§1 · Agent 1 · Literature and Table 1 audit · 2026-10-17T03:01:34Z

Proposition 1.4 · \(\GL_4/\Sp_4\): (SG) holds, and the uncovered label is reached by a cancellation

proved audit: error

Error. Conclusion correct, and every computed value confirmed: by hand for all q (c′ = 1, c = q(q+1), Lagrangian count (q+1)(q²+1), zero for (2,0,0,0) at ϖJ) and by an independent count of all lattices t³O⁴ ⊆ N ⊆ O⁴ over 𝔽₂[[t]] (gl4sp4_G1.py: 15 and 0 at ϖJ, c′ = 1 and c = 6 at ϖ²J₁₂ + J₃₄). Two false statements in step 2: (1) 'the labels ≥ µ with Pfaffian valuation 2 are only (2,2,0,0) and µ' — they are (1+k,1+k,1−k,1−k), k ≥ 0; (2) 'the translates reaching O_µ are T_λ with λ ∈ {(2,0,0,0),(1,1,0,0)}' — the dominant λ with T_λφ₀ ≠ 0 on O_µ are exactly (1+a,1+b,−b,−a), a ≥ b ≥ 0 (for q = 2, λ = (2,1,0,−1) and (2,2,−1,−1) give 360 and 960 at ϖJ), and (2,0,0,0) is not among them (its value there is 0, as the node itself computes). The proof needs only the true statement that the lower set of (2,2,0,0) in this component is {(2,2,0,0), µ}, so the formula for 1_{O_µ} and (SG) stand. Also, step 1 attributes the covered labels to Corollary C, which does not apply when uncovered labels lie below them; step 3 obtains everything correctly from Proposition B. The conclusion still holds by another argument.
Corrected by: Correction 10.3 · \(\GL_4/\Sp_4\): which Hecke translates reach the label \((1,1,1,1)\), and the corrected cancellation, Correction X1.1 · GL₄/Sp₄: every T_λ nonzero on the orbit (1,1,1,1), exact values, step 2 corrected.

For alternating forms in four variables, every label except \((a,a,a,a)\), \(a\) odd, has a dominant lift; the remaining orbit indicators are obtained by cancelling two Hecke translates on the top orbit. Hence (SG) holds over every non-archimedean field without Sakellaridis's hypotheses — the first case where H5H5 · Nonzero top splitting: for every orbit O_μ some Satake object V_λ ∗ IC_0 is supported on its closure with a nonzero split restriction to O_μ. fails and the function level still holds.

Proposition 1.4.

Let \(X=\GL_4/\Sp_4\) be the space of nondegenerate alternating forms on \(F^4\), with \(g\cdot J'=gJ'g^{\mathsf T}\) and \(X(\cO)=Kx_0\) the unimodular forms, over any non-archimedean local field \(F\). The \(K\)-orbits are the elementary-divisor classes \(\mathcal O_{(a_1,a_1,a_2,a_2)}\), \(a_1\ge a_2\); \(\varphi(\lambda)=(\lambda_1+\lambda_2,\lambda_1+\lambda_2,\lambda_3+\lambda_4,\lambda_3+\lambda_4)\); and the labels without a dominant lift are \((a,a,a,a)\), \(a\) odd. With \(\mu=(1,1,1,1)\) there are positive integers \(c,c'\) such that $$T_{(2,0,0,0)}\ast\varphi_0=c\,\mathbf 1_{\mathcal O_{(2,2,0,0)}},\qquad T_{(1,1,0,0)}\ast\varphi_0=c'\,\mathbf 1_{\mathcal O_{(2,2,0,0)}}+(q+1)(q^2+1)\,\mathbf 1_{\mathcal O_\mu},$$ $$\mathbf 1_{\mathcal O_\mu}=\frac{1}{(q+1)(q^2+1)}\Big(T_{(1,1,0,0)}-\frac{c'}{c}\,T_{(2,0,0,0)}\Big)\ast\varphi_0\in\mathcal H_K\cdot\varphi_0,$$ and (SG) holds for \(\GL_4/\Sp_4\) over every non-archimedean local field, with no hypothesis (A)“Under (A)”: assuming all of Sakellaridis's assumptions (A1)–(A5) for X at the given p-adic place, so that his Theorem 8.0.2 applies..

  1. Support bound, every \(F\). For \(k\in K\), \(\tau(t^\lambda kx_0)=t^\lambda\,(k\theta(k)^{-1})\,t^{-\theta\lambda}\in Kt^\lambda K\,t^{-\theta\lambda}K\), so the label \(\nu\) of \(t^\lambda kx_0\) satisfies \(\nu\le\lambda+(-\theta\lambda)^+\), and \((-\theta\lambda)^+-(-\theta\lambda)\) is a sum of positive coroots of \(L=\GL_2\times\GL_2\). Applying \(p=\tfrac12(1-\theta)\), which kills the coroots of \(L\), gives \(\varphi(\lambda)-\nu\in\Z\check\Phi\cap\mathbb Q_{\ge0}\check\Phi^+=\Z_{\ge0}\check\Phi^+\) (the argument of Proposition III, step 2). Hence \(\supp(T_\lambda\ast\varphi_0)=Kt^\lambda X(\cO)\subseteq\bigsqcup_{\nu\le\varphi(\lambda)}\mathcal O_\nu\), and Corollary C gives \(\mathbf 1_{\mathcal O_\mu}\in\mathcal H_K\varphi_0\) for every covered \(\mu\), with the lift \(\lambda=(\lceil a_1/2\rceil,\lfloor a_1/2\rfloor,\lceil a_2/2\rceil,\lfloor a_2/2\rfloor)\).
  1. The uncovered label \(\mu=(1,1,1,1)\). The labels \(\ge\mu\) with the same Pfaffian valuation \(2\) are only \((2,2,0,0)\) and \(\mu\), and \(\mathcal O_\mu\) is closed (minimal in its component). The translates reaching \(\mathcal O_\mu\) are \(T_\lambda\) with \(\lambda\in\{(2,0,0,0),(1,1,0,0)\}\), both with \(\varphi(\lambda)=(2,2,0,0)\). By Lemma A, \((T_\lambda\ast\varphi_0)(J')\) is the number of lattices \(M\supseteq\cO^4\) with \(\operatorname{inv}(\cO^4,M)=-\lambda\) on which \(J'\) is unimodular (\(M=g^{-\mathsf T}\cO^4\)). At \(J'=\varpi J\in\mathcal O_\mu\), unimodularity of \(\varpi J\) on \(M\) means \(M^{\vee_J}=\varpi M\), where \(M^{\vee_J}\) is the dual lattice for \(J\).
    • \(\lambda=(1,1,0,0)\): \(M=\cO^4+\varpi^{-1}\widetilde W\) for a plane \(W\subset\F_q^4\); then \(M^{\vee_J}=\{u\in\cO^4:\bar u\in W^\perp\}\) and \(\varpi M=\{u\in\cO^4:\bar u\in W\}\), equal exactly when \(W\) is Lagrangian. The count is \((q+1)(q^2+1)\) (\(40\) for \(q=3\), checked by enumeration).
    • \(\lambda=(2,0,0,0)\): \(M=\cO^4+\cO\varpi^{-2}v\) with \(v\) primitive; \(M^{\vee_J}\subseteq\cO^4\), while \(\varpi^{-1}v\in\varpi M\setminus\cO^4\). The count is \(0\).
    With the support bound, \(T_{(2,0,0,0)}\ast\varphi_0=c\,\mathbf 1_{\mathcal O_{(2,2,0,0)}}\) with \(c\ge1\) (Lemma A), and \(T_{(1,1,0,0)}\ast\varphi_0=c'\mathbf 1_{\mathcal O_{(2,2,0,0)}}+(q+1)(q^2+1)\mathbf 1_{\mathcal O_\mu}\); this gives the displayed formula for \(\mathbf 1_{\mathcal O_\mu}\).
  1. The other uncovered labels. Every \((a,a,a,a)\) with \(a=2k+1\) is a central translate: \(\mathbf 1_{\varpi^kK}\ast\mathbf 1_{\mathcal O_\mu}=\mathbf 1_{\mathcal O_{\mu+(2k,2k,2k,2k)}}\), because \(\varpi^kI\cdot J'=\varpi^{2k}J'\). Lower sets of the dominance order are finite inside each Pfaffian component, so steps 1–3 and Proposition B give (SG) for \(\GL_4/\Sp_4\) over every non-archimedean local field.

Consequences. (i) The cell "\(\GL_{2m}/\Sp_{2m}\): holds under (A)“Under (A)”: assuming all of Sakellaridis's assumptions (A1)–(A5) for X at the given p-adic place, so that his Theorem 8.0.2 applies." is unconditional for \(m=2\), in every characteristic. (ii) At the function level an orbit without a dominant lift is reached by cancelling two Hecke translates on the top orbit, which Corollary C cannot see. (iii) For \(\GL_{2m}/\Sp_{2m}\), \(m\ge3\), and \(\SL_{2n}/\Sp_{2n}\), \(n\ge3\), the same finite counting problem decides (SG) label by label (several \(\lambda\) with \(\varphi(\lambda)\ge\mu\) in the component); it is not carried out here.

Verified by computation: checks.py.

Builds on

  • Setup 1.1 · Table 1 of the Summary, re-derived row by row

Built on by

  • Summary 1.9 · Classification after the status audit, and the cells where a divergence can still hide
  • Audit 7.6 · \(\GL_4/\Sp_4\) lattice counts recomputed for \(q=2,3\): the values of four agents agree
  • Correction 10.3 · \(\GL_4/\Sp_4\): which Hecke translates reach the label \((1,1,1,1)\), and the corrected cancellation
  • Correction X1.1 · GL₄/Sp₄: every T_λ nonzero on the orbit (1,1,1,1), exact values, step 2 corrected
Original record, as the agent wrote it at 03:01 UTC

GL4/Sp4: (SG) holds unconditionally; the uncovered label is reached by a cancellation

Setting (families/symplectic.html, classical.html Prop F): \(X=\mathrm{GL}_4/\mathrm{Sp}_4\) = nondegenerate alternating forms, action \(g\cdot J'=gJ'g^{\mathsf T}\), \(X(\mathcal O)=\) unimodular forms \(=Kx_0\); \(K\)-orbits \(\mathcal O_{(a_1,a_1,a_2,a_2)}\) = elementary-divisor classes, \(a_1\ge a_2\); \(\varphi(\lambda)=(\lambda_1+\lambda_2,\lambda_1+\lambda_2,\lambda_3+\lambda_4,\lambda_3+\lambda_4)\). Uncovered labels: \((a,a,a,a)\), \(a\) odd (Lemma §5 of the family page). Any non-archimedean \(F\), any \(q\).

  1. Function-level support bound (same proof as criterion.html step 2, transcribed): for \(k\in K\), \(\tau(t^\lambda kx_0)=t^\lambda(k\theta(k)^{-1})t^{-\theta\lambda}\in Kt^\lambda Kt^{-\theta\lambda}K\), so the label \(\nu\) of \(t^\lambda kx_0\) satisfies \(\nu\le\lambda+(-\theta\lambda)^+\); \((-\theta\lambda)^+-(-\theta\lambda)\in\mathbb Z\check\Phi_L\); apply \(p=\tfrac12(1-\theta)\): \(\varphi(\lambda)-\nu\in\mathbb Z\check\Phi\cap\mathbb Q_{\ge0}\check\Phi^+=\mathbb Z_{\ge0}\check\Phi^+\). Hence \(\operatorname{supp}(T_\lambda\ast\varphi_0)=Kt^\lambda X(\mathcal O)\subseteq\bigsqcup_{\nu\le\varphi(\lambda)}\mathcal O_\nu\), and Corollary C gives \(\mathbf 1_{\mathcal O_\mu}\in\mathcal H_K\cdot\varphi_0\) for every covered \(\mu\) (lift \(\lambda=(\lceil a_1/2\rceil,\lfloor a_1/2\rfloor,\lceil a_2/2\rceil,\lfloor a_2/2\rfloor)\)).
  2. The uncovered label \(\mu=(1,1,1,1)\). Labels \(\ge\mu\) with the same Pfaffian valuation \(2\): only \((2,2,0,0)\) and \(\mu\); \(\mathcal O_\mu\) is closed (minimal in its component). Translates reaching \(\mathcal O_\mu\): \(\lambda\in\{(2,0,0,0),(1,1,0,0)\}\) (both with \(\varphi(\lambda)=(2,2,0,0)\)). Value formula: \((T_\lambda\ast\varphi_0)(J')=\#\{M\supseteq\mathcal O^4:\ \operatorname{inv}(\mathcal O^4,M)=-\lambda,\ J'\text{ unimodular on }M\}\) (\(M=g^{-\mathsf T}\mathcal O^4\)). At \(J'=\varpi J\) (\(\in\mathcal O_\mu\)): unimodularity of \(\varpi J\) on \(M\) is \(M^{\vee_J}=\varpi M\).

• \(\lambda=(1,1,0,0)\): \(M=\mathcal O^4+\varpi^{-1}\widetilde W\), \(W\subset\mathbb F_q^4\) a 2-plane; \(M^{\vee_J}=\{u\in\mathcal O^4:\bar u\in W^{\perp}\}\), \(\varpi M=\{u\in\mathcal O^4:\bar u\in W\}\); equality iff \(W\) Lagrangian. Count \((q+1)(q^2+1)\ne0\) (checked by enumeration for \(q=3\): 40). • \(\lambda=(2,0,0,0)\): \(M=\mathcal O^4+\mathcal O\varpi^{-2}v\), \(v\) primitive; \(M^{\vee_J}\subseteq\mathcal O^4\) but \(\varpi M\ni\varpi^{-1}v\notin\mathcal O^4\); count \(0\). So \(T_{(2,0,0,0)}\ast\varphi_0=c\,\mathbf 1_{\mathcal O_{(2,2,0,0)}}\), \(c\ge1\) (Lemma A), and \(T_{(1,1,0,0)}\ast\varphi_0=c'\mathbf 1_{\mathcal O_{(2,2,0,0)}}+(q+1)(q^2+1)\mathbf 1_{\mathcal O_\mu}\). Therefore $$\mathbf 1_{\mathcal O_{(1,1,1,1)}}=\frac{1}{(q+1)(q^2+1)}\Big(T_{(1,1,0,0)}-\frac{c'}{c}T_{(2,0,0,0)}\Big)\ast\varphi_0\in\mathcal H_K\cdot\varphi_0 .$$

  1. The other uncovered labels \((a,a,a,a)\), \(a\) odd, are central translates: \(\mathbf 1_{\varpi^kK}\ast\mathbf 1_{\mathcal O_\mu}=\mathbf 1_{\mathcal O_{\mu+(2k,2k,2k,2k)}}\) (\(\varpi^kI\cdot J'=\varpi^{2k}J'\)). With step 1 and Prop B (lower sets finite), (SG) holds for \(\mathrm{GL}_4/\mathrm{Sp}_4\) over every non-archimedean local field, with no hypothesis (A). ∎

Consequences. (i) The Table 1 cell "\(\mathrm{GL}_{2m}/\mathrm{Sp}_{2m}\): holds under (A)" is unconditional for \(m=2\), in every characteristic. (ii) The function-level mechanism for an orbit without a dominant lift is a cancellation of two Hecke translates on the top orbit — exactly what Corollary C cannot see. (iii) For \(m\ge3\) and \(\mathrm{SL}_{2n}/\mathrm{Sp}_{2n}\), \(n\ge3\), the same finite counting problem decides (SG) label by label (several \(\lambda\) with \(\varphi(\lambda)\ge\mu\) in the component); not carried out here (gap).

Sources