Machines Learn Number Fields, But How? The Case of Galois Groups

2026-05-17T18:26:22Z

By applying interpretable machine learning methods such as decision trees, we study how simple models can classify the Galois groups of Galois extensions over $\mathbb{Q}$ of degrees 4, 6, 8, 9, and 10, using Dedekind zeta coefficients. Our interpretation of the machine learning results allows us to understand how the distribution of zeta coefficients depends on the Galois group, and to prove new criteria for classifying the Galois groups of these extensions. Combined with previous results, this work provides another example of a new paradigm in mathematical research driven by machine learning.

Explicit cost analysis of Toom-4 multiplication for incomplete NTT in lattice-based cryptography

2026-05-17T15:34:27Z

Polynomial multiplication is fundamental in lattice-based cryptography. While the Number Theoretic Transform (NTT) enables fast multiplication, it imposes constraints on the modulus of the coefficient field. Hafiz et al. (2025) addressed this limitation by analyzing the incomplete NTT, which combines a truncated NTT with conventional multiplication methods In this work, we revisit Toom-4 multiplication in the context of incomplete NTT. Although Toom-4 is asymptotically faster than Karatsuba, its precise cost has not been expressed in a form compatible with the incomplete NTT framework. We present a concrete Toom-4 implementation and derive explicit operation counts that separate additions/subtractions and multiplications over the coefficient field. Our analysis based on addition chains yields a simple cost model for incomplete NTT. Using this model, we analyze hybrid strategies combining Toom-4, Karatsuba, and incomplete NTT. We identify parameter ranges where Toom-4 is advantageous and validate the predicted behavior experimentally.

Module Lattice Security (Part III): Structured CVP Distance on the Log-Unit Lattice

2026-05-17T12:00:59Z

We prove that the $L^2$ CVP distance from a random short ring element to the log-unit lattice of $\Q(ζ_{2^k})$ converges to $\fracπ{2\sqrt{6}}\sqrt{n}$ as $n=2^{k-1}\to\infty$. We then show that this target lies inside the Voronoi cell of the origin for $k\ge 4$. For the $L^\infty$ norm, the maximum over $n$ sub-Gaussian coordinates yields $O(\sqrt{\log n})$ which translates into a sub-polynomial approximation factor for the Short Generator Problem. We show a Coarse Lattice Theorem that Babai's algorithm returns zero for all structured targets, yet exactly recovers unit perturbations of arbitrary size. For module determinant ideals, we further prove the Trigamma Theorem that proves an intrinsic imbalance $σ_{g_0}=O(1)$ independent of the modulus $q$. Finally, combined with Parts I and II, we reduce the CDPR factor for ML-KEM from $\exp(\tO(\sqrt{n}))$ to a sub-polynomial value.

Modularity from $q$-series

2026-05-17T04:06:01Z

In 1975, G. E. Andrews challenged the mathematics community to address L. Ehrenpreis' problem, which was to directly prove the modularity of the Rogers-Ramanujan $q$-series' summatory forms. This question is important because many different $q$-series appearing in combinatorics, representation theory, and physics often seem to be mysteriously modular, yet there is no general test to confirm this directly from the exotic $q$-series expressions. In this note, we answer the challenge. We use $q$-series algebra, first-order $q$-differential systems, and analytic continuation with monodromy to give a criterion that decides when such series are modular. Specifically, we establish a necessary and sufficient condition for a vector of holomorphic $q$-series on $|q|<1$ to form a vector-valued modular function without modular input, providing a clear path to modularity for strange $q$-series.

On denominators of consecutive $\operatorname{SL}(2,{\mathbb N})$-saturated Farey fractions

2026-05-17T02:50:11Z

The sequence $({\mathscr S}_Q)_Q$ of $\operatorname{SL}(2,{\mathbb N})$-saturated Farey fractions was defined in our previous work by ${\mathscr S}_Q := \{ a/q \in {\mathbb Q} \cap (0,1]: q+a+\bar{a} \le Q\}$, where $\bar{a}$ is the multiplicative inverse of $a\pmod{q}$ in $[1,q)$. Here, we prove that the set of $Q$-scaled denominators of consecutive fractions in ${\mathscr S}_Q$ is dense in the region ${\mathcal V}:=\{ (x,y)\in [0,1]^2 : \max \{ (1-3x)/2,2x-1\} \le y \le \max \{ x,1-x\} \}$, and provide a formula for their distribution in ${\mathcal V}$ as $Q\rightarrow \infty$.

The 3-adic valuations of Stirling numbers of the first kind

2026-05-17T01:44:50Z

Let $v_3$ denote the usual $3$-adic valuation, and let $s(n, k)$ be the unsigned Stirling number of the first kind. In this paper, for $a\in\{1,2\}$, we determine the values of $v_3(s(a3^n, k))$ for all $1\le k\le a3^n$. More precisely, for each admissible pair $(m, k)$, we obtain an explicit formula for $v_3(s(a3^n, a3^m-k))$. The proof combines properties of the $m$-th Stirling numbers of the first kind with a detailed analysis of the relevant $3$-adic orders. As a consequence, we prove the case $p=3$ of a conjecture of Hong and Qiu proposed in 2020. We also derive formulas near the diagonal, comparison results for the adjacent orders $a3^n$ and $a3^n+1$, sharp upper bounds for the families $v_3(s(3^n, k))$ and $v_3(s(2\cdot3^n, k))$, and partial confirmations of conjectures of Lengyel and of Leonetti and Sanna.

A Dynamical Néron--Ogg--Shafarevich Criterion via Orbital Arboreal Representations

2026-05-16T23:05:55Z

Let $K$ be a non-archimedean local field and $\varphi : \mathbb{P}^1 \to \mathbb{P}^1$ a rational endomorphism of degree $d \geq 2$ over $K$. In the tame case ($p \nmid d$), we show that strict good reduction is equivalent to the existence of a nonempty Zariski open subset $U_k \subset \mathbb{P}^1_k \setminus \mathrm{PC}(\widetilde{\varphi})$ over which the canonical residual morphism is finite étale of degree $d$. The criterion separates two complementary local invariants of a normalized integral lift: $\mathrm{Res}(F,G)$ controls residual degree drop, while the fiber discriminants $\mathrm{Disc}(F_{n,x})$ control étaleness of the residual fibers once full residual degree is ensured. Consequently, for every finite $x \in \mathcal{O}_K$ with $\bar{x} \in U_k$, the extensions $K(X_n(x))/K$ are unramified for all $n \geq 1$. We introduce the orbital preimage tree $T_{O^+(x)} = \varinjlim_n X_\infty(\varphi^n(x))$, the colimit in $G_K$-sets along the forward orbit, and the orbital arboreal Galois image $\mathcal{G}_{O^+(x)} = \mathrm{Im}(G_K \to \mathrm{Aut}(T_{O^+(x)}))$. On the forward-invariant safe locus $U_k^{\mathrm{safe}} = \bigcap_{m \geq 0} \widetilde{\varphi}^{-m}(U_k)$, strict good reduction is captured by the bijectivity of the orbital reduction map $X_n(x_m) \to \widetilde{X}_n(\bar{x}_m)$. This canonical orbit-invariant framework connects with arboreal Galois representations (Boston-Jones, Jones, and others) and yields pointwise and orbit-level reformulations. Explicit examples over $\mathbb{Q}_p$ illustrate the criterion.

Gaps of Binary Numerical Semigroups and of Binary Inclusion-Exclusion Polynomials

2026-05-16T21:01:43Z

Let $p$ be a given modulus, let $u$ be prime to $p$, and consider the linear permutation $u\cdot n\pmod p$ of the residue system modulo $p$. Writing $\langle x\rangle_p$ to denote the least nonnegative residue of $x$ modulo $p$, we say that a pair of integers $(a,b)$ is a dominant pair of this permutation if either the inequality $\max(\langle ua\rangle_p,\langle ub\rangle_p)<\min_{a\max_{a

Multiple Gauss sums

2026-05-16T15:09:38Z

A multiple Gauss sum is a complete multiple exponential sum twisted by Dirichlet characters. We prove a new bound for multiple Gauss sums and, as an application, improve previous results in the Birch--Goldbach problem. Let $F_1, \ldots, F_R \in \mathbb{Z}[x_1, \ldots, x_s]$ be forms with differing degrees, with $D$ being the highest degree, and let $\boldsymbol{F} = (F_1, \ldots, F_R)$ be nonsingular. We prove that the system $\boldsymbol{F}(\boldsymbol{x})=\mathbf{0}$ is solvable in primes provided that $s \geq D^2 4^{D+2} R^5$.

One-level densities of large even and odd orthogonal families of automorphic L-functions

2026-05-16T14:22:24Z

We prove one-level density results for L-functions attached to primitive forms of level q, averaged over square-free q, conditional on the Generalized Riemann Hypothesis (GRH). We treat the even and odd orthogonal families separately and extend the support of the Fourier transform of the test function to (-3,3). This extended support yields the strongest known non-vanishing results for these families of L-functions and their derivatives at the central point, conditional on GRH.

n-ary elliptic groups, rings, and primes in arithmetic progressions

2026-05-16T12:51:51Z

I introduced the notion of an elliptic group in [Elliptic groups and rings. Beiträge zur Algebra und Geometrie 66(2), 497-529]. It is a quasi-group based on the tangent-chord law of elliptic curves and thus, becomes an abelian group upon singling out an element. This close proximity to abelian groups is reflected in the theory, and among other things, we can define elliptic rings, which are monoidal objects in elliptic groups. An other way of expressing this is to say that they are commutative monoids with an elliptic group structure that distributes over them. In this paper, we generalise this theory from the binary elliptic group structure to the $n$-ary structure, which we call $n$-ary elliptic groups and $n$-ary elliptic rings. The latter are once again (binary) commutative monoids with an $n$-ary operation that distributes over the monoidal structure in an $n$-ary sense. The key interest of these objects for us is their arithmetic properties, which are surprisingly pleasant. The key result is that Dirichlet's famous theorem on arithmetic progressions becomes simply Euclid's theorem in these $n$-ary rings, at least for progressions of the form $an + 1$. Motivated by the hope to eventually prove this $n$-ary Euclidean theorem purely algebraically using the theory of $n$-ary rings (and thus give an alternative and purely algebraic proof of Dirichlet's theorem), we start by exploring the first arithmetic facts of these objects, including introducing the $n$-ary class group and showing that it indeed captures the unique $n$-ary factorisation. We also obtain a type of Dedekinds theorem for our main $n$-ary ring of interest: $\mathsf{nEl}(\mathbb{Z})$.

Machine Learnability as a Measure of Order in Aperiodic Sequences

2026-05-16T12:44:24Z

Research on the distribution of prime numbers has revealed a dual character: deterministic in definition yet exhibiting statistical behavior reminiscent of random processes. In this paper we show that it is possible to use an image-focused machine learning model to measure the comparative regularity of prime number fields at specific regions of an Ulam spiral. Specifically, we demonstrate that in pure accuracy terms, models trained on blocks extracted from regions of the spiral in the vicinity of 500m outperform models trained on blocks extracted from the region representing integers lower than 25m. This implies existence of more easily learnable order in the former region than in the latter. Moreover, a detailed breakdown of precision and recall scores seem to imply that the model is favouring a different approach to classification in different regions of the spiral, focusing more on identifying prime patterns at lower numbers and more on eliminating composites at higher numbers. This aligns with number theory conjectures suggesting that at higher orders of magnitude we should see diminishing noise in prime number distributions, with averages (density, AP equidistribution) coming to dominate, while local randomness regularises after scaling by log x. Taken together, these findings point toward an interesting possibility: that machine learning can serve as a new experimental instrument for number theory. Notably, the method shows potential 1 for investigating the patterns in strong and weak primes for cryptographic purposes.

Prime detecting quasi-modular forms in higher level

2026-05-16T08:20:49Z

In a previous work, the authors resolved a conjecture about the structure of prime-detecting quasi-modular forms by studying sign changes occurring in quasi-modular cusp forms. In this paper, we extend the considerations to prime-detecting quasi-modular forms of higher level, in particular describing the structure of the space of quasi-modular forms that detect primes in various arithmetic progressions. We also provide an ``analytic'' proof of the level one case.

Explicit Formulas for the Casimir Eigenvalues of $SL(n,\mathbb{Z})$-Maass Forms

2026-05-16T04:15:50Z

Maass forms for $SL(n,\mathbb{Z})$ are defined to be eigenfunctions of the Casimir operators $\mathcal{D}_{m,n}$ of orders $1 \leq m \leq n$ for $GL(n,\mathbb{R})$. For any $1 \leq m \leq n$ and Maass form $φ$ for $SL(n,\mathbb{Z})$, we provide a formula for the eigenvalue of $\mathcal{D}_{m,n}$ associated with $φ$ in terms of the Langlands parameters of $φ$. In the case $m=2$, we recover the formula for the Laplace eigenvalue of a Maass form due to Terras, the Casimir differential operator of order $2$ being the Laplacian. Our proof takes a graph-theoretic approach, relating the action of every elementary differential operator of order $m$ for $GL(n,\mathbb{R})$ to the partitions of a directed, edge-ordered graph with $m$ edges and at most $m$ vertices.

Factorization of Additive Polynomials and van der Geer--van der Vlugt curves in characteristic 2

2026-05-16T00:42:22Z

In our previous work, we gave a formula for the Frobenius eigenvalues of van der Geer--van der Vlugt curves in characteristic 2 by considering suitable quotients of the curve. Although the formula is explicit, it depends on many choices, which makes the formula complicated. In this article, we take a different approach using a factorization of additive polynomials, and prove a new formula. The resulting formula is simpler and is useful for explicit computations. As applications, we provide a method for constructing maximal and minimal van der Geer--van der Vlugt curves, and show that every such curve arises from this construction. We also compute various examples of van der Geer--van der Vlugt curves and study their periods.