area. I would send it to peer review, asking a referee to verify the Sp6 tables and the congruence checks in Theorem 5.5 carefully, but I do not see a load-bearing flaw.","headline":"Careful, honest computation in Gross's program; the Sp6 stress-test worry dissolves on reading, but the paper remains conditional on Gross's conjectures.","tokens_in":27574,"tokens_out":8178,"would_cite":true,"duration_ms":66578,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["11F70","11G40","11R58","14G10"],"pacs":[],"model":"deepseek-v4-flash","headline":"Assuming Gross's trace-formula conjectures, the paper proves that for $G=\\mathrm{SL}_\\ell$ with $\\ell$ prime and for $G=\\mathrm{Sp}_4$ or $\\mathrm{Sp}_6$, the sum over semisimple conjugacy classes of the L-functions…","keywords":["cuspidal representations","Lefschetz type functions","base change","function fields","trace formula","Artin-Tate motives","simple supercuspidal representations","Steinberg representation"],"falsifier":"For the unconditional theorems, a direct check would be to compute $L(G,m)$ for $G=\\mathrm{Sp}_8$ or $G=\\mathrm{SL}_4$ over a concrete curve with $\\#S\\geq 2$ for several small $m$, and see whether the sequence is a finite integer-linear combination of exponentials; a negative answer disproves Conjecture 4.6. For the conditional claim, one could compute both sides of Gross's trace formula in an explicit case, for example $G=\\mathrm{Sp}_4$ over $\\mathbb{F}_q(t)$ with $S=\\{0,\\infty\\}$; a mismatch would show that the multiplicity-sum conclusion does not follow from the conjectures.","tokens_in":26421,"feed_emoji":"🧮","tokens_out":10406,"duration_ms":94511,"temperature":0.7,"pith_summary":"The paper is trying to establish a rigid law for how counts of cuspidal representations of a split simply connected group over a function field change when the base field $\\mathbb{F}_q$ is enlarged to $\\mathbb{F}_{q^m}$. The law, called Lefschetz type, says the count is a finite $\\mathbb{Z}$-linear combination of exponentials $\\alpha_i^m$; this is exactly the shape of a Frobenius trace on a finite-dimensional object. The route goes through Gross's conjectural trace formula, which converts a sum of representation multiplicities into a sum of L-functions attached to centralizers of semisimple elements. The paper proves unconditionally that for $G=\\mathrm{SL}_\\ell$ with $\\ell$ prime and for $G=\\mathrm{Sp}_4,\\mathrm{Sp}_6$, the relevant L-function sums are of Lefschetz type, and that the transfer to cuspidal multiplicities follows if Gross's conjectures are accepted. If true, this places representation counting in the same framework as counting $\\ell$-adic local systems, where the same Lefschetz law was previously known.","feed_headline":"Cuspidal counting obeys a Lefschetz law for SL_ℓ, Sp4, and Sp6","feed_subtitle":"If Gross's trace-formula conjectures hold, sums of cuspidal multiplicities vary as finite exponential combinations under base change.","key_machinery":"The central object is the Artin–Tate motive $M_G=\\bigoplus_{d\\geq 1}V_d(1-d)$ attached to a reductive group, where $V_d$ is the degree-$d$ part of the $\\Gamma_q$-representation on $X^*(T_0)\\otimes\\mathbb{Q}$, together with its global L-functions $L_{S,T}(M,s)$; the $S$-modified value $L_S(M_{G_\\gamma})$ is what the trace formula produces for each semisimple class. The argument is carried by three devices: the Springer–Steinberg classification of semisimple conjugacy classes by characteristic polynomial, which reduces each centralizer $G_\\gamma$ to products of Weil restrictions of $\\mathrm{GL}_a$, unitary groups, and symplectic groups; explicit determinant formulas $\\det(1-t\\,\\mathrm{Fr}_q\\mid M_{G_\\gamma})$; and cyclotomic-polynomial identities plus Möbius inversion that turn the base-change sum into a polynomial in $q^m$ and the Frobenius eigenvalue set $J^m$.","core_discovery":"The central discovery is Theorem 5.4: for $G=\\mathrm{SL}_\\ell$ with $\\ell$ prime, the function $m\\mapsto L(G,m)=\\sum_{[\\gamma]}L_S(M_{G_\\gamma})$, summed over semisimple conjugacy classes in $G(\\mathbb{F}_{q^m})$, is of Lefschetz type; and Theorems 6.5 and 6.6: the same holds for $G=\\mathrm{Sp}_4$ and $G=\\mathrm{Sp}_6$. Here $M_{G_\\gamma}$ is the Artin–Tate motive of the centralizer and $L_S$ is its $S$-modified L-function at $s=0$. In the case where the finite set $S$ has at least two places and the set $T$ is empty, the paper's Theorem 4.2 (conditional on Gross's Conjectures 1.1 and 1.2) identifies $1+(-1)^{\\# S\\cdot r(G)}\\sum_\\pi m(\\pi)$ with exactly this sum, so these theorems transfer the Lefschetz property to sums of multiplicities of cuspidal representations that are Steinberg at $S$ and unramified elsewhere. Theorems 5.4 and 6.5–6.6 are unconditional statements about L-functions; the representation-theoretic conclusion carries the conjectural trace formula as a hypothesis.","pith_inferences":["If the Lefschetz law holds for all split simply connected $G$, then the counting function $m\\mapsto\\sum m(\\pi)$ is morally the trace of $\\mathrm{Fr}_{q^m}$ on a virtual motive; this suggests a geometric construction of a motive or sheaf whose cohomology counts cuspidal representations, in the spirit of counting local systems.","Because Theorems 5.4, 6.5, and 6.6 are unconditional, they constrain any future trace formula: whatever the correct geometric side is, its sum over semisimple classes must reproduce the explicit Lefschetz identities proved here, so those identities can be used as test cases for proposed spectral expansions.","The methods suggest a concrete computational check of Conjecture 4.6: evaluate the rational function $P(x,J)$ constructed in Section 6 for $\\mathrm{Sp}_8$ or $\\mathrm{SL}_4$; if its denominator cannot be cancelled for all $q$, the conjecture fails for that group.","The sharp distinction between the $S,T\\neq\\emptyset$ case and the $T=\\emptyset$ case suggests that the hardest part of extending this work is not the Lefschetz law itself but the classification and counting of semisimple conjugacy classes that contribute to $L_S(M_{G_\\gamma})$."],"forward_implications":["For $G=\\mathrm{SL}_\\ell$ with $\\ell$ prime and for $G=\\mathrm{Sp}_4,\\mathrm{Sp}_6$, the function $m\\mapsto L(G,m)$ is a finite integer-linear combination of exponentials, so its values for all $m$ are determined by finitely many initial values.","If Gross's Conjectures 1.1 and 1.2 hold, the same Lefschetz law transfers to sums of multiplicities of cuspidal representations that are Steinberg at $S$ and unramified elsewhere, in the $T=\\emptyset$, $\\#S\\geq 2$ case.","When $S$ and $T$ are both nonempty, Proposition 4.4 gives the Lefschetz law conditionally for every split simply connected almost simple group, not only the groups in Theorems 5.4 and 6.5–6.6.","The explicit polynomial expressions for $L(G,m)$ mean the base-change counts can be computed in finite terms from $q$, the Frobenius eigenvalues on $H^1(X)$, and the places in $S$.","The truth of Conjecture 4.6 for all split simply connected $G$ is reduced to proving that the associated rational function $P(x,J)$ lies in $\\mathbb{Z}[x,J]$.","Editorial extension: if the Lefschetz law holds for all $G$, then the counting function $m\\mapsto\\sum m(\\pi)$ is morally the trace of $\\mathrm{Fr}_{q^m}$ on a virtual motive; this suggests that cuspidal counts should admit a geometric, motive-like realisation rather than only a trace-formula computation.","Editorial extension: because Theorems 5.4, 6.5, and 6.6 are unconditional, the polynomial identities they prove constrain any future spectral expansion: the geometric side of the trace formula must reproduce these explicit Lefschetz identities, so they can serve as test cases for proposed trace-formula refinements.","Editorial extension: a direct computational check of Conjecture 4.6 for the next open cases, such as $G=\\mathrm{SL}_4$ or $G=\\mathrm{Sp}_8$, would either extend the pattern or find the first counterexample; the paper's tables and polynomial recipes make such a check concrete."],"supporting_citations":[{"why":"Supplies the conjectural trace formula and the identity (Theorem 4.2) expressing sums of cuspidal multiplicities as finite sums of orbital integrals and L-functions.","marker":"[Gro11]"},{"why":"Defines the Artin–Tate motive $M_G$ and its L-functions, the objects whose sums are shown to be of Lefschetz type.","marker":"[Gro97]"},{"why":"Gives the conjugacy criteria for semisimple elements in simply connected groups, used to classify centralizers by characteristic polynomial.","marker":"[SS70]"},{"why":"Provides the count of self-reciprocal irreducible polynomials over $\\mathbb{F}_q$, needed for the numbers $N_\\tau(q)$ in the $\\mathrm{Sp}_{2n}$ computation.","marker":"[Car67]"},{"why":"Constructs the affine generic characters and simple supercuspidal representations whose local conditions define the set $\\mathrm{Irr}^{S,T}_0(G,\\chi)$ and whose count enters Remark 4.3.","marker":"[GR10]"}],"fun_headline_variants":["Lefschetz law for cuspidal multiplicities under base change","Cuspidal sums obey finite exponential combinations in base change","SL_ℓ, Sp4, Sp6: cuspidal counts follow Lefschetz type","Base change of cuspidal multiplicities is Lefschetz for SL_ℓ, Sp4, Sp6","Cuspidal representation sums: a Lefschetz property under base change"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The whole transfer from L-function sums to representation counts depends on an unproved trace-formula conjecture of Gross, which says that a certain global test function can be decomposed into finitely many orbital integrals; if that fails, the representation-counting conclusion fails, though the L-function theorems survive.","fun_headline_variants_meta":{"raw":{"variants":["Lefschetz law for cuspidal multiplicities under base change","Cuspidal sums obey finite exponential combinations in base change","SL_ℓ, Sp4, Sp6: cuspidal counts follow Lefschetz type","Base change of cuspidal multiplicities is Lefschetz for SL_ℓ, Sp4, Sp6","Cuspidal representation sums: a Lefschetz property under base change"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000228,"raw_usage":{"total_tokens":1481,"prompt_tokens":957,"completion_tokens":524,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":573,"completion_tokens_details":{"reasoning_tokens":413}},"tokens_in":573,"tokens_out":524,"duration_ms":4627,"temperature":1.0,"reasoning_tokens":413,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-16T05:27:27.458325+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"For the unconditional theorems, a direct check would be to compute $L(G,m)$ for $G=\\mathrm{Sp}_8$ or $G=\\mathrm{SL}_4$ over a concrete curve with $\\#S\\geq 2$ for several small $m$, and see whether the sequence is a finite integer-linear combination of exponentials; a negative answer disproves Conjecture 4.6. For the conditional claim, one could compute both sides of Gross's trace formula in an explicit case, for example $G=\\mathrm{Sp}_4$ over $\\mathbb{F}_q(t)$ with $S=\\{0,\\infty\\}$; a mismatch would show that the multiplicity-sum conclusion does not follow from the conjectures.","supporting_citations":[],"review_version":1}