{"id":"18770446-c6b6-4488-937d-9c76a51e3359","arxiv_id":"2505.17422","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"For principal and mediate series of p-adic G2, the Aubert-Zelevinsky duality matches the Kato involution on affine Hecke algebra modules, confirming several cases of the Bernstein unitarity conjecture.","lead":"This paper computes how the Aubert-Zelevinsky duality functor acts on principal and intermediate series representations of the p-adic group G2, and matches the results with an involution on the associated Hecke algebras. It also verifies new cases of the Bernstein conjecture that duality preserves unitarity for G2.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The paper's Hecke-side conclusions hinge on unproved transcriptions of [AX23]/[Ram03] indexing triples; checking the Kazhdan-Lusztig condition for each quoted triple would settle whether misquotations undermine the central claim.","rationale":"Good-faith reading: the paper is a case-by-case computation following [Mui97] and [AX23]; the representation-side derivations (e.g., Propositions 3.9, 3.11, 3.15, 3.17, 3.21, 3.23, 3.28, 3.30) are long and I did not find a specific internal algebraic error. The main results, however, are not self-contained: the Hecke-side conclusions in Propositions 3.12, 3.18, 3.24, 3.25, 3.31, and 3.32 are all obtained by reading the representation-side objects off [AX23] tables and the Hecke-module structure off [Ram03] tables. Because [AX23] is a preprint (arXiv:2208.12391) and the paper does not define the labels t_a,...,t_g or the nilpotent-orbit notation e_..., the central claim inherits any error in those sources. The self-referential note in the acknowledgements about an ambiguity in Proposition 3.28 and the correction of a typo in [Mui97] show the author is attentive, but also underscore that the external-indexing step needs independent checking. A concrete, finite check is to verify the defining Kazhdan-Lusztig relation Ad(s)n = qn for every quoted triple; this is independent of [AX23] and would catch transcription errors. The Section 4 'Proof' of Conjecture 4.1 verifies cases rather than proving the conjecture, and one non-unitarity argument cites Proposition 4.1(3) in a direction that does not logically apply; however, that is a minor, repairable flaw relative to the triple-identification dependence. The reader's CONDITIONAL verdict is therefore appropriate; no verdict change is needed from this pass.","tokens_in":30390,"tokens_out":15724,"duration_ms":120337,"concrete_test":"For each indexing triple (t_*, e_*, rho) appearing in Tables 1 to 5, reconstruct the semisimple element s and nilpotent element n from the definitions in [Ram03] and verify the Kazhdan-Lusztig condition Ad(s)n = qn, and check that rho is an irreducible representation of the component group A(s,n). This is a finite computation requiring no input from [AX23], and it would detect any misquotation or misindexing in the paper's transcription of Ram's tables; if all quoted triples satisfy the condition, the Hecke-side labels are at least internally valid, isolating the remaining risk to the [AX23] representation-to-module correspondence.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim of the paper, that the Aubert-Zelevinsky involution on Bernstein blocks of G2 matches the Kato involution on the associated Hecke algebras, is established only by translating the representation-side computations through the explicit local Langlands correspondence of [AX23] and the Kazhdan-Lusztig/Ram indexing of irreducible Hecke modules by triples (s,n,rho). These identifications are quoted from tables in Sections 3.3 to 3.8 (e.g., Propositions 3.12, 3.18, 3.24, 3.25, 3.31, 3.32) rather than proved, and the labels t_a,...,t_g, e_alpha, e_beta, p21, and so on are never defined in the paper, so a reader cannot independently audit the transcription. If any triple is misindexed or the correspondence in [AX23] is incorrect for these blocks, the stated identities D_Hs([M_{t,*,*,*}]) = [M_{t,0,1}] do not follow from the internally consistent Jacquet-module computations. The paper also assumes, without stating a theorem, that the Bernstein equivalence Rep_s(G2) = H_s-mod intertwines DG2 with the Hecke-side duality D_Hs of Section 3.1.3; if this compatibility is not a known theorem, the 'deduction' of the Hecke-side involution is circular. Section 4 labels a case verification as a 'Proof' of Conjecture 4.1; moreover, its non-unitarity claim for I_alpha(delta(nu^{pm 1/2} xi_2)) cites Proposition 4.1(3), which only states that unitarity of the inducing representation implies unitarity of the induction, not the converse needed there; this gap is repairable via central characters but as written weakens the Bernstein-conjecture verification. The load-bearing issue remains the external triple identifications.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper computes the Aubert-Zelevinsky duality functor D_G on several Bernstein blocks of the p-adic exceptional group G2, including the two maximal-parabolic intermediate blocks and the principal series blocks corresponding to various choices of ramified/unramified and quadratic/cubic characters. For each block, the author determines D_G on the irreducible subquotients through explicit Jacquet-module computations based on the geometric lemma, following the method of Muić. The representation-side results are then matched, via the explicit local Langlands correspondence of Aubert–Xu [AX23] and the Kazhdan–Lusztig/Ram indexing of standard modules, with the Kato involution D_Hs on the associated affine Hecke algebra modules. The paper closes with a verification of several instances of the Bernstein conjecture on unitarity under Aubert–Zelevinsky duality for G2.","tokens_in":30717,"tokens_out":8795,"duration_ms":65900,"significance":"If the statements are correct, the paper provides a useful set of explicit, reproducible computations confirming the compatibility of Aubert–Zelevinsky duality with the Kato involution for a range of Bernstein blocks of G2, and gives new evidence for the Bernstein conjecture. The Jacquet-module computations are detailed and are cross-checked through Grothendieck-group equations such as (3.26)–(3.28), (3.41), (3.56), and (3.71); the author also carefully flags a typo in a result of Muić (Remark after Proposition 3.28). The main limitations are that the Hecke-side identifications are quoted rather than proved, the indexing-triple labels are not defined in the paper, and one non-unitarity argument in Section 4 rests on a misapplied converse. These issues are local and repairable.","major_comments":[{"comment":"The paper never states or proves that the Bernstein equivalence Rep_s(G2) ≅ H_s-mod introduced in Section 2.5 intertwines the Aubert–Zelevinsky duality D_G with the Hecke-side Kato involution D_Hs defined in Section 3.1.3. The Hecke-side identities in the cited propositions are presented as consequences of the G2-side computations, but they follow only if such a compatibility is a known theorem or is proved here. Please cite a precise reference for this compatibility or add a proof for the blocks considered; otherwise the 'deduction' of the Hecke involution is not established.","section":"§2.5, §3.1.3, and Propositions 3.12, 3.18, 3.24, 3.25, 3.31, 3.32"},{"comment":"The standard-module labels t_a, t_b, t_c, t_d, t_e, t_g, e_α_, e_β_, p21, p3, and similar expressions are used throughout the tables and propositions without being defined. Section 2.5 recalls the general notion of an indexing triple (s,n,ρ) from [KL87], but the specific labels are neither introduced nor keyed to the tables in [Ram03] and [AX23]. Because the central comparison depends on the correct transcription of each triple, please add a table defining all labels used, with explicit references to the relevant sources.","section":"§2.5 and §§3.4–3.8"},{"comment":"In the proof of Conjecture 4.1, the claim that Iα(δ(ν^{±1/2}ξ2)) and its dual Iα(ν^{±1/2}ξ2∘det) are not unitarizable is attributed to Proposition 4.1(3). That proposition states only that unitarity of the inducing representation implies unitarity of the induced representation; it does not give the converse needed here. Since this is an essential step in the claimed verification of the Bernstein conjecture, please replace the citation with a correct argument (e.g., using the central character, non-Hermiticity, or a suitable result from [Mui97]).","section":"§4, non-unitarity paragraph"}],"minor_comments":[{"comment":"The phrase 'mediate series' should be 'intermediate series'.","section":"Abstract and Introduction"},{"comment":"The label 'Mte,eα_+eα_+2β_,p21q' contains a repeated 'eα_'; please clarify whether this is a typo for 'Mte,eα_+2β_,p21q' (the same issue appears for 'Mte,eα_+eα_+2β_,p3q').","section":"Eq. (3.29), Prop. 3.12, Table 5"},{"comment":"In the definition of the convolution product, 'ϕi, ϕ2' should be 'ϕ1, ϕ2'.","section":"Definition 2.1"},{"comment":"The notation 'rϕ' in the displayed formula for the Jacquet module of π(χ) is undefined; it should presumably be r_T or r_H, and should be corrected.","section":"Proposition 3.21(1)"},{"comment":"The phrase 'all reduces' should be 'all reduce' (e.g., the sentence before Corollary 3.7).","section":"§3.6.1 and §3.7"}],"recommendation":"major_revision","confidential_remarks":"The manuscript relies heavily on the unpublished preprint [AX23] for the explicit local Langlands correspondence and for the tables of standard modules; the editor may wish to confirm that the relevant results of [AX23] are stable. The contribution is a careful case-by-case computation, likely of interest to specialists, provided the compatibility theorem for the Bernstein equivalence and the labeling definitions are added, and the Section 4 non-unitarity argument is corrected."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThis is a chapter-length worked example. For several Bernstein blocks of p-adic G2, the paper computes the Aubert-Zelevinsky involution on the Grothendieck group by explicit Jacquet-module calculations, then matches the result with Kato's involution on the associated affine Hecke algebra modules using Kazhdan-Lusztig indexing triples from Aubert-Xu and Ram. The genuinely new part is the matching computation itself and the resulting case checks of the Bernstein unitarity conjecture for G2. Muic already did much of the structural work, and the paper says so plainly, attributing Propositions 3.9, 3.21, and 3.28 to him. That is honest.\n\nThe representation-side computations are internally consistent and detailed. Equations (3.26)-(3.28), (3.41), (3.56), and (3.71) show real work, not hand-waving. The paper also acknowledges an ambiguity in the original proof of Proposition 3.28 and flags a typo in Muic's Proposition 4.2. That is the right way to handle inherited results.\n\nThe soft spot is exactly where the reader put it: the Hecke-side identifications are quoted, not proved. Propositions 3.12, 3.18, 3.24, 3.25, 3.31, and 3.32 all depend on the triple transcription from [AX23] and Ram's tables, and a reader cannot audit labels like t_a...t_g, e_alpha, or p21 without those sources. Since [AX23] is a preprint, the paper's central claim is conditional on an external unverified list. I would not call it circularity: the G2-side computations stand on their own, and Kato's involution is defined independently. What is missing is an explicit theorem statement or reference that the Bernstein equivalence intertwines D_G with the Hecke-side involution; the paper uses that compatibility implicitly. That needs to be added.\n\nSection 4 has a real but repairable gap. It labels a case verification as a \"Proof\" of the Bernstein conjecture, and the non-unitarity claim for I_alpha(delta(nu^{pm 1/2} xi_2)) cites Proposition 4.1(3), which only gives unitarity of induction from unitarity of the inducing representation, not the converse. The intended argument can be repaired via central characters, but as written the section overclaims.\n\nThis paper is for specialists in p-adic representation theory who want concrete matching examples for G2 and for anyone working on explicit LLC for exceptional groups. It deserves a serious referee. I would send it to review and ask for three things: cite or prove the compatibility theorem, point precisely to the [AX23] table entries for each triple, and fix the Section 4 argument.","headline":"A careful, honestly attributed set of explicit Aubert-Zelevinsky duality computations for G2 Bernstein blocks; the Hecke-side matching rests on quoted [AX23] tables, but the case checks are real and the paper deserves refereeing.","tokens_in":31318,"tokens_out":2145,"would_cite":true,"duration_ms":17642,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["22E50","20C08"],"pacs":[],"model":"deepseek-v4-flash","headline":"The Aubert–Zelevinsky duality for the p-adic group G2, computed on the principal and intermediate series, coincides with the Kato involution on the associated affine Hecke algebras, verifying several cases of the Bernstein conjecture.","keywords":["Aubert-Zelevinsky duality","G2","affine Hecke algebras","Bernstein blocks","Kazhdan-Lusztig triples","Bernstein conjecture","unitarizability","p-adic reductive groups"],"falsifier":"Compute $D_{G2}(\\pi(\\chi))$ for one of the tabulated cases, for example the ramified-cubic case with $s=1/2$ and $\\chi^3=1$, directly from the alternating-sum formula (2.1) using only the Jacquet modules listed in Propositions 3.22 and 3.29, and check whether the result matches the Hecke-side module $M_{t,0,1}$ predicted by the table; a mismatch in any single case would disprove the claimed compatibility.","tokens_in":30124,"feed_emoji":"🧮","tokens_out":8082,"duration_ms":58632,"temperature":0.7,"pith_summary":"This paper tries to establish that a known duality on representations of the p-adic exceptional group G2—the Aubert–Zelevinsky involution—is the same operation, in the principal and intermediate-series Bernstein blocks, as the Kato involution on the associated affine Hecke algebras. The author computes the duality explicitly on both sides, case by case, using parabolic induction and Jacquet restriction on the representation side and Kazhdan–Lusztig indexing triples on the Hecke side. If the computations are correct, the paper confirms several instances of the Bernstein conjecture for G2: duality preserves unitarizability in the listed cases. A sympathetic reader should care because the result gives concrete evidence for the general principle that such dualities on p-adic groups are visible through their Hecke algebras.","feed_headline":"Aubert–Zelevinsky duality matches Hecke involution for G2","feed_subtitle":"Explicit computations on principal and intermediate series confirm several cases of the Bernstein conjecture for G2.","key_machinery":"The central object on the representation side is the Aubert–Zelevinsky duality functor $D_G$ on the Grothendieck group, defined as the alternating sum over subsets of simple roots of $i_{P_I} \\circ r_{P_I}$; it is an involution that exchanges parabolic induction and Jacquet restriction. On the Hecke side, the matching mechanism is the Kato involution: Theorem 3.3 defines a twisted action $h \\mapsto h^*$ on the extended affine Hecke algebra, and $D[M] = [M^*]$ in the Grothendieck group of finite-dimensional modules. The bridge between the two sides is the indexing of irreducible Hecke modules by Kazhdan–Lusztig triples $(t,e,\\rho)$, where $t$ is semisimple, $e$ is nilpotent with $\\mathrm{Ad}(t)e = qe$, and $\\rho$ is a representation of the component group; the standard modules $M_{t,e,\\rho}$ are extracted from the equivariant K-theory of the variety of Borel subgroups containing $t$ and the corresponding unipotent element.","core_discovery":"The paper computes the Aubert–Zelevinsky duality functor on the Bernstein blocks of G2 labelled by the principal series and the two intermediate series (whose Levi factors are isomorphic to GL2(F)), and shows that in every computed case the representation-side duality D_{G2} matches the Kato involution on the associated Hecke algebra H(J_s,1). Under the indexing by Kazhdan–Lusztig triples (t,e,ρ), the standard module M_{t,e,ρ} is sent to the corresponding module M_{t,0,ρ}, with explicit identities listed in Propositions 3.12–3.32. Using these pairs together with earlier unitarity results, the paper verifies several cases of the Bernstein conjecture: each unitarizable representation in the listed blocks has a unitarizable Aubert–Zelevinsky dual, and each non-unitarizable one has a non-unitarizable dual.","pith_inferences":["One implication the author leaves implicit is that if the compatibility extends to the remaining Bernstein blocks of G2, the Aubert–Zelevinsky involution for the whole group would be determined by affine Hecke algebra data, making the Bernstein conjecture a finite table-check for each block.","A testable extension would be to carry out the same case-by-case strategy for other exceptional p-adic groups once the explicit local Langlands correspondence and triple tables are available; the structure of the argument does not use anything special to G2 beyond the tables.","The pairing of standard modules under the Hecke involution suggests a direct test of unitarity: for the blocks studied, checking whether a module carries a definite Hermitian form amounts to verifying positivity of the corresponding standard module data, which could be automated."],"forward_implications":["For each Bernstein block treated, the representation-theoretic Aubert–Zelevinsky involution is compatible with the Kato involution on the corresponding affine Hecke algebra.","The Bernstein conjecture holds for all the listed representations: unitarizable representations have unitarizable duals, and non-unitarizable ones have non-unitarizable duals.","The tables give explicit identifications of standard modules of $H(J_s,1)$ with the irreducible subrepresentations and Langlands quotients of parabolic inductions, such as $\\pi(\\chi) \\leftrightarrow M_{t,e,1}$ and the corresponding quotients with $M_{t,0,1}$.","The duality computations reproduce known facts, such as $D_{G2}(\\mathrm{St}_{G2}) = 1_{G2}$, serving as a check on the method."],"supporting_citations":[{"why":"Supplies the explicit local Langlands correspondence for G2 and the tables of indexing triples and standard modules used throughout Sections 3.4–3.8.","marker":"[AX23]"},{"why":"Identifies the Hecke algebras $H(J_s,1)$ attached to the Bernstein blocks and the four possible series for the Levi $L_\\beta$.","marker":"[AX24]"},{"why":"Provides the unitarity results, the reducibility criteria, and the Jacquet-module computations on which the representation-side cases are built.","marker":"[Mui97]"},{"why":"Establishes the indexing of standard modules by triples $(t,e,\\rho)$ via equivariant K-theory, the language in which the Hecke-side results are stated.","marker":"[KL87]"},{"why":"Gives the explicit tables of rank-two affine Hecke algebra modules indexed by triples, from which the identities such as $M_{t,e,1} \\to M_{t,0,1}$ are read.","marker":"[Ram03]"},{"why":"Defines the Aubert–Zelevinsky duality and proves the involution properties used throughout the computations.","marker":"[Aub95]"},{"why":"Constructs the types for principal series and the isomorphism between $H(G,\\rho)$ and $H(J_s,1)$.","marker":"[Roc98]"},{"why":"Proves that the Hecke-side duality $D[M]$ equals the Kato twisted action $[M^*]$, the exact statement matched with the representation side.","marker":"[Kat93]"}],"fun_headline_variants":["G2 duality matches Hecke involution in computed blocks","Bernstein conjecture cases confirmed by G2 duality computations","Aubert-Zelevinsky and Kato involution align for G2 blocks","Explicit G2 computations link representation duality to Hecke involution","AZ duality equals Hecke involution for G2 principal and intermediate series"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing assumption is that the explicit local Langlands correspondence for G2 and the Kazhdan–Lusztig indexing tables, quoted from the preprint and earlier work, are correct and complete for the blocks treated; if a triple is mislabeled or a block is missing, the Hecke-side identities do not follow from the representation-side computations.","fun_headline_variants_meta":{"raw":{"variants":["G2 duality matches Hecke involution in computed blocks","Bernstein conjecture cases confirmed by G2 duality computations","Aubert-Zelevinsky and Kato involution align for G2 blocks","Explicit G2 computations link representation duality to Hecke involution","AZ duality equals Hecke involution for G2 principal and intermediate series"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000665,"raw_usage":{"total_tokens":2970,"prompt_tokens":811,"completion_tokens":2159,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":427,"completion_tokens_details":{"reasoning_tokens":2069}},"tokens_in":427,"tokens_out":2159,"duration_ms":13222,"temperature":1.0,"reasoning_tokens":2069,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-07T14:46:58.632222+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute $D_{G2}(\\pi(\\chi))$ for one of the tabulated cases, for example the ramified-cubic case with $s=1/2$ and $\\chi^3=1$, directly from the alternating-sum formula (2.1) using only the Jacquet modules listed in Propositions 3.22 and 3.29, and check whether the result matches the Hecke-side module $M_{t,0,1}$ predicted by the table; a mismatch in any single case would disprove the claimed compatibility.","supporting_citations":[],"review_version":1}