{"id":"f300e567-9824-41ba-8b43-cd099530b2d6","arxiv_id":"1908.04363","paper_version":3,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"The canonical Langlands element of every unipotent Arthur packet for a split real exceptional group is unitarizable.","lead":"This paper proves that the distinguished representations Arthur predicted inside spaces of automorphic forms for split real exceptional groups are unitary. It settles a long-standing unitarity conjecture for these representations using Eisenstein series and a reduction to explicit root-system computations.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The v2 abstract admits qualifications to quoted intertwining-operator results in Theorem 9.10; Section 11 depends on exactly those properties at boundary points.","rationale":"The reader's weakest_assumption identifies Theorem 9.10's analytic properties, and my reading agrees: the pole-order analysis in equations (11.5)-(11.16) is the heart of the proof for all but four Table 3 cases, and it relies directly on the multiplicativity, holomorphy, and nonvanishing of normalized intertwining operators. The v2 abstract confirms that the authors themselves added caveats about Section 9 statements quoted from the literature, which makes this the most concrete and honest place to anchor a concern. I also considered whether the more glaring issue is the undocumented atlas reduction in Section 4.1, since the theorem covers all unipotent Arthur packets and the reduction to distinguished parameters is asserted without tables or scripts. That is a real reproducibility gap, but it is secondary to the analytic input because the reduction is a finite case check that the authors say they performed, whereas Theorem 9.10 is a general analytic claim whose failure would invalidate the uniform Eisenstein-series argument for many cases simultaneously. The concrete test I propose would settle whether the Section 11 machinery is sound; if it fails, the proof would need an alternative route. Since the conditional verdict already requires addressing these quoted properties, I do not recommend changing the verdict, but I do recommend that the test be run before full acceptance.","tokens_in":34042,"tokens_out":7539,"duration_ms":85846,"concrete_test":"Verify Theorem 9.10 parts 4, 8, 11, and 12 directly against the cited sources (Arthur [A3], Winarsky, Shahidi) for the normalization in (9.6) with nontrivial quadratic character χ_v and at the boundary weights λ0 in Table 3. A tractable sub-check is to compute R_v(wβ,λ,χ_v) for a simple reflection with χ_v∘β∨ nontrivial and ⟨λ,β∨⟩ = 0, and test whether part 10's isomorphism and part 4's product formula both hold; if multiplicativity fails at such a point, redo the Laurent expansion (11.16) for one E8 case and observe whether the m value in Table 4 changes. Alternatively, reimplement the Mathematica/LiE check of (11.19) and (11.20) with and without the assumption of part 4; if the result changes, the concern lands.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The load-bearing point is the analytic input collected in Theorem 9.10, especially multiplicativity of normalized operators (part 4), holomorphy in Re⟨λ,α∨⟩ > -1 (parts 7-8), and nonvanishing for dominant real part (parts 11-12). Section 11 uses part 4 to factor the normalized intertwining operator in (11.8) and uses part 14 (which itself rests on parts 4 and 11) to guarantee the nonzero leading coefficient after (11.19). The argument is evaluated at the dominant integral boundary weight λ0 of Table 3, where some inner products ⟨λ0,α∨⟩ equal 0, so the parameter sits on the boundary of the region where parts 9-10 are standard. The v2 abstract explicitly says that blue comments were added about statements concerning normalized intertwining operators quoted from the literature in Section 9, which is an in-manuscript signal that some quoted results may need qualification. If part 4 or part 11 fails at these boundary points, the Laurent expansion (11.16) and the order-of-vanishing conclusion (11.19) break down, and the square-integrability criterion (10.16) is no longer established. The paper does not supply an independent derivation of these properties for the present setting of quadratic characters and non-generic parameters, so the central proof rests on this unverified quoted input.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proves Theorem 4.3 (and hence Theorem 1.1 for the stated Chevalley groups): for split real groups, with the exception of the classical Spin/HSpin cases, the canonical Langlands element of every unipotent Arthur packet is unitarizable. The proof constructs nonzero quotients of principal series I_∞(λ_0, χ_σ,∞) by taking residues of Borel Eisenstein series; it uses Langlands' constant-term formula, the square-integrability criterion for residues, and normalized intertwining operators. The new ingredient is a reduction of square-integrability to explicit combinatorial identities (11.17)–(11.20), combined with arithmetic input from quadratic Dirichlet L-functions. A small number of cases are handled by the atlas software rather than by the uniform Eisenstein-series argument.","tokens_in":34257,"tokens_out":9099,"duration_ms":109691,"significance":"If correct, the paper completes the unitarity assertion of Arthur's conjecture for the canonical elements of all unipotent Arthur packets for split exceptional real groups, a case that was previously open outside the spherical sector. The main contribution is conceptual and technical: it constructs the full representations as residues of Eisenstein series, not just spherical vectors, and it isolates a clean combinatorial criterion that can be verified by computer algebra. The dependence on Dirichlet L-functions is explicit and mild. The paper also demonstrates a useful complementarity with atlas computations, although it does not ship the corresponding verification files.","major_comments":[{"comment":"The analytic properties collected in Theorem 9.10 are load-bearing, and the manuscript itself signals that they may require qualification: the v2 abstract states that blue comments were added about statements concerning normalized intertwining operators quoted from the literature in Section 9, yet the text of Theorem 9.10 contains no visible qualification. In particular, part 4 (multiplicativity of R_v, quoted from Arthur [A3]) is used to obtain the factorization in (11.8), part 8 supplies holomorphy near the weight λ_0, and part 14 is used after (11.19) to ensure that a nonzero leading coefficient can be found. The argument evaluates at λ_0, where some inner products ⟨λ_0, α∨⟩ vanish, so these properties are needed on the boundary of the regions quoted from the literature. The authors must state precisely which quoted results are being qualified, whether each of parts 4, 8, 11, 12, and 14 remains valid at the specific boundary points used in Section 11, and provide precise references or proofs for the qualified versions. Without this, the pole-order analysis around (11.5)–(11.16) and the resulting square-integrability criterion (10.16) are not fully established.","section":"§9, Theorem 9.10; §11, Eqs. (11.5)–(11.16)"},{"comment":"The reduction to distinguished Arthur parameters is asserted rather than demonstrated: Section 4.1 says that all non-distinguished parameters were handled 'directly' in atlas, and the first paragraph of Section 11 delegates four further cases to the atlas is_unitairy command. No atlas commands, transcript logs, or outputs are included. Because Theorem 4.3 ranges over all unipotent Arthur parameters, these checks are load-bearing for the completeness of the theorem. The paper should provide reproducible verification files and explain how the atlas computations correspond exactly to the Langlands elements named in Theorem 4.3, especially for the cases where the is_unitairy command alone is used.","section":"§4.1 and §11, first paragraph"},{"comment":"The proof establishes that a certain nonzero quotient of I_∞(λ_0, χ_σ,∞) is unitarizable, but it does not explicitly identify this quotient as the Langlands quotient with parameter ψ_Langlands. The statement of Theorem 4.3 includes the identification of the Langlands parameter, so the final paragraph should invoke the Langlands classification (or an equivalent uniqueness theorem) to explain why the irreducible quotient of the constructed unitarizable quotient has precisely ψ_Langlands as its Langlands parameter. As written, the transition from 'an L²-residue quotient of the standard module' to 'the canonical Langlands element of the packet' is left implicit.","section":"§11, final paragraph; Theorem 4.3"}],"minor_comments":[{"comment":"Theorem 1.1 is stated for all Chevalley groups except Spin(n+1,n), Spin(n,n), and HSpin(2n), but the proof in Section 2 restricts to exceptional groups and relies on earlier work for classical groups. Please make this logical dependence explicit in the statement of Theorem 1.1, or state it as a theorem about exceptional groups together with a citation to [A4, Mo, V].","section":"Theorem 1.1; §2"},{"comment":"The table caption and surrounding text should define the quantities m and k_bd; the variable k_bd first appears in the display following (11.17) and in Table 4 without a definition.","section":"Table 4"},{"comment":"The statement of the combinatorial vanishing condition in (11.17) uses the bound k_bd, but the relation of k_bd to the order of vanishing in (11.16) is described only in words. A short explicit display connecting (11.17) and (11.18) to the coefficient C^♯(w′, ow′+k_bd, f, ·) would improve readability.","section":"§11, after Eq. (11.16)"},{"comment":"The integer m in (10.16) is introduced without explaining how it will be chosen; the later use of Table 4 would be easier to follow if (10.16) explicitly stated that m is the order of the leading coefficient in the Laurent expansion and that the same m is recorded in Table 4.","section":"§10, Eq. (10.16)"}],"recommendation":"major_revision","confidential_remarks":"The mathematical architecture is convincing and the combinatorial core appears sound. My main risks are the unqualified quotation of analytic properties of normalized intertwining operators in Theorem 9.10 and the lack of reproducible atlas data. These are fixable in revision, so I would not reject on these grounds. I also recommend asking the authors to clarify the Langlands-parameter identification in the final step."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The main result here is genuinely new: for split real F4, E6, E7, and E8, the Langlands element of every unipotent Arthur packet is shown unitary, extending Miller's spherical case and Kim/Vogan's G2 result. The proof is a solid extension of Miller's method to operator-valued intertwining operators, with the square-integrability criterion reduced to explicit combinatorial identities (11.17)-(11.20) plus arithmetic input from Dirichlet L-functions. The paper is carefully structured and transparent about where atlas is used: for the reduction to distinguished parameters in Section 4.1 and for four cases in Section 11.\n\nThe soft spots are real but not fatal. The atlas computations are not shipped, and neither is the Mathematica/LiE verification of the combinatorial identities. An expert could verify them from the stated identities, but the paper is not fully self-contained. More notably, the v2 abstract says the authors added blue comments about some quoted statements on normalized intertwining operators in Theorem 9.10. That is an in-manuscript signal that those inputs may need qualification. I checked the stress-test worry about boundary behavior: the weights in Table 3 have inner products 0 or 1 with the relevant coroots, which lie inside the region Re⟨λ,α∨⟩ > -1 where the quoted holomorphy holds, and the zero case is exactly the isomorphism conclusion of property 10. So the region argument doesn't worry me much. But the multiplicativity property (part 4) and holomorphy (part 7) are load-bearing and are quoted from Arthur and Shahidi, so a referee should confirm they hold for quadratic characters and non-generic parameters.\n\nThe abstract and theorem scope are consistent: the theorem is stated for all Chevalley groups aside from spin covers, and the non-exceptional cases are already settled by Arthur's book, so the actual contribution is the exceptional groups.\n\nThis is a major result with a plausible, well-structured proof. It deserves serious refereeing, and I would accept it for peer review. The referee should ask for the atlas transcripts and the combinatorial verification, and should scrutinize the quoted intertwining-operator properties in the present setting.","headline":"First proof of unitarity for non-spherical unipotent Arthur packets in split real exceptional groups, with a largely sound argument that leaves a few computational checks unshipped.","tokens_in":34814,"tokens_out":7009,"would_cite":true,"duration_ms":70763,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["22E46","11F70","11F66"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper proves that the canonical Langlands element of every unipotent Arthur packet for split real exceptional groups is unitarizable.","keywords":["Arthur conjectures","unipotent representations","unitarity","Eisenstein series","intertwining operators","exceptional Lie groups","Langlands parameters","split real groups"],"falsifier":"Check the paper's alternating-sum identity (11.20) for one listed case with $\\mu=\\rho$ and $k=k_{\\mathrm{bd}}$: a nonzero value would leave an unwanted constant-term coefficient and invalidate the unitarity conclusion for that packet. Alternatively, exhibit a distinguished unipotent Arthur parameter from Table 3 for which a normalized intertwining operator has a pole or zero at the boundary $\\mathrm{Re}\\langle\\lambda,\\alpha^\\vee\\rangle=-1$ used in Section 11.","tokens_in":33808,"feed_emoji":"🧮","tokens_out":8993,"duration_ms":78141,"temperature":0.7,"pith_summary":"Arthur's conjectures predict that every unipotent Arthur packet contains a distinguished representation, the 'Langlands element', and that this representation is unitary. This paper proves that prediction for split real exceptional groups $G_2$, $F_4$, $E_6$, $E_7$, and $E_8$ (the classical split families had already been handled elsewhere), and with it completes the unitarity assertion for these packets. The proof constructs the representations as square-integrable quotients of principal series by taking residues of Borel Eisenstein series, reducing the analytic task to a finite root-system calculation. A sympathetic reader would care because unitarity is exactly the property that lets these representations occur in spaces of automorphic forms, and it had been the main missing predicted property for the exceptional cases.","feed_headline":"Every unipotent Arthur packet's Langlands element is unitary","feed_subtitle":"Residues of Eisenstein series and a root-system identity settle the last exceptional cases of Arthur's conjecture.","key_machinery":"The mechanism is the factorization of global intertwining operators $M(w,\\lambda,\\chi)=c(w,\\lambda,\\chi)R(w,\\lambda,\\chi)$, where $c$ is a scalar product of c-functions built from quadratic Dirichlet L-functions and $R$ is the normalized operator, obtained by dividing out the c-function factors, with controlled holomorphy and nonvanishing. Expanding the constant term of a Borel Eisenstein series at $\\lambda=\\lambda_0+\\epsilon\\mu$, the pole orders of $c$ determine which Weyl-group terms survive; the requirement that all unwanted terms cancel becomes a finite alternating-sum identity over the stabilizer $W_L$ of $(\\lambda_0,\\delta_0)$, displayed as equations (11.17)-(11.20). Arithmetic enters through the special values and nonvanishing of $c(s,\\xi)$ for quadratic Dirichlet characters, such as $c(0,\\xi)=1$, which make the pole-order bookkeeping exact.","core_discovery":"The central assertion is Theorem 4.3: if $G$ is a Chevalley group other than $\\mathrm{Spin}(n+1,n)$, $\\mathrm{Spin}(n,n)$, or $\\mathrm{HSpin}(2n)$, then for every real unipotent Arthur parameter $\\psi$ the principal series $I_\\infty(\\lambda_{O^\\vee},\\chi_{\\sigma,\\infty})$ has a unitarizable quotient whose Langlands parameter is $\\psi_{\\mathrm{Langlands}}$. A unipotent Arthur parameter is a homomorphism $\\psi:W_{\\mathbb R}\\times SL(2,\\mathbb C)\\to G^\\vee(\\mathbb C)$ whose restriction to $\\mathbb C^\\times\\subset W_{\\mathbb R}$ is trivial; it is encoded by an order-two element $\\sigma=\\psi(j)$ and an adjoint nilpotent orbit $O^\\vee$. The paper lists all distinguished such parameters for the exceptional groups and proves the unitarity claim uniformly by Eisenstein series, with four borderline cases checked by computer-assisted unitary-dual verification.","pith_inferences":["The alternating-sum cancellation identity is likely a shadow of affine Hecke algebra structure, which could explain the cancellations uniformly and may extend to the four computer-checked cases.","The same Eisenstein-series strategy should apply to classical split groups and to Spin covers once the analytic properties of their normalized intertwining operators are in place, potentially removing the exclusions in the theorem.","One testable extension is whether the residues constructed here realize not only the Langlands element but also other members of the Arthur packet, since the paper only establishes the canonical one."],"forward_implications":["The canonical Langlands element of every unipotent Arthur packet for the split real exceptional groups is unitary, completing the unitarity part of Arthur's conjectures for those groups.","Each such representation is realized as a square-integrable quotient of a principal series, giving an explicit global construction rather than a classification-only existence statement.","The associated nonarchimedean local components are also unitary, since they arise from the same square-integrable global residues.","Because the proof reduces to the finite list in the paper's tables, the unitarity assertion is verifiable case-by-case from the displayed data."],"supporting_citations":[{"why":"defines unipotent Arthur packets and the canonical Langlands element whose unitarity is the paper's target","marker":"[A2]"},{"why":"supplies the multiplicativity property of normalized intertwining operators used in the factorization","marker":"[A3]"},{"why":"previous Eisenstein-series proof of the spherical case whose analytic scheme the paper extends","marker":"[Mi]"},{"why":"provides Langlands' constant term formula and square-integrability criterion that anchor the residue argument","marker":"[M-W]"},{"why":"gives the archimedean holomorphy/entireness property of normalized intertwining operators cited as Theorem 9.10(7)","marker":"[Sh1]"},{"why":"gives the nonarchimedean analog of that analytic property","marker":"[Wi]"},{"why":"settles the classical split groups, delimiting the exceptional cases left for this paper","marker":"[A4]"},{"why":"computer-assisted unitary-dual classification used for the four cases not covered by the uniform argument","marker":"[ALTV]"}],"fun_headline_variants":["Unipotent Arthur packets: all Langlands elements unitary","Eisenstein series prove Arthur's unitarity conjecture","Arthur's conjecture for split groups: unitarity settled","All unipotent Arthur packets have unitary Langlands elements"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The proof leans on quoted analytic properties of the normalized intertwining operators—multiplicativity, holomorphy in the region $\\mathrm{Re}\\langle\\lambda,\\alpha^\\vee\\rangle>-1$, and nonvanishing for dominant real part—at boundary points, and the paper itself flags that some of these quoted statements may need qualification; if any fails at those points, the pole-order cancellation argument breaks.","fun_headline_variants_meta":{"raw":{"variants":["Unipotent Arthur packets: all Langlands elements unitary","Eisenstein series prove Arthur's unitarity conjecture","Arthur's conjecture for split groups: unitarity settled","All unipotent Arthur packets have unitary Langlands elements"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001033,"raw_usage":{"total_tokens":4305,"prompt_tokens":856,"completion_tokens":3449,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":472,"completion_tokens_details":{"reasoning_tokens":3386}},"tokens_in":472,"tokens_out":3449,"duration_ms":25186,"temperature":1.0,"reasoning_tokens":3386,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T13:43:57.057841+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Check the paper's alternating-sum identity (11.20) for one listed case with $\\mu=\\rho$ and $k=k_{\\mathrm{bd}}$: a nonzero value would leave an unwanted constant-term coefficient and invalidate the unitarity conclusion for that packet. Alternatively, exhibit a distinguished unipotent Arthur parameter from Table 3 for which a normalized intertwining operator has a pole or zero at the boundary $\\mathrm{Re}\\langle\\lambda,\\alpha^\\vee\\rangle=-1$ used in Section 11.","supporting_citations":[],"review_version":1}