{"id":"2d204253-30ea-4716-aadc-819d4d715eaf","arxiv_id":"2501.12127","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"For simple Shimura varieties whose local group is an arbitrary inner form of a product of Weil-restricted general linear groups, the Scholze test functions satisfy the expected vanishing and base-change properties, yielding the semisimple cohomology description.","lead":"This paper proves that certain local test functions for bad-reduction Shimura varieties with non-quasi-split local groups satisfy a vanishing property and match explicit elements of the stable Bernstein center. As a consequence it gives the Langlands-Kottwitz description of the semisimple cohomology and local Hasse-Weil zeta functions for a broad class of simple Shimura varieties.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Corollary 4.4.2's fixed-point count may overcount: the naive parahoric model need not be flat, and the partition of Fix_{j,L}(g_p) by Isog(A,u,lambda) excludes points outside the closure of the generic fiber without proving their contribution to Rpsi pi_* F_xi vanishes.","rationale":"The reader's weakest assumption and the most load-bearing concern coincide: the paper explicitly notes in footnote 3 that the naive integral model need not be flat, yet the proof of Corollary 4.4.2 uses a partition of the fixed-point set by Isog(A,u,lambda), which only includes points in the closure of the generic fibre. For a non-flat model, special-fiber components not meeting the generic fibre can contain fixed points whose nearby-cycle stalks vanish but whose associated local test-function value phi_{tau,h}(delta) is not visibly zero. If such points exist with nonzero contribution, the expression for Tr(tau x h f^p | H^*_xi) that feeds the pseudo-stabilization argument is incorrect. This is an internal gap in the trace-formula step, not merely a disagreement with the literature, and it is directly checkable. Other potential concerns, such as the reliance on [Shi12, Prop. 3.1] or on deep stable base-change results, are external and standard in this subject; the non-flatness issue is where the paper's own argument is most exposed. The verdict remains CONDITIONAL: the overall strategy is coherent and the gap is addressable by a careful closure-restriction and a vanishing argument for non-liftable fixed points, but the current text does not supply that argument.","tokens_in":59055,"tokens_out":46373,"duration_ms":515686,"concrete_test":"Choose a concrete EL datum for which Pappas-Rapoport prove the naive parahoric model is non-flat, such as a ramified parahoric for an inner form of a Weil restriction of GL_n, and enumerate the F_p-fixed points of the Frobenius-Hecke correspondence. For each fixed point outside the closure of the generic fibre, compute phi_{tau,h}(delta) from the local deformation space. If all such values are zero, the concern is resolved and Corollary 4.4.2's implicit restriction to the closure is harmless; if any is nonzero, the displayed sum must be corrected by summing only over Isog(A,u,lambda) with nonempty intersection with the closure.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Corollary 4.4.2 is the exact bridge from the Grothendieck-Lefschetz trace formula to the sum over isogeny classes feeding pseudo-stabilization. Its proof partitions Fix_{j,L}(g_p) as the disjoint union of Isog(A,u,lambda), but Isog was defined earlier as the subset of Fix consisting of points lying in the closure of the generic fibre; footnote 3 concedes that the naive parahoric model M_{K_L K^p} need not be flat (Pappas-Rapoport). A non-flat scheme has special-fiber components with empty generic fibre; fixed points on such components are missing from the partition. Their local terms involve Rpsi pi_* F_xi, and the nearby-cycle sheaf is supported on the closure, so such terms should vanish; however the displayed formula in Corollary 4.4.2 assigns to every isogeny class the value vol(I(Q)\\I(A_f)) O_gamma(f^p) TO_{delta sigma}(phi_{tau,h}) tr_xi(gamma_ell). The test function phi_{tau,h}(delta) is nonzero for many delta because it is computed from the formal deformation space of the p-divisible group, whose generic fibre is nonempty even if the global point does not lift. Thus, unless the sum is explicitly restricted to isogeny classes whose fixed points lie in the closure, or unless a proof is supplied that non-closure points have zero phi-values, the trace formula (4.6) and Theorem 5.5.1 inherit a real gap. This is the most load-bearing step because it converts geometry to the test functions on which Theorems 3.4.3 and 3.4.4 rest.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper studies the Scholze test functions attached to the bad reduction of Kottwitz-type simple Shimura varieties at a prime p where the local group is an inner form of a product of Weil restrictions of general linear groups, hence not necessarily quasi-split. The main local theorems are the vanishing property of twisted orbital integrals of the test functions (Thm 3.4.3) and the assertion that z_{τ,-μ} * h is a base-change transfer of φ_{τ,h} (Thm 3.4.4). From these the authors deduce a semisimple description of the cohomology (Thm 4.3.1), explicit semisimple local Hasse-Weil zeta factors (Cor 4.3.3), and the Haines-Kottwitz test-function conjecture in this setting (Cor 4.6.2). The proof is global: the authors introduce companion unitary similitude groups G_β and G', compute the cohomology of the G_β-Shimura variety via Galois representations and Chebotarev, and compare the resulting simple trace formulas with the Langlands-Kottwitz-Scholze fixed-point count. The paper also contains a new twisted local Jacquet-Langlands theorem (Thm 2.4.1) and explicitly repairs a gap in [She18, Lemma 5.3].","tokens_in":59429,"tokens_out":12261,"duration_ms":128170,"significance":"If correct, this is a substantial contribution: it removes the quasi-splitness assumption for a large class of groups, establishes the expected base-change transfer to the stable Bernstein center, and verifies a special case of the Haines-Kottwitz test-function conjecture. The global method is a genuine methodological advance and appears to avoid circularity: the cohomology of the companion Shimura variety is computed from Galois representations and Chebotarev rather than from the local test functions. The paper is unusually explicit about a gap in earlier work and provides a new local theorem of independent interest. The main correctness risk identified below concerns the fixed-point count on the non-flat naive integral model; this is a localizable and likely repairable issue rather than a fundamental flaw.","major_comments":[{"comment":"The statement of Corollary 4.4.2 uses the equality Fix_{j,L}(g_p) = ⨆_{(A,u,λ)} Isog(A,u,λ). However, Isog(A,u,λ) was defined just above as the subset of Fix_{j,L}(g_p) consisting of points lying in the closure of the generic fibre, and footnote 3 explicitly concedes that the naive integral model M_{K_L K^p} need not be flat. For a non-flat model there can be special-fibre fixed points that are not in the closure of the generic fibre, and these points are contained in none of the sets Isog(A,u,λ). Their local terms vanish because Rψπ_*F_ξ is supported on the closure of the generic fibre, but the proof of Corollary 4.4.2 does not say this; instead it breaks the sum according to the displayed partition. Since this partition feeds directly into Eq. (4.6) and Theorem 5.5.1, the authors must either replace Fix_{j,L}(g_p) by Fix_{j,L}(g_p) ∩ closure in the trace formula and justify omission of the remaining points, or prove that every point with nonzero φ_{τ,h}(g^{-1}δσ(g)) lies in the closure. As written, the displayed formula can overcount.","section":"§4.4.3, Corollary 4.4.2"},{"comment":"Lemma 5.4.3 is the key replacement of Kottwitz triples by generalized Kottwitz triples, but its proof is largely a reference to [Kot92b, pp. 420-422] with 'slight modifications'. In particular, the construction of the global algebra embedding N → C' compatible with the Rosati involution is not written out; the reader needs to see how the use of quasi-splitness of G(Q_p) in Kottwitz's argument is replaced by the splitness of B' at the places above p. This is load-bearing for Theorem 5.5.1 and should be expanded.","section":"§5.4.2, Lemma 5.4.3"},{"comment":"The proof of Theorem 3.4.4 passes from the statement that f^*_{τ,h} is a Jacquet-Langlands transfer of z_{τ,-μ}*h to the conclusion that z_{τ,μ}*h (sic) is a base-change transfer of φ_{τ,h}. The sign of the character in the final display is written as μ rather than -μ. This is likely a typo, but because the whole point of Theorem 3.4.4 is the precise element z_{τ,-μ}, the sign should be checked carefully throughout the last paragraph and corrected consistently.","section":"§5.5, final step"}],"minor_comments":[{"comment":"In the paragraph after Proposition 4.5.2, 'combined with the vanishing property Theorem 3.4.4' should cite Theorem 3.4.3 (or both Theorems 3.4.3 and 3.4.4), since Theorem 3.4.4 is the matching statement, not the vanishing statement.","section":"§4.5.1"},{"comment":"The displayed formula has an unbalanced parenthesis in H^*(Sh_{K_pK^p} ⊗_E \\bar{Q}, F_{K_pK^p}) and the subscript notation on the local system is inconsistent with the notation used elsewhere.","section":"Corollary 4.6.2"},{"comment":"There are many extraction artifacts such as 'W e ', 'V arieties', and 'A f' in the abstract and introduction; these should be cleaned before the final version.","section":"Abstract and front matter"},{"comment":"The notation tr(τ|(r_{-μ}∘φ|W_E)|·|^{-⟨ρ,μ⟩}_E) is overloaded and hard to parse; the role of the '|·|' factor and the choice of √p mentioned in Remark 3.4.2 should be made explicit in the displayed formula.","section":"§3.4, Definition 3.4.1"},{"comment":"In the proof of Lemma 5.1.1, the notation 'H^1(F_0, G^{ad}_{β_0})' is used for both the global cohomology set and local cohomology sets; a short sentence fixing the usual restrictions would eliminate ambiguity.","section":"§5.1.3, Lemma 5.1.1"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is clearly within scope for a serious number theory journal, and the overall strategy is coherent and promising. My recommendation of major revision is driven by the non-flat model gap in §4.4.3 and the terseness of Lemma 5.4.3; both are localizable and likely repairable. The editor may also wish to ask the authors to state precisely which facts from the forthcoming work [HZZ] are used in Corollary 4.4.2, since that is a delicate point in the proof."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"You should take a look at this paper. It proves a real extension of the Scholze-Shin results to non-quasi-split inner forms of Weil-restricted general linear groups, which is a genuinely new step. The main theorems — the vanishing property of twisted orbital integrals for Scholze test functions and the base change transfer to the stable Bernstein center — are new for these non-quasi-split groups, and the authors get a nice application: a semisimple local Hasse-Weil zeta function for simple Shimura varieties of Kottwitz type at primes where previous arguments stopped. The global method is clever: they construct a companion group G_beta and an auxiliary group G' that is quasi-split at p, compute the cohomology of the companion Shimura variety via Galois representations and Chebotarev, and then use the trace formula to transfer information back. The circularity burden is low; they do not use the Langlands-Kottwitz-Scholze method to prove the local claims.\n\nThe paper is honest about a gap in Shen's lemma and fixes it. The proof of the twisted local Jacquet-Langlands via global methods is substantial and seems to be a genuine new result. Reliance on a private communication from Waldspurger and unpublished ideas of Kottwitz is a minor auditability issue, but that is common in this area.\n\nThe soft spot, and it is a real one, is in the fixed-point counting. The stress-test note is right: Corollary 4.4.2 asserts the partition Fix_{j,L}(g_p) = Isog(A,u,lambda), but Isog was defined as the subset of fixed points lying in the closure of the generic fiber. The integral model is not flat — the authors know this, they put it in a footnote — and there can be special-fiber components with no generic point. For those points the nearby-cycle stalk should vanish, but the authors do not provide the argument that the contribution is zero, and the test function phi_{tau,h} is nonzero on many delta that come from such non-liftable points. So the displayed trace formula may overcount. This is load-bearing: everything downstream, including Theorem 5.5.1 and the deduction of the local results, uses this formula. This is not a fatal flaw in the strategy; it looks like a missing argument that a referee should ask for. But it needs to be fixed, not just asserted.\n\nOverall: this is a serious paper with substantial new content, and it deserves a careful referee. I would not cite it myself in the next year, but I would send it out. The right verdict is conditional: the main structure holds, but the fixed-point partition needs a rigorous justification.","headline":"Genuinely new extension of Scholze-Shin to non-quasi-split inner forms, but the fixed-point counting in Corollary 4.4.2 has an unproven partition that needs referee attention.","tokens_in":60018,"tokens_out":8986,"would_cite":false,"duration_ms":90815,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["11G18","11F70","11F80","22E50"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper proves that the local test function for bad reduction of simple Shimura varieties matches an explicit stable-Bernstein-center element, yielding the semisimple cohomology and Hasse-Weil factors.","keywords":["Shimura varieties","bad reduction","test functions","stable Bernstein center","twisted orbital integrals","base change","cohomology","non-quasi-split groups"],"falsifier":"Take an explicit non-quasi-split local group such as the unit group of a central division algebra over $\\mathbb{Q}_p$, choose a $\\sigma$-semisimple $\\delta$ in $G(\\mathbb{Q}_{p^r})$ whose naive norm is not conjugate into $G(\\mathbb{Q}_p)$, and compute the twisted orbital integral $TO_{\\delta\\sigma}(\\varphi_{\\tau,h})$ directly from the deformation-space definition; a single nonzero value would contradict the vanishing theorem.","tokens_in":58806,"feed_emoji":"🔢","tokens_out":12893,"duration_ms":119955,"temperature":0.7,"pith_summary":"The paper's aim is to describe the p-adic cohomology of simple Shimura varieties at a prime of bad reduction where the local group is an arbitrary inner form of a product of Weil restrictions of general linear groups, and hence need not be quasi-split. It proves that the local test function $\\varphi_{\\tau,h}$ entering the Lefschetz fixed-point count can be moved to the spectral side: its twisted orbital integrals match those of $z_{\\tau,-\\mu}*h$, an explicit element of the stable Bernstein center. It also proves a vanishing property: twisted orbital integrals of $\\varphi_{\\tau,h}$ are zero unless the naive norm of the relevant element lies in the local group up to conjugacy. Together these results give a formula for the semisimple cohomology and for the local Hasse-Weil zeta function in terms of Langlands parameters, extending earlier results from the quasi-split case.","feed_headline":"Bad-reduction test functions matched to stable center","feed_subtitle":"The geometric point count is pinned to an explicit spectral distribution, giving cohomology and zeta factors.","key_machinery":"The engine is the pair consisting of the local test function $\\varphi_{\\tau,h}$ and the stable Bernstein center. The function $\\varphi_{\\tau,h}$ on $G(\\mathbb{Q}_{p^r})$ is defined by the trace of $\\tau\\times h$ on the \\'etale cohomology of the deformation space of a $p$-divisible group with EL structure (an action of a maximal order of a semisimple algebra together with lattice-chain level data); it packages the geometric contribution of each fixed point. The element $z_{\\tau,-\\mu}$ of the stable Bernstein center acts on an irreducible smooth representation by the trace of $\\tau$ on the representation $r_{-\\mu}$ composed with the semisimple $L$-parameter. The proof that the two match runs through a twisted local Jacquet-Langlands correspondence and a base-change theorem for the stable Bernstein center, together with a global construction of two companion groups whose simple trace formulas are compared. The matching is what converts a geometric test function into a spectral quantity.","core_discovery":"On the paper's own terms, the central discovery is an orbital-integral identity. For any cut-off function $h$ on the parahoric subgroup $G_L(\\mathbb{Z}_p)$, the function $z_{\\tau,-\\mu}*h$ on the quasi-split inner form $G^*(\\mathbb{Q}_p)$ is a base change transfer of the local test function $\\varphi_{\\tau,h}$ on $G(\\mathbb{Q}_{p^r})$, and $z_{-\\mu}*h$ is the corresponding transfer of the semisimple variant $\\varphi^{(r)}_h$. The identity implies the vanishing property: $TO_{\\delta\\sigma}(\\varphi_{\\tau,h}) = 0$ whenever the naive norm of $\\delta$ is not conjugate into $G(\\mathbb{Q}_p)$, so only isogeny classes admitting a global triple $(\\gamma_0; \\gamma, \\delta)$ contribute to the Lefschetz trace formula. Via the trace formula the same identity yields the cohomology description: the restriction of $H^*_\\xi$ to $W_{E_p}$ is a sum of terms $\\pi_f \\otimes (r_{-\\mu}\\circ\\phi_{\\pi_p}|_{W_{E_p}})\\,|\\cdot|^{-\\dim Sh/2}$ with the multiplicities $a(\\pi_f)$; consequently the semisimple local Hasse-Weil factors are products of local $L$-factors. This extends the known quasi-split results to any inner form of a product of Weil restrictions of general linear groups.","pith_inferences":["Beyond the paper: because the transfer is expressed through explicit Bernstein-center elements, the local Hasse-Weil factors for these varieties are in principle computable by evaluating characters rather than by counting points, suggesting direct numerical checks in small cases.","Beyond the paper: the same mechanism should apply at any parahoric level for which the deformation spaces have controlled cohomology; if a flat integral model replaces the naive one, the method's caveat about non-flatness disappears and the argument may extend to neighboring groups beyond Weil restrictions of general linear groups.","Beyond the paper: the vanishing property itself can be tested locally in a rank-one division algebra setting, where the deformation spaces are concrete; a nonzero twisted orbital integral for a non-transferable $\\delta$ would contradict the paper's main local theorem."],"forward_implications":["$H^*_\\xi$ restricted to $W_{E_p}$ equals $\\sum_{\\pi_f} a(\\pi_f)\\,\\pi_f \\otimes (r_{-\\mu}\\circ\\phi_{\\pi_p}|_{W_{E_p}})\\,|\\cdot|^{-\\dim Sh/2}$ in the relevant Grothendieck group.","The semisimple local Hasse-Weil factor of $Sh_K$ at $p$ equals $\\prod_{\\pi_f} L^{ss}(s-\\dim Sh_K/2, \\pi_p, r_p)^{a(\\pi_f)\\dim \\pi_f^K}$.","The vanishing property means that only isogeny classes for which the naive norm is conjugate into $G(\\mathbb{Q}_p)$ contribute to the Lefschetz trace formula, removing the obstruction to forming global triples $(\\gamma_0; \\gamma, \\delta)$.","The semisimple trace of Frobenius at level $K_pK^p$ is expressed as a sum over such triples involving the Bernstein-center element $z^{(r)}_{-\\mu}*e_{K_{p^r}}$, verifying the test-function conjecture in this setting."],"supporting_citations":[{"why":"supplies the definition of the local test functions via deformation spaces of p-divisible groups and the point-counting method the paper generalizes.","marker":"[Sch13b]"},{"why":"proved the quasi-split case of the matching and cohomology description that the paper extends to non-quasi-split groups.","marker":"[SS13]"},{"why":"provides the stable Bernstein center formalism and states the test-function conjecture later verified by the paper's Corollary 4.6.2.","marker":"[Hai14]"},{"why":"defines the simple Shimura varieties and the multiplicities $a(\\pi_f)$ appearing in the main cohomology formula.","marker":"[Kot92a]"},{"why":"supplies the point-counting and norm-triple framework used to convert fixed points into orbital integrals.","marker":"[Kot92b]"},{"why":"gives the base change and stable base change transfer theory underlying the local harmonic analysis in Section 2.","marker":"[AC89]"},{"why":"provides the global Galois representations and base change results used in the proof of the local theorems.","marker":"[HT01]"},{"why":"supplies the base change of unit idempotents and the vanishing property used in the semisimple trace formula of Corollary 4.6.2.","marker":"[Kot86a]"}],"fun_headline_variants":["Vanishing orbital integrals pin down Shimura cohomology","Inner forms yield explicit stable center base change","Orbital identity gives local zeta factors for Shimura varieties","Non-quasi-split local groups: cohomology via orbital integrals","Test functions satisfy vanishing property beyond quasi-split"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The argument assumes the fixed-point count in the special fiber of the integral model is valid, even though the parahoric model used at $p$ is only shown to be a scheme and may fail to be flat.","fun_headline_variants_meta":{"raw":{"variants":["Vanishing orbital integrals pin down Shimura cohomology","Inner forms yield explicit stable center base change","Orbital identity gives local zeta factors for Shimura varieties","Non-quasi-split local groups: cohomology via orbital integrals","Test functions satisfy vanishing property beyond quasi-split"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000591,"raw_usage":{"total_tokens":2775,"prompt_tokens":954,"completion_tokens":1821,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":570,"completion_tokens_details":{"reasoning_tokens":1750}},"tokens_in":570,"tokens_out":1821,"duration_ms":13042,"temperature":1.0,"reasoning_tokens":1750,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T17:29:31.219145+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take an explicit non-quasi-split local group such as the unit group of a central division algebra over $\\mathbb{Q}_p$, choose a $\\sigma$-semisimple $\\delta$ in $G(\\mathbb{Q}_{p^r})$ whose naive norm is not conjugate into $G(\\mathbb{Q}_p)$, and compute the twisted orbital integral $TO_{\\delta\\sigma}(\\varphi_{\\tau,h})$ directly from the deformation-space definition; a single nonzero value would contradict the vanishing theorem.","supporting_citations":[],"review_version":1}