{"id":"d7b5c954-ce6d-42f2-bc99-6f35544a37c3","arxiv_id":"2505.11381","paper_version":3,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"The paper proves the characterization of complementary Arthur representations for p-adic Sp_2n and split SO_{2n+1}: unitarity fails exactly when a reducible Speh factor appears an odd number of times.","lead":"This paper proves a conjecture that describes exactly which complementary Arthur representations of p-adic symplectic and split odd orthogonal groups are unitary. The result is a step toward a full classification of the unitary dual and gives new constraints on local pieces of automorphic forms on GL_N.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Lemma 4.10(ii)/(iii), Lemma 4.17, and Lemma 4.19 are asserted with essential case checks omitted; Proposition 4.20 — hence Theorems 6.6, 6.8 and the non-unitarity half of Theorem 6.3 — rests on them. The proof is conditional until these combinatorial statements are independently verified.","rationale":"The reader's weakest assumption is the same as mine: the combinatorial core of Section 4 is only partially verified. I agree with that assessment. I did not find an actual contradiction or a circular argument; the surrounding representation-theoretic framework, including Mœglin's and Atobe's packet construction, Arthur's local intertwining relation, and Tadić's unitarity criteria, is applied coherently. The main theorem is also corroborated by the concurrent Atobe-Minguez preprint, which the authors disclose honestly, but that does not remove the need to verify the proof in this paper. The central concern is therefore not that the theorem is likely false, but that the submitted proof of the non-unitarity direction is not fully established because Proposition 4.20 and the lemmas feeding into it contain multiple explicitly omitted case checks. Those checks are concrete and bounded, so the appropriate verdict remains CONDITIONAL; no change from the reader's verdict is needed.","tokens_in":38475,"tokens_out":6920,"duration_ms":70769,"concrete_test":"Implement Definitions 4.3, 4.7, 4.9, and 4.11 verbatim and exhaustively enumerate all extended Z-segments with supports contained in, say, [-6,6] and all valid l and eta. Check every admissible pair and triple for Lemma 4.10(i)-(iii), Lemma 4.13(2), and Lemma 4.19(1)-(3), and every interval pair satisfying the hypotheses of Proposition 4.20(2). A single counterexample would invalidate Theorems 6.6 and 6.8; if none is found, the omitted cases are likely routine and the conditional verdict can be upgraded.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Section 4 is the load-bearing combinatorial engine for the non-unitarity direction of Theorem 6.3, and its core statements are not fully proved in the manuscript. Lemma 4.10(ii) (adjacency is preserved under NV(e,−) and NV(−,e)) is dismissed in every case as a case-by-case straightforward computation which is omitted, and the whole verification for NV(−,e2) is said to be omitted. Lemma 4.10(iii) ends with \"We omit the rest of the verification\" in Case (c). Lemma 4.17's most difficult case ends with a direct computation which is omitted, and Lemma 4.19(1)-(3) are all deferred to direct computations. Proposition 4.20, which is the key reduction statement, uses these results, and Proposition 4.20(2) is invoked by Theorem 6.8. Theorem 6.8 is in turn essential to the induction step that proves non-unitarity in Theorem 6.3, via Corollary 3.3. If any of the omitted interval or adjacency assertions fails, the induction step loses its proof; the text itself marks these cases as unverified, and no external verification or machine-checked proof is supplied. This does not show the theorem is false, but it makes the submitted proof of the harder direction conditional.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proves an explicit characterization of the unitary complementary Arthur representations for split symplectic groups Sp_2n(F) and split odd special orthogonal groups SO_{2n+1}(F) over a non-Archimedean local field F of characteristic zero. The main theorem (Theorem 6.3) states that a representation in the enlarged Arthur-type set Π_{A+}(G_n) is unitary if and only if it lies in the set Σ_{A+,u}(G_n), i.e. the multiplicities of certain generalized Speh factors whose induced representation is reducible are even. The proof combines Arthur's local intertwining relation, Mœglin's construction of Arthur packets as reformulated by Atobe, a non-unitarity criterion of Muić and Tadić, and a new combinatorial theory of intervals and adjacency on extended Z-segments developed in Section 4. As applications, the paper derives constraints on the local components of irreducible self-dual cuspidal automorphic representations of GL_N, with concrete consequences for N=2,3,4.","tokens_in":38749,"tokens_out":3435,"duration_ms":33792,"significance":"If the proof is correct, the paper proves Conjecture 1.2 from the authors' program [HJLLZ24], giving a precise description of Π_{A+,u}(G_n) for the two largest families of split classical groups. This is a substantive step toward the broader conjecture that the unitary dual is the closure of the Arthur-type representations. The paper also yields falsifiable constraints on automorphic local components that do not require the generalized Ramanujan conjecture, which is a genuinely useful application. The combinatorial framework of intervals and adjacency on extended Z-segments is a new technical contribution that is likely to be reusable. However, the decisive combinatorial lemmas in Section 4 are not fully proved in the manuscript; several key statements are dismissed as straightforward or deferred to omitted direct computations. Since these lemmas are load-bearing for the non-unitarity direction of Theorem 6.3, the present version is conditional on their verification.","major_comments":[{"comment":"Lemma 4.10(ii) and the NV(−,e2) half of (i)–(iii) are not proved. The proof explicitly says 'Part (ii) follows from a case-by-case straightforward computation, which we omit' and 'We omit the analogous verification of these statements for NV(−,e2)(Δ1)'; Lemma 4.10(iii) Case (c) is closed with 'We omit the rest of the verification, which is similar to Case (b).' These statements are not optional: Lemma 4.10 is used in the proof of Proposition 4.20(1) and hence in Theorem 6.6, which underlies the base case of the non-unitarity argument. The manuscript itself flags these as omissions, so the proof of Theorem 6.3 is conditional on assertions that remain unchecked.","section":"Lemma 4.10, §4.2"},{"comment":"The proof of Lemma 4.17, after reducing to the case e3 = ([A3,A3],0,η3), ends with 'the proof proceeds by a direct computation, which we omit.' Lemma 4.19(1)-(3) are all deferred with the same phrase: 'These statements follow from direct computations, which we omit.' These lemmas are used in the proof of Proposition 4.20 through Observations (i) and (ii), and Proposition 4.20(2) is invoked by Theorem 6.8. Thus the induction step for the non-unitarity direction of Theorem 6.3 rests on unverified combinatorial identities. A rigorous submission needs either complete proofs of these case checks or a machine-checkable verification.","section":"Lemmas 4.17 and 4.19, §4.4"},{"comment":"In the proof of Proposition 4.20(2), the text states: 'It is possible that [Ee,e′]†† is empty' and then immediately asserts 'Suppose [Ee,e′]†† is empty for some (equivalently, for all) (e,e′) ∈ NVE(S) × NVE(S′).' The asserted equivalence is not justified. If emptiness occurs only for some pairs, the contradiction derived from Conditions (b) and (c) does not follow, and one would need an additional argument to reduce to the non-empty case. Since Proposition 4.20(2) is essential for Theorem 6.8, this is a second load-bearing gap in the same combinatorial engine.","section":"Proposition 4.20(2), §4.4"}],"minor_comments":[{"comment":"In the proof of Theorem 6.3, the text 'for any i ∈ IInu,nsd' has an extra 'I': it should be 'i ∈ Inu,nsd'.","section":"§6.3"},{"comment":"The final sentence of the proof says 'This gives a contradiction to the existence of Π + k−1 or Π + k−1'; the second expression should be Π − k−1.","section":"Corollary 6.7"},{"comment":"Lemma 4.6(c) is used in Proposition 4.20(2), but its proof is only one line: 'These are straightforward consequences of the definitions.' Since the proposition is load-bearing, expanding the proof of Lemma 4.6(c) would improve verifiability.","section":"Lemma 4.6"},{"comment":"The definition of ǫ in Definition 4.7(2) uses a lift ηi ∈ {±1}; Remark 4.8(1) asserts independence of the lift, but an explicit verification would help the reader, especially because Lemma 4.10 relies on this convention.","section":"Definition 4.7(2)"}],"recommendation":"major_revision","confidential_remarks":"The authors acknowledge a concurrent independent proof by Atobe and Mínguez [AM25]; this is not itself a problem. The main risk is that the omitted computations in Section 4 are not merely cosmetic: if any of the interval or adjacency claims in Lemmas 4.10, 4.17, or 4.19 fails, the non-unitarity direction of Theorem 6.3 loses its proof. The manuscript is built on a credible and well-cited framework, so the gaps appear fixable, but the submitted version is conditional. I recommend requesting a complete write-up of the omitted case checks, or a machine-checked verification, before publication."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Dear [Colleague],\n\nThe thing to know: this is the first proof of the complementary Arthur representation conjecture for Sp_2n and split SO_{2n+1}. That's a real theorem, and the authors mean it. The proof strategy is clear: handle the base case |I_nu|=1 with Arthur's intertwining relations and a Muić–Tadić non-unitarity criterion, then reduce to that base case using a new combinatorial formalism—intervals and adjacency on extended Z-segments. The row-exchange combinatorics in Section 4 is genuinely new, and the applications to automorphic forms (constraints on local components of self-dual cuspidal representations of GL_2 and GL_3) are nice and concrete. They also disclose the concurrent independent proof by Atobe and Mínguez, which is the right thing to do.\n\nNow the soft spot. The non-unitarity direction of Theorem 6.3 rests on Proposition 4.20, which rests on Lemmas 4.10, 4.17, and 4.19. Several of the key cases in those lemmas are not actually proved. Lemma 4.10(ii) is dismissed as a 'case-by-case straightforward computation' that is omitted, and the analogous verification for NV(−,e2) is simply not there. Lemma 4.10(iii) has 'we omit the rest of the verification' in the most delicate case. Lemma 4.17's hardest case ends with 'direct computation, which we omit,' and Lemma 4.19 says all three parts follow from direct computations. The text itself marks these as unverified. That is not a fatal flaw in the theorem—I suspect the claims are true, and the concurrent work gives independent support—but it does mean the paper as written is not a complete proof. A referee cannot check the main argument without reconstructing a pile of case analyses.\n\nI did not find any sign of circularity. The main proof does not use the conjecture it proves. Self-citations are confined to motivation, definitions, and applications, and the relevant external theorems (Arthur, Mœglin, Atobe, Xu, Tadić) are cited appropriately.\n\nBottom line: this is a serious paper for the representation-theory community. It deserves full peer review. But the referee instructions should explicitly ask for Section 4 to be expanded or for a companion note with the omitted computations. I would not accept it in its current form; I would accept after the combinatorial gap is closed. If I worked in this area I would cite it for the theorem and for the interval/adjacency technique, and I'd bring it to a reading group to see how the combinatorics works.\n\nBest,\n[Name]","headline":"A substantial conjecture proven with a credible strategy, but the combinatorial core leaves enough 'direct computations' omitted that the proof is conditional as written.","tokens_in":39294,"tokens_out":2419,"would_cite":true,"duration_ms":23036,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["11F70","22E50","11F85","22E55"],"pacs":[],"model":"deepseek-v4-flash","headline":"For p-adic symplectic and split odd orthogonal groups, a complementary Arthur representation is unitary precisely when each reducible generalized Speh factor occurs with even multiplicity.","keywords":["Admissible Representations","Local Arthur Packets","Local Arthur Parameters","Arthur Representation","Unitary Dual","p-adic Classical Groups","Complementary Series","Generalized Speh Representations"],"falsifier":"A single explicit adjacent pair $e_1,e'_1$ and a segment $\\Delta_2$ for which $NV(e_1,-)(\\Delta_2)$ and $NV(e'_1,-)(\\Delta_2)$ are not adjacent would contradict Lemma 4.10(ii). Since Theorem 6.6, and then the non-unitarity direction of Theorem 6.3, rests directly on that statement, such a pair would invalidate the proof of the parity criterion.","tokens_in":38270,"feed_emoji":"🧮","tokens_out":11095,"duration_ms":96756,"temperature":0.7,"pith_summary":"This paper proves a precise criterion for unitarity of the complementary Arthur representations of $\\mathrm{Sp}_{2n}(F)$ and split $\\mathrm{SO}_{2n+1}(F)$ over a p-adic field $F$. Writing such a representation as $\\times_{i\\in I_{\\mathrm{nu}}} u_{\\rho_i}(a_i,b_i)|\\cdot|^{x_i} \\rtimes \\pi_A$ with $0<x_i<\\frac12$, the criterion says it is unitary if and only if every generalized Speh factor whose induction is reducible occurs with even multiplicity. The proof works by tracking the component-group characters of the summands produced by the unitary induction $u_\\rho(a,b) \\rtimes \\pi(E)$, using a new combinatorial calculus of intervals and adjacency for extended Z-segments. A consequence is that the unitary part of Arthur packets is now explicitly known for these groups, and constraints are obtained on local components of self-dual cuspidal automorphic representations of $\\mathrm{GL}_N$, especially $N=2,3$.","feed_headline":"Even multiplicities decide unitarity of Arthur-type representations","feed_subtitle":"Complementary Arthur representations are unitary exactly when each reducible factor appears with even multiplicity.","key_machinery":"The central object is the extended Z-segment, a triple $([A,B],l,\\eta)$ where $[A,B]$ is a consecutive block of integers, $l$ is a nonnegative integer at most half the block length, and $\\eta$ is a sign (with a relation when $l$ equals half the length). These segments parameterize the rows of an extended multi-segment, and the paper defines an interval as a consecutive set of such segments under a total order, with two intervals adjacent when they meet at a boundary. The key machinery is the non-vanishing set $NV(e_1,-)(\\Delta_2)$ of segments $e'$ for which the ordered pair $(e_1,e')$ satisfies the combinatorial non-vanishing criterion, together with the row-exchange operator $R$ that swaps comparable segments. Lemma 4.10 shows these sets are intervals and that adjacency is preserved; this lets the proof decompose the unitary induction $u_\\rho(a,b)\\rtimes\\pi(E)$ into summands whose component-group characters alternate sign, so a reducible induction produces both characters and hence a non-scalar intertwining operator.","core_discovery":"Theorem 1.3 (announced as Theorem 6.3) asserts that for $G_n=\\mathrm{Sp}_{2n}(F)$ or $\\mathrm{SO}_{2n+1}(F)$, a representation $\\pi\\in\\Pi_{\\psi}$ with $\\psi\\in\\Psi^+_{\\mathrm{unit}}(G_n)$ is unitary exactly when $\\pi$ lies in $\\Sigma_{A+,u}(G_n)$. In concrete terms, decomposing $\\pi=\\times_{i\\in I_{\\mathrm{nu}}} u_{\\rho_i}(a_i,b_i)|\\cdot|^{x_i}\\rtimes\\pi_A$, unitarity is equivalent to the parity condition: for each $i$ such that $u_{\\rho_i}(a_i,b_i)\\rtimes\\pi_A$ is reducible, the number of $j$ with $\\rho_j\\cong\\rho_i$, $a_j=a_i$, and $b_j=b_i$ is even. If all such inductions are irreducible, the representation is automatically unitary.","pith_inferences":["The interval/adjacency calculus is likely reusable: any family of representations whose reducibility is read off from extended multi-segments, not only Arthur packets of symplectic and odd-orthogonal groups, may obey the same parity principle as long as the non-vanishing sets are intervals.","If a proof of Lemma 4.10 can be given by explicit closed formulas instead of omitted case checks, the non-unitarity direction would become more robust and the criterion would be easier to verify computationally for larger parameters.","For self-dual cuspidal automorphic representations of $\\mathrm{GL}_N$ with $N>3$, the same localization argument should force even-multiplicity constraints on more non-tempered forms; the $N=4$ list in the paper is a natural place to test the pattern."],"forward_implications":["For $G_n=\\mathrm{Sp}_{2n}(F)$ or $\\mathrm{SO}_{2n+1}(F)$, the unitary complementary Arthur representations are exactly $\\Sigma_{A+,u}(G_n)$, so $\\Pi_{A+,u}(G_n)=\\Sigma_{A+,u}(G_n)$.","Every local component at a finite place of a discrete automorphic representation of split $\\mathrm{Sp}_{2n}$ or $\\mathrm{SO}_{2n+1}$ obeys the parity constraint, with no appeal to the generalized Ramanujan conjecture.","For irreducible self-dual cuspidal automorphic representations of $\\mathrm{GL}_N$ of orthogonal or symplectic type, any non-tempered Speh factor of the corresponding type that is not in the tempered part must occur with even multiplicity among the non-tempered factors.","In the low-rank cases $N=2$ and $N=3$, this forces temperedness: orthogonal-type $\\mathrm{GL}_2$ cuspidal representations have tempered local components at every finite place, and self-dual ramified $\\mathrm{GL}_3$ local components of the form $\\chi|\\cdot|^x \\times 1 \\times \\chi|\\cdot|^{-x}$ must have $x=0$ unless $\\chi=1$.","As a consequence, the closure of Arthur representations equals the unitary Arthur representations for these groups, $\\Pi_{A}(G_n)=\\Pi_{A+,u}(G_n)$, completing one step toward the unitary-dual conjecture."],"supporting_citations":[{"why":"States Conjecture 1.2, the characterization proved here, and the larger unitary-dual conjecture that motivates it.","marker":"[HJLLZ24]"},{"why":"Provides Arthur's local intertwining relation and local Arthur packets, used in the base case to show the intertwining operator's eigenvalues are component-group characters.","marker":"[Art13]"},{"why":"Gives the non-unitarity criterion: a reducible induced representation at s=0 with non-scalar intertwining operator has non-unitary complementary series.","marker":"[MT11]"},{"why":"Supplies the extended multi-segment construction of local Arthur packets, the row-exchange operator, the non-vanishing theorem, and the character formula.","marker":"[Ato22a]"},{"why":"Gives the decomposition of $u_\\rho(a,b) \\rtimes \\pi(E)$ into summands indexed by extended segments, which is the target of the interval machinery.","marker":"[Ato22b]"},{"why":"Shows the relevant parabolic inductions are irreducible, so every packet member has the form used in the theorem and the reducibility question is well-posed.","marker":"[Mœ11b]"},{"why":"Classifies the unitary dual of general linear groups, used to define $\\Psi^+_{\\mathrm{unit}}(G)$ and to recognize unitary factors in reductions.","marker":"[Tad86]"},{"why":"Provides the non-vanishing criteria for extended multi-segments that motivate Definition 4.7 and support the proof of Theorem 4.16.","marker":"[Xu21b]"}],"fun_headline_variants":["Even multiplicity rule for unitary Arthur representations","Parity condition sets unitarity for p-adic groups","Reducible factors: unitarity when counted evenly","Even count of same-type factors implies unitarity"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the non-vanishing sets $NV(e_1,-)(\\Delta_2)$ and $NV(-,e_2)(\\Delta_1)$ are always intervals and that adjacency of extended Z-segments is preserved under these set-valued maps and under row exchanges; the text verifies this through case checks, several of which are described as straightforward or omitted.","fun_headline_variants_meta":{"raw":{"variants":["Even multiplicity rule for unitary Arthur representations","Parity condition sets unitarity for p-adic groups","Reducible factors: unitarity when counted evenly","Even count of same-type factors implies unitarity"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000684,"raw_usage":{"total_tokens":3056,"prompt_tokens":852,"completion_tokens":2204,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":468,"completion_tokens_details":{"reasoning_tokens":2144}},"tokens_in":468,"tokens_out":2204,"duration_ms":16965,"temperature":1.0,"reasoning_tokens":2144,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T20:53:52.770369+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A single explicit adjacent pair $e_1,e'_1$ and a segment $\\Delta_2$ for which $NV(e_1,-)(\\Delta_2)$ and $NV(e'_1,-)(\\Delta_2)$ are not adjacent would contradict Lemma 4.10(ii). Since Theorem 6.6, and then the non-unitarity direction of Theorem 6.3, rests directly on that statement, such a pair would invalidate the proof of the parity criterion.","supporting_citations":[],"review_version":1}