{"id":"67c64a43-4987-4654-a3b2-0973511a1bf5","arxiv_id":"2505.09991","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"For p-adic split SO_{2n+1} and Sp_{2n}, irreducible representations of good parity are unitary if and only if they are of Arthur type.","lead":"For p-adic split orthogonal and symplectic groups, the paper proves that every representation of good parity is unitary exactly when it comes from an Arthur parameter. The result supplies an algorithm to test unitarity and settles a conjecture of Tadic for these groups.","discovery_kind":"unification","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Proposition 3.5's use of [5, Theorem 1.1] is the point where unitarity is converted into Arthur-type structure; the paper neither states nor verifies the hypotheses of that theorem for the data in Proposition 3.5(2), notably for the Aubert-dual range B < 0.","rationale":"The reader's weakest_assumption already identifies [5, Theorem 1.1] as the step where unitarity is converted into Arthur-type structure, and my reading agrees. The central claim is otherwise supported by a long, carefully structured argument that builds on Arthur's and Mœglin's published classifications and on the first author's published algorithms [4, 6]. Theorem 3.13 and Lemma 3.7 are technical but are argued in the text; Remark 3.6 honestly documents a corrected false expectation, which is evidence of care rather than of a hidden flaw. The external dependency on [5] is published and peer-reviewed, so it is not a soundness failure by itself, but in this paper it is used as a black box at the exact load-bearing junction: without the explicit extended multi-segment form of every socle summand, Lemma 3.11 has no input and Proposition 3.5 cannot conclude that π is Arthur type. The paper should state the hypotheses of [5, Theorem 1.1] and verify them for the data of Proposition 3.5, including the range B < 0 that the proposition explicitly allows. This is a request for clarification and verification, not a demonstration of an error, so the appropriate verdict remains the reader's CONDITIONAL and I recommend no change to that verdict.","tokens_in":36310,"tokens_out":10069,"duration_ms":98711,"concrete_test":"Obtain the statement of [5, Theorem 1.1] and check it verbatim against Proposition 3.5(2). In particular verify: (a) whether [5, Theorem 1.1] requires A+B ≥ 0 and, if so, whether the Aubert-duality reduction in Section 3.6 explicitly routes every B < 0 case back to Proposition 3.5(1); (b) whether its conclusion is 'any irreducible summand has the displayed extended multi-segment' or only 'the socle is a single Arthur-type representation'; and (c) whether the admissible order asserted in Section 3.6 for E'' coincides with the order produced by [5, Theorem 1.1]. If any of these checks fails, attempt to prove Lemma 3.7 directly from [5, Theorem 1.1]'s proof; if the repair is not immediate, the conditional verdict is justified.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Theorem 1.1 (Corollary 3.2) is proved by the induction in Theorem 3.3, whose engine is Proposition 3.5. In Section 3.6 the proof of Proposition 3.5(2) invokes [5, Theorem 1.1] as a black box: every irreducible summand π'' of soc(soc(S' × τ × S) ⋊ σ) is of Arthur type and has the explicit extended multi-segment form E'' = E ∪ {([A,B]_ρ, l, η_i), ([A+t_i, B+t_i]_ρ, l, η'_i)} with the stated admissible order. This is the single step where the common-subrepresentation information extracted by Lemma 3.7 is upgraded to Arthur-type structure; Lemma 3.11 and Proposition 3.12 then operate only on that explicit form. If [5, Theorem 1.1] does not apply verbatim, the proof of Proposition 3.5 collapses and with it the induction proving Theorem 3.3. The paper gives no statement of the hypotheses of [5, Theorem 1.1], so the reader cannot check that they match the present data: σ of Arthur type, τ = Z_ρ[B+1, A+1]^k, S = Speh(ρ, A+B+1, A-B+1)^k, S' as in Lemma 3.7(2), t_i ≫ 0, and the two derivative-vanishing conditions. A specific concern is that Proposition 3.5(2) allows A ≥ B > -1, so A+B can be negative, whereas Definition 3.9 requires A+B ≥ 0 for every extended segment of a good-parity Arthur representation; the text does not explain whether [5, Theorem 1.1] covers this range or whether the Aubert-duality reduction always returns to the A+B ≥ 0 case. Since [5] is the first author's own published work, this is not an accusation of error but a load-bearing unverified premise that should be made explicit and checked.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper proves that for F a p-adic field and G a split SO_{2n+1} or Sp_{2n}, an irreducible smooth representation of G of good parity is unitary if and only if it is of Arthur type (Theorem 1.1, Corollary 3.2). The authors prove the stronger Theorem 3.1: for any irreducible unitary representation π, the good-parity part π(φ_good, ε) is of Arthur type. The proof uses an SZ-decomposition of π, reducing to an induction step (Theorem 3.3) whose engine is Proposition 3.5 and a technical Jacquet-module inequality (Lemma 3.7); the base case is handled by Theorem 3.13 via Mœglin's construction. The paper also determines unitarity for representations of the form ×_i Sp(ρ_i,c_i,d_i)|·|^{x_i} ⋊ π_0 with π_0 of Arthur type of good parity and 0 ≤ x_i < 1/2 (Theorem 4.1), confirming a conjecture from [21], and discusses phenomena beyond this range.","tokens_in":36666,"tokens_out":8368,"duration_ms":73027,"significance":"If correct, Theorem 1.1 is a major advance: it gives a complete description of the good-parity part of the unitary dual for split SO_{2n+1} and Sp_{2n}, identifies it with local Arthur packets, and, combined with the algorithms of [6] and [22], yields an explicit unitarity test. The treatment is honest and careful: Remark 3.6 explicitly records a false initial expectation (with credit to Gurevich) and works around it, and the paper states a stronger result (Theorem 3.1) than the title promises. The proof is long and technically sophisticated, using ρ-derivatives and Mœglin's classification. The verification of the central claim, however, depends on an external theorem [5, Theorem 1.1] whose hypotheses are not checked in the paper; this is the main weakness.","major_comments":[{"comment":"In the proof of Proposition 3.5(2) (Section 3.6), the paper applies [5, Theorem 1.1] as a black box to conclude that every irreducible summand π'' of soc(soc(S' × τ × S) ⋊ σ) is of Arthur type and has the displayed extended multi-segment form. However, the hypotheses of [5, Theorem 1.1] are never stated, and the data in Proposition 3.5(2) allow A ≥ B > −1, so that A+B can be negative; Definition 3.9(3) requires A_i+B_i ≥ 0 for every extended segment. The text does not explain whether [5, Theorem 1.1] covers the range A+B < 0 or how the Aubert-duality reduction (asserted at the start of Section 3.6) returns to the non-negative case. This is the single step where the common-subrepresentation information from Lemma 3.7 is converted into Arthur-type structure, so the induction in Theorem 3.3 depends on it. Please state the theorem being cited, verify its hypotheses for the present data (including the case B < 0), or prove the needed variant.","section":"Section 3.6, proof of Proposition 3.5(2)"},{"comment":"Lemma 3.4(1) is stated with the comment 'the proof of (1) is similar' and no proof is given. Lemma 3.4 is used for the τ^- steps in the SZ-decomposition, and part (1) is an essential input to the induction behind Theorem 3.3. Since the proof of part (2) is long and relies on delicate Jacquet-module inequalities, 'similar' is not sufficient for the reader to verify the claim; please provide the full argument or a precise reduction to part (2).","section":"Lemma 3.4(1)"}],"minor_comments":[{"comment":"In the display 'DS(L(m|·|−1) × L(m|·|1)) = C2', the notation C2 is unexplained; it presumably denotes the constant two-dimensional representation, but it should be defined or replaced with a clearer symbol such as ℂ².","section":"Remark 3.6"},{"comment":"The sentence 'By the weak Ramanujan bound, which is know, it takes the form' contains a typo: 'know' should be 'known'.","section":"Section 5, first paragraph"},{"comment":"The phrase 'the unitary induction Sp(ρ,c,d)⋊π0 is irreducible' is misleading: 'unitary induction' is a method (Proposition 2.3(1)), not an object. The condition should read 'the parabolically induced representation Sp(ρ,c,d)⋊π0 is irreducible'.","section":"Theorem 4.1 and its proof"},{"comment":"The phrase 'τ^-_bad × Δ_{ρ_i}[x_i,y_i] is irreducible by [48, Proposition 8.6]' is ambiguous: it should clarify that this is the normalized parabolic induction of the product of the two representations, not a product in the Grothendieck group.","section":"Section 3.2"}],"recommendation":"major_revision","confidential_remarks":"The central claim is likely correct and the paper is a serious contribution, but the proof has a load-bearing gap: Proposition 3.5(2) applies [5, Theorem 1.1] without verifying its hypotheses for the range A+B<0. Since [5] is the first author's own published work, this is not a suspicion of error, but a rigor issue that must be resolved before publication. I recommend major revision with a request to either prove the needed variant, state the hypotheses and verify them, or restrict the statement to the case where A+B≥0 (which suffices for Lemma 3.4(2)) and explain the Aubert-duality reduction explicitly."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Dear [Colleague],\n\nThis is the real thing: Atobe and Minguez prove that for p-adic split SO_{2n+1} and Sp_{2n}, an irreducible good-parity representation is unitary if and only if it is of Arthur type. That settles the good-parity case of Tadic's unitary dual conjecture, gives an effective unitarity test via earlier algorithms, and pinpoints the local components of the discrete automorphic spectrum. Genuinely new, not a routine step.\n\nThe proof is structurally impressive. The SZ-decomposition breaks any good-parity representation into a chain of socle steps, and the key idea -- induce a unitary GL representation, use semisimplicity to force common subrepresentations, then read off Arthur-type structure from the Langlands data -- is coherent. The authors are also honest: Remark 3.6 records a false initial expectation supplied by Gurevich's counterexample, and they work around it. Lemma 3.7 is the technical engine; I didn't verify every inequality, but the surrounding argument is plausible and carefully laid out.\n\nSoft spots, in proportion. The most significant is that Proposition 3.5(2) invokes [5, Theorem 1.1] as a black box without stating its hypotheses. This is the step where the shared-subrepresentation information is upgraded to an explicit extended multi-segment, so it is load-bearing. I don't think this is an error -- the setup appears designed to match [5], and the cited paper is the first author's own published work -- but a referee should ask for a statement of the hypotheses and a check that they cover the Aubert-dual situation. The stress-test worry about A+B<0 looks like a non-issue, because the Speh representation S in the proposition is only defined when -A ≤ B ≤ A, so A+B ≥ 0 automatically; still, the written condition 'A ≥ B > -1' is sloppier than it should be. Second, Lemma 3.4(1) is dismissed with 'similar' -- probably fine, but it leaves the reader to fill in a symmetric argument. Third, Theorem 4.1 (the slightly-beyond result) relies on the unreviewed preprint [17]; that is a weaker dependency and should be flagged explicitly. The citation pattern is clean; the self-citations point to genuinely relevant earlier results.\n\nThis paper is for specialists in p-adic representation theory, and they will get substantial value from it. It deserves a serious referee, and I'd bring it to a reading group. My recommendation: send it out; ask the authors to state and verify the hypotheses of [5, Theorem 1.1] and to expand a few lines on Lemma 3.4(1). If those are satisfied, the result is solid.","headline":"Settles the good-parity unitary dual conjecture for p-adic split SO/Sp; the proof is dense and coherent, but the referee should verify the hypotheses of the [5, Theorem 1.1] black box and a couple of skipped details.","tokens_in":37283,"tokens_out":13910,"would_cite":true,"duration_ms":125858,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["22D10","22E50","11S37"],"pacs":[],"model":"deepseek-v4-flash","headline":"For p-adic split SO(2n+1) and Sp(2n), good-parity unitary representations are exactly the Arthur-type ones.","keywords":["unitary dual","p-adic classical groups","Arthur type representations","good parity","A-packets","parabolic induction","Jacquet modules","Speh representations"],"falsifier":"Run the known Arthur-type detection algorithm on every irreducible good-parity representation of a small-rank group such as $\\mathrm{Sp}_4(F)$ or $\\mathrm{SO}_5(F)$ and compare with a direct unitarity test: a good-parity representation that is unitary but fails the Arthur-type test would refute the main theorem.","tokens_in":36032,"feed_emoji":"📐","tokens_out":10021,"duration_ms":85323,"temperature":0.7,"pith_summary":"This paper claims that, for a $p$-adic field $F$ and $G$ either split $\\mathrm{SO}_{2n+1}(F)$ or $\\mathrm{Sp}_{2n}(F)$, an irreducible smooth representation of $G$ of 'good parity' is unitary exactly when it is of Arthur type. Good parity isolates the part of a representation's Langlands parameter that is self-dual of the same type as $G$, and Arthur type means the representation appears in an $A$-packet attached to the endoscopic classification of automorphic representations. The result settles a conjecture about the unitary dual that had been refined in recent work, and it makes the good-parity unitary dual computable: existing algorithms that test whether a representation is of Arthur type now double as unitarity tests. The paper also classifies unitarity slightly beyond good parity, for representations induced from unitary Speh representations with small positive exponents over an Arthur-type good-parity base.","feed_headline":"Unitary equals Arthur type for p-adic SO and Sp","feed_subtitle":"For split SO(2n+1) and Sp(2n), the good-parity unitary dual is decidable and matches the automorphic spectrum.","key_machinery":"The carrying object is the SZ-decomposition, which writes any irreducible representation $\\pi$ as the socle of a parabolic induction $\\tau_r \\times \\cdots \\times \\tau_1 \\rtimes \\pi_0$, where each $\\tau_i$ is a product of copies of a segment representation and $\\pi_0$ satisfies an 'almost no derivatives' condition: its $\\rho$-derivatives vanish unless the exponent lies in $\\{0, \\tfrac12\\}$. The proof runs an induction along the sequence of socles $\\pi_i = \\mathrm{soc}(\\tau_i \\rtimes \\pi_{i-1})$. The key mechanism is a shared-subrepresentation argument: unitarity of the whole representation forces $S \\rtimes \\pi_i$ and $\\mathrm{soc}(\\tau_i \\times S) \\rtimes \\pi_{i-1}$ to have a common irreducible subrepresentation for a suitably chosen unitary Speh representation $S$; a Jacquet-module inequality then transfers Arthur-type structure from $\\pi_{i-1}$ to $\\pi_i$. The base case, where no derivatives occur outside $\\{0, \\tfrac12\\}$, is handled by the explicit construction of $A$-packets in terms of extended multi-segments, which the paper uses to show that the base representation is of Arthur type.","core_discovery":"The central discovery is a two-way identification: for irreducible representations of good parity of split $\\mathrm{SO}_{2n+1}(F)$ and $\\mathrm{Sp}_{2n}(F)$, being unitary and being of Arthur type are the same property. The forward direction is the established theorem that $A$-packets consist of unitary representations. The reverse direction is proved by showing that the good-parity part of any irreducible unitary representation is of Arthur type (Theorem 3.1); this is obtained by an inductive argument through a canonical decomposition of the representation into a socle of parabolically induced pieces, in which each step preserves Arthur type under the hypothesis of unitarity. A direct corollary is that the good-parity part of the unitary dual coincides with the set of local components of discrete automorphic representations in that case.","pith_inferences":["If the main theorem is correct, the full unitary dual now splits into a known good-parity part and a residual bad-parity part; the bad-parity behaviour is governed by complementary series attached to Speh representations, and a complete classification would need analytic control of first reducibility points, as the paper itself notes.","The shared-subrepresentation technique, where unitarity forces a common irreducible subrepresentation of two differently induced representations, looks transferable; a similar mechanism may work for real groups or for quasi-split unitary groups once the corresponding $A$-packet construction is available.","Theorem 4.1 suggests a simple combinatorial rule for unitarity just beyond good parity: matching of exponents between $\\rho$ and $\\rho^\\vee$ and a parity condition on packet sizes. A testable extension would be to verify the same rule for larger exponents $x_i$ or for other classical groups.","The example discussed in Remark 3.6 and Section 5.2 shows that Langlands data alone do not make unitarity visible; if the main theorem is right, such unitary representations are still Arthur type, so their extended multi-segment description carries the unitarity information that the Langlands data hide."],"forward_implications":["Every irreducible unitary representation of good parity of split $\\mathrm{SO}_{2n+1}(F)$ or $\\mathrm{Sp}_{2n}(F)$ is a local component of a discrete automorphic representation, and conversely such local components are unitary.","Unitarity of a given good-parity representation becomes algorithmically decidable: apply an Arthur-type test; the answer coincides with unitarity.","For any irreducible unitary representation (not necessarily good parity), its good-parity part is of Arthur type, so the remaining unknown in the full unitary dual is concentrated in the bad-parity part.","Just beyond good parity, for representations of the form $\\times_i \\mathrm{Sp}(\\rho_i,c_i,d_i)|\\cdot|^{x_i} \\rtimes \\pi_0$ with $0 \\le x_i < \\tfrac12$ and $\\pi_0$ Arthur-type of good parity, unitarity is characterized by matching exponents for non-self-dual $\\rho_i$ and by irreducibility of $\\mathrm{Sp}(\\rho,c,d) \\rtimes \\pi_0$ when the relevant index set has odd size.","The good-parity result is conjectured to extend to all quasi-split classical groups once the underlying classification of Arthur-type representations is extended accordingly."],"supporting_citations":[{"why":"The endoscopic classification of A-parameters: supplies the A-packets, the unitarity of Arthur-type representations, and the intertwining-operator formalism used in Section 4.","marker":"[2]"},{"why":"Construction of local A-packets via extended multi-segments: gives Theorem 3.10, the classification of Arthur-type good-parity representations used throughout.","marker":"[4]"},{"why":"Theorem 1.1 on socles of certain parabolic inductions: the black box giving Arthur-type summands with explicit extended multi-segments in Proposition 3.5.","marker":"[5]"},{"why":"Algorithm for deciding whether a representation is of Arthur type: used in Section 3.6 to conclude that the target representation is Arthur type and later to turn unitarity into an explicit test.","marker":"[6]"},{"why":"Explicit Zelevinsky–Aubert duality and derivative constraints: used in the SZ-decomposition and in Theorem 3.13, including the ruling-out of irreducibility for certain inductions.","marker":"[9]"},{"why":"Irreducibility criterion for parabolic induction of essentially Speh representations over Arthur-type pieces: used in Section 4 to prove and apply Theorem 4.1.","marker":"[17]"},{"why":"The refined conjecture on the unitary dual and Arthur type, together with the example and Conjecture 5.9 that Theorem 4.1 confirms.","marker":"[21]"},{"why":"An independent algorithm for detecting Arthur type representations: combined with [6] it gives the explicit unitarity check promised in the abstract.","marker":"[22]"},{"why":"Parabolic induction theory for general linear groups via derivatives: supplies the socle irreducibility and related facts used throughout Lemma 3.4 and Section 3.4.","marker":"[25]"}],"fun_headline_variants":["Good parity unitary = Arthur type for p-adic SO/Sp","Unitary dual decidable for good parity p-adic SO/Sp","p-adic SO and Sp: unitarity iff Arthur type","Arthur type characterizes unitarity for p-adic SO/Sp","Good parity: unitarity iff Arthur type for SO/Sp"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is a previously proved classification result, used as a black box, that certain induced representations built from an Arthur-type representation have only Arthur-type summands of an explicitly described shape. The paper does not reprove it; if that result were false, the induction from unitarity to Arthur type would not go through.","fun_headline_variants_meta":{"raw":{"variants":["Good parity unitary = Arthur type for p-adic SO/Sp","Unitary dual decidable for good parity p-adic SO/Sp","p-adic SO and Sp: unitarity iff Arthur type","Arthur type characterizes unitarity for p-adic SO/Sp","Good parity: unitarity iff Arthur type for SO/Sp"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000533,"raw_usage":{"total_tokens":2525,"prompt_tokens":864,"completion_tokens":1661,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":480,"completion_tokens_details":{"reasoning_tokens":1572}},"tokens_in":480,"tokens_out":1661,"duration_ms":12175,"temperature":1.0,"reasoning_tokens":1572,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T21:19:34.625155+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the known Arthur-type detection algorithm on every irreducible good-parity representation of a small-rank group such as $\\mathrm{Sp}_4(F)$ or $\\mathrm{SO}_5(F)$ and compare with a direct unitarity test: a good-parity representation that is unitary but fails the Arthur-type test would refute the main theorem.","supporting_citations":[{"cited_title":"Arthur, The endoscopic classification of representations","cited_arxiv_id":null,"evidence_quote":"The endoscopic classification of A-parameters: supplies the A-packets, the unitarity of Arthur-type representations, and the intertwining-operator formalism used in Section 4."},{"cited_title":"Atobe, Construction of local A-packets","cited_arxiv_id":null,"evidence_quote":"Construction of local A-packets via extended multi-segments: gives Theorem 3.10, the classification of Arthur-type good-parity representations used throughout."},{"cited_title":"Atobe, On the socles of certain parabolically induced representat ions of p-adic classical groups","cited_arxiv_id":null,"evidence_quote":"Theorem 1.1 on socles of certain parabolic inductions: the black box giving Arthur-type summands with explicit extended multi-segments in Proposition 3.5."},{"cited_title":"Atobe, The set of local A-packets containing a given representation","cited_arxiv_id":null,"evidence_quote":"Algorithm for deciding whether a representation is of Arthur type: used in Section 3.6 to conclude that the target representation is Arthur type and later to turn unitarity into an explicit test."},{"cited_title":"Atobe and A","cited_arxiv_id":null,"evidence_quote":"Explicit Zelevinsky–Aubert duality and derivative constraints: used in the SZ-decomposition and in Theorem 3.13, including the ruling-out of irreducibility for certain inductions."},{"cited_title":"Lapid and A","cited_arxiv_id":null,"evidence_quote":"Parabolic induction theory for general linear groups via derivatives: supplies the socle irreducibility and related facts used throughout Lemma 3.4 and Section 3.4."}],"review_version":1}