{"id":"bbab3065-db66-4782-8ca9-8eec6f1c5c03","arxiv_id":"2608.06273","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Random hypersurface sections over finite fields are shown to avoid positive-dimensional singular loci with probability at least 1 - O((d+1)^r p^{-ceil(d/2)}), proving Poonen's arithmetic Bertini conjecture.","lead":"This paper proves a strong bound on how often a random degree-d hypersurface over a finite field has a positive-dimensional singular locus on a fixed smooth variety: the probability decays like p^{-d/2} up to a polynomial factor. It settles an open conjecture of Poonen about Bertini theorems over finite fields, using a new technique called Frobenius-orbit slicing.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The only concrete defect found is the localized algebraic-closure slip in Prop. 2.15; the intended fix is clear, so the central estimate is not threatened by this pass.","rationale":"The reader's weakest_assumption was Proposition 2.1 on uniform controlled charts, which is indeed structurally load-bearing but appears to be proved correctly. The reader also flagged, in the rationale, the same localized error in Proposition 2.15 that I regard as the only concrete defect: the set M_C must be taken in \\bar F_q, not F_q, because irreducible polynomials of degree e>1 have roots in extensions and because the curves in Proposition 3.1 are extracted over the algebraic closure. This is a genuine proof-level slip in a key estimate, but the surrounding text and Lemma 2.13 make the intended correction unambiguous. I checked the filtered Q-adic decomposition (Lemma 2.5), the exact preservation of uniform randomness (Cor. 2.6), the normal-tower entropy (Lemma 2.18), the recursive step (Prop. 3.1), and the moving-chart version (Cor. 4.3); I did not find a second gap that would move the verdict. In particular, the apparent omission of mixed jet derivatives in the inclusion (3.4) is harmless because the left-hand side is contained in the right-hand side, and the union bound over closed points only uses the right-hand side as a superset. The complete-intersection reduction through Jacobian-pivot charts also appears internally consistent. Since the only identified defect is localized and fixable, I do not recommend changing the reader's CONDITIONAL verdict; the paper should be accepted only after the algebraic-closure correction is made explicit.","tokens_in":29019,"tokens_out":62754,"duration_ms":640597,"concrete_test":"Rewrite Prop. 2.15 with M_C = \\bar F_q \\setminus t_j(C(\\bar F_q)) for an arbitrary integral curve C ⊆ U_{\\bar F_q}, and re-verify Prop. 3.1's inequality (3.5) using that version. In particular, confirm that every curve C_F supplied by Lemmas 2.10–2.11 is interpreted over \\bar F_q and that no step silently requires C to be defined over F_q. If the corrected proof fails at the averaging over J and Q, the recursion breaks; if it passes, the theorem stands as written modulo this typo.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The load-bearing point is the orbit-hitting estimate Prop. 2.15. In its proof, M_C is defined as F_q \\ t_j(C(F_q)). This cannot control intersection with V(Q(t_j)) for an irreducible Q of degree e>1, whose roots lie in \\bar F_q; and in Prop. 3.1 the curve C_F is extracted from W_I^{(m)}(F)_{\\bar F_q}, so it is only defined over \\bar F_q. If the statement were read literally, the counting of missed Q by #M_C/e would be meaningless and Prop. 3.1 would lose its auxiliary-slice estimate. Lemma 2.13 is already formulated over \\bar k, and the text immediately notes that M_C need not be Frobenius-stable, so the intended correction is to take M_C = \\bar F_q \\setminus t_j(C(\\bar F_q)). With that replacement the proof goes through: root orbits of distinct monic irreducibles are disjoint subsets of \\bar F_q, each of size e, and the lower bound I_e(q) ≥ q^e/(2e) gives (2.28). This is a genuine localized error in a key estimate, but not an obstruction to the claimed theorem once the algebraic closure is used consistently. I found no other load-bearing gap after checking the filtered decomposition, the normal-tower entropy, the moving-chart estimates, and the global union bound.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper develops a new finite-field sieve, termed Frobenius-orbit slicing, to estimate the probability that the m-th principal-parts jet of the restriction of a random degree-d form to a fixed arithmetic variety has a positive-dimensional zero scheme. The main result, Theorem 1.1, gives the bound C(d+1)^{N_m} p^{-\\lambda_m(d)} for the probability that dim J_m(f)≥1, with N_m=binom(r+m,m) and lambda_m(d)=floor(m(d+1)/(m+1)). Corollary 1.2 derives Poonen's arithmetic Bertini Conjecture 5.2 and, after increasing the degree threshold, a p^{-A} bound for every fixed A>0. Theorem 1.3 extends the estimate to complete intersections, with both a termwise bound and a uniform bound in terms of the minimum degree. The proof is based on uniform affine-étale charts with controlled projective closures, an exact filtered Q-adic decomposition of the random polynomial spaces, a normal Hasse-derivative tower on orbit slices, and Jacobian-pivot charts for the complete-intersection reduction.","tokens_in":29188,"tokens_out":21477,"duration_ms":244692,"significance":"If the proof is correct after the localized repair discussed below, this is a substantial advance: it settles Poonen's conjecture in a stronger form and extends the statement to higher jets and complete intersections. The method is genuinely new: the same irreducible coordinate slice Q(t_j)=0 serves both as a geometric hitting set for a controlled curve and as the algebraic source of independent normal Taylor conditions. The proof is largely elementary and self-contained, avoiding Weil conjectures, Lang-Weil estimates, and fitted constants; the exact preservation of uniform randomness through the filtered decomposition is a particular strength. The only serious defect I found is a localized algebraic-closure slip in the orbit-hitting estimate, whose intended fix is clear and does not appear to threaten the main argument.","major_comments":[{"comment":"The proof of Proposition 2.15 defines M_C = F_q \\setminus t_j(C(F_q)). This cannot control the event C∩V(Q(t_j))=∅ when deg Q=e>1, because the roots of Q generally lie in \\bar F_q, not in F_q. Moreover, in Proposition 3.1 the curve C_F is extracted from W_I^{(m)}(F)_{\\bar F_q}, so it is only defined over \\bar F_q. As written, the counting of missed polynomials by #M_C/e is therefore meaningless for e>1, and Proposition 3.1 loses its auxiliary-slice estimate. The intended fix is clear and is already signaled by the sentence that M_C need not be Frobenius-stable: one must take M_C = \\bar F_q \\setminus t_j(C(\\bar F_q)), where Lemma 2.13 gives #M_C≤b deg C. With that replacement, disjointness of root orbits and Lemma 2.14 give (2.28). The same correction must be applied consistently in the moving-chart orbit-slicing step of Corollary 4.3, where the curve is again extracted from a geometric reduced support. Since this is the load-bearing orbit-hitting estimate, the proof needs this correction before the main claims can be accepted as written.","section":"§2.5.4, Proposition 2.15, and its use in §3.1"}],"minor_comments":[{"comment":"After inverting N0, the text uses connectedness of Spec Z[1/N0] to conclude that the Hilbert polynomial is constant; this is true, but a one-line justification would help the reader.","section":"§2.1, Proposition 2.1"},{"comment":"The notation for the affine open chart \\mathcal U_\\alpha and its projective closure is easy to confuse, especially because the closure is later denoted by the same symbol with an overline but also appears in other local arguments; a distinct symbol for the closure would improve readability.","section":"§2.1, Proposition 2.1"},{"comment":"The warning that a general polynomial étale coordinate can fail the degree filtration condition is important, but the assertion is made without an example; a short concrete illustration would make the limitation easier to verify.","section":"§4.1.6"},{"comment":"The manuscript contains several typographical and formatting artifacts, such as the arrow symbols in Lemma 2.4 and missing spaces in the references; a careful copyedit is needed.","section":"Throughout"}],"recommendation":"major_revision","confidential_remarks":"This is a strong paper with a clear main idea and an internally coherent long proof. The sole substantive issue I found is the algebraic-closure slip in Proposition 2.15; the correction is straightforward and the text itself hints at it, but because the estimate is load-bearing, the manuscript should be revised before acceptance. I would be willing to look at a revised version. The self-citation [23] is not used in the proof, and the generative-AI disclosure is transparent and does not affect the mathematical assessment."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Dear [Colleague],\n\nThis paper is worth taking seriously. It proves Poonen's arithmetic Bertini conjecture from 2004 and extends it to higher jets and complete intersections. The method is genuinely new: instead of characteristic-p derivative decoupling, which degrades when d is comparable to p, they slice by irreducible coordinate polynomials Q(t_j) of degree e, use the Q-adic filtered decomposition of the random polynomial, and control the probability that a slice misses a curve via a boundary-value lemma. The balance gives exponent floor(m(d+1)/(m+1)), which for m=1 is ceil(d/2). I checked the central recursion and it holds together: Lemma 2.18's normal-tower entropy and Lemma 3.3's balance are consistent, and the complete-intersection reduction via Jacobian-pivot charts is structurally sound. The paper is self-contained and does not rely on unproven conjectures or fitted constants; the only self-citation [23] is not load-bearing.\n\nThe one concrete defect I found is in Proposition 2.15. As written, M_C is F_q minus t_j(C(F_q)), but the curve C may only be defined over the algebraic closure, and an irreducible Q of degree e has roots in \\bar F_q. The counting of missed polynomials by #M_C/e is meaningless unless M_C is taken in \\bar F_q. The intended correction is M_C = \\bar F_q \\setminus t_j(C(\\bar F_q)), and with that replacement the estimate (2.28) follows from Lemma 2.13 and the disjointness of root orbits. This is a localized slip, not a load-bearing gap. I did not find another issue of that kind, though the proof is long and no formalization or code is supplied, so a referee should go through Lemmas 2.5, 2.18, 3.1, and 4.2 in detail. The paper also transparently discloses use of AI in generating drafts; that is not itself a problem, but it raises the bar for careful checking.\n\nIf the proof holds up, the result is important: it settles a 20-year-old open problem and gives a sharper bound. This deserves a serious referee. I'd bring it to the reading group and would cite it if it passes review.\n\nBest.","headline":"Strong paper that plausibly settles Poonen's conjecture with a genuinely new slicing method; one localized algebraic-closure slip in Prop. 2.15 is easily fixed.","tokens_in":29792,"tokens_out":3292,"would_cite":true,"duration_ms":33889,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["14G15","14J70","11T06","14B05"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper proves that for any fixed arithmetic variety of relative dimension r, the probability that the m-th jet-zero locus of a random degree-d form is positive-dimensional is at most C(d+1)^{N_m} p^{-\\lambda_m(d)}, with N_m =…","keywords":["Bertini theorem","finite fields","Frobenius orbits","higher jets","singular loci","complete intersections","arithmetic Bertini conjecture","jet schemes"],"falsifier":"For a concrete check of the core entropy bound, take U = $A^{1}$ over F_5, m=2, e=2, d=7, fix an irreducible quadratic Q, and enumerate all F in the filtered space V_∅(7) to count, for each F_0, how many higher blocks make both normal derivatives $partial^{{[1]}}$F and $partial^{{[2]}}$F vanish at a closed point of the slice Q(t)=0; Lemma 2.18 predicts the conditional probability is at most $5^{{-4}}$ = 1/625, so any computed fraction above that for some choice of F_0 would indicate a failure of the recursion feeding the main theorem.","tokens_in":28723,"feed_emoji":"🎲","tokens_out":13120,"duration_ms":148601,"temperature":0.7,"pith_summary":"The paper establishes a uniform probability bound for the event that the m-th principal-parts jet of a random degree-d hypersurface section, restricted to a fixed smooth arithmetic variety, has a positive-dimensional zero scheme. The bound is C(d+1)^{N_m} $p^{{-lambda_m(d)}}$, and for m=1 it becomes C(d+1)^{r+1} $p^{{-ceil(d/2)}}$. This settles the arithmetic Bertini conjecture, which asked for a bound below c/$p^{2}$, and after raising the degree threshold the estimate is $p^{{-A}}$ for every fixed A>0. The same method gives a termwise and a minimum-degree bound for the probability of a positive-dimensional Jacobian degeneracy locus in complete intersections of up to r hypersurfaces.","feed_headline":"Random hypersurface singular loci vanish at rate p^{-d/2}","feed_subtitle":"For every fixed arithmetic variety, the chance of a positive-dimensional singular locus is uniformly bounded, settling the arithmetic…","key_machinery":"Frobenius-orbit slicing: for a monic irreducible polynomial Q of degree e over F_q, slice by the equation Q(t_j)=0, where t_j is one of the original degree-one étale coordinates. The key identities are the exact filtered decomposition V_I(d) = direct sum_a Q(t_j)^a V_{I∪{j}}(d-ae) (Lemma 2.5), which preserves uniformity of the random blocks, and the triangular normal Taylor expansion $partial^{{[a]}}$_j F ≡ Q'(t_j)^a F_a (mod Q(t_j)), giving per-block entropy. The recursion is controlled by a mixed-dimensional Bézout bound and a boundary-value lemma that bounds how many coordinate values a bounded-degree affine curve can omit.","core_discovery":"The central discovery is that slicing a positive-dimensional jet-zero locus by the zero set of a degree-e irreducible polynomial in one of the original degree-one étale coordinates serves two purposes at once. Geometrically, a random such slice hits any controlled curve with probability at least 1 - O(Gamma_m(d) $q^{{-e}}$), so the recursion can force the locus to drop dimension; algebraically, every closed point of the slice has residue field containing F_{q^e}, so the Q-adic block decomposition F = sum_a Q(t_j)^a F_a yields an upper-triangular system of normal Taylor coefficients whose a-th diagonal contributes min{e, d-ae+1} independent evaluation conditions. Balancing the two linear magnitudes gives the exponent lambda_m(d) = min_e min{me, d-e+1}, optimized at e = ceil((d+1)/(m+1)). The paper establishes this mechanism uniformly over the arithmetic family by constructing finitely many controlled étale charts with bounded complexity.","pith_inferences":["The same slicing mechanism may extend to other singular loci defined by higher-order differential conditions, such as the m-th jet schemes of arbitrary codimension, by choosing a normal tower in several variables; the entropy count would then be a sum over multi-indices rather than a single interval length.","Because the method never uses characteristic-p derivative decoupling, it may yield non-asymptotic bounds in the regime d ≍ p, where the p-power method stalls at a bounded auxiliary degree; this could be tested numerically for d near p on small charts.","The envelope argument behind the minimum-degree bound suggests that only the minimum degree of the equations matters for uniform degree thresholds; one could try to sharpen the termwise envelope with the complete-intersection pivot charts.","A natural testable extension is to ask whether the exponent lambda_m(d) is optimal for the present one-orbit-degree strategy, or whether mixing several orbit degrees e_1,...,e_s within one recursion can raise the exponent beyond m/(m+1); the paper's Lemma 3.3 shows optimality only among constant e at each step."],"forward_implications":["For m=1, the probability that the Bertini singular locus of H_f ∩ X_p is positive-dimensional is at most C(d+1)^{r+1} p^{-ceil(d/2)}, which implies a bound below c/p^2 and, after a degree threshold adjustment, p^{-A} for every fixed A>0.","For complete intersections of c <= r independent forms, the probability of a positive-dimensional Jacobian degeneracy locus is bounded termwise by C sum_i (d_i+1)^{r+1} p^{-ceil(d_i/2)} and hence by C'(d_min+1)^{r+1} p^{-ceil(d_min/2)}, uniformly over all degree vectors once d_min is large.","The higher-jet estimate yields exponential decay in degree: for any eta < m/(m+1), the probability of a positive-dimensional m-jet zero locus is at most p^{-eta d} for sufficiently large d uniformly in p.","The result holds for quasiprojective X, not just projective, because the orbit-slice hitting estimate replaces the deterministic intersection with a fixed ample divisor by a randomized orbit slice that misses a bounded-degree affine curve with probability O(Gamma_m(d) q^{-e}).","The proof gives a finite-field local theorem: any smooth affine chart with degree-one étale coordinates, bounded derivations, and a principal boundary of bounded degree satisfies the same local estimate, uniformly over fibers of a fixed integral chart."],"supporting_citations":[{"why":"States the arithmetic Bertini conjecture and the singular-locus formulation that the paper settles.","marker":"[13]"},{"why":"Provides the projective p^{-2} estimate that the quasiprojective result extends.","marker":"[21]"},{"why":"Supplies the elementary count of monic irreducible polynomials used in the orbit-slice hitting bound.","marker":"[11]"},{"why":"Gives the étale criteria, dimension formula, and regular-local quotient theorem used in the chart construction and the Jacobian criterion.","marker":"[19]"},{"why":"Supplies the EGA étale and dimension facts invoked for base change and relative dimension.","marker":"[8]"},{"why":"Gives the probability that a random complete intersection is smooth, the baseline for the complete-intersection theorem.","marker":"[5]"},{"why":"Shows that Galois orbits impose expected numbers of conditions on hypersurfaces, the nearest analogue to the orbit-slice entropy step.","marker":"[2]"}],"fun_headline_variants":["Poonen's Bertini conjecture resolved for positive-dimensional singular loci","Frobenius-orbit slicing yields uniform singular-locus bounds","Random hypersurface singularities: probability drops as p^{-d/2}","Positive-dimensional singular loci eliminated at rate p^{-d/2}","Arithmetic Bertini: uniform bounds for jet-zero schemes"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The entire argument rests on being able to cover the fixed arithmetic variety by finitely many affine charts where the original degree-one ambient coordinates are étale coordinates, with uniformly bounded derivation degrees and boundary degrees; if these uniform controlled charts did not exist, the filtered Q-adic decomposition and the recursive probability estimate would not go through.","fun_headline_variants_meta":{"raw":{"variants":["Poonen's Bertini conjecture resolved for positive-dimensional singular loci","Frobenius-orbit slicing yields uniform singular-locus bounds","Random hypersurface singularities: probability drops as p^{-d/2}","Positive-dimensional singular loci eliminated at rate p^{-d/2}","Arithmetic Bertini: uniform bounds for jet-zero schemes"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.0009,"raw_usage":{"total_tokens":3911,"prompt_tokens":1015,"completion_tokens":2896,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":631,"completion_tokens_details":{"reasoning_tokens":2805}},"tokens_in":631,"tokens_out":2896,"duration_ms":25003,"temperature":1.0,"reasoning_tokens":2805,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-07T10:49:08.827530+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"For a concrete check of the core entropy bound, take U = $A^{1}$ over F_5, m=2, e=2, d=7, fix an irreducible quadratic Q, and enumerate all F in the filtered space V_∅(7) to count, for each F_0, how many higher blocks make both normal derivatives $partial^{{[1]}}$F and $partial^{{[2]}}$F vanish at a closed point of the slice Q(t)=0; Lemma 2.18 predicts the conditional probability is at most $5^{{-4}}$ = 1/625, so any computed fraction above that for some choice of F_0 would indicate a failure of the recursion feeding the main theorem.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Gives the étale criteria, dimension formula, and regular-local quotient theorem used in the chart construction and the Jacobian criterion."},{"cited_title":"Grothendieck and J","cited_arxiv_id":null,"evidence_quote":"Supplies the EGA étale and dimension facts invoked for base change and relative dimension."},{"cited_title":"Bucur and K","cited_arxiv_id":null,"evidence_quote":"Gives the probability that a random complete intersection is smooth, the baseline for the complete-intersection theorem."},{"cited_title":"Hypersurfaces passing through the Galois orbit of a point","cited_arxiv_id":"2501.01906","evidence_quote":"Shows that Galois orbits impose expected numbers of conditions on hypersurfaces, the nearest analogue to the orbit-slice entropy step."}],"review_version":1}