{"id":"06f2556b-a290-4b6a-92f6-793a512f7b8e","arxiv_id":"2602.11343","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"The excursion algebra of a scheme over a finite field is canonically isomorphic to the ring of functions on the geometrically semisimple locus, yielding reduced/normal components, finite generation, GL_n surjectivity, and ell-independence.","lead":"This paper proves that the excursion algebra—a central object in the Langlands program over finite fields—coincides with the functions on the semisimple part of the space of Galois representations. This makes it a product of classical invariant rings, proves finiteness and surjectivity properties, and gives an ell-independent rational model for smooth varieties.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 3.7.2 hinges on the unproved weight-factorization input Theorem 2.5.8; the cited [LLaf, Cor. VII.8] must be checked for arbitrary smooth finite-type U, not just curves.","rationale":"The reader's weakest-assumption diagnosis is on target: Theorem 2.5.8 is the main imported input on which the contraction and hence Theorem 3.7.2 rest. I refine the statement in two respects. First, the paper actually needs ℓ to be an arbitrary line in Shv^Weil(pt), not necessarily a Weil-number line; the reader's paraphrase is slightly stronger and would be false already for rank-one pullbacks from the arithmetic Galois group. Second, the relevant scope is a locally closed smooth U of arbitrary dimension, so the cited result must be valid for smooth finite-type F_q-schemes, not only for curves. I do not claim the theorem is false; the concern is that this is a load-bearing external input stated without verification of its hypotheses. I also considered the 'well known' lift in §5.7.8; that is less load-bearing because it affects only the ℓ-independence/rationality portion and the surrounding argument is largely supplied. Since the reader's conditional verdict already reflects this dependency, no adjustment is needed.","tokens_in":51314,"tokens_out":33589,"duration_ms":351704,"concrete_test":"Check the precise statement of [LLaf, Cor. VII.8] and its correction [De3, §1.7–1.9]: does it yield, for every irreducible lisse Weil sheaf on an arbitrary smooth finite-type F_q-scheme U, an isomorphism F ≅ F0 ⊗ ℓ with F0 of integral weight and ℓ an arbitrary rank-one Weil character? If the cited theorem is curve-specific or imposes a hypothesis absent in §2.5 (finite determinant, bounded ramification, etc.), re-run the GM-extension argument in Theorem 2.5.8 and see whether the counterexample on that U survives; any such counterexample invalidates (2.12) and hence Theorem 3.7.2.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central isomorphism (0.1) is obtained by contracting Shv^relev(X) to Shv^relev,0(X); the contraction (Theorem 2.8.2) first needs the de-equivariantization equivalence (2.12), whose essential content is Theorem 2.5.8. The entire proof of Theorem 2.5.8 is one paragraph: an irreducible Weil perverse sheaf is reduced to an irreducible lisse Weil sheaf on a locally closed smooth U by GM extension, and then [LLaf, Cor. VII.8] (with [De3]) is invoked. This is the load-bearing point because every later statement — including the identification LS^arithm,0 with (LS^restr,0)^Frob and the product decomposition of Exc(X,G) — uses this factorization. The manuscript does not spell out the exact hypotheses of the cited result. In particular, [LLaf] is fundamentally about curves, and the reduction to U of arbitrary dimension is not justified in the text; if Cor. VII.8 applies only to curves or to local systems with extra conditions (e.g. finite determinant or bounded ramification), Theorem 2.5.8 can fail exactly in the regime needed. Note also that the paper requires ℓ to be only a line in Shv^Weil(pt), not necessarily with Weil-number eigenvalues; the citation must be verified at that weaker level, since the stronger reading (ℓ of Weil-number type) is false even for rank-one characters pulled back from \\hat{Z}.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the excursion algebra Exc(X,G) of a scheme X over F_q with a reductive group G, defined as the algebra of global functions on the stack of arithmetic G-local systems on X. Its central theorem (Thm 3.7.2) asserts that the restriction map to the Frobenius-fixed semisimple locus is an isomorphism, obtained by constructing an A^1-contraction on a category of 'relevant' Weil sheaves. From this isomorphism the authors deduce structural properties of Exc(X,G): each connected component is classical, integral, and normal; the whole algebra is a product of invariant-theoretic rings O_{G_α//Ad_g(G_α)}; it is finitely generated over each local Hecke algebra; for G=GL_n the global Hecke algebra surjects onto it; and, for smooth X, there is a canonical Q-form independent of ℓ. The paper is written as a sequel to [AGKRR V] and depends heavily on that work, on Lafforgue's [VLaf]/[LLaf], and on Drinfeld's [Dr].","tokens_in":51638,"tokens_out":25454,"duration_ms":267598,"significance":"If correct, the main theorem gives a very clean and powerful description of the excursion algebra in complete generality, not just for curves. The contraction mechanism in Sects. 1–3 is original, and the deduction of the structural properties from the semisimple-locus description is elegant. The group-theoretic finiteness statement (Prop. 4.7.4) has a complete, self-contained proof. The paper is also honest about its limitations, e.g., Remark 5.1.6 states that for general G the rational-structure result is 'rationality in name only.' The main risk is the unverified external input in Theorem 2.5.8, which is load-bearing for the entire contraction argument. If that input is confirmed in the required generality, the paper would be a major contribution to the geometric Langlands program.","major_comments":[{"comment":"This theorem is the single load-bearing external input: it gives the equivalence (2.12), which via Corollary 2.5.11 underlies the contraction Theorem 2.8.2 and hence Theorem 3.7.2. The proof is a one-paragraph reduction to the lisse case followed by an invocation of [LLaf, Cor. VII.8] (with [De3]). The hypotheses of the cited result are not stated. In particular, [LLaf] is concerned with curves over F_q, whereas the lisse sheaf here lives on an arbitrary locally closed smooth finite-type U⊂X; no reduction to the curve case is provided. Moreover, the theorem is asserted with an arbitrary line ℓ∈Shv^Weil(pt), i.e., an arbitrary Frobenius eigenvalue, not necessarily a Weil number; the citation must be verified at that level, since a weaker version with ℓ of Weil-number type does not suffice for the de-equivariantization used in (2.12). If Cor. VII.8 carries extra hypotheses (finite determin","section":"§2.5.7 (Theorem 2.5.8)"}],"minor_comments":[{"comment":"There are typographical slips: 'theclassical stackunderlying' (§3.3.2) and 'will notchange the notation' (§4.1.2). Please fix throughout.","section":"§3.3.2, §4.1.2"},{"comment":"The sentence 'Since the operation of Goresky-MacPherson extension preserves weights' would benefit from a reference or a brief argument for objects in Shv^Weil,loc.fin, not only for pure sheaves. The reduction to the lisse case is standard, but the weight preservation in this generality is not obvious.","section":"§2.5.7"},{"comment":"Remark 3.7.12 gives a nice alternative proof of a key step; however, the notation [0]^* in (3.22) is introduced without a formal definition. Please clarify that it denotes the action of 0∈A^1 on global functions.","section":"§3.7.12"},{"comment":"Remark 5.1.6 explicitly states that for general G the rational-structure result is 'rationality in name only.' This is an honest caveat, but the introduction (preamble item (5) and (6)) presents the rational form as a main result; consider adding a caveat there so readers are not misled.","section":"§5.1.6"}],"recommendation":"major_revision","confidential_remarks":"The paper is well written and the contraction mechanism is compelling. My main concern is the unverified external input [LLaf, Cor. VII.8] as used in Theorem 2.5.8; the authors should confirm that the cited statement holds for lisse Weil sheaves on arbitrary smooth finite-type U, with ℓ an arbitrary line in Shv^Weil(pt). The paper also relies heavily on the unpublished [AGKRR V]; editors may want to ensure that the cited results are available in a verifiable form. The manuscript otherwise appears internally sound."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Two things to know. This is a serious paper with a real main theorem: Exc(X,G), the algebra of functions on the stack of arithmetic G-local systems, is canonically isomorphic to functions on the Frobenius-fixed semisimple locus (Theorem 3.7.2). If it holds, each factor is O(G_α)//Ad_g(G_α), hence reduced and normal, finite over local Hecke algebras; for GL_n the global Hecke algebra surjects onto it, and for smooth X there is a canonical Q-form independent of ℓ. None of that was in VLaf, Xue, Dr, or AGKRR V. The novelty is genuine.\n\nWhat is good: the A^1-contraction machinery (Sects. 1–2) is carried out in the text, not black-boxed. Theorem 1.1.7 (filtration ↔ A^1-comodule) is proved in detail, and the group-theoretic Proposition 4.7.4 — a twisted Vinberg — has a complete five-step proof. The GL_n surjectivity proof via Chebotarev is a clean upgrade of Laumon's argument, with credit to Lafforgue. The ℓ-independence section builds honestly on Drinfeld's rational model, and the authors flag that for general G it is 'rationality in name only' (footnote 12). The paper is unusually clear about what is imported versus proved.\n\nThe soft spot, and it is the load-bearing one: Theorem 2.5.8, the weight-factorization statement that every irreducible Weil perverse sheaf is F0 ⊗ ℓ with F0 of Weil-number type and ℓ an arbitrary line in Shv^Weil(pt). The whole contraction, and therefore 3.7.2, depends on it. The proof is one paragraph: GM extension reduces to the lisse case on a locally closed smooth U, then 'follows from [LLaf, Cor. VII.8]' with a correction in [De3]. Two things need checking. First, [LLaf] is fundamentally a curve paper (chtoucas), and the reduction from arbitrary smooth finite-type U to a curve is not justified in the text; restriction to a curve does not obviously globalize the twist parameter. Second, ℓ here is only required to be a line in Shv^Weil(pt) — an arbitrary scalar Frobenius twist — not a Weil-number twist. The citation must be verified at exactly that weaker level. This is probably fillable, but as written it is a missing check on the hinge, not a proof.\n\nMinor: one step in the ℓ-independence proof (the lift of σ to an étale local system on Drinfeld's group, §5.7.8) is asserted 'well known' without a reference; and the paper leans on the AGKRR V preprint for the definition of LS^arithm and Theorem 24.1.4. The contraction idea is traced to the third author's thesis and a forthcoming paper [LR] — provenance worth noting, not a flaw, since the needed theorems are proved here.\n\nVerdict: the central argument holds up as far as I can see; the geometry of the contraction is convincing and the structural payoffs are derived cleanly. The one citation whose hypotheses are not verified is the difference between 'convincing' and 'fully demonstrated.' Send it to a serious referee — this deserves refereeing, not a desk reject. The referee should check [LLaf, Cor. VII.8]'s exact hypotheses and the §5.7.8 lift. For anyone working on global Langlands over function fields, excursion operators, or independence of ℓ, it is exactly the paper they want.","headline":"Real main theorem — excursion algebra equals functions on the Frobenius-fixed semisimple locus — with clean new consequences, but the proof hinges on a one-paragraph citation ([LLaf, Cor. VII.8]) whose exact hypotheses are never checked; referee it, don't desk-reject it.","tokens_in":52177,"tokens_out":12988,"would_cite":true,"duration_ms":120202,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["14D23","14F20","11R39","14G17"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper proves that for a scheme X over a finite field and a reductive group G, the excursion algebra Exc(X,G) is canonically isomorphic to the algebra of functions on the Frobenius-fixed semisimple locus of the stack of arithmetic local","keywords":["excursion algebra","arithmetic local systems","Weil sheaves","semisimple locus","reductive group","Hecke algebra","independence of ℓ","automorphic forms"],"falsifier":"A test is to look for an irreducible lisse Weil sheaf on a smooth curve over a finite field whose Frobenius eigenvalues, at some closed point, cannot be written as a Weil number times a single constant independent of the point. Such a sheaf would violate the factorization input used to build the contraction, and would thereby falsify Theorem 3.7.2's conclusion that global functions on the arithmetic stack equal those on the semisimple locus. A direct stack-theoretic check would be to compute Γ(LS^arithm_G(X), O) on a Frobenius-fixed component with a nontrivial unipotent part and see whether fu","tokens_in":51157,"feed_emoji":"","tokens_out":11541,"duration_ms":113547,"temperature":0.7,"pith_summary":"The paper studies the excursion algebra Exc(X,G)—the global functions on the stack of arithmetic G-local systems on a finite-field scheme X—and proves that it is computed entirely by the semisimple locus. The main theorem identifies Exc(X,G) with the functions on the Frobenius-fixed semisimple part of the stack. From this single identification, each connected component of the algebra becomes an invariant ring O(G_α)/ /Ad_g(G_α), hence reduced and normal; the algebra is a finite module over every local Hecke algebra; for G=GL_n the global Hecke algebra maps onto it; and for smooth X there is a canonical rational model over Q independent of ℓ. These are exactly the structural facts needed to use the algebra in automorphic-function applications.","feed_headline":"Collapse the excursion algebra to the semisimple locus","feed_subtitle":"A contraction argument proves the full local-system stack has the same global functions as its Frobenius-fixed semisimple part.","key_machinery":"The central mechanism is a contraction of Shv^relev(X), the category of relevant Weil sheaves, by the monoid A^1. The paper first proves a general principle (Theorem 1.1.7): giving an A^1-action on a category is the same as giving a Z-invariant filtration on its G_m-equivariantization. Applying this to the weight filtration on integral-weight Weil sheaves produces the contraction, whose attracting category Shv^{relev,0}(X) is the semisimple category of weight-zero sheaves—the semi-simplification of Shv^relev(X). This contraction descends to the stack of relevant local systems, and the Frobenius fixed points of that contraction are the arithmetic local systems studied by the paper.","core_discovery":"The paper's main theorem is Theorem 3.7.2: for a connected scheme X of finite type over F_q and a reductive group G, the restriction map from the global functions on the arithmetic local-system stack to the global functions on its Frobenius-fixed semisimple locus is an isomorphism. The proof builds an action of the affine line A^1 on the category of relevant Weil sheaves—sheaves whose irreducible perverse pieces are invariant under some power of Frobenius—that contracts the category to the semisimple category of weight-zero sheaves. Functoriality carries this contraction to the stack of relevant local systems, contracting it onto the semisimple locus. After passing to Frobenius fixed points,","pith_inferences":["If the theorem is right, global functions on the arithmetic local-system stack cannot see unipotent or non-semisimple variation: the full stack and its semisimple locus have identical function algebras, stronger than the usual relation of a stack to its coarse space.","The whole construction rests on a single cited factorization statement about irreducible Weil perverse sheaves; a counterexample to that statement would not just leave a gap but would remove the contraction and with it the main theorem.","The A^1-contraction template may be reusable: any moduli problem whose category of sheaves admits a weight filtration and whose Frobenius action on functions is trivial would acquire the same 'functions come from the semisimple part' phenomenon.","For general G the rational model is, in the paper's own words, 'rationality in name only'—it exists but is not controlled by Hecke operators; making it useful would require a substitute for the GL_n surjectivity, perhaps through pro-semisimple completions of the fundamental group."],"forward_implications":["Exc(X,G) splits as a product over connected components of invariant rings O(G_α)/ /Ad_g(G_α), each of which is reduced and normal.","Every component of Exc(X,G) is a finite module over each local Hecke algebra H_x(G), so the algebra is finitely generated there.","For G=GL_n, the global Hecke algebra H_X(G) surjects onto Exc(X,G); the proof uses density of Frobenius conjugacy classes and trace comparisons.","For smooth X, a canonical Q-algebra Exc(X,G)_Q exists with Q_ℓ ⊗_Q Exc(X,G)_Q ≅ Exc(X,G) for every ℓ≠p, and the Hecke operators are rational; for GL_n this rational structure is unique.","The action of the excursion algebra on automorphic functions factors through this arithmetic version, so the structural results apply to the automorphic side."],"fun_headline_variants":["Excursion algebra collapses to semisimple locus","Contraction proves excursion algebra equals semisimple part","Frobenius-fixed semisimple locus captures global functions","Excursion algebra isomorphic to semisimple locus functions","Contracting local systems yields excursion algebra equality"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The load-bearing premise is the cited, unproved fact that every irreducible Weil perverse sheaf is a tensor product of an integral-weight Weil sheaf and a rank-one Weil sheaf with Weil-number Frobenius eigenvalues; if any irreducible Weil sheaf escapes this shape, the contraction construction and the main isomorphism do not follow.","fun_headline_variants_meta":{"raw":{"variants":["Excursion algebra collapses to semisimple locus","Contraction proves excursion algebra equals semisimple part","Frobenius-fixed semisimple locus captures global functions","Excursion algebra isomorphic to semisimple locus functions","Contracting local systems yields excursion algebra equality"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000184,"raw_usage":{"total_tokens":1065,"prompt_tokens":566,"completion_tokens":499,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":310,"completion_tokens_details":{"reasoning_tokens":423}},"tokens_in":310,"tokens_out":499,"duration_ms":5263,"temperature":1.0,"reasoning_tokens":423,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-03T00:10:29.010817+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A test is to look for an irreducible lisse Weil sheaf on a smooth curve over a finite field whose Frobenius eigenvalues, at some closed point, cannot be written as a Weil number times a single constant independent of the point. Such a sheaf would violate the factorization input used to build the contraction, and would thereby falsify Theorem 3.7.2's conclusion that global functions on the arithmetic stack equal those on the semisimple locus. A direct stack-theoretic check would be to compute Γ(LS^arithm_G(X), O) on a Frobenius-fixed component with a nontrivial unipotent part and see whether fu","supporting_citations":[],"review_version":1}