{"id":"044b6f83-eb81-490c-bcf2-7a0547258a8a","arxiv_id":"2412.16575","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Irreducibility of local models is classified by double quotients of relative Weyl groups, and all fibers of level-changing maps are single Schubert varieties.","lead":"This paper classifies when the geometric special fibers of local models of Shimura varieties are irreducible, and proves that fibers of the level-changing map between different parahoric levels are always single Schubert varieties. These are foundational structural results for the geometry of local models in the Langlands program.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"No significant objection identified: Theorem 2.1's proof is coherent; the cited acute-cone criterion [8, Cor. 4.4] is a published theorem that exactly supplies the needed input, and Remark 2.3 independently corroborates Proposition 2.2.","rationale":"The reader's weakest_assumption identifies the acute-cone criterion as the key external dependency of Proposition 2.2, and that is indeed the single most load-bearing input for Theorem 2.1. However, this is a published theorem invoked with matching hypotheses, not an internal gap; the proof of Proposition 2.2 is otherwise logically consistent. The alternative proof mentioned in Remark 2.3 further supports the proposition. The unreported brute-force classification in §1.7 affects Theorem 1.4/Proposition 1.5, not the strongest claim Theorem 2.1. I therefore keep the reader's CONDITIONAL verdict unchanged: the paper's central fiber theorem is mathematically sound as far as the stress-test shows, while the classification's computational component remains a reproducibility concern that the authors should address.","tokens_in":17104,"tokens_out":22727,"duration_ms":198934,"concrete_test":"Verify the original statements of [8, Cor. 4.4] and [7, Cor. 5.6] to confirm they cover arbitrary connected reductive G over k((t)), including non-reduced relative root systems; if they do, Proposition 2.2 and hence Theorem 2.1 stand as written.","verdict_should_be":"UNCHANGED","load_bearing_attack":"I looked for a load-bearing weakness in the proof of the fiber theorem. The reduction of Theorem 2.1 to Proposition 2.2 via (2.2) is valid: w is a minimal double-coset representative in K1˜W K2, so left multiplication by w identifies Bruhat order on W_{K2} with Bruhat order on the left coset wW_{K2}; since Adm(μ) is a lower ideal, the indexing set {x∈W_{K2}: wx∈Adm(μ)} is a principal Bruhat ideal under its maximum. Proposition 2.2 is proved from two external inputs: [7, Cor. 5.6] (every alcove lies in some acute cone) and Theorem 2.6 ([8, Cor. 4.4], admissible alcoves in C(a,z) are ≤ t_z(μ)). The quoted hypotheses match the paper's setup, including non-transitive W̃-action on alcoves and possibly non-reduced relative root systems. I found no misstatement or missing hypothesis in this chain. The classification in §1.7 uses an unreported brute-force computation for F4/G2, which is a real reproducibility weakness, but it is not needed for Theorem 2.1. Thus the strongest claim has no demonstrated soft spot.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the geometric special fibers of local models of Shimura varieties and G-shtuka moduli, encoded as the closed subvarieties A_K(G,\\mu) of partial affine flag varieties. The first main result, Theorem 1.4, gives a natural bijection between the irreducible components of A_K(G,\\mu) and the double quotient W_{pr(K)}\\setminus W_0 / W_{I(\\mu)}, together with an explicit irreducibility criterion for quasi-simple G in terms of W_{pr(K)}, W_0, and the short-root subgroup W_short. The second main result, Theorem 2.1, asserts that for any spherical subsets K_1\\subseteq K_2, every fiber of the level-changing map \\pi: A_{K_1}(G,\\mu)\\to A_{K_2}(G,\\mu) is isomorphic to a single Schubert variety in the partial flag variety \\breve{K}_2/\\breve{K}_1, hence irreducible. The proof is reduced to Proposition 2.2, a combinatorial statement about unique maximal elements in wW_K\\cap \\operatorname{Adm}(\\mu), which is proved using the acute-cone criterion of Haines--He and a geometric description of the admissible set. Section 3 gives an explicit algorithm, based on the quantum Bruhat graph and Demazure products, for computing the maximal element in the hyperspecial case, with several applications and corollaries.","tokens_in":17368,"tokens_out":33580,"duration_ms":279082,"significance":"Theorem 2.1 is a striking and surprising result: although A_K(G,\\mu) is usually reducible, the fibers of the level-changing map are always irreducible Schubert varieties. The parametrization of irreducible components in Theorem 1.4 is clean and potentially useful for applications to local models and affine Deligne--Lusztig varieties. The paper is largely self-contained after standard external inputs, and the proof of Theorem 2.1 is coherent and carefully structured. Section 3 provides a concrete algorithmic tool that is likely to be of independent interest. The main weakness is that the classification of exceptional types F_4 and G_2 in §1.7 is not documented: it relies on an unstated 'computer program' with no code, input, or output, which affects the full strength of the irreducibility classification in Theorem 1.4 and Proposition 1.5.","major_comments":[{"comment":"The classification table for exceptional types is not reproducible: the sentence 'For exceptional types, we use the computer program to compute the set supp(pr(K)W0) by brutal force' gives no description of the computation, no pseudocode, no input data, and no output. Since the 'moreover' part of Theorem 1.4 and Proposition 1.5 explicitly rely on this classification for types F_4 and G_2, this is a load-bearing step. The authors should supply either a human-checkable case-by-case proof, or a complete reproducible computation (code and output) covering all spherical subsets K of the affine Dynkin diagrams of types F_4 and G_2.","section":"§1.7"},{"comment":"The proof of Theorem 2.1 establishes, by the displayed cell decompositions in (2.1) and (2.2), that the underlying set of the fiber \\pi^{-1}(m) is a union of Schubert cells corresponding to the set {x\\in W_{K_2}: wx\\in \\operatorname{Adm}(\\mu)}. Proposition 2.2 then shows that this indexing set has a unique maximal element, so the fiber is set-theoretically a single Schubert variety. However, the theorem states that the fiber is 'isomorphic' to a Schubert variety, which is a scheme-theoretic assertion. The argument does not address whether the scheme-theoretic fiber has nilpotents or embedded components; the displayed equalities are set-theoretic except for the final identification, which is written as an isomorphism. The authors should clarify whether the isomorphism is meant at the level of reduced varieties, and if so, justify that the scheme-theoretic fiber coincides with the reduced Schubert variety, or add the missing reducedness argument.","section":"§2.2, Theorem 2.1"}],"minor_comments":[{"comment":"There is a typo: 'brutal force' should be 'brute force'.","section":"§1.7"},{"comment":"In the proof of Corollary 3.4, the phrase 'left-right symmetry (3.2)' is inaccurate: the left-right symmetry is displayed in (3.1), while (3.2) is the triangle inequality for the weight function.","section":"§3, Corollary 3.4"},{"comment":"The algorithm defining z_\\gamma does not handle the case \\gamma=0. Since the set {x\\in W_0: \\operatorname{wt}(x,1)\\leq 0} is {1}, the definition should be initialized with z_0=1, and the induction should treat \\gamma=0 as the base case or exclude it explicitly.","section":"§3.3, Theorem 3.7"},{"comment":"The introduction states k=\\mathbb{F}_q, while §1.1 sets k to be an algebraically closed field. The relationship between the finite field of the original local model and the algebraically closed base field used for geometric fibers should be stated consistently.","section":"§0.1 and §1.1"}],"recommendation":"major_revision","confidential_remarks":"The central fiber theorem, Theorem 2.1, appears sound and the proof is convincing; I found no gap in the reduction to Proposition 2.2 or in the subsequent use of the acute-cone criterion. The main issue is the undocumented exceptional-type computation in §1.7, which is load-bearing for the irreducibility classification. This is fixable within the scope of the paper by providing a reproducible computation or a written case-by-case check. The scheme-theoretic wording in Theorem 2.1 should also be tightened. I therefore recommend major revision rather than rejection."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: this is a substantial, honest paper. The two headline theorems are real advances and the proofs hold up under scrutiny. The only genuine problem is a small reproducibility gap in the classification of exceptional cases.\n\nThe classification in Theorem 1.4—irreducible components of A_K(G, μ) parameterized by W_pr(K)\\W0/W_I(μ)—is clean and useful. The \"moreover\" criterion for irreducibility, with the explicit table in Proposition 1.5, is exactly what people in this area will use. Theorem 2.1 is the real prize: fibers of level-changing maps are single Schubert varieties, so even when source and target are reducible, the fibers stay irreducible. I checked the reduction to Proposition 2.2; the double-coset argument is valid and the use of [7, Cor 5.6] plus [8, Cor 4.4] is legitimate. The stress-test note is right: no missing hypothesis there. The argument that wW_K ∩ Adm(μ) has a unique maximal element is sound.\n\nThe algorithm in Section 3 is a nice extra. It computes max(wW_K ∩ Adm(μ)) in the hyperspecial case using the quantum Bruhat graph and Demazure products, with a proof written down (Theorem 3.7). That part is new as a written proof.\n\nThe real caveat is §1.7. The classification for exceptional types F4 and G2 is settled by \"brutal force\" on an unstated computer program. No code, no data, no description of the algorithm. That means the classification part is not independently verifiable as written. It is a genuine but minor issue: the fiber theorem and Proposition 2.2 don't depend on that computation, and the classical types are done by hand. Still, for a theorem whose conclusion is a complete case table, I'd want the authors to either give a verifiable certificate or a documented script before publication.\n\nThe cited acute cone criterion is a published result, so leaning on it is not a flaw. It might be worth asking the authors to reassure readers that the non-transitive action case is covered, but the stress-test confirms it is.\n\nWho is this for? Anyone working on local models, affine Deligne-Lusztig varieties, or admissible sets. It deserves a serious referee. My recommendation: send it out, and ask for an appendix documenting the brute-force check.","headline":"Genuinely new results on local models; the fiber theorem holds up, but the exceptional-type classification needs a documented computation.","tokens_in":17880,"tokens_out":2122,"would_cite":true,"duration_ms":32698,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["11G25","20G25"],"pacs":[],"model":"deepseek-v4-flash","headline":"Every fiber of a level-changing map between local models is a single Schubert variety, and the irreducible local models are classified by a relative Weyl group double quotient.","keywords":["local models","Schubert varieties","admissible sets","parahoric level structures","Shimura varieties","quantum Bruhat graph","Bruhat order"],"falsifier":"Searching the finite set $wW_K \\cap \\mathrm{Adm}(\\mu)$ for a pair of incomparable maximal elements in a small non-simply-laced root system such as $C_2$ or $G_2$, where the admissible set can be listed explicitly via the quantum Bruhat graph formula, would settle Proposition 2.2; if such a pair exists, the unique-maximal-element claim, and with it the fiber irreducibility theorem, is false.","tokens_in":16908,"feed_emoji":"📐","tokens_out":10339,"duration_ms":76018,"temperature":0.7,"pith_summary":"Local models are flat projective schemes that capture the singularity structure of Shimura varieties and moduli of $G$-shtukas at a chosen prime, and their geometric special fibers are unions of affine Schubert varieties. This paper answers two questions about those fibers. First, it classifies exactly when the geometric special fiber of a local model is irreducible, in terms of a double quotient of the relative Weyl group, and gives an explicit list of the exceptional non-special cases. Second, it proves that every fiber of the level-changing map between local models with different parahoric level structures is isomorphic to a single Schubert variety in a partial flag variety, hence is itself irreducible. The geometric rigidity is reduced to a combinatorial claim about the $\\mu$-admissible set: every left coset $wW_K$ intersects the admissible set in a unique Bruhat-maximal element.","feed_headline":"Local-model fibers are always irreducible Schubert varieties","feed_subtitle":"Even when source and target are reducible, every fiber between parahoric levels is one Schubert variety.","key_machinery":"The central object is the $\\mu$-admissible set $\\mathrm{Adm}(\\mu) = \\{w \\in \\widetilde{W} \\mid w \\leq t_{x(\\mu)} \\text{ for some } x \\in W_0\\}$, a finite set in the Iwahori–Weyl group whose Bruhat order controls the closure relations among the Schubert cells making up $A_K(G,\\mu)$. The load-bearing mechanism for the fiber theorem is the acute cone $C(a,z)$, defined as the set of alcoves reachable from the base alcove by a gallery that crosses each wall on the side of the deep alcoves in the chamber $z(C^+)$; the acute-cone criterion says that any $w$ whose alcove $w(a)$ lies in $C(a,z)$ satisfies $w \\leq t_z(\\mu)$. This lets the authors contain an entire left coset $wW_K$ in a single acute cone, so that $wW_K \\cap \\mathrm{Adm}(\\mu)$ becomes the Bruhat interval $\\{w' \\in wW_K \\mid w' \\leq t_z(\\mu)\\}$, which has a unique maximal element by a classical splitting criterion. For the classification of irreducible components, the machinery is the description of the double quotient $W_{\\mathrm{pr}(K)}\\backslash W_0/W_{I(\\mu)}$, together with chamber geometry of the cones $C_K$ and $C^+_K$ used to list all non-special $K$ for which the relative supports stay proper. In the hyperspecial case the paper adds the quantum Bruhat graph and the Demazure product to compute the unique maximal element explicitly.","core_discovery":"The paper proves two theorems about the varieties $A_K(G,\\mu)$ that model the geometric special fibers of local models. Theorem 1.4 establishes a natural bijection between the irreducible components of $A_K(G,\\mu)$ and the double quotient $W_{\\mathrm{pr}(K)}\\backslash W_0/W_{I(\\mu)}$, where $\\mu$ is the dominant cocharacter, $W_0$ the relative Weyl group, $W_{\\mathrm{pr}(K)}$ the image of the parabolic subgroup generated by $K$, and $I(\\mu)$ the stabilizer of $\\mu$. It further shows, when $G$ is quasi-simple and $\\mu$ is non-central, that $A_K(G,\\mu)$ is irreducible exactly when either $K$ is maximal special or $W_0 = W_{\\mathrm{pr}(K)}W_{\\mathrm{short}}$ with $W_{\\mathrm{short}}$ stabilizing $\\mu$; Proposition 1.5 lists the exceptional cases (certain $K$ in types $B_n$, $C_n$, $F_4$, $G_2$). Theorem 2.1 then shows that for any spherical subsets $K_1 \\subseteq K_2$, every fiber of the level-changing map $\\pi_{K_1,K_2}: A_{K_1}(G,\\mu) \\to A_{K_2}(G,\\mu)$ is isomorphic to a single Schubert variety in the partial flag variety $\\breve{K}_2/\\breve{K}_1$, and is therefore irreducible. The fiber statement is reduced to Proposition 2.2, a combinatorial assertion that for every $w \\in \\mathrm{Adm}(\\mu)$ and every spherical $K$, the intersection $wW_K \\cap \\mathrm{Adm}(\\mu)$ has a unique maximal element in Bruhat order. The proof encloses the whole left coset $wW_K$ in one acute cone, applies the acute-cone criterion to bound the intersection by a Bruhat interval, and concludes by a splitting criterion for Bruhat order that such an interval has a unique maximum.","pith_inferences":["The paper leaves implicit that the level-changing maps are flat families with known fiber type; one could read off local invariants of each fiber, such as nearby cycles or intersection cohomology, directly from Schubert calculus.","The unique-maximal-element property of $wW_K \\cap \\mathrm{Adm}(\\mu)$ in Proposition 2.2 suggests a hidden convexity of the admissible set; testing whether analogous intersections with other $K$-double cosets satisfy the same property would delimit how far the phenomenon extends.","The classification's reliance on $W_{\\mathrm{short}}$ suggests that non-simply-laced root systems are essential for non-special irreducible level structures; a natural next step is to ask whether the analogous list controls smoothness of local models in mixed characteristic."],"forward_implications":["Every level-changing map between parahoric local models has irreducible fibers, so all fibers are connected and equidimensional, and their singularities are exactly the singularities of a single Schubert variety.","The irreducible components of any $A_K(G,\\mu)$ are in bijection with $W_{\\mathrm{pr}(K)}\\backslash W_0/W_{I(\\mu)}$, so irreducibility is a purely group-theoretic condition; the classification shows irreducible non-special levels occur only in non-simply-laced types.","In the split regular case, every Schubert variety in the flag variety $\\breve{K}_0/\\breve{I}$ occurs as a fiber of the map from Iwahori level to hyperspecial level, so fiber geometry can be as singular as arbitrary Schubert varieties.","When $K$ is maximal special, the unique maximal element of $wW_K \\cap \\mathrm{Adm}(\\mu)$ is computable by an efficient algorithm using greedy root decompositions and Demazure products, without evaluating quantum Bruhat weights."],"supporting_citations":[{"why":"Supplies the acute cone criterion (Corollary 4.4) that bounds $w$ by $t_z(\\mu)$ whenever $w(a)$ lies in $C(a,z)$, the key step in proving Proposition 2.2.","marker":"[8]"},{"why":"Provides the fact that every alcove lies in some acute cone and the gallery characterization used to show a whole left coset is contained in one acute cone.","marker":"[7]"},{"why":"Gives the splitting criterion for Bruhat order used to conclude the interval $\\{w' \\in wW_K \\mid w' \\leq t_z(\\mu)\\}$ has a unique maximal element.","marker":"[2]"},{"why":"Establishes the identity $\\mathrm{Adm}(\\mu)_K \\cap \\widetilde{W}^K = \\mathrm{Adm}(\\mu) \\cap \\widetilde{W}^K$, which reduces union-of-Schubert-varieties descriptions to the admissible set.","marker":"[11]"},{"why":"Identifies the maximal elements of $\\mathrm{Adm}(\\mu) \\cap \\widetilde{W}^K$ as the translations $t_x(\\mu)$, the starting point for the irreducible-component bijection.","marker":"[13]"},{"why":"Provides the quantum Bruhat graph description of $\\mathrm{Adm}(\\mu) \\cap W_0 t_\\lambda W_0$, used to compute the maximal element in the hyperspecial case and to derive Corollary 3.4.","marker":"[14]"},{"why":"Supplies a generalized version of the admissible-set description and, per the paper's remark, an independent route to Proposition 2.2.","marker":"[22]"},{"why":"Gives the greedy decomposition and the element $z_\\gamma$ used in the algorithm of Theorem 3.7 for the maximal element in the hyperspecial case.","marker":"[1]"}],"fun_headline_variants":["Every fiber between parahoric levels is a single Schubert variety","All parahoric level-change fibers are irreducible Schubert varieties","Fiber irreducibility proven: each level-change fiber is one Schubert variety","Level-changing maps have irreducible fibers: always one Schubert variety","Even reducible local models have irreducible level-change fibers"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The proof of Proposition 2.2 rests on the acute-cone criterion quoted from [8], namely that any alcove lying in an acute cone $C(a,z)$ corresponds to an element bounded above by $t_z(\\mu)$, and this criterion is cited rather than proved in the present paper.","fun_headline_variants_meta":{"raw":{"variants":["Every fiber between parahoric levels is a single Schubert variety","All parahoric level-change fibers are irreducible Schubert varieties","Fiber irreducibility proven: each level-change fiber is one Schubert variety","Level-changing maps have irreducible fibers: always one Schubert variety","Even reducible local models have irreducible level-change fibers"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001517,"raw_usage":{"total_tokens":6133,"prompt_tokens":1056,"completion_tokens":5077,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":672,"completion_tokens_details":{"reasoning_tokens":4987}},"tokens_in":672,"tokens_out":5077,"duration_ms":29222,"temperature":1.0,"reasoning_tokens":4987,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T10:27:07.218330+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Searching the finite set $wW_K \\cap \\mathrm{Adm}(\\mu)$ for a pair of incomparable maximal elements in a small non-simply-laced root system such as $C_2$ or $G_2$, where the admissible set can be listed explicitly via the quantum Bruhat graph formula, would settle Proposition 2.2; if such a pair exists, the unique-maximal-element claim, and with it the fiber irreducibility theorem, is false.","supporting_citations":[{"cited_title":"Vertexwise criteria for admissibility of alco ves","cited_arxiv_id":null,"evidence_quote":"Supplies the acute cone criterion (Corollary 4.4) that bounds $w$ by $t_z(\\mu)$ whenever $w(a)$ lies in $C(a,z)$, the key step in proving Proposition 2.2."},{"cited_title":"Alcoves associated to special fibe rs of local models","cited_arxiv_id":null,"evidence_quote":"Provides the fact that every alcove lies in some acute cone and the gallery characterization used to show a whole left coset is contained in one acute cone."},{"cited_title":"A splitting criterion for the Bruhat orderings on Coxeter groups","cited_arxiv_id":null,"evidence_quote":"Gives the splitting criterion for Bruhat order used to conclude the interval $\\{w' \\in wW_K \\mid w' \\leq t_z(\\mu)\\}$ has a unique maximal element."},{"cited_title":"Kottwitz-Rapoport conjecture on unions of affine Delig ne-Lusztig varieties","cited_arxiv_id":null,"evidence_quote":"Establishes the identity $\\mathrm{Adm}(\\mu)_K \\cap \\widetilde{W}^K = \\mathrm{Adm}(\\mu) \\cap \\widetilde{W}^K$, which reduces union-of-Schubert-varieties descriptions to the admissible set."},{"cited_title":"On the µ-ordinary locus of a Shimura variety","cited_arxiv_id":null,"evidence_quote":"Identifies the maximal elements of $\\mathrm{Adm}(\\mu) \\cap \\widetilde{W}^K$ as the translations $t_x(\\mu)$, the starting point for the irreducible-component bijection."},{"cited_title":"Dimension formula for the affine Deligne-Lusztig variety X(µ, b)","cited_arxiv_id":null,"evidence_quote":"Provides the quantum Bruhat graph description of $\\mathrm{Adm}(\\mu) \\cap W_0 t_\\lambda W_0$, used to compute the maximal element in the hyperspecial case and to derive Corollary 3.4."},{"cited_title":"Affine Bruhat order and Demazure products","cited_arxiv_id":null,"evidence_quote":"Supplies a generalized version of the admissible-set description and, per the paper's remark, an independent route to Proposition 2.2."},{"cited_title":"Curve neighborhoods of Schuber t varieties","cited_arxiv_id":null,"evidence_quote":"Gives the greedy decomposition and the element $z_\\gamma$ used in the algorithm of Theorem 3.7 for the maximal element in the hyperspecial case."}],"review_version":1}