Papers

This paper aims to employ the weak Galerkin method to solve a class of nonlinear eigenvalue problems. We proved the weak Galerkin scheme produces lower bound for the energy. Moreover, by the post-processing technique, we obtain lower bound for the ground state eigenvalue. Finally, numerical experiments are provided to validate the theoretical analysis.

We prove the existence and uniqueness of weak solutions for the generalized Monge-Ampère equation and the supercritical deformed Hermitian-Yang-Mills equation in cohomology classes lying on the boundary of the solvable region. Moreover, we prove that the associated geometric flows converge to the weak solutions in the sense of currents. The proof combines viscosity-theoretic and pluripotential-theoretic techniques.

We classify knot traces with trisection genus at most 2. We give infinitely many torus knots whose traces have trisection genus 3, and infinitely many hyperbolic knots whose traces have trisection genus 4. We also show that there exist infinite families of knots whose traces have arbitrarily large trisection genus. In addition, we determine or give sharp bounds for the trisection genus of the traces of several well-known knots, such as the figure-eight knot, the $(p, pq+1)$-torus knots, and the $(-2, 3, 2n+1)$-pretzel knots.

This paper introduces a Laws of Form version of the Quaternions. We call this the Q-Calculus, a 16-valued extension of Laws of Form (LoF) which is closely related to the BF Calculus (where we have a single square root of the mark) and the concept of the square root of negation (related to the the square root of minus one). We construct Q as a system of LoF mark operators acting on 4-tuples, and prove that the set of eight operators in Q is isomorphic to the quaternion group, which is non-commutative. We give a novel proof of several of Q's distribution laws using non-commutative logic gates. We indicate how to represent Q as braids by associating elementary braids to square roots of negation. This results in a very concise representation of Q as LoF braids. We end the paper with an indication of how we can represent the Artin braid group in LoF and how we can generalize our work with the quaternions to Clifford algebras.

We investigate constrained quantum motion on curves and surfaces using connection factorization methods. We show that Laplace operators admit an exact half-connection factorization generated by connection one-forms. The first-order part of the Laplacian is identified with the Darboux rotational connection of the orthogonal frame. Elimination of this rotational connection naturally produces quadratic geometric invariants analogous to supersymmetric Riccati potentials. Using orthonormal moving frames, we derive effective geometric contributions for planar curves, spatial curves, and embedded surfaces. We further analyze Dirac reductions using structure equations and show that for Fermi-type reductions the scalar Jensen–Koppe–da Costa contribution is cancelled in the reduced Dirac sector, leaving a residual first-order spinorial derivative structure. The resulting framework suggests the existence of hidden nilpotent covariant differential complexes naturally generated by geometric connection structures.

This paper studies the $\alpha$-stability property of differentially flat nonlinear dynamical systems. The results build off the recently introduced notion of $\alpha$-stability, which is particularly amenable to characterize the ability of a system to track dynamic output reference signals. We consider systems controlled using the Newton-Raphson tracking controller, which results in closed-form control policies, therefore it is computationally efficient, and it has been shown to be effective to control a large variety of mobile robots, including autonomous vehicles. The main results of the paper consist in sufficient conditions for the $\alpha$-stability of differentially flat systems and for the equivalence between the proposed control algorithm and the Newton-Raphson tracking controller applied directly to the nonlinear dynamics. We demonstrate the behavior of the proposed controller applied to the kinematic unicycle and dynamic bicycle models.

This paper introduces a multifidelity formulation that reduces the computational cost of the proper orthogonal decomposition (POD) of a high-fidelity model by leveraging data from cheaper, lower-fidelity models. POD is a prevalent technique for extracting a low-dimensional basis from training data to achieve subsequent dimension reduction or reduced-order modeling. In scientific and engineering applications, the training data are typically numerical snapshot solutions of a high-fidelity model, and computation of a sufficiently rich snapshot set can be prohibitively expensive, especially when sampling over a high-dimensional parameter space. Insufficient snapshot training data risks overfitting and poor generalizability of the POD basis to outside the training regime. Our multifidelity POD (MFPOD) formulation reallocates computational budget to cheaper, low-fidelity models that can be sampled more extensively. MFPOD then weights high- and low-fidelity snapshot data via a control-variate formulation to guarantee an unbiased estimate of the expected high-fidelity least-squares projection error. The MFPOD subspace is chosen to minimize the estimate of this projection error, and converges in probability to the same subspace as single-fidelity POD in the limit of an arbitrarily large budget. For restrictive computational budgets, the MFPOD cost function has (under some assumptions) lower variance than the POD cost function, which makes the MFPOD subspace more robust against variations in the training data and thus less prone to overfitting. For a numerical example modeling the velocity of the Pine Island glacier, MFPOD achieves the same accuracy as single-fidelity POD with an order of magnitude reduction in the offline computational cost of snapshot generation.

The hull of a linear code is the intersection of the code with its dual and depends on the choice of inner product. We study how the numbers of linear codes with prescribed Euclidean, Hermitian, and symplectic hull dimensions vary with the hull dimension. Building on closed enumeration formulae, we derive exact ratio decompositions for the Hermitian and symplectic cases, determine the parameter ranges in which monotonicity holds, and identify the exceptional families. We also compare the three inner products and complement the theoretical analysis with computational classification results for quaternary codes.

Timelike Liouville field theory is a candidate model for positive curvature two-dimensional quantum gravity, but a mathematically controlled Lorentzian formulation has remained elusive. In this paper we construct the theory on the cylinder $\mathbb{R}\times \mathbb{S}^1$ in the integer screening sector for a natural algebra of renormalized exponential observables. Starting from a renormalized finite-volume torus regularization, we construct infinite-volume Euclidean correlation functions, prove analytic continuation in the time variables, and identify the resulting Lorentzian boundary values by explicit contour formulas. This yields exact Lorentzian correlators for a natural class of exponential observables. We then prove locality: spacelike separated vertex operators commute in the Lorentzian theory. For smeared observables generated by the integer-charge fields $e^{2nb\phi}$, these Lorentzian expectation values define a vacuum functional on an ordered $*$-algebra and support an AQFT-type quantization without positivity. More precisely, we obtain isotone local algebras, a complete locally convex space $\mathcal H$ with dense algebraic subspace $\mathcal H_0$ carrying a nondegenerate Hermitian form (shown to be indefinite for $b<8^{-1/2}$), a continuous cyclic representation, operator-topologically closed represented local algebras, an action of cylinder translations by continuous linear homeomorphisms, and locality for the represented local net. The construction does not produce a Hilbert space or a Haag-Kastler net of local von Neumann algebras in the usual sense, but it shows that a substantial part of the Euclidean-to-Lorentzian and algebraic reconstruction mechanism survives in this nonpositive setting for timelike Liouville theory on the cylinder.

Many stochastic physical systems evolve smoothly over time in the sense that the distribution of states changes regularly across time steps. The transition from current state to the next state can often be modeled as the combination of a smooth map and an explicit source of randomness. Stochastic Lifting exploits this structure by attaching an independent, high-dimensional random label to each state transition in the training data and fitting a transition map from the current state and label to the next state using a standard regression loss. The labels act as auxiliary coordinates that let the model represent multiple plausible next states from similar current states, avoiding collapse to a mean prediction in the finite-sample size regime. At inference, fresh labels are sampled at each time step and the learned map is rolled forward autoregressively, generating diverse trajectories with a single network evaluation per time step.

Non-uniform scaling control of formation enables multi-agent systems to adjust their shape by scaling with different ratios along different coordinate axes, offering enhanced flexibility in complex environments. However, like most existing formation maneuver strategies, it typically assumes a fixed set of agents, limiting its applicability in scenarios requiring dynamic team expansion. This paper introduces a distributed control framework that enables a formation to incorporate new agents during non-uniform scaling maneuvers in arbitrary dimensions while preserving the spectral properties of the graph Laplacian. Simulation examples validate the effectiveness of the theoretical results.

Multiple Access (MA) technology has consistently served as the core driving force behind the evolution of mobile communications. As a promising paradigm for next-generation communications, Semantic Communication explores entirely new semantic spatial resources by mining the deep meaning of information. However, the inherent spatial correlation and importance heterogeneity of semantic features often cause semantic collisions and semantic collapse in multi-user concurrent transmission scenarios. To address these challenges, this paper proposes a Semantic Interleaved Division Multiple Access (SIDMA) technique. By utilizing a permutation operator to perform structural whitening on semantic features and combining it with an Importance-aware Power Allocation (ImpPA) module for differentiated protection, SIDMA scatters core features across the interleaving domain and adaptively optimizes power levels based on real-time channel conditions. Simulation results demonstrate that, compared with traditional MA techniques and advanced semantic multiple access schemes including Orthogonal-Model Division Multiple Access (OMDMA), Deep Multiple Access (DeepMA), and Shared Embedding (SE), the proposed SIDMA exhibits superior reconstruction fidelity and scalability in multi-user concurrent transmissions, effectively enhancing the communication quality and robustness in resource-constrained environments.

How much meaning does a text carry? Shannon's theory measures uncertainty over symbols and is intentionally indifferent to meaning, while pairwise metrics such as BERTScore compare two texts rather than characterizing one. We develop a geometric framework that measures semantic content from the structure of a text's sentence embeddings. The framework has three parts. First, within a fixed embedding and baseline, six natural axioms uniquely determine a scalar measure up to scale, a frame-conditional uniqueness theorem. The resulting scalar is empirically too coarse, motivating a richer representation. Second, we propose a three-coordinate semantic profile capturing novelty (displacement from generic discourse), breadth (diversity of distinct ideas), and integration (connectedness among them), together with a discrete minimal unit (the semantic quantum) whose resolution is fixed by a clustering threshold $\tau$. Third, we prove a no-go theorem: no scalar summary of the profile can simultaneously satisfy analytic stability under paraphrase and concatenation, ordinal robustness across text scales, and cross-representation comparability. We exhibit two practical scalars, $S_{\mathrm{minmax}}$ and $S_{\mathrm{rank}}$, each occupying a distinct corner of this trade-off triangle. Validation across 23 synthetic categories, 5 Project Gutenberg novels, and 3 embedding models confirms the trade-off. The recommended rank-normalized configuration passes 25 of 28 ordinal checks as point estimates (21 of 28 after Benjamini-Hochberg correction), outperforming seven baselines including unigram entropy and a BERTScore-based novelty signal. A separate variational result connects the breadth coordinate to the log-determinant of a determinantal point process (Spearman $\rho = 0.985$ over 507 Gutenberg chapters), giving an optimization-theoretic foundation for breadth.

We consider the black-box optimization problem on a sphere. Two information-geometric optimization flows (IGO flows) are designed with rigorous calculation of natural search gradients based on hyperbolic (information) geometry of Poincaré and Bergman balls. We demonstrate that ensembles of generalized Kuramoto oscillators on spheres compute natural search gradients and realize IGO algorithms on both manifolds. The relationship between natural gradient policies in Bergman balls and quantum decision making is pointed out.

We develop an analytic umbral approach to Bessel moments, using them as a concrete testbed justifying the passage from formal indicial umbral calculus to Mellin–Barnes umbral transmutation theory. [...] While the formal procedure reproduces the correct results in suitable convergence chambers, it may lead to non-admissible hypergeometric expansions at physically relevant parameter values. The cubic moment provides the basic example [...] We show that this obstruction is removed by replacing the purely formal expansion with an analytic umbral transmutation. In this setting, exponential umbral pairings are interpreted through Mellin–Barnes integrals, and Ramanujan's Master Theorem acts as an inverse selection principle for the spectral ground state, or clock, associated with a given Bessel product. The factorisation \(J_0^3=J_0J_0^2\) produces two distinct clocks and reduces the cubic full-line moment to a one-dimensional Barnes integral, equivalently to a Meijer \(G\)-function. This gives the classical value of the cubic Bessel moment and clarifies why the divergent Appell realisation is only a local representation of a globally meaningful umbral identity. The same mechanism is then applied to scaled cubic products and to the fourth Bessel moment. [...] The fifth moment marks the first genuinely higher-rank case: the natural umbral grouping leads to a bivariate Barnes transmutation rather than to an ordinary Meijer \(G\)-function. Finally, we discuss real fractional powers \(J_0^\alpha\), \(\alpha>2\), showing that the same interpretation persists beyond integer moments. [...] The resulting picture identifies Bessel moments as values of effective umbral transmutations and separates the global analytic meaning of the umbral representation from the local convergence properties of its hypergeometric residue expansions.

We propose a systematic way to construct trigonometric-like functions beyond the classical sine–cosine pair by factorising rational umbral operators. The guiding idea is simple: the usual trigonometric functions may be viewed as cyclic components arising from a finite factorisation, and the same principle can be extended to an $n$-fold decomposition of rational umbral expressions. For each integer $n\geq 2$, the construction produces $n$ functions which play the role of higher-order trigonometric components: their sum reconstructs the corresponding umbral function, while the individual components isolate the different cyclic sectors of its expansion. The construction is developed first in the formal umbral setting. The quadratic case $n=2$ gives the Gaussian trigonometric functions, in which the cosine-like component is a Gaussian and the sine-like component is its natural umbral companion. The cubic case $n=3$ yields a three-component cyclic system and shows how the same idea extends beyond the usual even–odd decomposition. These examples suggest that trigonometric factorisation is not restricted to ordinary rotations, but belongs to a broader cyclic principle in umbral calculus. We then reinterpret the same formal identities through the recently developed analytic umbral framework. In this second step, the cyclic components are realised by Mellin–Barnes pairings, and the root-of-unity decomposition is related to the splitting of the corresponding spectral kernel. This analytic formulation provides contour representations, local expansions, and sectorial asymptotics for the functions obtained formally. Finally, we indicate how the same cyclic kernels act on Fourier transforms. The resulting framework presents higher-order umbral trigonometric functions as natural cyclic components of factorised rational or exponential umbral operators.

Image semantic communication is a critical component in next-generation wireless communication systems. However, such systems typically suffer from large memory footprints and high computational complexity, making them difficult to deploy on resource-constrained devices. To address these challenges, we propose a vision transformer (ViT)-enabled image semantic communication system. In this system, a recursive structure is introduced to iteratively refine semantic features and reduce the parameter count. In addition, three dynamic adjustment strategies are designed to adaptively reduce computational complexity: dynamic depth adjustment, dynamic width adjustment, and joint width-depth optimization. Dynamic depth adjustment adaptively determines the number of recursive modules according to image content and channel conditions, while dynamic width adjustment selectively preserves important neurons and attention heads. The joint width-depth optimization further enables flexible computation configurations. Simulation results verify that the proposed recursive ViT-based system, combined with the three dynamic adjustment strategies, reduces the parameter count by 48.7% and achieves higher reconstruction quality than existing baselines under comparable computational complexity.

We consider finite-horizon convolution equations with matrix-valued coefficients and their use in Markov renewal computations. A sequence is inverted in a truncated noncommutative series algebra, and explicit coefficient formulae are combined with zero-padded fast Fourier transform (FFT) multiplication, Newton iteration and Gauss–Jordan elimination. We prove exactness of finite-horizon inversion, deterministic perturbation identities, and left and right a posteriori residual bounds. The FFT analysis includes transform errors and rounding in the frequency-domain matrix products. For continuous-time semi-Markov equations, an endpoint mean-value rule converts matrix Stieltjes convolutions into discrete matrix convolutions. Error estimates are obtained under bounded-variation and smoothness assumptions, and a weighted resolvent argument yields fixed-horizon convergence, with second-order convergence for smooth kernels. The same inversion framework computes transition probabilities, first-entrance distributions, reliability, availability, renewal visits and reward-type quantities. Numerical experiments examine scaling in the horizon and state dimension, residual accuracy, first-entrance probabilities, convergence against an exact Markov benchmark, and a heavy-tailed Lognormal model. The accelerated methods preserve the probabilistic calculations while reducing the cost of long-horizon convolutional inversion.

We provide a lower bound for a weighted ratio of $\Gamma-$functions and its $\log{}-$derivative. Then we apply this result to estimate the norm of the operator associated with the Cesáro operator in the Hardy space $H^p({\mathbb C}_+)$, $1<p<\infty$

We develop a natural Bayesian multiplicity-correcting prior distribution within the probabilistic forward stepwise representation of model space priors for regression problems. The proposed prior, obtained from making an analogy to the Holm procedure, exhibits behavior closely aligned with that of the Matryoshka doll prior. We compare both priors to several other priors, including some recently put forward as objective choices for model space prior probabilities. Our comparisons indicate that adequate multiplicity correction requires a degree of sparsity that many recommended priors do not provide, and we argue that multiplicity correction itself offers a principled and transparent criterion for specifying model space priors in regression.

Centered finite-difference discretizations of convection–diffusion equations may oscillate when convection dominates at the mesh scale. For homogeneous Dirichlet problems with constant coefficients on uniform Cartesian grids, we derive ADSC (Adaptive Directional Sparse Correction), a local directional edge-diffusion correction guided by modal rectification of the centered-stencil Fourier symbol. The ideal modal reference damps modes independently, but its exact nodal action is nonlocal; ADSC replaces it by a nearest-neighbor positive semidefinite correction. For a regularized operator with activation fixed by an auxiliary sequence, we prove consistency, fixed-epsilon energy stability, and conditional discrete H^1-seminorm convergence. The implemented iteration instead uses activation generated by the computed solution. For that fully coupled nonlinear problem we prove existence and qualitative L^2 compactness/convergence only; uniqueness, convergence of activation updates, and energy-norm rates remain open. Numerical tests show selective extrema control, reduced modal-dominance indicators, and a low-cost few-shot variant. Comparisons with upwinding, SUPG, and AFC-inspired strategies are diagnostic rather than claims of uniform superiority.

The longest increasing subsequence (LIS) of a random walk has been studied mainly for zero-mean, symmetric step increments. We numerically investigate the LIS of biased Gaussian random walks, with unit-variance increments and positive drift $\mu_{p} = \Phi^{-1}(p)$, where $p = P(\xi>0)$. In contrast with the symmetric case, we find that for every fixed $p>1/2$ the mean LIS length grows linearly, $\langle L_{n}(p)\rangle \sim a(p)n$, with $a(p)$ increasing from $0$ at $p=1/2$ to $1$ as $p \to 1$. The record count is also linear, with coefficient $\lambda(p)$ fixed by Spitzer's formula for the ascending ladder epoch, and the LIS becomes increasingly aligned with this record skeleton as $p$ grows. At the symmetric point $p=1/2$, the record skeleton collapses to the Sparre Andersen $\sqrt{n}$ scale, while the LIS returns to the finite-variance $\sqrt{n}\log{n}$ regime. Near this limit the record rate has the closed-form small-drift slope $\lambda(\mu_{p}) \simeq \sqrt{2}\,\mu_{p}$, whereas the excess $a(\mu_{p})-\lambda(\mu_{p})$ vanishes more slowly than linearly in the drift, although our data do not resolve a single power law. The empirical distribution of $L_{n}$ also changes across this point, from lognormal-like at $p=1/2$ to Gaussian-like for every sampled $p>1/2$.

This manuscript explores a variational quantum formulation for nonlinear elasticity problems arising from hyperelastic material models. The approach leverages the potential energy structure of hyperelasticity and employs a hybrid quantum classical framework in which the energy functional is evaluated using parameterized quantum circuits and optimized through classical routines. To enable a hybrid (classical quantum implementation), polynomial approximations of the nonlinear terms in strain energy density are introduced, yielding a representation compatible with variational quantum algorithms. The methodology is demonstrated on a special case of the NeoHookean material model in a one dimensional setting using finite element discretizations with first and second order shape functions and nonhomogeneous boundary conditions. Numerical experiments investigate the influence of the polynomial approximation order on the accuracy and efficiency of the proposed approach, illustrating its feasibility for near-term quantum devices.

Recently, Balla, Janzer, and Sudakov showed a lower bound on the MaxCut in terms of the vector chromatic number, recovering known results on the MaxCut of $H$-free graphs. In this note, we show that their bound is tight, providing a construction that achieves a value arbitrarily close to the optimal constant. This answers a question raised by Elphick. Our construction is a modification of the geometric graph used by Feige and Schechtman to establish the integrality gap for the Goemans–Williamson semidefinite relaxation of the MaxCut.

We prove existence of solutions to a nonlinear degenerate elliptic equation of the form \[ \begin{cases} -\Delta_{1} u+ \frac{|D u|}{(1-u)^{\gamma}}=g & \mbox{in $\Omega$,}\\ u=0 \hfill & \mbox{on $\partial\Omega$,} \end{cases} \] in a suitable sense, where $\Omega $ is a bounded open set of $\mathbb{R}^n$, $\gamma>0$ is a fixed parameter, $g\geq 0 $ is a function in some Lebesgue space.

Coupled oscillator networks underlie many biological systems, from cardiac cycles to circadian rhythms. Phase-reduced models such as the Kuramoto model have been widely used to study synchronization, but they typically assume that oscillators remain continuously responsive to inputs and often produce tightly clustered phase distributions. Biological oscillators, however, commonly exhibit phase intervals during which inputs have little or no effect, called “dead zones." Here, we extend the Kuramoto model by introducing receiver-gated dead zones, in which oscillators transiently ignore incoming signals. Using analytical and numerical approaches, we show that dead zones can reduce synchronization rates, modulate the distribution of phase-locked solutions, and modify the stability of phase-locked states. For identical oscillators, full synchrony remains locally exponentially stable and numerically dominant, although convergence times depend sensitively on the dead-zone width. For heterogeneous oscillators that admit phase-locked solutions when the coupling strength satisfies $K>K^*$, dead zones broaden the long-term phase distributions. Numerical exploration across dead zone widths reveals a transition in which steady phase gaps lose stability, resulting in dead zone-induced phase drifting. Overall, these results identify phase-response dead zones as a biologically plausible mechanism for flexible phase organization beyond classical Kuramoto dynamics.

We study a finite form of the classical interval discrepancy problem. Starting from the unit interval, one repeatedly splits an existing interval into two until $n$ intervals have been produced. The discrepancy of such a process is the maximum, over all intermediate stages, of the ratio between the longest interval and the shortest interval. A theorem of de Bruijn and Erdős from 1949 shows that this ratio must approach $2$ as $n\to\infty$, and they give a sharp construction achieving this bound. For fixed $n$, their construction gives the upper bound $\operatorname{disc}(n)\leq 2-\frac{3}{2n}+O\bigl(\frac 1{n^2}\bigr)$. In this paper, we prove that $\operatorname{disc}(n)=2^{1-1/\lceil n/2\rceil}=2-\frac{4\ln 2}{n}+O\bigl(\frac 1{n^2}\bigr)$ for every $n$.

Ropelength is usually studied as a minimization problem for one knot at a time. We instead use ropelength to filter the space of all realizations of a knot type. For a knot type $K$ and a budget $\Lambda$, let $Y_\Lambda(K)$ be the space of unit-thickness $C^{1,1}$ configurations of length at most $\Lambda$, modulo rigid motions and constant-speed reparametrization. We prove compact capture for smooth knot families, yielding homotopical and homological exhaustion of the ordinary knot space by finite ropelength levels. In the normalized Euclidean model we prove compactness of every $Y_\Lambda(K)$, right continuity of connected components, and attainment of connected merge levels; on components of the ideal stratum these levels define an ultrametric. We also prove that $Y_L(K)$ strongly deformation retracts onto its exact-length shell. This separates an attained connected merge scale from the path merge scale, the genuine min-max quantity for constrained deformation paths. For higher homotopy we introduce ropelength widths and relate them to known models of spaces of knots. We record mirror symmetry and the complete minimal-layer BFACF merge trees of $4_1$ and $6_3$ as certified discrete benchmarks, without identifying discrete and continuum barriers. Finally, we relate the path filtration to the companion swept-area construction: at fixed ropelength its infimal trace area defines a genuine extended metric on the corresponding moduli space, with flat-current and projected-area lower bounds. The paper also formulates open problems concerning Gordian phenomena, regularization, and quantitative topology of thick-knot spaces.

Randomly initialized neural networks induce a prior over functions, but the predictor used in practice is produced only after training. We ask how much of this initial bias survives the training pipeline. To make the question measurable, we introduce initialization memory: the dependence of the validation-selected predictor on the scale of the random initialization. We perform controlled CIFAR-10 experiments on ResNets where initialization memory already sharply separates training regimes. Low-learning-rate SGD can interpolate while still remembering its initialization: on ResNet-9 with batch size $b=128$, test accuracy varies by $26.5$ percentage points across initialization scales despite $\ge99.5\%$ training accuracy. This is not undertraining: extending the same low-learning-rate regime to $5{,}000$ epochs leaves the spread essentially unchanged. In contrast, Adam-family methods largely erase the dependence. SGD can also be made to forget when larger learning rates are paired with explicit $L_2$ norm control. We interpret these findings in terms of the time scale of forgetting: gradient-flow-like dynamics can preserve initialization memory, whereas stochastic finite-step effects, explicit norm decay, and adaptive preconditioning erase it on scales governed by the size of explicit or implicit regularization. The practical inductive bias of a trained network is therefore not the architectural prior alone, but the architectural prior after being filtered by the forgetting dynamics of the training pipeline; and the same regularizers that improve generalization are precisely those that erase memory of initialization.

We prove real-rootedness for the Poincaré polynomial \[ P_n(t)=\sum_{i=0}^{n-3} \dim H^{2i}(\overline{\mathcal M}_{0,n};\mathbb{Q})t^i \] of the Deligne–Mumford moduli space $\overline{\mathcal M}_{0,n}$ of stable $n$-pointed rational curves, proving a conjecture of Aluffi–Chen–Marcolli. The proof starts from the Keel–Manin–Getzler recurrence, but its main new idea is a bivariate deformation $F_m(y,t)$ of the Poincaré polynomial. This deformation reveals a hidden interlacing structure not visible in the one-variable recurrence. For fixed $t<0$, the zero set of $F_m$ in the $y$-direction is controlled by a Sturm–Rolle argument on the interval $0<y<1-t$. The original polynomial is recovered on the slice $y=1$, and the ordered crossings of the moving roots through this slice give both real-rootedness and strict interlacing. Consequently, the Betti numbers of $\overline{\mathcal M}_{0,n}$ form an ultra-log-concave sequence. We further prove real-rootedness and ultra-log-concavity for the Poincaré polynomial of the Fulton–MacPherson space $\mathbb{P}^1[n]$ of $n$ ordered points in degenerations of the complex projective line. The proof for $\overline{\mathcal M}_{0,n}$ was obtained through an iterative AI-assisted workflow with Co-Mathematician, an agentic frontier-model system developed by Google DeepMind. Our role was to formulate the problem, evaluate the proposed proof attempts, identify gaps and request corrections, compare the developing argument with the literature, and refine the presentation of the final proof. Our additional human contribution was to observe that a similar residual deformation strategy applies to the Fulton–MacPherson spaces $\mathbb P^1[n]$, yielding the corresponding real-rootedness theorem.