{"id":"1a36e7f0-3f83-4a5c-8d15-21e687847b12","arxiv_id":"2504.12160","paper_version":2,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"The number of cubic extensions of F_q(T) of discriminant q^M is C_1 q^M - C_2(M) q^(5M/6) + O(q^((2/3+epsilon)M)), with explicit constants and an unconditional omega-result for refined counts.","lead":"Cubic extensions of the rational function field F_q(T) are counted by discriminant size. The paper proves an asymptotic with error O(X^(2/3+epsilon)), matching the best known error for cubic number fields, plus refined counts with prime splitting conditions and an unconditional lower bound on the error.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The O(Y^{2/3+eps}) error term rests on an unproved function-field analogue of [BTT, Proposition 4.5] quoted before (5.15); without this r-saving bound, the sieve estimate collapses.","rationale":"I read the paper in good faith as a substantial, technically detailed contribution. The exact orbit-counting formula in Theorem 1.4, the explicit secondary-term table in Proposition 4.5, and the unconditional one-level density argument for Theorem 1.3 are independently valuable and are presented with care, including the honest retraction concerning Zhao's work. The central asymptotic in Theorem 1.1 is not internally inconsistent, but its advertised error term O(Y^{2/3+eps}) is obtained only after an unproved quoted bound: the function-field analogue of [BTT, Proposition 4.5] in Section 5.2. This is exactly the weakest assumption identified by the reader, and I agree that it is the single most load-bearing point. The proof of Theorem 1.3 does not depend on this bound, since it works under the assumed shape (1.7), but Theorems 1.1 and 7.2 do depend on it. Because the issue is a genuine gap in support rather than evidence of falsity, the appropriate status remains conditional: the paper should not be accepted in its current form until the analogue is either proved in the function-field setting or traced to a precise published source. My stress-test therefore does not move the reader's verdict.","tokens_in":54562,"tokens_out":4810,"duration_ms":52482,"concrete_test":"Supply a proof, or a precise citation, of the exact function-field analogue used before (5.15): for R = F_q[T], uniformly in squarefree r and q-powers Y, the number of GL_2(R)-orbits of binary cubic forms with Disc(y) = Y and r^2 | Disc(y) is O_{q,eps}(Y/|r|^(2-eps)). A minimal analytical check is to adapt [BTT, Proposition 4.5] using the explicit fundamental domain of Proposition 4.1 and Lemma 4.2; if the claimed saving requires a nontrivial geometry-of-numbers argument that is not present in the paper, the conditional verdict stands. As a robustness check, recompute (5.15) with the trivial orbit-count bound Y: the dyadic sum over |Disc(y)| <= |r|^4 then contributes |r|^3 log X instead of |r|^(1+eps), so the final error term would exceed X^{2/3+eps} after summation over F.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The decisive step is in Section 5.2, immediately before equation (5.15). To bound the contribution of non-degenerate dual forms y in the maximality sieve, the paper invokes 'the function field analogue of [BTT, Proposition 4.5]': the number of GL_2(R)-orbits of binary cubic forms with r^2 | Disc(y) and |Disc(y)| = Y is << Y/|r|^(2-eps). This bound is not proved in the paper, and no precise function-field source is cited. It is genuinely load-bearing: after the change of variables to GL_2(K_infty), the displayed bound in (5.15) is reduced to a sum over r|fg of |r|^3 times the orbit count; the quoted r-saving turns this into << |fg| << |F|, and the final O(X^{2/3+eps}) error follows by taking delta = 2/3. If the r^{-2+eps} factor is unavailable, the trivial orbit count Y leaves a factor |r|^3 per divisor, producing a contribution of order |fg|^3, which after summing over |F| <= X^delta is far larger than X^{2/3}. The same unproved input is reused in Section 7.2.1 when handling splitting conditions, so Theorem 7.2 inherits the same gap. The bound may well be true and provable by adapting [BTT] or [BST, Lemma 34], but as written the paper contains only an assertion at the exact point where the advertised improvement over the earlier O(X^{5/6+eps}) quality is obtained.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the asymptotic counting function of cubic extensions of F_q(T), with q not divisible by 2 or 3. Theorem 1.1 gives #F(Y) = ((q^2-1)(q^3-1)/(q^4(q-1)))Y - ((q^2-1)/q) C_2(M)Y^{5/6} + O(Y^{2/3+epsilon}) for Y = q^M, M even, with C_2(M) depending on M mod 3; this matches the error-term quality of Bhargava-Taniguchi-Thorne over Q. Theorem 1.2 and Theorem 7.2 refine the count by imposing splitting types at finitely many primes, with error O(X^{2/3+epsilon}|P_1...P_n|^{2/3}), and Theorem 1.3 proves an omega-result omega+theta >= 1/2 for such refined counts, unconditional because the Riemann Hypothesis holds over function fields. The proofs use geometry-of-numbers methods: an exact orbit-count formula for binary cubic forms with fixed discriminant (Theorem 1.4), a sieve for maximality based on finite Fourier transforms, and a one-level density computation for Artin L-functions. The paper also acknowledges a gap in earlier work of Zhao and compares with independent work of Kural.","tokens_in":54853,"tokens_out":13397,"duration_ms":122535,"significance":"The results, if correct, are significant. The error term O(Y^{2/3+epsilon}) in Theorem 1.1 matches the best-known number-field result of Bhargava-Taniguchi-Thorne and improves Kural's O(Y^{3/4+epsilon}); the explicit secondary-term constants depending on M mod 3 and the refined counting theorem with splitting conditions are new in the function-field setting. The paper is careful in several respects: the main constants are computed from volumes and integrals rather than fitted; Theorem 1.4 is an exact orbit count; the one-level density computation yields an unconditional lower-bound result; and the author is transparent about the prior gap in Zhao's work and the relation to Kural's preprint. The main caveat is that the decisive r-saving estimate used in the sieve is currently asserted rather than proved, so the advertised error-term improvement is conditional until that input is supplied.","major_comments":[{"comment":"Section 5.2, immediately before Eq. (5.15): the paper invokes 'the function field analogue of [BTT, Proposition 4.5]', asserting that the number of GL_2(R)-orbits of binary cubic forms with r^2|Disc(y) and |Disc(y)|=Y is << Y/|r|^{2-epsilon}. This estimate is neither proved nor traced to a precise function-field reference. It is load-bearing: after the change of variables, (5.15) bounds the non-degenerate contribution by a sum over r|fg of |r|^3 times this orbit count, and the r^{-2+epsilon} saving is what reduces the total to << |fg| and hence to the final O(X^{2/3+epsilon}) after summing over |F| <= X^delta. With only the trivial bound on the orbit count, the |r|^3 factor gives a contribution of size |fg|^3 per divisor, which is far larger than X^{2/3}. The same unproved input is reused in Section 7.2.1, so Theorem 7.2 inherits the gap. I recommend that the authors supply a full proof of this function-field analogue (for instance by adapting the argument in [BTT, Section 4] or [BST, Lemma 34]) or give a precise citation.","section":"5.2, before (5.15); 7.2.1"},{"comment":"Section 5, Eq. (5.1): the estimate N(W_F cap V(R)_sigma;X) << X/|F|^{2-epsilon} is stated as being 'proven in the same way as [BST, Lemma 34]', without a proof or a function-field citation. This bound is used to truncate the inclusion-exclusion sum at |F| <= X^delta and to produce the O(X^{1-delta+epsilon}) tail error, so it is part of the central error analysis. If the adaptation is straightforward, please provide a brief proof or a precise reference; as written, the reader cannot verify the tail truncation from the cited number-field result alone.","section":"5, Eq. (5.1)"},{"comment":"Section 5.2, paragraph immediately before the non-degenerate case: the error terms are listed as having exponents 1-delta, 5/6-2delta/3, 2/3 and 2delta, and it is stated that these are minimized at delta = 1/3. However, the concluding sentence of the same section says the sum over |F| <= X^delta is evaluated 'using delta = 2/3'. If delta = 2/3 is used globally, the X^{2delta} term would become X^{4/3}, contradicting the claimed O(X^{2/3+epsilon}); if the intended value is delta = 1/3, the conclusion is consistent. Please correct the inconsistency and state the final choice of delta explicitly.","section":"5.2, choice of delta"}],"minor_comments":[{"comment":"The proof says that the independence of I^sigma_1(lambda_0) from the choice of v_sigma is 'postponed to the end of the section', but Section 4.5 does not explicitly identify where this fact is proved; the argument in Proposition 4.5 shows an analogous invariance for I'_sigma, so please add a pointer or a short direct argument for I^sigma_1.","section":"4.3, end of Proposition 4.3"},{"comment":"The formal product symbol with phi_ell(n) and the subsequent identification of expressions with their evaluated values is not fully specified for infinite products; please clarify the definition of the infinite formal product and justify the interchanges used in the displayed identities.","section":"7, around (7.25) and Theorem 7.1"},{"comment":"The symbol C_2(ell) appears in Theorem 1.4 without a definition in the statement; it would help to state explicitly that C_2(ell) is the value I'_sigma(q^ell) computed in Proposition 4.5 and related to C_2(M) by the summation over sigma in Section 5.3.","section":"1, Theorem 1.4, Eq. (1.8)"},{"comment":"The removal of reducible maximal rings is summarized as 'cf. [BTT, Lemma 8.1]'; for the refined splitting-condition theorem the same reduction is asserted with 'Removing the reducible forms as before'. Please indicate how the splitting conditions are preserved under this reduction, or give the needed quadratic-field estimate.","section":"5.3 and 7, after Proposition 5.1"}],"recommendation":"major_revision","confidential_remarks":"To the editor: the manuscript is substantial and likely correct in substance, but the main advertised improvement over Kural rests on an unproved r-saving bound. I would ask the authors to provide a complete proof of the function-field analogue of [BTT, Proposition 4.5] and to fix the delta inconsistency before publication. If these points are addressed, the paper would be a strong addition to the literature."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Here's the short version. Ahlquist proves an asymptotic for cubic function fields with a secondary term and error O(X^{2/3+ε}), matching the best number-field result of Bhargava-Taniguchi-Thorne; he also handles splitting-type refinements and gives an unconditional omega-result via the one-level density. The geometry-of-numbers arguments are self-contained and clean, and the exact orbit-count formula (Thm 1.4) is a real new tool. Section 1.2's retraction of the Zhao attribution is honest.\n\nThe soft spot is exactly where the stress-test points: at (5.15) the paper invokes 'the function field analogue of [BTT, Proposition 4.5]' — the orbit count bound O(Y/r^{2-ε}) for forms with r^2|Disc — without proof or a precise citation. That bound is genuinely load-bearing. Without the r^{-2+ε} saving, the sieve error term would only come out as something like |fg|^3 times Y, which is far too large; the advertised exponent 2/3 depends on having that saving. The same unproved input is reused in Section 7.2.1, so Theorem 7.2 inherits the issue. It may well be provable by adapting [BST, Lemma 34] or [BTT], but as written it's an assertion at the exact point where the improvement over Kural's O(X^{3/4+ε}) is obtained.\n\nThere's also a smaller presentational bug: in the final summation of Section 5.2, the text says 'using δ=2/3', but the correctly tuned parameter is δ=1/3 (the value picked by an earlier exponent balance). With δ=1/3 the final error is still X^{2/3+ε}; with δ=2/3 the X^{2δ} term from the first case blows up. It looks like a typo, but it should be fixed.\n\nThe reader's skepticism is otherwise matched: the main and secondary constants are computed from volumes and integrals with no fitted parameters; the final C_2(M) table matches simpler cases; the number-field benchmarks are cited. The omega-result is a genuine unconditional analogue of CFLS. A couple of other bounds are imported without proof (e.g., (5.1)), but those are standard and less dangerous.\n\nFor whom: anyone working in arithmetic statistics over function fields, or on Davenport-Heilbronn-type theorems. It's a serious paper that deserves refereeing. If the referee forces a proof or explicit citation of the function-field analogue of [BTT, Prop 4.5], the paper should be accepted without too much further pain. I'd engage with it.","headline":"Strong function-field analogue of Bhargava-Taniguchi-Thorne with a secondary term and O(X^{2/3+ε}) error; the advertised exponent rests on a single unproved 'function field analogue of [BTT, Prop 4.5]' quoted before (5.15), which is load-bearing and needs proof or precise citation.","tokens_in":55414,"tokens_out":9095,"would_cite":true,"duration_ms":76918,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":null,"created_at":"2026-08-16T12:36:30.817290+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":null,"supporting_citations":[],"review_version":1}