{"id":"ce1206c1-d110-4fcc-9076-8a6fecb7cce2","arxiv_id":"2505.02165","paper_version":1,"verdict":"UNVERDICTED","confidence":"UNKNOWN","novelty_score":8.0,"correctness_risk":"unknown","formal_verification":"none","parameter_count":0,"one_line_summary":"For an abelian variety over a number field with semistable reduction at v, the Mumford-Tate group valued Weil-Deligne representation at v is defined over Q and is independent of the auxiliary prime l.","lead":"This paper proves that the l-adic Galois representations attached to an abelian variety over a number field form a strongly compatible system valued in the Mumford-Tate group, including at places of semistable reduction and with l equal to the residue characteristic. It extends the authors' earlier good-reduction theorem and supplies a hypothesis needed for van Hoften's Hecke orbit conjecture.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Proposition 5.3.8 only places Frobenius in H(Z_p), but Theorem 5.3.10 requires the full Galois image, including the inertial unipotent subgroup, to lie in K_p; the stated proof does not close this gap.","rationale":"The paper’s central claim is substantial and the overall strategy is coherent: integral models, toroidal compactifications, boundary curves, and Lafforgue/Abe companion comparisons are combined in a natural way, and I found no internal inconsistency in the companion-argument chain formed by Propositions 4.1.9, 4.2.6, Corollary 5.1.6, and Lemma 4.3.12. The least secure point is exactly the auxiliary construction in Proposition 5.3.8, which the reader also flagged. My inspection sharpens the objection: condition (A) and its proof control only the Frobenius element, whereas the proof of Theorem 5.3.10 needs the entire Galois image, in particular the inertial pro-p unipotent subgroup, to lie in K_p. Without a lemma covering inertia, the interpretation of A_F as a Shimura point and the ensuing appeal to the boundary-curve theorem 3.2.3 are not justified as written. This is not a disagreement with existing consensus but a request for a missing support step. If the missing containment follows from a stronger reading of [KZ, Lemma 6.2.1] together with p-adic monodromy integrality, the theorem can stand; I therefore recommend CONDITIONAL rather than REJECT. I credit the paper’s substantial independent support, including Theorem 2.1.12 from [PR21], the boundary-curve construction in Theorem 3.2.3, and the careful comparison of Weil–Deligne representations via tame fundamental groups and isocrystals.","tokens_in":43110,"tokens_out":19011,"duration_ms":275728,"concrete_test":"Read [KZ, Lemma 6.2.1] and the proof of Proposition 5.3.8 to determine whether the lemma provides a special parahoric G_F(O_F) containing the full image ρ^G_{A,p}(Γ_{E'}) for some finite extension E', or only the single element ρ^G_{A,p}(Frob). Then test a concrete semistable case, such as a Tate curve over Q_p with p > 2 and G = GL_2, by computing the inertial unipotent generator exp(tN) and checking whether N lies in Lie(H(Z_p)) for the parahoric H constructed in Proposition 5.3.8. If it does not, the assertion that ρ^G_{A,p}|_{Γ_{E'}} factors through K_p is false for that example, and Theorem 5.3.10 Case (2) lacks a valid reduction as written.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The proof of Theorem 5.3.10 depends on the deduction, immediately after Proposition 5.3.8, that “By construction, ρ^G_{A,p}(eσ_q) lies in K_p := H(Z_p). Hence there is a finite extension E'/E such that ρ^G_{A,p}|_{Γ_{E'}} factors through K_p.” Proposition 5.3.8(A) controls only a single Frobenius lift eσ_q. For a semistable p-adic Galois representation, Γ_{E'} is topologically generated by Frobenius together with the inertia group, and the inertia image is typically an infinite pro-p unipotent subgroup. An integral Frobenius element does not force this unipotent subgroup to lie in H(Z_p); one additionally needs an integrality condition on the p-adic monodromy operator N with respect to the parahoric H. Since K_p = H(Z_p) is the level structure used to view A_F as a point x_A ∈ Sh_K(E'), and since Case (2) of Theorem 5.3.10 then needs x_A to extend to an O_{E'}-point with special fiber on the boundary of S^Σ_K, the missing containment of the inertia image is load-bearing. The proof of Proposition 5.3.8 cites [KZ, Lemma 6.2.1] only for the Frobenius element and gives no argument for N or for the full Galois image; this is a concrete gap, possibly repairable, in the reduction to Shimura varieties.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proves a motivic refinement of the Weil–Deligne–Raynaud theorem on compatible systems for abelian varieties. For an abelian variety A over a number field E ⊂ C with Mumford–Tate group G, it establishes that after a finite base change the ℓ-adic Galois representations on H^1_{ét}(A, Q_ℓ) factor through G(Q_ℓ) and that, at every place v of semistable reduction, the associated G-valued Weil–Deligne representations ρ^{WD,G}_{A,ℓ,v} are all equivalent over G to a single representation ρ^{WD,G}_{A,v} defined over Q; this independence is claimed for all ℓ, including ℓ | v (Theorem 1.2 and Theorem 5.3.10). The strategy is to embed the Mumford–Tate group in an auxiliary quasi-split group H = Res_{F/Q} G_F, to view A as a point of a Hodge-type Shimura variety with parahoric level structure, to pass through a boundary curve in a toroidal compactification, and then to reduce the comparison to equal-characteristic function fields where Lafforgue's theorem and its p-adic analogue by Abe apply. Along the way, the paper proves new results on CM lifts in isogeny classes, local monodromy at the boundary, and G-valued Weil–Deligne representations, and it derives a version of ℓ-independence for points in the special fibers of Hodge-type Shimura varieties.","tokens_in":43432,"tokens_out":17014,"duration_ms":210162,"significance":"If the main theorem is correct, it is a substantial advance: it extends the authors' previous good-reduction result [KZ] to semistable reduction and to the p-adic place, includes the case p = 2, and verifies [vH24, Hypothesis 2.3.1], thereby yielding new instances of the Chai–Oort Hecke-orbit conjecture. The paper is organized as a coherent chain of theorems, and many auxiliary statements are proved in detail rather than merely cited: Proposition 1.4 on GL_n-detection of G-conjugacy, Theorem 3.2.3 on existence of boundary curves, and the log-crystalline comparison in §4.3 are notable examples. The use of Lafforgue–Abe companions and of function-field local–global compatibility is a convincing route to the desired independence. The main reservation is that two load-bearing points in §5.3 are not fully justified as written: the level-structure reduction for ℓ = p controls only a single Frobenius element, and the comparison at the boundary for ℓ = p is asserted rather than proved. These are concrete gaps, but they appear repairable within the scope of the paper's methods.","major_comments":[{"comment":"In the proof of Theorem 5.3.10, immediately after Proposition 5.3.8, the authors write: “By construction, ρ^G_{A,p}(eσ_q) lies in K_p := H(Z_p). Hence there is a finite extension E′/E such that ρ^G_{A,p}|_{Γ_E′} factors through K_p.” This inference is not justified. Proposition 5.3.8(A) controls only the single Frobenius lift eσ_q. For a semistable p-adic Galois representation, Γ_E′ is topologically generated by Frobenius together with inertia, and the inertia image is typically an infinite pro-p unipotent subgroup determined by the monodromy operator N. An integral Frobenius element does not force this unipotent subgroup to lie in H(Z_p): one additionally needs an integrality condition on N, or on the full image of inertia, with respect to the parahoric H. This missing containment is load-bearing because K_p = H(Z_p) is the level structure used to view A_F as a point of Sh_K(E′), and Case (2) of Theorem 5.3.10 then requires this point to extend to an O_{E′}-point of S^Σ_K whose special fiber lies on the boundary. The proof of Proposition 5.3.8 cites [KZ, Lemma 6.2.1] only for the Frobenius element and gives no argument for the inertia image. This is a concrete gap, possibly repairable, in the reduction to Shimura varieties.","section":"5.3, Theorem 5.3.10"},{"comment":"At the end of the proof of Theorem 5.3.10, the equality [ξ ∘ ρ^{WD}_{A,p,v}] = [ρ^{log}_{E^H_C}] is asserted to be “part of Lemma 5.2.3”. Lemma 5.2.3, however, only states an isomorphism x^*(E^G_C) ≃ E^G_x for points x ∈ C̄(O_{E″}) ∩ C(E″) in the interior; the representation ρ^{log}_{E^H_C} is formed at the boundary point c of the special fiber. The needed comparison between D_st of A over E′ at the specialization δ and the log-isocrystal stalk at c is neither stated nor proved. Since this equality is precisely the ℓ = p part of Theorem 1.2 for places of bad reduction, the proof requires an explicit log-crystalline specialization or monodromy comparison along C̄ from δ to c, or a precise reference for it. As written, the ℓ = p case of Case (2) rests on an unproved assertion.","section":"5.3, Case (2), ℓ = p"}],"minor_comments":[{"comment":"The paragraph after the announcement of Theorem 3.2.3 contains a visibly corrupted fragment: “Thm. 3.3.10]KZ, Mloc G,{μ} satisfies the Scholze–Weinstein conjecture ... fundamental group π1(Gder)...”. This appears to be an editing artifact and should be removed or rewritten.","section":"3, opening paragraph"},{"comment":"The last sentence of §5.3.4 contains the duplicated phrase “good reduction reduction at v”; it should read “good reduction at v”.","section":"5.3.4"},{"comment":"In Proposition 5.3.8 and Theorem 5.3.10, the same symbol H is used both for the reductive group H = Res_{F/Q} G_F and for the parahoric group scheme H (for example, “there exists a parahoric group scheme H for H”). This makes the statements harder to read; a distinct symbol such as H_p or H for the parahoric would help.","section":"5.3.7–5.3.10"}],"recommendation":"major_revision","confidential_remarks":null},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Punchline: the main theorem is significant, but the proof has a gap in §5.3 that is load-bearing and needs repair. The paper deserves a serious referee, not a desk reject.\n\nWhat's new: the G-valued strong compatibility for abelian varieties at places of semistable reduction, including ℓ=p and p=2. Noot only had a coarser equivalence and extra conditions; Raynaud did the GL_{2g} version. The method—using p-adic shtukas and toroidal compactifications, producing a curve through the boundary point, then comparing local systems via Lafforgue/Abe—is thoughtful and more powerful than earlier work. Many auxiliary propositions are proved rather than cited, and the organization is clear.\n\nThe soft spot, and it's real: in §5.3, Proposition 5.3.8 only tells you that the geometric Frobenius σ_q maps into H(Z_p). The proof then states that after finite extension the whole representation ρ_{A,p} factors through H(Z_p). For a semistable p-adic representation, inertia maps to exp(t_p N); controlling one Frobenius element does not put the inertial unipotent subgroup into H(Z_p). You need an integrality condition on the monodromy operator N with respect to the parahoric, and no such condition is established. The subsequent use of K=H(Z_p) as the level structure for the Shimura variety depends on exactly this containment, so the reduction to the Shimura variety doesn't start as written. The stress-test note is right.\n\nIs this fatal? Probably not. A repair could choose the base field so that both Frobenius and N are integral, or first perform a base change to make the representation crystalline, then argue. But it is not a typo; it is a missing argument in the central proof. The rest of the paper—the boundary monodromy, the equal-characteristic comparison, the GL_n-detection criterion—looks solid and is genuinely useful.\n\nBottom line: this is a serious paper for specialists in Shimura varieties and compatible systems. It should go to an expert referee who can either verify the missing integrality or suggest the repair. I would not cite the semistable theorem until the gap is fixed, but the geometric machinery in §§2–4 is worth knowing. Bring it to reading group only if your group can stomach a 40-page chain with a few to-appear references.","headline":"Substantial, significant result with a real gap in §5.3: the reduction to Shimura varieties needs an integrality condition on the inertial monodromy that is not proven.","tokens_in":43957,"tokens_out":4174,"would_cite":false,"duration_ms":50177,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["11G10","11G18","14G35","11F80"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper proves that for an abelian variety over a number field, the local Weil–Deligne representation at a semistable place is independent of the auxiliary prime ℓ and is defined over Q, after a finite base extension.","keywords":["abelian varieties","semistable reduction","Mumford–Tate group","Weil–Deligne representations","independence of ℓ","compatible systems","Shimura varieties","toroidal compactifications"],"falsifier":"For a concrete semistable abelian variety at a place $v$ lying over $p=2$, compute the pair $(s,N)$ attached to the $\\ell=2$ crystalline Weil–Deligne representation and the pair attached to an $\\ell\\neq2$ Tate module in a non-standard representation of $G$; if their images in $\\Phi(q,G,\\mathbb{C})$ are not $G(\\mathbb{C})$-conjugate, Theorem 1.2 fails. A more direct check would be to find a point in the special fiber of a strongly admissible Shimura variety whose isogeny class contains no lift to a special point, contradicting Theorem 2.2.7.","tokens_in":42909,"feed_emoji":"🔢","tokens_out":12984,"duration_ms":112259,"temperature":0.7,"pith_summary":"The paper establishes a motivic version of the classical result of Weil, Deligne, and Raynaud on strongly compatible systems. For an abelian variety $A$ over a number field, it shows that, after a finite extension of the base field, the $\\ell$-adic Galois representations on $H^1_{\\mathrm{ét}}(A_{\\overline{E}},\\mathbb{Q}_\\ell)$ form a strongly compatible system valued in the Mumford–Tate group $G$ of $A$. The new content is at places of semistable reduction: the local Weil–Deligne representation attached to $A$ at such a place $v$ is, up to $G$-conjugacy, independent of the auxiliary prime $\\ell$, and it is defined over $\\mathbb{Q}$. This includes the case $\\ell\\mid v$, where the comparison uses crystalline data. The interest is that the local Galois invariant at $v$ is a property of the abelian variety itself, not of the prime chosen to compute it.","feed_headline":"Semistable abelian varieties get prime-independent local data","feed_subtitle":"At semistable places the Weil–Deligne class is ℓ-independent inside the Mumford–Tate group.","key_machinery":"The machinery has three layers. First, a $G$-valued Weil–Deligne representation that is unipotently ramified and Frobenius-semisimple (URFS) is determined by the pair $(s,N)$ with $s\\in G(\\mathbb{C})_{\\mathrm{ss}}$ semisimple and $N\\in\\mathrm{Lie}(G)$ nilpotent satisfying $\\mathrm{Ad}(s)N=qN$; the equivalence class depends only on the residue-field size $q$, not on the characteristic of the local field. Second, Proposition 4.1.9 shows that for URFS representations, $G$-conjugacy is detected by composing with all representations $r:G\\to \\mathrm{GL}_n$, reducing the comparison to $\\mathrm{GL}_n$. Third, the geometric bridge is Theorem 3.2.3, which produces, from a semistable abelian variety, a smooth curve mapping into a toroidal compactification of an integral Shimura variety with the given point on the boundary and generic fiber in the interior; pullback of the $G$-local systems to this curve lets the authors invoke Lafforgue's and Abe's theorems on compatible systems over finite fields.","core_discovery":"The central claim is Theorem 1.2: if $v$ is a place where $A$ has semistable reduction, there exists a $G$-valued Weil–Deligne representation $\\rho^{\\mathrm{WD},G}_{A,v}$ defined over $\\mathbb{Q}$ such that $\\rho^{\\mathrm{WD},G}_{A,v}\\sim_G \\rho^{\\mathrm{WD},G}_{A,\\ell,v}$ for every prime $\\ell$, including $\\ell$ dividing $v$. Here $G$ is the Mumford–Tate group of $A$, and $\\sim_G$ is conjugacy by an element of $G(\\mathbb{C})$ after fixing isomorphisms $\\overline{\\mathbb{Q}}_\\ell\\cong\\mathbb{C}$. The proof passes from the mixed-characteristic local field at $v$ to an equal-characteristic field $\\mathbb{F}_q((u))$ by placing the abelian variety on a smooth curve inside a toroidal compactification, then uses compatible systems on curves over finite fields to compare the resulting Weil–Deligne representations for every $\\ell$. It also shows that each isogeny class in the special fiber of a strongly admissible Shimura variety contains a point lifting to a special point, and that Frobenius conjugacy classes on such integral models are independent of $\\ell$.","pith_inferences":["Editorial inference: if Proposition 5.3.8 could be replaced by a more direct construction, the same curve-on-the-boundary argument might prove the overarching compatibility conjecture for broader classes of motives, not just abelian varieties.","Editorial inference: the failure of Proposition 4.1.9 without the unipotent-ramified condition suggests that any extension to places of non-semistable reduction will need genuinely new input, since element-conjugate non-conjugate representations already occur for groups such as $\\mathrm{SO}_6$.","Editorial inference: the equal-characteristic comparison via log geometry suggests a testable pattern: semistable degeneration over a $p$-adic field and over the function field $\\mathbb{F}_q((u))$ should carry identical local Galois data, so either side could be computed in small examples."],"forward_implications":["At every place of semistable reduction, the local $G$-valued Weil–Deligne representation is defined over $\\mathbb{Q}$ and its class in $\\Phi(q,G,\\mathbb{C})$ is the same for all primes $\\ell$, including the prime below $v$.","The global $\\ell$-adic representations $\\rho^G_{A,\\ell}$ form a strongly compatible system valued in the Mumford–Tate group, giving the motivic refinement of the Weil–Deligne–Raynaud theorem.","The good-reduction case is extended to $p=2$ and to $\\ell=p$, and the verification of the van Hoften hypothesis implies instances of the Chai–Oort Hecke orbit conjecture for Shimura varieties of Hodge type.","For strongly admissible triples with $G_{\\mathbb{Q}_p}$ quasi-split, each isogeny class in the special fiber contains a point lifting to a special point, confirming a conjecture stated in the paper.","For arbitrary Hodge-type Shimura varieties, sufficiently divisible powers of the Frobenius conjugacy classes at points of the special fiber are independent of $\\ell$, including $\\ell=p$."],"supporting_citations":[{"why":"This is the authors' earlier proof at good reduction, which the present paper extends to semistable places and to $p=2$.","marker":"[KZ]"},{"why":"Lafforgue's function-field Langlands correspondence supplies the local-global compatibility that compares Weil–Deligne representations for different primes.","marker":"[Laf02]"},{"why":"Abe's crystalline companions provide the corresponding comparison for the $\\ell=p$ isocrystal side.","marker":"[Abe18a]"},{"why":"Pappas–Rapoport's $p$-adic shtuka models give the integral Shimura varieties and the strong-admissibility framework used throughout.","marker":"[PR21]"},{"why":"Gleason–Lim–Xu's description of isogeny classes via affine Deligne–Lusztig sets underlies the existence of CM lifts in Theorem 2.2.7.","marker":"[GLX]"},{"why":"Madapusi Pera's toroidal compactifications supply the boundary geometry on which the curve argument rests.","marker":"[Mad19]"},{"why":"Deligne's theorem that Hodge cycles are absolutely Hodge is what makes the Galois representations factor through the Mumford–Tate group.","marker":"[Del82]"},{"why":"Steinberg's conjugacy theorem detects semisimple conjugacy classes and is used in Proposition 4.1.9.","marker":"[Ste65]"}],"fun_headline_variants":["Semistable abelian varieties: ℓ-independent local invariants","At semistable places, abelian varieties get ℓ-stable local data","Motivic proof: semistable AVs have ℓ-independent Weil–Deligne reps","Semistable abelian varieties: local ℓ-independence from motivic methods","ℓ-independence at semistable reduction: a motivic refinement"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The proof depends on the auxiliary construction in Proposition 5.3.8: for each semistable abelian variety and place $v$ one must find a totally real field $F$ making $H=\\mathrm{Res}_{F/\\mathbb{Q}}G_F$ quasi-split at $p$ with a strongly admissible integral group model whose Frobenius element lies in its $\\mathbb{Z}_p$-points; if that construction is impossible for some $A$, the reduction to Shimura varieties does not go through.","fun_headline_variants_meta":{"raw":{"variants":["Semistable abelian varieties: ℓ-independent local invariants","At semistable places, abelian varieties get ℓ-stable local data","Motivic proof: semistable AVs have ℓ-independent Weil–Deligne reps","Semistable abelian varieties: local ℓ-independence from motivic methods","ℓ-independence at semistable reduction: a motivic refinement"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000848,"raw_usage":{"total_tokens":3703,"prompt_tokens":970,"completion_tokens":2733,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":586,"completion_tokens_details":{"reasoning_tokens":2630}},"tokens_in":586,"tokens_out":2733,"duration_ms":21068,"temperature":1.0,"reasoning_tokens":2630,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-16T00:59:37.154963+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"For a concrete semistable abelian variety at a place $v$ lying over $p=2$, compute the pair $(s,N)$ attached to the $\\ell=2$ crystalline Weil–Deligne representation and the pair attached to an $\\ell\\neq2$ Tate module in a non-standard representation of $G$; if their images in $\\Phi(q,G,\\mathbb{C})$ are not $G(\\mathbb{C})$-conjugate, Theorem 1.2 fails. A more direct check would be to find a point in the special fiber of a strongly admissible Shimura variety whose isogeny class contains no lift to a special point, contradicting Theorem 2.2.7.","supporting_citations":[],"review_version":1}