{"id":"3f5bf815-b273-48dc-a187-9d10ea6b441b","arxiv_id":"2504.15505","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Explicit variance formulas for point counts of one-parameter families of cubic curves over F_p are derived, with several residual sums evaluated using Jacobsthal sums.","lead":"This paper computes the variance of the number of solutions to certain cubic equations over a finite prime field as one coefficient is varied. The formulas, expressed through quadratic character sums and Jacobsthal sums, generalize earlier moment computations for families of elliptic curves.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Unproved Jacobsthal-sum identities from the author's forthcoming monograph [8] carry the explicit values advertised in Theorem 2.4 and the examples.","rationale":"I read the full text and re-derived the key steps of Sections 4–7. Theorems 2.1–2.3 are proved using only standard character sum identities (5)–(7); the computations check out, and I verified Example 4.2 and Example 5.1 against direct enumeration at p=5 with exact agreement. Theorem 2.4's derivation up to the unsimplified Σ is elementary, and the simplification to the table relies on the author's forthcoming monograph [8] for three types of evaluations: Lemmas 3.1/3.2 (classical Jacobsthal sums), the cubic-to-ϕ2(1) conversion for x^3−6x^2+x, and the palindromic quintic identity plus the ρ-symmetry identities. These are not proved in the preprint, so the advertised 'explicit computations' are conditional on an unavailable source. That is a genuine load-bearing concern, but not evidence of error: my small-prime spot checks are consistent with the stated formulas. The appropriate disposition is to keep the conditional verdict, requesting either proofs or a citable source for the imported identities before the explicit table is relied upon.","tokens_in":15847,"tokens_out":15673,"duration_ms":113461,"concrete_test":"Verify the imported identities computationally for small primes: for all primes p ≤ 200 (p>3) and a few parameter choices (e.g., α,β in the palindromic quintic identity with α≠±4, plus the cubic evaluations σ(x^3−6x^2+x)=σ(2)ϕ2(1) and ρ(1/36)=σ(−3)ϕ2(1)), compute the left-hand and right-hand character sums directly in a short Sage script. Any mismatch at p ≤ 200 would show that the (1,−1) row of Theorem 2.4 and the related Table 2.4 entries are not supported. Alternatively, ask the author to include proofs of Lemmas 3.1, 3.2 and the Section 7 identities in an appendix or to make the relevant pages of [8] available for review.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central variance derivations in Sections 4–7 are internally coherent and self-contained. The load-bearing weakness is the import of Jacobsthal-sum evaluations from the author's unpublished monograph [8]: Lemmas 3.1, 3.2, the evaluation σ(x^3−6x^2+x)=σ(2)ϕ2(1), and the palindromic quintic identity [8, Thm.5.20] with the ρ-symmetry [8, Thm.5.3] and [8, Cor.5.4]. These are used to turn the residual character sums in Theorem 2.4's table and in Examples 5.1, 5.2, 6.1, 6.2 into explicit numbers. If any of these identities is incorrect, those advertised explicit values—especially Σ for (b,c)=(1,−1) and the rows involving ϕ2(1)—are wrong, even though the un-evaluated variance formulas in Theorems 2.1–2.3 remain valid. For a paper whose abstract promises 'explicit computations,' this unverified reliance on a forthcoming work is the principal risk.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper computes the variance of the number of F_p-points for one-parameter families of cubic curves of the form y^2=f_lambda(x), for primes p>3. The main results are Theorem 2.1, giving the exact closed form Var = p-1-sigma(-3)-sigma(a^2-3b) (with an alternative case when a^2=3b) for the family y^2=x^3+ax^2+bx+lambda; Theorem 2.2, reducing the variance for y^2=x^3+ax^2+lambda x+c to a residual quadratic character sum and a root count n_3; Theorem 2.3, giving analogous formulas for y^2=x^3+lambda x^2+bx+c; and Theorem 2.4, providing a table of values for y^2=x^3+bx+c+lambda(x^2-x). The method rewrites #C_lambda = p+S_lambda with S_lambda a quadratic character sum, expands the second moment via sum_lambda sigma(lambda+r(x))sigma(lambda+r(y)) = p[x=y]-1, and converts the variance into counts of coincidences r(x)=r(y). Several examples evaluate residual sums with cubic Jacobsthal sums phi_2 and psi_3, yielding explicit formulas. A central limitation is that Lemma 3.1, Lemma 3.2, and several identities in Section 7 are quoted from the author's forthcoming monograph [8] and are not proved in the manuscript.","tokens_in":15951,"tokens_out":14631,"duration_ms":124509,"significance":"If the imported Jacobsthal-sum identities are correct, the paper gives clean, exact variance formulas that generalize earlier first- and second-moment computations of Birch, He-McLaughlin, Miller, Hopf, and Yamauchi, and it introduces a useful variance-centred framing. The core reductions in Sections 4-6 are written out in detail and are internally coherent; the counting-of-coincidences technique is elegant and yields genuinely explicit formulas for several families. The paper is honest about the residual character sums that remain unevaluated, and the examples showing how Jacobsthal sums make them explicit are valuable. The principal weakness is that the advertised explicit values in Theorem 2.4 and in Examples 5.1, 5.2, 6.1, and 6.2 depend on unproved identities from a forthcoming monograph [8]. This is a load-bearing gap, not merely a stylistic issue, because the abstract promises 'explicit computations' and the table in Theorem 2.4 advertises those values as results of the present paper.","major_comments":[{"comment":"The evaluations of the Jacobsthal sums phi_2(c) and psi_3(c) are quoted from the author's forthcoming monograph [8] with no proof. These lemmas are used directly to produce the explicit formulas in Example 5.1 (formulas (19) and (20)), Example 5.2 (formula (22)), Example 6.1 (formulas (29) and (30)), and Example 6.2 (formula (32)). Since [8] is not available to the reader, the claimed explicit computations in these examples are not verifiable from the manuscript. The paper should either prove Lemmas 3.1 and 3.2, or give a complete, accessible reference, or explicitly state that those examples are conditional on [8]. As written, the 'explicit' status of the computations is overstated.","section":"Section 3.2, Lemmas 3.1 and 3.2"},{"comment":"The last row of the table in Theorem 2.4 depends on three unproved quotes from [8]: the palindromic quintic identity [8, Thm.5.20], the twisted symmetry [8, Thm.5.3], and the evaluation [8, Cor.5.4]. These identities are used to turn the residual palindromic quintic sum into sigma(-1)rho(2)+sigma(2)phi_2(1). Similarly, the third row (b,c)=(1,0) uses the unproved evaluation sum sigma(x^3-6x^2+x)=sigma(2)phi_2(1), cited to [8, Ch.5]. Without proofs of these identities, the values in these rows are not established by the paper. The author should include proofs or a self-contained appendix for these identities, since they are load-bearing for the advertised table.","section":"Section 7, case (b,c)=(1,-1) and case (b,c)=(1,0)"},{"comment":"The paper's framing that the table in Theorem 2.4 gives 'explicit computations' is not fully accurate even after the quoted identities are accepted: the row for (b,c)=(1,-1) contains n_4 and rho(2), where rho(2) is itself an unevaluated character sum, as the paper acknowledges. More importantly, the variance formula Var #C = p-2-1/p+Sigma(b,c) is derived for general (b,c), but the table covers only seven selected pairs. The reader should be told clearly which parts of the paper are unconditional reductions and which parts are intended as explicit evaluations; the current terminology conflates the two.","section":"Section 7, paragraph before Theorem 2.4"}],"minor_comments":[{"comment":"The word 'explicit' is used loosely: several results, such as Theorem 2.2(ii) and Theorem 2.3(ii), are reductions to residual character sums that are not evaluated in general. Suggest clarifying the distinction between 'closed-form variance' and 'evaluation of the residual sum'.","section":"Abstract and Introduction"},{"comment":"The claim that 'Most of our results are intractable through their exponential viewpoint' is not substantiated. It would be more appropriate to say that the author's method applies to these families, rather than asserting intractability for other methods.","section":"Section 2, last paragraph"},{"comment":"The signs in the four residue classes p mod 12 are not fully determined in the displayed formula: the terms '+-2A_2' and '+-2B_2' depend on the sign convention from Lemma 3.1, but the connection is not spelled out at that point. A short note would help the reader use Lemma 3.1 correctly.","section":"Example 5.2, formula (22)"},{"comment":"The sentence explaining why the first double sum can be extended to all of F_p^* is slightly compressed; the displayed values q(x,1)=-x and q(1,y)=-y are correct, but the reader has to reconstruct them. Expanding this one line would improve readability.","section":"Section 7, equation (39)"},{"comment":"Reference [8] is listed as 'forthcoming' with no date or arXiv identifier. If the monograph is not yet available, the cited identities cannot be checked; providing a preprint link or a published volume number would be helpful.","section":"References"}],"recommendation":"major_revision","confidential_remarks":"The perceived weakness is not a mathematical error in the derivations I checked; the reductions in Sections 4-6 are coherent and the counting arguments are sound. The obstacle is the unproved dependence on the author's own forthcoming monograph, which appears both in the Jacobsthal-sum lemmas and in the Section 7 character-sum identities. If the author can supply proofs or verifiable references for those identities, the paper should be suitable for publication. If the monograph is published before the revision is completed, citing the published version would largely resolve the concern."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"This is a solid, self-aware computation paper. What's actually new: Theorem 2.1 gives a genuinely general closed-form variance for y^2 = x^3 + ax^2 + bx + lambda over all a,b, extending the special families in Jacobsthal, He-McLaughlin, Miller, Hopf, and Yamauchi. The variance framing itself is a modest repackaging of first and second moments, but the paper doesn't pretend otherwise, and the specific formulas in Theorems 2.2-2.4 do go beyond what was published before. Theorem 2.4's table, with the explicit Jacobsthal-sum entries, is the kind of thing people in arithmetic statistics will find useful. The proofs, as far as I checked, are careful and complete for the central reductions. The exceptional cases (a^2 = 3b, c = 0, the dual-pair structure in Section 7) are handled consistently, and the author is candid about limits: he flags that the (b,c) = (1,-1) entry is not fully explicit because of the unresolved sum rho(2), and he notes he cannot do higher moments. That honesty is worth credit. The soft spot is exactly what the stress-test says: Lemmas 3.1 and 3.2, the evaluation sigma(x^3 - 6x^2 + x) = sigma(2)phi2(1), and the palindromic quintic-sum identities in Section 7 are imported from the author's forthcoming monograph [8] without proof. The un-evaluated variance formulas in Theorems 2.1-2.3 are self-contained and would survive even if an imported identity failed, but the advertised explicit values in Theorem 2.4 and Examples 5.1, 5.2, 6.1, 6.2 depend on those unproved identities. For a paper whose abstract promises 'explicit computations,' that is a real risk. It's fixable, either by including short proofs or by restating the identities as conditional on the monograph, but as it stands, a referee cannot fully verify the headline numbers without chasing a forthcoming book. The citation pattern is not a problem: citing your own monograph is fine when the results are genuinely needed, and the reader's complaint about repackaging is minor because the paper generalizes earlier moment computations rather than simply rewriting them. Bottom line: this deserves a serious referee. I'd send it to someone who knows Jacobsthal sums and character sums, and specifically ask that referee to check the imported [8] identities. If those hold up, the paper is publishable with at most minor revisions. It's not a breakthrough, but it's a clean, useful contribution to a subfield. I'd probably not cite it myself in the next year unless I worked directly on these families, but for someone in that niche it's citable.","headline":"A careful, honest computation paper with genuinely new variance formulas; the main caveat is that several explicit evaluations lean on the author's forthcoming monograph without proof.","tokens_in":820,"tokens_out":1124,"would_cite":false,"duration_ms":35848,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["11G20","11L10"],"pacs":[],"model":"deepseek-v4-flash","headline":"For one-parameter families of cubic curves over finite fields, the variance of the point count is computed exactly in terms of quadratic character values and Jacobsthal sums.","keywords":["variance of point counts","cubic curves over finite fields","Jacobsthal sums","quadratic character","character sums","one-parameter families of curves","point counting","Legendre family"],"falsifier":"For a small prime $p$, enumerate all solutions to $y^2=x^3+ax^2+bx+\\lambda$ for every $\\lambda\\in\\mathbb{F}_p$, compute the sample variance of the point counts, and compare with the formula $p-1-\\sigma(-3)-\\sigma(a^2-3b)$; for example $p=5$, $a=0$, $b=1$ would already settle the correctness of Theorem 2.1 for that case. Similarly, numerically evaluating the Jacobsthal sums $\\phi_2(1)$ and $\\psi_3(1)$ for primes up to 100 and comparing with Lemmas 3.1 and 3.2 would test those quoted identities.","tokens_in":15543,"feed_emoji":"🧮","tokens_out":8927,"duration_ms":70527,"temperature":0.7,"pith_summary":"This paper proves that for one-parameter families of cubic curves over the finite field $\\mathbb{F}_p$, the variance of the number of $\\mathbb{F}_p$-points is governed by explicitly computable character sums, and in many cases by exact closed formulas. The key observation is that the variance of the point count equals the variance of the quadratic character sums $S_\\lambda=\\sum_{x\\in\\mathbb{F}_p}\\sigma(f_\\lambda(x))$, and translation invariance makes this variance depend on first and second moments. Theorem 2.1 gives the cleanest result: for the family $y^2=x^3+ax^2+bx+\\lambda$, the variance is $p-1-\\sigma(-3)-\\sigma(a^2-3b)$ unless $a^2=3b$, where it becomes $(1+\\sigma(-3))(p-1)$. Theorems 2.2 and 2.3 give formulas for the other two one-parameter deformations of the general cubic, leaving explicit Jacobsthal sums that are evaluated in examples, and Theorem 2.4 tabulates the variance for the family $y^2=x^3+bx+c+\\lambda(x^2-x)$ for special coefficient pairs. These results make precise how the number of points on cubic curves fluctuates as the parameter varies.","feed_headline":"Exact variance formula for point counts on cubic curves","feed_subtitle":"For one-parameter families over F_p, the fluctuation of point counts is pinned down exactly.","key_machinery":"The engine is the translation of point counts into quadratic character sums: $\\#C_\\lambda = p+\\sum_{x\\in\\mathbb{F}_p}\\sigma(f_\\lambda(x))$, so the variance of the point count equals the variance of the character-sum vector. This variance is computed through the identity $\\sum_{\\lambda}\\sigma(\\lambda+r(x))\\sigma(\\lambda+r(y))=p[x\\approx y]-1$ (with Iverson brackets), which reduces the second moment to counting coincidences $r(x)=r(y)$ for a rational function $r$ associated to the family. The residual sums that cannot be evaluated by elementary character summation are cubic Jacobsthal sums $\\phi_2(c)=\\sum_x\\sigma(x^3+cx)$ and $\\psi_3(c)=\\sum_x\\sigma(x^3+c)$, whose values are determined by representations of $p$ as $A^2+B^2$ or $A^2+3B^2$ (Lemmas 3.1 and 3.2). These evaluations convert the variance formulas into explicit numbers.","core_discovery":"The central claim is that variance of point-counts for the families in question is not just asymptotically something but exactly computable. For instance, when $a^2\\neq 3b$, the variance over $\\lambda\\in\\mathbb{F}_p$ of $\\#\\{(x,y): y^2=x^3+ax^2+bx+\\lambda\\}$ is $p-1-\\sigma(-3)-\\sigma(a^2-3b)$, where $\\sigma$ is the quadratic character; when $a^2=3b$ it is $(1+\\sigma(-3))(p-1)$. The paper establishes analogous explicit formulas for the families $y^2=x^3+ax^2+\\lambda x+c$ and $y^2=x^3+\\lambda x^2+bx+c$ (Theorems 2.2 and 2.3), and for $y^2=x^3+bx+c+\\lambda(x^2-x)$ it obtains a table of explicit values $\\Sigma(b,c)$ for certain pairs, several in terms of the Jacobsthal sum $\\phi_2(1)$ and one involving an unevaluated sum $\\varrho(2)$.","pith_inferences":["The same variance-to-character-sum reduction should apply to one-parameter families of higher-degree curves, although computing higher moments via quadratic character sums will likely be harder; the paper explicitly notes it cannot compute moments beyond the second.","For the family $y^2=x^3+bx+c+\\lambda(x^2-x)$ with $(b,c)=(1,-1)$, the paper leaves the variance in terms of the unevaluated sum $\\varrho(2)$; a testable extension is to evaluate $\\varrho(2)$ using other Jacobsthal-sum identities, which would complete the table.","Since $\\sigma(-3)$ is $1$ when $p\\equiv 1\\pmod{3}$ and $-1$ when $p\\equiv 2\\pmod{3}$, Theorem 2.1 shows that for fixed $a,b$, the variance distinguishes primes by their congruence class modulo 3, an observation not highlighted in the paper.","The explicit variance formulas give a distributional interpretation of the 'negative bias' phenomenon: for those families, the variance is slightly below $p-1$, corroborating earlier observations from a different perspective."],"forward_implications":["For the family $y^2=x^3+ax^2+bx+\\lambda$, the variance is completely determined by $p$, $\\sigma(-3)$, and $\\sigma(a^2-3b)$; no residual sums remain.","In the twisted family $y^2=x^3+\\lambda^2(bx+1)$, the variance is $p-5-\\sigma(-3)-\\sigma(-3b)-\\phi_2(b)^2/p$, which becomes fully explicit when $p\\equiv 3\\pmod{4}$ or via Lemma 3.1 when $p\\equiv 1\\pmod{4}$.","For $y^2=x^3+\\lambda x+c$ and $y^2=x^3+\\lambda x^2+c$, the residual $\\psi_3$ term is evaluated by whether $c$ is a cube, giving variance in terms of $A_3,B_3$ from $p=A_3^2+3B_3^2$.","For the dual family in Theorem 2.4, the table covers several dual pairs, recovering the Legendre family when $(b,c)=(-1,0)$ and earlier moment computations when $(-3,1)$.","The variance formulas reproduce and generalize earlier first- and second-moment results, but from a character-sum perspective that handles families inaccessible to exponential-sum methods."],"supporting_citations":[{"why":"Supplies the Jacobsthal-sum evaluations used in Lemmas 3.1, 3.2 and the palindromic quintic identities in Section 7.","marker":"[8]"},{"why":"Provides earlier first- and second-moment evaluations that the variance formulas in Examples 4.1 and 5.1 recapture.","marker":"[2]"},{"why":"Contains the moment computation for the model family that Theorem 2.4 recovers in its $(-3,1)$ entry.","marker":"[6]"},{"why":"Raises a difficulty that Example 5.1 resolves, and contributes further moment computations for related families.","marker":"[7]"},{"why":"Supplies the Legendre-family distribution that Theorem 2.4 recovers in the $(-1,0)$ case.","marker":"[3]"},{"why":"Provides character-sum identities for the Legendre family that support the $(-1,0)$ row of Theorem 2.4.","marker":"[9]"}],"fun_headline_variants":["Exact variance of cubic curve point counts","Jacobsthal sums pin down curve variance","Variance formula for cubic families over F_p","Exact fluctuation of point counts on cubics"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The proof depends on Jacobsthal-sum evaluations quoted without proof from a forthcoming monograph by the same author; if any of these quoted identities is incorrect, the explicit numerical values in the examples and the table in Theorem 2.4 would be wrong, although the variance formulas expressed as character sums would still hold.","fun_headline_variants_meta":{"raw":{"variants":["Exact variance of cubic curve point counts","Jacobsthal sums pin down curve variance","Variance formula for cubic families over F_p","Exact fluctuation of point counts on cubics"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000347,"raw_usage":{"total_tokens":1825,"prompt_tokens":795,"completion_tokens":1030,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":411,"completion_tokens_details":{"reasoning_tokens":973}},"tokens_in":411,"tokens_out":1030,"duration_ms":7268,"temperature":1.0,"reasoning_tokens":973,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-16T11:25:31.854278+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"For a small prime $p$, enumerate all solutions to $y^2=x^3+ax^2+bx+\\lambda$ for every $\\lambda\\in\\mathbb{F}_p$, compute the sample variance of the point counts, and compare with the formula $p-1-\\sigma(-3)-\\sigma(a^2-3b)$; for example $p=5$, $a=0$, $b=1$ would already settle the correctness of Theorem 2.1 for that case. Similarly, numerically evaluating the Jacobsthal sums $\\phi_2(1)$ and $\\psi_3(1)$ for primes up to 100 and comparing with Lemmas 3.1 and 3.2 would test those quoted identities.","supporting_citations":[{"cited_title":"Nica: Jacobsthal Sums , Monographs in Number Theory, World Scientific (forth- coming)","cited_arxiv_id":null,"evidence_quote":"Supplies the Jacobsthal-sum evaluations used in Lemmas 3.1, 3.2 and the palindromic quintic identities in Section 7."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides earlier first- and second-moment evaluations that the variance formulas in Examples 4.1 and 5.1 recapture."},{"cited_title":"Miller: 1 - and 2-level densities for families of elliptic curves: evidence for the underlying group symmetries , PhD thesis Princeton University (2002)","cited_arxiv_id":null,"evidence_quote":"Contains the moment computation for the model family that Theorem 2.4 recovers in its $(-3,1)$ entry."},{"cited_title":"Miller: Variation in the number of points on elliptic curves and appl ications to excess rank, C","cited_arxiv_id":null,"evidence_quote":"Raises a difficulty that Example 5.1 resolves, and contributes further moment computations for related families."},{"cited_title":"Hopf: ¨Uber die Verteilung quadratischer Reste , Math","cited_arxiv_id":null,"evidence_quote":"Supplies the Legendre-family distribution that Theorem 2.4 recovers in the $(-1,0)$ case."},{"cited_title":"Yamauchi: Some identities on the character sum containing x(x − 1)(x − λ ), Nagoya Math","cited_arxiv_id":null,"evidence_quote":"Provides character-sum identities for the Legendre family that support the $(-1,0)$ row of Theorem 2.4."}],"review_version":1}