{"id":"5ba6a94c-7890-484c-87d1-f79cd545ad3f","arxiv_id":"2501.04022","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"An explicit hybrid level-aspect subconvexity bound for triple product L-functions over number fields, with unconditional saving exponent 1/60, extends to joint ramification and conductor dropping.","lead":"This paper proves new explicit hybrid subconvexity bounds for triple product L-functions over number fields, securing a P_f^{-1/60} saving over the convexity bound in the level aspect. A generalist reader may care because explicit subconvexity bounds drive progress on equidistribution and special value problems in analytic number theory.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The explicit saving in Theorem 1.3 rests on the imported local lower bound I_T^v ≫ Q_v^{-1/4} (Prop 4.2); if that bound degrades at jointly ramified or small-residue places, the diagonal extraction fails and the claimed exponent shrinks.","rationale":"I read the paper as an explicit hybrid level-aspect subconvexity result for triple product L-functions, obtained by a spectral-reciprocity first moment and amplification. The internal arithmetic of the proof is broadly consistent: the amplifier lower bound A(π3) ≫ L^2/(log L)^2 follows from the Hecke relation λ(p)^2 = λ(p^2)+1 via Landau's prime ideal theorem, and balancing L = P_f^{(1/4−θ/2)/(7−2θ1−2θ2)} reproduces the stated saving. The main weakness is not a visible internal contradiction but a dependence on imported local period bounds. The lower bound (4.1) of Proposition 4.2 is used exactly to make the diagonal term π ≃ π3 survive the amplification: without it, the first moment with amplifier only gives an upper bound, not an individual L-value bound. Everything downstream — the Q_f^{-1/4} factor in Section 6.1 and the final exponent — inherits any failure of this local bound. The paper also explicitly allows bounded residue-field cardinalities and conductor-dropping phenomena, so the imported theorem must be checked in precisely those regimes. This is the same load-bearing assumption flagged by the reader, and I agree with the conditional verdict: the stated exponent should not be accepted as final until Proposition 4.2 is verified in the needed local generality or proved directly. I do not find grounds to reject the argument, only to require this verification.","tokens_in":26761,"tokens_out":18446,"duration_ms":189108,"concrete_test":"Independently verify Proposition 4.2 in the exact local configurations used in Theorem 1.3: v nonarchimedean, π1,v unramified, π2,v ramified of conductor c, π3,v arbitrary including the conductor-dropping range c(π2⊗π3) < c(π2)+c(π3), and the test vectors from Section 4 (φ2,v, φ3,v newvectors; φ1,v = π1,v(diag(1, ϖ^s))φ1^0 with s = c(π2⊗π3)/2 plus the bounded b-shift for small residue fields). Compute the normalized local integral I_T^v from the explicit Whittaker/Jacquet data of Lemmas 4.7, 4.8, 4.10 and the double-coset Lemma 4.4, checking that the lower bound (4.1) holds uniformly in q_v and conductor exponents. If a counterexample with I_T^v ≪ Q_v^{-1/4−α} is found, recompute the exponent in Theorem 1.3 with diagonal weight Q_f^{-1/4−α}; if the bound is uniform, the conditional objection is withdrawn.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Section 5.3 concludes ℓ(π3, m, n, a) ≫ Q_f^{-1/4} from Proposition 4.2, and Section 6.1 then lower-bounds the diagonal contribution by C Q_f^{-1/4} A(π3) L(1/2, π1⊗π2⊗π3)/Λ(1, π3, Ad). This is the only mechanism that turns the upper-bound first moment into an individual subconvex bound. Proposition 4.2 is not proved here; it is quoted from [HMN23, Thm 3.22]. The quoted statement suppresses the local test-vector hypotheses: the paper's own choices in Section 4 shift φ1,v by diag(1, ϖ^s) with s = c(π2⊗π3)/2, plus a bounded shift b for small residue cardinality, and the conductor-dropping cases c(π2⊗π3) < c(π2)+c(π3) are exactly where such lower bounds can fail. If the true lower bound is Q_v^{-1/4−α} with α > 0, the diagonal term is smaller by Q_f^{-α}; the final bound in Theorem 1.3 becomes Q_f^{1/4+α+ε} P_f^{-δ}, which can destroy subconvexity when P_f is only a small power of Q_f. This is a verification gap, not an observed contradiction, but it is load-bearing and should be closed before the stated exponent is advertised.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper develops a spectral reciprocity formula for twisted first moments of triple product L-functions over a fixed number field and uses it, together with an amplifier, to prove a hybrid level-aspect subconvexity bound for L(1/2, π1⊗π2⊗π3). The main theorem gives an explicit saving P_f^{-δ} with δ > 1/60 unconditionally when θ=θ1=θ2=7/64, and allows joint ramification and conductor dropping. The argument builds on the period integral approach of Ichino-Watson, the GL2 subconvexity machinery of Michel–Venkatesh, and the local estimates of Hu–Michel–Nelson and Blomer–Brumley–Khayutin, with the latter imported from preprints.","tokens_in":27116,"tokens_out":13595,"duration_ms":120850,"significance":"If the proof is correct, the paper provides a genuinely explicit hybrid subconvexity bound for triple product L-functions over a general number field, improving on the qualitative results in [HMN23] and being new even over Q. The allowance for joint ramification and conductor dropping is a valuable technical feature. The paper is not fully self-contained, but it identifies the key local and global inputs and connects them in a plausible way, which is a useful contribution to the literature.","major_comments":[{"comment":"The lower bound I_T^v(φ1,v, φ2,v, φ3,v) ≫ Q_v^{-1/4} is imported verbatim from [HMN23, Theorem 3.22] and is used in §5.3 to conclude ℓ(π3,m,n,a) ≫ Q_f^{-1/4}, which is the only mechanism that makes the diagonal term in the amplifier argument of §6.1 non-negligible. However, the test vectors fixed in §4 are not the same as those in the quoted theorem: φ1,v is replaced by π1,v(diag(1, ϖ^s))φ0_{1,v} with s = c(π2⊗π3)/2 (plus a bounded shift in small residue cardinality), and the conductor-dropping cases c(π2⊗π3) < c(π2)+c(π3) are exactly where local lower bounds are delicate. The paper does not verify that the hypotheses of [HMN23, Thm 3.22] hold for these choices. Since the final saving exponent in Theorem 1.3 is proportional to the exponent -1/4 in this lower bound, this is a load-bearing gap.","section":"§4.1, Prop. 4.2; §5.3; §6.1"},{"comment":"The symmetric relation (5.6) is asserted with the single sentence 'By the Hecke relation (3.5), we have the following symmetric relation' and no derivation. The relation involves a specific constant q^{1/2}ζ_q(1)/ζ_q(2) and a weighted sum over k of periods with Hecke operators, and it is the foundation of the reciprocity formula used throughout the paper. Please provide a complete derivation or a precise reference to a source that proves this exact identity.","section":"§5.2, Eq. (5.6)"},{"comment":"The bound for the generic term G_{p^{v-2k}}(q, Ψ1, Ψ2) is compressed into a single sentence: the paper states that 'From above discussion, especially Proposition 4.2, Proposition 5.1, Remark 5.2, convexity bound..., we obtain an upper bound' and then gives the final estimate. The combination of the local triple product bounds, the spectral expansion, the Cauchy-Schwarz step, and the Weyl law is not shown. This is the core analytic estimate of the paper and should be written out in detail.","section":"§5.2, between (5.8) and (5.17)"},{"comment":"In the amplification step, the paper uses a 'slightly stronger version of Theorem 1.1' in which the exponent on ℓ is 3/2 rather than 3/2 + 2θ1 + 2θ2. The justification in Remark 5.2 is a sketch that covers k=1,2,4, but the cases with θ_i appearing in the exponent are not fully handled, and the reduction to k=1,2,4 is not justified rigorously. Since the final saving exponent depends on this strengthened bound, this step needs a complete proof.","section":"§6.1 and Remark 5.2"},{"comment":"The paper states that the best exponent toward Ramanujan-Petersson for GL(2) over a fixed number field F satisfies 0 ≤ θ ≤ 7/64. The bound 7/64 is the Kim-Sarnak bound for F=Q; to my knowledge it is not established for a general number field (the standard uniform bound is 1/4 − 1/9, due to Blomer–Brumley). If 7/64 is not available for the number fields considered, the 'unconditional' claim in Theorem 1.3 with θ=7/64 is not justified. Please provide a reference for the 7/64 bound over arbitrary number fields, or restrict the statement accordingly.","section":"§1, paragraph defining θ"}],"minor_comments":[{"comment":"In the abstract, 'coprimes' should be 'coprime', and the sentence 'The estimation becomes a reciprocity formula between different moments of L-functions' is unclear; the intended meaning is that the estimation is based on a reciprocity formula.","section":"Abstract"},{"comment":"The notation 'a' is overloaded: on page 4 it denotes the ideal a, while in Section 4 the phrase 'we may have a ≤ 1' seems to refer to a different quantity. Please disambiguate.","section":"§4"},{"comment":"The sentences 'This is the phenomenon of the spectral reciprocity formula' and 'We get a close and interesting relation between different type of L-functions with different spectral length' are informal and should be replaced by precise mathematical statements.","section":"§5.2"},{"comment":"The paper refers to 'Section 6.3 Choice of test vectors and Proposition 6.5 in [HMN23]' in the introduction and to 'Section 6.4 and Assumption 5.3 in [HMN23]' in Section 6.1, but the present paper's Section 6 contains only Section 6.1. Please update the cross-references to avoid confusion.","section":"§1 and §5.3"},{"comment":"The reduction assuming (mna)^4 ≥ Qf ≥ (mna)^{1/2} is stated as following from 'Section 6.4 and Assumption 5.3 in [HMN23]' without explanation. Please spell out this reduction or provide a precise reference.","section":"§6.1"}],"recommendation":"major_revision","confidential_remarks":"The paper depends critically on two preprints ([HMN23] and [BBK22]) for the key local estimates, and the applicability of one of them to the chosen test vectors is not verified. The editor may wish to check the status of these preprints. Additionally, the claimed unconditional use of θ=7/64 over a general number field in Theorem 1.3 deserves scrutiny; if the 7/64 bound is only known over Q, the advertised 'unconditional' bound is overstated."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: this is a legitimate extension of the MV10/HMN23 spectral reciprocity machinery. The genuinely new output is the explicit hybrid subconvexity exponent (delta > 1/60 unconditionally, covering conductor dropping and joint ramification), not a fundamentally new mechanism. If the imported local estimates are correct, the main theorem follows from the established template in HMN23 and Zacharias. The paper deserves a serious referee, but the referee should be told that the advertised exponent depends on a load-bearing lower bound that is not proved here.\n\nThe presentation is genuinely useful: the moment bound, the reciprocity relation, and the amplification step are arranged so a careful reader can see exactly where each estimate enters. The paper is honest about its debts, quoting HMN23 and BBK22 rather than silently copying. The sketch in Remark 5.2 for the k=4 local Rankin-Selberg bound is a real attempt to fill one gap, and it gives the reader something concrete to check. The citation pattern is fine; the self-reference to Miao24 covers an unramified special case and is not masking circularity.\n\nThe soft spot is the one the stress-test note identifies. Section 5.3 concludes l(pi3,m,n,a) >> Q_f^{-1/4} from Proposition 4.2, and that lower bound is what turns the upper-bound first moment into an individual subconvexity bound. Proposition 4.2 is quoted from [HMN23, Thm 3.22], and the test-vector hypotheses matter exactly in the conductor-dropping and small-residue-characteristic cases that this paper claims to cover. If the true lower bound is Q_v^{-1/4-alpha}, the final bound becomes Q_f^{1/4+alpha+epsilon} P_f^{-delta}, which destroys subconvexity when P_f is only a small power of Q_f. This is a verification gap, not a detected contradiction, but it is load-bearing. The symmetric relation (5.6) is asserted rather than proved, and the Eisenstein Holder step is summarized by reference to Blomer; both should be spelled out. None of these is obviously wrong, but together they mean the paper is not self-contained.\n\nThis paper is for specialists in analytic number theory working on subconvexity via period integrals. I would send it to referees, with a clear request to verify the two imported propositions and the symmetric relation, and to confirm that the local lower bound survives the conductor-dropping cases. The final acceptance should be conditional on those checks.","headline":"Solid extension of the HMN23 reciprocity program with explicit hybrid exponents; the advertised saving rests on an imported local lower bound that needs verification before the exponent is trusted.","tokens_in":27617,"tokens_out":2321,"would_cite":true,"duration_ms":23603,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["11F70","11M41","11F72"],"pacs":[],"model":"deepseek-v4-flash","headline":"Triple product L-functions admit an explicit hybrid level-aspect subconvexity bound over any number field.","keywords":["subconvexity","triple product L-functions","spectral reciprocity","level aspect","automorphic forms","analytic conductor","amplification method"],"falsifier":"Evaluate the local triple product integral $I_v^T$ for the paper's test vectors at a non-archimedean place where the residue field has small characteristic, for instance characteristic 2, and compare it with $Q_v^{-1/4}$; a decay faster than $Q_v^{-1/4}$ would invalidate the diagonal lower bound and collapse the subconvex saving.","tokens_in":26587,"feed_emoji":"🧮","tokens_out":7708,"duration_ms":69702,"temperature":0.7,"pith_summary":"Over a fixed number field, the paper proves a subconvexity bound at the central point for triple product L-functions of three automorphic representations of PGL$_2$. The bound takes the form $Q_f^{1/4+\\varepsilon}P_f^{-\\delta}$ with an explicit positive saving exponent $\\delta$; unconditionally one may take $\\delta=1/60$, and under the Ramanujan conjecture $\\delta=1/28$. The saving is measured against a finite conductor parameter $P_f$ that also tracks the self-twists, so the result remains meaningful when the three conductors are jointly ramified and the analytic conductor drops. This makes the hybrid level-aspect saving explicit and covers the conductor-dropping range, which is the situation previous methods left open.","feed_headline":"Triple product L-functions beat convexity with explicit saving","feed_subtitle":"Unconditional saving of $P_f^{-1/60}$ over fixed number fields, covering the conductor-dropping range.","key_machinery":"The machinery is a spectral reciprocity formula for the twisted first moment $$\\mathcal M(\\pi_1,\\pi_2,a,m,n,\\mathfrak l)=\\sum_{\\pi} \\lambda_\\pi(\\mathfrak l)\\,\\frac{L(\\tfrac12,\\pi\\otimes\\pi_1\\otimes\\pi_2)}{\\Lambda^*(1,\\pi,\\mathrm{Ad})} f(\\pi_\\infty)H(\\pi,a,m,n),$$ together with the analogous Eisenstein contribution. The formula is derived by studying a symmetric period $P_{\\mathfrak q}(\\mathfrak l,\\Phi,\\Phi)=\\langle T_{\\mathfrak l}\\Phi,\\Phi\\rangle$ with $\\Phi=\\varphi_1\\varphi_2^{\\mathfrak q}$; expanding it in the level aspect via the Plancherel formula and regrouping the same inner product as $\\langle\\varphi_1^{\\mathfrak p}\\varphi_1,\\varphi_2^{\\mathfrak p}\\varphi_2\\rangle$ moves the Hecke action to the dual side and produces a moment of self triple products. Bounding that dual moment with local period-integral estimates and then applying the amplifier yields the subconvex bound.","core_discovery":"Let $F$ be a fixed number field, $\\pi_1,\\pi_2$ unitary cuspidal representations of $\\mathrm{PGL}_2(\\mathbb{A}_F)$, and $\\pi_3$ a unitary automorphic representation, with finite conductors $m,n,a$. Theorem 1.3 asserts that $$L\\big(\\tfrac12,\\pi_1\\otimes\\pi_2\\otimes\\pi_3\\big) \\ll_{\\varepsilon,F,\\pi_{i,\\infty}} $Q_f^{{1/4+\\varepsilon}}$ $P_f^{{-(1/4-\\theta/2)(1-2\\theta_1-2\\theta_2)/(7-2\\theta_1-2\\theta_2)}}$,$$ where $Q_f$ is the finite part of the analytic conductor of the triple product, $P_f$ is the finite part of the parameter $\\prod_v Q_v^{1/2}\\max_{i=1,2}C_v(\\pi_i\\otimes\\pi_i)$, and $\\theta,\\theta_i$ are the best exponents toward the Ramanujan conjecture available for $\\mathrm{GL}(2)$ over $F$. With $\\theta=\\theta_1=\\theta_2=7/64$ the bound becomes $Q_f^{1/4+\\varepsilon}P_f^{-1/60}$ unconditionally; under Ramanujan it becomes $Q_f^{1/4+\\varepsilon}P_f^{-1/28}$. The paper also gives explicit corollaries for pairwise-coprime squarefull levels and for the case where $\\pi_3$ is an Eisenstein series.","pith_inferences":["A testable extension is to let the archimedean parameters grow together with the levels; the archimedean weight in the reciprocity formula suggests the saving should persist, but that range is not carried by the present statement.","The reciprocity identity should also yield lower bounds or non-vanishing statements for the first moment by evaluating the dual side asymptotically, a direction the paper does not pursue.","Because the bound is uniform in the fixed number field, it may feed into equidistribution statements for Hecke eigenvalues against triple product periods, where a subconvex exponent in the level aspect is the standard input."],"forward_implications":["For any fixed number field, the central value of a triple product $L$-function has an explicit power saving in the level-aspect parameter $P_f$, not just a saving in the full conductor $Q_f$.","The saving remains available when the three finite conductors are jointly ramified, i.e. in the conductor-dropping range where $Q_f$ can be much smaller than the product of the individual conductor powers.","Unconditionally the exponent is $\\delta=1/60$; assuming the Ramanujan conjecture for $\\mathrm{GL}(2)$ over $F$ it improves to $\\delta=1/28$.","With pairwise coprime conductors the bound becomes an explicit expression in the norms of the three level ideals, and a stronger form holds when the two cuspidal levels are squarefull.","The same method covers $\\pi_3$ equal to an Eisenstein series, giving subconvexity for $L(\\tfrac12,\\pi_1\\otimes\\pi_2\\otimes\\chi)$."],"supporting_citations":[{"why":"Supplies the quoted local lower bound $I_v^T(\\varphi_{1,v},\\varphi_{2,v},\\varphi_{3,v})\\gg Q_v^{-1/4}$ that keeps the diagonal term large, together with the overall hybrid subconvexity strategy.","marker":"[HMN23]"},{"why":"Supplies the period-integral framework, the analytic-conductor normalizations, and the amplification method used to turn moment bounds into subconvex bounds.","marker":"[MV10]"},{"why":"Provides the Plancherel expansion and spectral-reciprocity technique for the symmetric period in the level aspect.","marker":"[Zac20]"},{"why":"Gives the triple product formula relating the global period to the central value $L(\\tfrac12,\\pi_1\\otimes\\pi_2\\otimes\\pi_3)$ for cuspidal representations.","marker":"[Ich08]"},{"why":"Provides the local Whittaker-function computations and the level-aspect triple product subconvexity results that the present paper extends.","marker":"[Hu17]"},{"why":"Supplies the amplifier construction and the amplified-moment technique used in Section 6.","marker":"[BK19]"},{"why":"The author's preceding reciprocity formula for the first moment of triple product L-functions; the present paper generalizes it to the ramified level aspect.","marker":"[Miao24]"},{"why":"Gives the regularized adjoint-value identity and the comparison of local Rankin-Selberg and triple product integrals used for the Eisenstein contribution.","marker":"[BJN24]"}],"fun_headline_variants":["Hybrid subconvexity: triple L-functions get explicit saving","Triple L-functions beat convexity via spectral reciprocity","New bound: triple L-functions subconvex with P_f^{-1/60}","Unconditional saving for triple L-functions in level aspect","Spectral reciprocity yields explicit saving for triple L-functions"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is the imported lower bound on the local triple product integral at ramified places, namely $I_v^T(\\varphi_{1,v},\\varphi_{2,v},\\varphi_{3,v})\\gg Q_v^{-1/4}$; if the true decay is worse for some ramified or small-residue-characteristic place, the diagonal term would shrink and the final saving exponent would degrade or disappear.","fun_headline_variants_meta":{"raw":{"variants":["Hybrid subconvexity: triple L-functions get explicit saving","Triple L-functions beat convexity via spectral reciprocity","New bound: triple L-functions subconvex with P_f^{-1/60}","Unconditional saving for triple L-functions in level aspect","Spectral reciprocity yields explicit saving for triple L-functions"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000989,"raw_usage":{"total_tokens":4272,"prompt_tokens":1103,"completion_tokens":3169,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":719,"completion_tokens_details":{"reasoning_tokens":3089}},"tokens_in":719,"tokens_out":3169,"duration_ms":22764,"temperature":1.0,"reasoning_tokens":3089,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T23:49:01.859321+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Evaluate the local triple product integral $I_v^T$ for the paper's test vectors at a non-archimedean place where the residue field has small characteristic, for instance characteristic 2, and compare it with $Q_v^{-1/4}$; a decay faster than $Q_v^{-1/4}$ would invalidate the diagonal lower bound and collapse the subconvex saving.","supporting_citations":[],"review_version":1}