Mathematical Physics
Papers
Fresh from arXiv, typeset properly. Highlighted words trade on the market.
All of arXiv
Thursday 27 August
In this thesis, we study the spectral properties of dynamically defined aperiodic metric and discrete graphs. Our goal is to determine to what extent spectral properties of discrete one-dimensional ergodic Schrödinger operators persist when the aperiodicity is manifested through the geometry rather than through a potential. The graphs considered here are inspired by one-dimensional aperiodic tilings, and are called tiling graphs and decorated $\mathbb{Z}$-graphs. For a large family of metric tiling graphs equipped with the standard Laplacian, we show that the spectrum is of zero Lebesgue measure, and is a generalized Cantor set up to a possible discrete set of energies. For decorated $\mathbb{Z}$-graphs, we further show that for a Baire-generic and Lebesgue almost-sure choice of the decoration edge lengths, the spectrum is a generalized Cantor set. We then study the integrated density of states (IDS) for metric and discrete decorated $\mathbb{Z}$-graphs. We prove a gap labelling theorem, which characterizes the set of possible values taken by the IDS inside spectral gaps. We show that the gap labels are contained in the Schwartzman group associated with the dynamical system generating the graph, up to a geometric scaling factor. Lastly, we consider the Dry Ten Martini Problem for discrete Sturmian decorated $\mathbb{Z}$-graphs, asking whether all possible values predicted by the gap labelling theorem are indeed attained by the IDS inside spectral gaps. We answer this question negatively, by identifying a large set of gap labels which are not attained due to jump discontinuities of the IDS. We then show that away from these jump discontinuities, the periodic approximants for Sturmian graphs display the same combinatorial structure as the standard Sturmian Hamiltonians, and use this to obtain an explicit characterization of the realized gap labels for Sturmian comb graphs.
Multimedia
A Mixed-Behavior Vote Model for Multimedia Subjective Quality Votes, Means, and Variances
The relationship between subjective test vote variance and vote mean (or MOS) is well-studied, and the mathematically admissible vote variance region has been previously defined. We propose a reduced admissible variance region called the Unimodal Variance Region (UVR) that better describes real subjective rating behavior of multimedia. Further, subjective vote variance is often modeled as parabolic. We explain that, in practice, the parabolic model often violates the admissible region in the variance vs. MOS plane and we propose alternatives that respect the admissible region. We also present a parametrized random process to model votes that mixes voting processes and produces a realistic range of vote variances within the UVR at any desired MOS. This process was inspired by and comports with voting behavior that is observed in many subjective tests. By modeling vote variance from a subjective experiment, this vote model offers additional interpretable insights into voting behavior observed in a given experiment. We present example results from 16 datasets spanning speech, image, and video subjective quality experiments.
Read the full abstract
Game Theory
Horizon-Independent Contraction for Continuous-Time Discounted Regularized Mean-Field Games
We study contraction properties of non-stationary continuous-time mean-field games (MFGs) under discounting and entropy regularization. The state of the representative agent evolves according to a controlled continuous-time Markov chain, and both the state and action spaces are finite. In contrast to the undiscounted case, we show that, under a sufficiently large discount rate, finite-horizon MFGs admit a horizon-independent contraction condition, which also coincides with the corresponding infinite-horizon non-stationary contraction condition. As a byproduct, we obtain an explicit convergence rate between finite- and infinite-horizon mean-field equilibria. For each finite horizon, we further derive a refined contraction criterion from the spectral radius of a positive operator that majorizes the propagation of policy errors, and show that its large-horizon limit agrees with the horizon-independent contraction factor. Finally, we provide an explicit error bound between discounted and undiscounted finite-horizon regularized equilibria.
Read the full abstract
Materials Science
Dilute-Limit Defect Displacements Enabled by Brillouin-Zone Sampling
Defects and their interaction with the semiconductor host lattice play an essential role in a variety of technologies. When a defect transitions between two electronic states, the lattice distorts in response. Large displacements occur near the defect, inducing displacements on neighboring atoms, and so on, producing small displacements on atoms hundreds of Ångströms from the defect. Describing these displacements accurately is challenging for first-principles calculations due to limited supercell sizes. Here we demonstrate a procedure to efficiently obtain dilute-limit atomic displacements, using information obtained in typical modest supercells. In our approach, the atomic displacements are first obtained in a small supercell and converted into a force difference under a harmonic potential energy surface. The force difference and phonon modes at different ${\bf q}$-points are then unfolded into the Born-von Kármán supercell to obtain the dilute-limit atomic displacements. We critically analyze the convergence behavior of the force difference and study possible electron density differences that give rise to those forces, arguing that modest supercells are sufficient for internal transitions and bound-exciton transitions. Two example applications of our approach are given: (1) we calculate the luminescence spectrum of the NV center in diamond and the T center in Si and (2) we obtain dilute-limit configuration coordinate diagrams for these defects. In particular, we find that coupling to acoustic phonon modes reduces the accepting-mode frequency in the configuration coordinate diagram. Our work provides the missing ingredients necessary to address truly dilute-limit transitions at defects.
Read the full abstract
General Relativity and Quantum Cosmology
Thermodynamic Criticality in FLRW Cosmology with Non-extensive Loop Quantum Gravity Entropy
We investigate thermodynamic criticality in a spatially flat Friedmann-Lemaître-Robertson-Walker universe with a non-extensive Loop Quantum Gravity inspired entropy on its apparent horizon. Using the full Kodama-Hayward temperature and the unified first law, we derive the modified Friedmann dynamics and construct the corresponding horizon equation of state. We see that a finite physical critical point only appears in the non-extensive branch $q>1$, while in the Bekenstein-Hawking limit $q\to1$, the critical point is continuously pushed to $v_c\to\infty$ and $T_c\to0$. Along the same spinodal curve, both the constant-pressure heat capacity and the isothermal compressibility become unbounded, and the extremum of this curve is located at the critical point. Below the critical temperature, the Gibbs free energy develops several thermodynamic branches, and the phase coexistence for different horizon states is verified by the equality of temperature, pressure, and Gibbs free energy. The critical exponents are $(\alpha_{cr},\beta_{cr},\gamma_{cr},\delta_{cr}) =(0,1/2,1,3)$, indicating that the system belongs to the standard mean-field universality class despite the non-algebraic form of the equation of state. The normalized Ruppeiner curvature diverges precisely on the spinodal curve and shows critical scaling $R_N\sim-|v-v_c|^{-4}$ on the critical isotherm, and $R_N\sim-|t|^{-2}$ on the critical isochore. Finally, the critical expansion scale is given by $H_c^2=2(\sqrt{5}-2)|\beta|$, which implies that the thermodynamic critical point occurs when the entropy deformation is of order unity. These results establish a self-consistent critical structure for the effective thermodynamic state space of the cosmological apparent horizon and retain the mean-field critical universality.
Read the full abstract
Digital Libraries
Measuring the Installed Base: Nordic Health Dataset Catalogues Against HealthDCAT-AP Release 7
The European Health Data Space requires member states to publish machine readable descriptions of the health datasets available for secondary use, and the European Commission publishes HealthDCAT-AP as the metadata profile those descriptions are meant to satisfy. The profile has been designed and validated against curated examples, never against the catalogues already live. We report that measurement for the Nordic region. On 25 August 2026 the 11 Nordic national catalogues harvested by the European data portal held 2,811 dataset descriptions carrying the EU health theme, and none satisfies all eight properties HealthDCAT-AP Release 7 makes mandatory on a dataset. Three of the eight are present on exactly zero records across five countries. Set beside the portal's own quality assessment, which validates DCAT-AP and never mentions the health profile, this is not a health extension skipped on top of sound generic practice: no Nordic catalogue reaches the assessment's top rating band, 5 of the 11 are reported at zero per cent DCAT-AP compliant, and the properties surviving in both layers are the ones a human types into a form, not the ones needing a value bound to a controlled vocabulary. Two further results follow. Finland contributes 2,259 descriptions to the European portal of which 1,146 carry a theme, and not one uses the EU theme authority vocabulary, so a European health filter returns no Finnish dataset at all. Separately, the authority namespace answers HTTP 200 with a well formed empty document for terms it never defined, letting 1,238 datasets across the wider portal carry theme IRIs that resolve to nothing while passing any status code check. We publish the vocabulary, the shapes and the harvesters, record every verdict as a dated observation rather than a property of the dataset, and report the five errors this discipline caught before publication.
Read the full abstract
Machine Learning
Leveraging a Foundation Model for the EEG-Based Diagnosis of Alzheimer's Disease
Biological heterogeneity in Alzheimer's Disease (AD) poses a critical diagnostic challenge, particularly for traditional linear methods that fail to capture non-linear neural dynamics. To address this, we propose a diagnostic framework utilizing the Large Brain Model (LaBraM), pretrained on over 2,500 hours of EEG data. By integrating these high-dimensional latent embeddings with a non-linear Random Forest classifier, our approach effectively isolates robust disease markers. Under a rigorous subject-independent 5-fold cross-validation protocol, the method achieves an ROC-AUC of 89.36% +/- 3.49%, PR AUC of 81.45% +/- 4.43%, and Balanced Accuracy of 82.44% +/- 4.34% in distinguishing dementia patients from healthy controls. Notably, this performance uses only 8-second EEG segments, surpassing traditional spectral baselines, including band-power and parameterized oscillatory features (FOOOF). Post-hoc occlusion analysis confirms the model captures clinically validated biomarkers, specifically occipital-frontal Alpha and Theta rhythm degradation. Additional neurophysiological alignment analysis demonstrated that higher LaBraM-predicted dementia probability significantly correlated with worse cognitive performance, greater clinical severity, increased theta and alpha relative power, and higher aperiodic exponent. These findings demonstrate that deep latent representations extract clinically relevant signatures from noisy signals, enabling precise, rapid, and data-efficient diagnosis.
Read the full abstract
Combinatorics
Zero-free columns in character tables of symmetric groups
The rows and columns of the character table of the symmetric group $S_n$ are both naturally indexed by partitions of $n$. Let $D(n)$ denote the number of conjugacy classes of $S_n$ whose column contains no zero entry. The identity column is always zero-free, so $D(n)\geq 1$. It is known that $D(n)\ll n^2$. We prove that $D(n)\ll n^{3/4}$. Second, we prove for almost all positive integers $n$ that $D(n)\ll_B n^{1/2}(\log n)^B$ for every $B>5/6$, with a quantitative bound for the exceptional set, using work of Matomäki and Radziwill. Finally, we offer a heuristic supporting our conjecture that $D(n)\ll_{\varepsilon} n^{\varepsilon}$. AxiomProver formalized the results in this paper in Lean assuming preexisting literature.
Read the full abstract
General Mathematics
Dirichlet Beta Analogues, Sign Alternation, and Lerch Unification of Hyperbolic and Logarithmic Tangent Integrals
We develop a Dirichlet-beta counterpart and a structural extension of the coefficient-extraction framework for hyperbolic and logarithmic tangent integrals. For integers $m\geq n\geq1$ with $m+n$ even, the zeta-type formula \[ \int_0^\infty\frac{\tanh^{m+1}x}{x^{n+1}}\,dx = (-1)^{(m-n)/2} \sum_{p=\lceil n/2\rceil}^{(m+n)/2} \binom{2p}{n}(2^{2p+1}-1) \frac{\zeta(2p+1)}{\pi^{2p}} [u^{m+n-2p}](u\cot u)^{m+1} \] admits the beta-type analogue \[ \int_0^\infty \frac{\tanh^m x}{x^n\cosh x}\,dx = (-1)^{(m-n)/2} \sum_{p=\lceil n/2\rceil}^{(m+n)/2} 2^{2p}\binom{2p-1}{n-1} \frac{\beta(2p)}{\pi^{2p-1}} [u^{m+n-2p}] \frac{u}{\sin u}(u\cot u)^m. \] The proof uses derivative polynomials, a beta kernel, and a coefficient-collapse argument. We then establish strict sign alternation for several coefficient families. In the boundary shifted-hyperbolic case the sign is controlled by generalized Bernoulli polynomials, while the arctanh and logarithmic tangent families are identified with continuous dual Hahn and continuous Hahn polynomials. Their zero distributions turn the observed alternation into structural statements valid for all admissible indices. Finally, with $\lambda(s)=(1-2^{-s})\zeta(s)$, we show that the parallel zeta–lambda and beta identities are the two binary specializations of a single Lerch-transcendent scheme with $\varepsilon\in\{0,1\}$. The same convention $\varepsilon=0$ for the beta case and $\varepsilon=1$ for the zeta–lambda case is used throughout. Besides the shifted hyperbolic, odd-$\sinh$, logarithmic tangent, and $\tanh$ families, this framework yields a further unified pair of reciprocal-arctanh integral formulae on $(0,1)$.
Read the full abstract
Artificial Intelligence
PCFBench: A Diagnostic Benchmark for Product Carbon Footprint Estimation
AI systems are being deployed on high-stakes, domain-specific workflows that demand correctness not just in the final output, but at every intermediate step. One such workflow is estimating a product carbon footprint (PCF), the greenhouse-gas emissions attributable to a physical product. AI agents are increasingly being used to generate PCFs, but existing evaluations score either total emissions (hiding error sources and cancelling mistakes) or sub-tasks in isolation (missing compositional interactions). We introduce PCFBench, the first benchmark to carve PCF modeling into independently-evaluable tasks that require decomposition, retrieval, ontology matching, and numerical extraction. It comprises 614 expert-labelled items across six tasks. Together they probe reasoning under under-specification, conflicting context, and numerical constraints. Across eight frontier LLMs from four providers, no single model dominates. Although the strongest models estimate total product emissions within 2 times of declared totals on 77% of products, this rate drops to 37-58% when the PCF is generated step by step, with only 45-75% obeying mass conservation. These failures undermine the transparency practitioners need to compare products and drive decarbonization. We release the dataset and evaluation harness to support targeted progress.
Read the full abstract
Machine Learning
Beyond Non-IID: Learner–Client Distribution Mismatch in Federated Learning
Federated learning systems are increasingly deployed to facilitate collaborative model training across a heterogeneous client population. Existing practice mostly implicitly assumes that the aggregated client data distribution is representative of the learner's target distribution or that learning from all available clients is uniformly beneficial for the learner distribution. However, such an assumption often does not hold in reality. Traditional client selection strategies in FL literature largely overlook such misalignment, while most existing work on multi-source transfer learning either requires direct access to local data or uses one-shot model/feature aggregation. In this paper, we take the initiative to understand and mitigate the impacts of such learner-client population misalignment. In particular, we consider the practical setting where the learner keeps a small proxy dataset. We observe that client contributions vary significantly across training rounds, and traditional technology is insufficient to identify beneficial sources under multi-source transfer diversity. Then, we propose a dynamic, influence-aware client selection framework that estimates each client's potential utility to the learner's optimization objective using proxy influence signals on a learner-specific proxy set. Via using leave-one-out evaluations, we prioritize the most informative sources of knowledge while controlling the negative impacts of statistical noise and data heterogeneity. Experiments on CIFAR-10 under heterogeneous data partitions demonstrate that our approach consistently outperforms static and dynamic baselines, achieving faster convergence and higher accuracy.
Read the full abstract
Functional Analysis
On Lipschitz Spaces, Semigroup Transference, and Heat Kernel Derivative Estimates in Dunkl–Schrödinger Analysis
In this paper, we study Lipschitz spaces adapted to Dunkl–Schrödinger operators $\mathcal{L} = -\Delta_k + V$, where $\Delta_k$ is the Dunkl Laplacian and $V \ge 0$ belongs to the reverse Hölder class $\mathrm{RH}^q(dw)$ with $q > \max(1, \mathbf{N}/2)$. For $0 < \beta < 2$, we establish the full equivalence between the heat semigroup-based Lipschitz spaces $\widetilde{\Lambda}_{\mathcal{L},k}^{\beta/2}(\mathbb{R}^N)$ and the pointwise weighted Zygmund classes $\Lambda_{\mathcal{L},k}^{\beta}(\mathbb{R}^N)$ defined via the critical radius function $m(\mathbf{x})$. Remarkably, this characterization shows that the spatial smoothness component decouples completely from the underlying reflection group symmetries and root systems. As an intermediate result, we establish new equivalences for inhomogeneous weighted Lipschitz spaces associated with the unperturbed Dunkl Laplacian $\Delta_k$. A central difficulty in this context — arising from the necessity to disentangle the Euclidean distance from the orbit metric $d(\mathbf{x},\mathbf{y})$ — is resolved through a detailed geometric analysis. Finally, we provide pointwise Gaussian upper bounds for the time derivatives of the Dunkl–Schrödinger heat kernel, which are of independent interest for future work in harmonic analysis on Dunkl structures. Crucially, we do not assume the potential $V$ to be $G$-invariant under the action of the Weyl group, which introduces major geometric difficulties resolved here via a local analysis of the orbit metric.
Read the full abstract
Machine Learning
DART-FL: Burst-Aware Multitask Federated Learning under Dynamic Inference Demand at the Edge
Edge intelligence systems increasingly require model training and online inference to coexist on resource-constrained devices, while inference demand can vary substantially across tasks over time. This creates two coupled challenges: sufficient computation must be reserved for inference to maintain service-level objectives (SLOs), while the remaining training capacity should adapt to task-specific demand so that frequently requested tasks can improve earlier during training. We propose an SLO-aware, demand-driven multitask federated learning framework (DART-FL) that jointly adapts the inference-training resource split and task-level training emphasis. At each scheduling interval, DART-FL uses the inference backlog and profiled service capacity to determine the minimum resource allocation required for inference. The remaining training capacity is then distributed across tasks using a queue-aware DPP-inspired scheduler, and the resulting task allocations are mapped to dynamic loss weights. This allows tasks experiencing higher inference demand to receive greater training emphasis in earlier communication rounds. Clients train a shared backbone with task-specific heads, and the complete multitask model is aggregated through FedAvg. We evaluate DART-FL using Stanford Cars and Oxford Flowers 102 under both synthetic and real Alibaba trace-derived workloads. Results show that DART-FL dynamically adapts the inference-training resource split to time-varying inference demand and shifts the learning progress of high-demand tasks toward their burst periods, improving model accuracy when those tasks are frequently requested while maintaining comparable long-term multitask performance.
Read the full abstract
Quantum Physics
Efficient Quantum Simulation of Variable-Coefficient Transport with Continuous Source Injection
Quantum time-marching algorithms for transport PDEs often represent variable coefficients and forcing through register-expanding dilations, block-encoding oracles, or repeated postselection. We present an alternative algorithm for a forced variable-coefficient advection-diffusion equation in flow-inspired skew-symmetric form that incorporates spatially varying velocity, viscous dissipation, and persistent source injection with a peak logical requirement of $n_q+1$ qubits. A centered skew-symmetric discretization makes the advection operator strictly skew-Hermitian for arbitrary velocity profiles, enabling an ancilla-free unitary realization using a Gray-code Trotter sequence of controlled-$R_y$ rotations. Diffusion is applied in the Fourier basis through a uniformly controlled rotation on one postselected ancilla, which is measured, reset, and reused between the two diffusion half-steps, while the source is incorporated classically through second-order Strang splitting. Statevector simulations for $N=16$ and $32$ recover second-order temporal convergence against high-accuracy classical solutions, while Richardson extrapolation gives fourth-order accuracy and reduces kernel calls by factors of four to fourteen. Independent tests through $N=256$ confirm second-order spatial consistency. We further show that the per-step ancilla failure probability is proportional to the instantaneous viscous dissipation rate, making postselection cost self-regulating over a fifty-fold viscosity range. Stable evolution is demonstrated for $5\times10^4$ time steps without observable secular error growth, while Gray-code advection accounts for $71$–$95\%$ of transpiled controlled-NOT gates. The fixed-width kernel provides a qubit-efficient building block for near-term hardware studies, although classical readout and state re-preparation remain the main obstacles to coherent multistep evolution.
Read the full abstract
Biomolecules
Dark energy: the cost of function in protein evolution
The evolutionary fate of proteins is driven by both folding stability and biological function, dual constraints that often conflict, creating frustration and imposing functional costs beyond stability. These costs can be captured by a "dark energy": the difference between the evolutionary energy of protein sequences and their physical folding energy. Recent advances in deep mutational scanning, protein language models, and inverse-folding models have enabled the quantification of dark energy across the protein universe. We review the computational and experimental approaches that disentangle folding and function at scale, revealing a dark energy component and providing new insights into how biological information flows from sequence to structure to function and back to sequence.
Read the full abstract
Performance
Adaptation Fidelity of SPEC CPU2026
Standardized benchmarks are often criticized for not being "real workloads," but this critique is rarely backed by data. This paper provides the first systematic, quantitative analysis of the "fidelity gap" between the SPEC CPU2026 suite and its original, upstream open-source counterparts. We compile both the SPEC benchmarks and their upstream applications and execute them with official input workloads under two scenarios: a single-copy latency run and a 192-copy throughput run. Our findings show that most benchmarks exhibit high fidelity in single-copy runs, while a few outliers reveal the impact of SPEC's adaptation process. The multi-copy results further highlight the necessity of this adaptation: several benchmarks become significantly more efficient than their upstream versions under heavy load, underscoring the importance of I/O reduction. This work offers data-driven validation of SPEC's methodology, showing that the fidelity gap is not a flaw but a quantifiable consequence of enforcing portability, determinism, and CPU-centric measurement.
Read the full abstract
Exactly Solvable and Integrable Systems
A new family of Darboux integrable partial differential equations
A new family of Darboux integrable partial differential equations is constructed. The minimal orders of integrals in both characteristics can simultaneously be arbitrarily high for equations within this family. This suggests that the complete list of Darboux integrable equations may contain substantially more equations than are currently known.
Read the full abstract
Probability
Proof of the Lyons–White Conjecture
Let $D_n$ be the dihedral group of order $2n$. Consider a continuous-time random walk on $D_n$ driven by arbitrary symmetric rates whose support generates $D_n$. For $p\in[1,\infty]$, we say the pair $(D_n,p)$ is rate-monotonic if for each fixed time $t$, the $\ell^p$-distance between the random walk's distribution at time $t$ and the uniform distribution is monotonically decreasing as a function of the rates. Lyons and White proved that $(D_n,2)$ and $(D_n,\infty)$ are rate-monotonic. Somewhat counterintuitively, they found several pairs $(D_n,p)$ with ${p\in[1,1.997]\cup[2.001,3.999]\cup[4.001,5.995]}$ that are not rate-monotonic, and they asked whether any such pairs exist with $p=4$ or $p=6$. We resolve their question, proving that $(D_n,2m)$ is rate-monotonic for all positive integers $m$ and $n$. In fact, we prove a generalization of this result to a broader family of groups that includes generalized dihedral groups, dicyclic groups, and generalized quaternion groups. In the other direction, we prove that for every real $p\geq 1$ that is not an even integer, there exists a positive integer $n$ such that $(D_n,p)$ is not rate-monotonic. The results of this paper were formally verified in Lean by AxiomProver assuming standard literature.
Read the full abstract
Optimization and Control
Solution Methods for Infinite-Dimensional Generalized Disjunctive Programming
Generalized disjunctive programming (GDP) expresses mixed discrete-continuous decisions through Boolean indicators and disjunctions, and can be systematically solved via a library of methods proposed in the literature. The recent InfiniteGDP abstraction lifts this modeling layer to continuous domains such as time, space, and uncertainty, but only the big-M and hull reformulations, the two endpoints of the relaxation spectrum, have been generalized to the infinite setting. This work closes this gap by generalizing four other GDP solution methods to infinite-dimensional optimization: the multiple big-M reformulation, P-split reformulation, cutting plane reformulation, and the logic-based outer approximation algorithm. It further proposes MBM-GP, a novel Gaussian-process variant of multiple big-M that learns the big-M function over the infinite domain from a small subset of the subproblem solves. Moreover, these approaches are implemented in the Julia package InfiniteDisjunctiveProgramming.jl. The methods are benchmarked on case studies arising in dynamic and stochastic optimization. The results demonstrate how the generalized solution methods can outperform big-M and hull, with MBM-GP retaining the tightness of multiple big-M at a fraction of its reformulation cost.
Read the full abstract
Dynamical Systems
Infinitely many graph manifolds with unique geometrical piece that admit arbitrarily many Anosov flows
Anosov flows have a long and rich history, firstly motivated by the study of geodesic flows in negative curvature surface by Anosov and Sinai. Not every closed manifold admits an Anosov flow for well-known reasons: the fundamental group of a 3-manifold \(M\) admitting an Anosov flow must have exponential growth, and \(M\) must be universally covered by \(\mathbb{R}^{3}\). Nevertheless, there are sufficient mechanisms for constructing distinct Anosov flows on admissible 3-manifolds, such as Dehn-Goodman-Fried surgery or playing with hyperbolic building blocks. A central problem in the field has been to determine the number of Anosov flows that can be supported by a single manifold. The question of whether there exists an infinite set of pairwise non-equivalent Anosov flows on a 3-manifold remains open to this day. However, there are several papers proving the existence of a manifold $M_n$ that admits $n$ pairwise inequivalent Anosov flows for any natural number $n$. In all known examples, the manifolds $M_n$ are composed of several geometric pieces. In the present paper, we prove the existence of a countable number of graph manifolds $M_{k,n}$, $k \in \mathbb{N}$ with a single geometric piece, each of which admits $n$ pairwise non-equivalent transitive Anosov flows. All previously known constructions of different flows on the same graph manifold were based on gluing geodesic flows. The nature of the flows constructed in this paper is completely different; they are constructed from a single hyperbolic plug, which is a suspension over a Morse-Smale diffeomorphism on a surface.
Read the full abstract
Optimization and Control
On the Computational and Statistical Efficiency of the Empirical Maximum Entropy on the Mean Method
The Maximum Entropy on the Mean (MEM) method provides a flexible computational framework for solving inverse problems by combining data fidelity with entropy-based regularization. In practice, however, the prior distribution is typically unknown but can be estimated from data, giving rise to the empirical MEM method. We establish a parametric convergence rate of $O(n^{-1/2})$ in expectation for empirical MEM, improving upon the previously established $O(n^{-1/4})$ guarantee by King-Roskamp et al. (2026). Our proof is based on a novel stability analysis of the primal and dual optimization problems under perturbations of the underlying probability measure, relying only on foundational tools from convex analysis and probability. We further show that the MEM dual problem admits a reformulation as an expected risk minimization problem, thereby placing MEM within the modern framework of stochastic optimization and enabling scalable stochastic gradient algorithms for large-scale inverse problems. Together, these results place empirical MEM as a statistically and computationally efficient methodology for data-driven inverse problems.
Read the full abstract
Machine Learning
RiskBlend: A Multi-Signal Framework for Test Input Prioritization in Machine Learning Regression Testing
When machine learning classifiers are retrained, inputs correctly classified by the previous model version may be misclassified by the updated version, creating regression faults that are costly to detect because verifying predictions against ground truth may require human annotation, expert review, or expensive simulation rather than inexpensive model inference. Test input prioritization addresses this problem by ranking inputs so that a limited verification budget reveals as many regression faults as possible. Existing approaches rely predominantly on single-model confidence scores and do not exploit how predictions, decision boundaries, and local neighborhoods change between model versions. We propose RiskBlend, a classifier-agnostic prioritization framework that combines four complementary risk signals: historical failure patterns, prediction shift, decision-boundary shift, and neighborhood change. These signals are combined using validation-learned APFD-squared weighting. Across four datasets, five classifiers, four regression-update scenarios, and 15 random seeds, totaling 1,200 experimental configurations, RiskBlend achieves the highest average APFD in all 80 dataset-classifier-scenario combinations, with improvements of up to 0.32 APFD over the strongest baseline. Confidence-based methods remain competitive primarily for linear classifiers on sparse categorical features, which we attribute to feature-space geometry. The results show that cross-version behavioral signals provide important complementary information for prioritizing regression faults in machine learning systems.
Read the full abstract
Software Engineering
Operationalizing Regulations into Code: A Model to Enhance Governance and Compliance in LLM Selection for Software Engineering
Integrating Large Language Models (LLMs) into the Software Development Life Cycle (SDLC) can improve developer productivity, but it also introduces security, privacy, and compliance risks during model selection. Regulations and frameworks such as the EU AI Act, the NIST AI Risk Management Framework (RMF), the General Data Protection Regulation (GDPR), the Lei Geral de Proteção de Dados (LGPD), and ISO/IEC 42001 establish obligations that are often difficult to translate into operational criteria for technical decision-making. This paper proposes a model to support governance and compliance in LLM selection for software engineering projects. The model is developed through Design Science Research (DSR) and is structured in three layers: (i) regulatory requirements, (ii) organizational governance capabilities, instantiated by a multi-criteria decision matrix with knock-out and weighted scoring criteria, and (iii) productivity and sustainability outcomes, operationalized by the LLM governance assessment protocol (PAG-LLM). A regulatory feedback loop connects operational results back to the normative layer, enabling iterative refinement of the model. A pilot evaluation with 20 adversarial scenarios based on Common Weakness Enumeration (CWE) and the OWASP Top 10 suggests distinct risk profiles between commercial cloud-based LLMs and local open-source LLMs. The results provide preliminary evidence that regulatory disqualification logic, particularly K.O. criteria, can prevent the selection of technically competitive models that nonetheless pose unacceptable compliance risks, demonstrating the feasibility of governance-oriented LLM selection in software engineering projects.
Read the full abstract
Optics
Extending Silicon Avalanche Photodetection Beyond $2~\mu$m by Direct GeSn Integration
Silicon avalanche photodiodes provide a technologically mature platform for sensitive photodetection, but their spectral response is intrinsically limited by the silicon bandgap. Extending their operation into the infrared requires the integration of narrow-bandgap absorbers while preserving efficient avalanche multiplication and compatibility with silicon processing. Here, we propose and demonstrate a monolithic approach that combines direct, buffer-free growth of GeSn on silicon with a lateral thin-junction separate-absorption-multiplication architecture. The GeSn layer, with a Sn composition reaching 6 at.%, extends optical absorption to a wavelength of $2.6~\mu$m, while avalanche multiplication is spatially confined to an ion-implanted silicon lateral junction. This separation enables independent control of infrared absorption and carrier multiplication without the thick Ge virtual substrates conventionally used for GeSn epitaxy. GeSn-on-Si avalanche photodiodes exhibit low pre-breakdown dark current, stable breakdown at 72 V independent of device diameter, and clear infrared photoresponse extending beyond $2~\mu$m. At 78 K, the devices exhibit external quantum efficiency exceeding 100%, reaching 163% at $1.55~\mu$m, providing direct evidence of avalanche multiplication. These results establish direct integration of narrow-bandgap group-IV absorbers with silicon multiplication regions as a scalable strategy for extending the spectral reach of silicon avalanche photodetection, opening a route toward monolithic infrared detectors for sensing, imaging, communications, LiDAR, and quantum photonics
Read the full abstract
High Energy Physics (Theory)
Universal and Transcendental Structures in Protected ABJM Two-Point Functions
We consider two-loop corrections to two-point functions of protected scalar operators in ABJM theory. We infer a compact finite-rank formula valid for operators of arbitrary dimension and multi-trace structure. The result is governed by the exact tree-level metric and a simple kernel on the Young lattice, while the planar limit reduces to a simple partition-theoretic rule. The required integrals organize into uniformly transcendental combinations, providing evidence for uniform transcendentality of these protected two-point functions.
Read the full abstract
Atomic Physics
Quantified absorption of laser light by silver atoms in a hollow-cathode lamp
We present a quantitative study of laser absorption on the $5{^{2}}\mathrm{S}_{1/2} \rightarrow 5{^{2}}\mathrm{P}_{3/2}$ transition of silver in a hollow-cathode lamp. Spectra obtained using a weak probe are measured and compared to a simple theoretical model. We fit our results to extract an effective temperature and number density associated with the silver in the lamp and confirm that the spectra are dominated by Doppler broadening. We consider the broadening effect of collisions between silver and the neon buffer gas in the lamp, and establish that these effects are small but modify the optimum effective parameters for the model. Our work results in a simple, quantitative, and predictive model for laser absorption in the lamp.
Read the full abstract
High Energy Astrophysical Phenomena
Magnetospheric birefringence breaks the rotating-vector model and depolarises X-ray pulsars
Accreting X-ray pulsars have long been expected to show large linear polarisation below the cyclotron resonance. In the standard picture, this follows from two facts: radiative opacities in the strongly magnetised neutron star environment depend strongly on polarisation mode, and the emission comes from compact polar regions. The emerging radiation should therefore be locally polarised and add almost coherently, with a polarisation angle tracing the projected magnetic field as in the rotating-vector model. Recent X-ray polarimetric observations, instead, show modest polarisation fractions and position angle swings that often depart from this simple picture. Here we show that propagation through the magnetosphere of an accreting neutron star can naturally produce both effects. Using general-relativistic ray tracing and Stokes-parameter transport, we find that photons emitted from compact polar regions acquire ray-dependent polarisation angle offsets in two ways. First, nearly field-aligned trajectories can cross local non-adiabatic pockets inside the QED birefringent region. Second, photons that intersect the plasma-loaded accretion flow accumulate additional phase shifts between plasma normal modes. Both effects broaden the distribution of polarisation angles across the ray bundle and reduce the hotspot-integrated linear polarisation, while also producing systematic deviations from rotating-vector-model expectations. Magnetospheric propagation should therefore be treated as part of the polarimetric transfer problem in accreting X-ray pulsars, and may help explain the observed low polarisation fractions and anomalous angle swings.
Read the full abstract
Sound
Evaluating Loss Functions in Differentiable Out-of-Domain Sound-Matching with Partial Parameter Distance
In out-of-domain (OOD) sound-matching, a synthesizer is optimized to mimic a sound it did not generate. OOD evaluation of loss functions is underexplored in part because the standard "parameter loss" metric requires a shared parameter space between target and imitator, which OOD settings lack. We introduce Partial Parameter Distance (PPD), which applies parameter loss only to the critical parameters that mismatched synthesizers share (e.g., filter cutoffs), enabling automatically evaluated OOD experiments; we verify its results with blinded listening tests. Across seven scenarios involving band-pass filtering, amplitude modulation, and pitch-bending, we evaluate four differentiable loss functions (SIMSE_Spec, L1_Spec, JTFS, DTW_Envelope). Loss-function effectiveness remains tightly coupled to the method of synthesis: SIMSE_Spec excels at filter-cutoff recovery, DTW_Envelope at amplitude-modulation recovery, and JTFS at smooth pitch trajectories. Parameter-based evaluation agrees with listening tests on the top-ranked loss function in five of seven scenarios, demonstrating its utility as a diagnostic tool.
Read the full abstract
Number Theory
Diophantine m-tuples of Triangular Numbers
A $m$-tuple with the property $D(n)$ is a tuple of $m$ positive integers $(a_1, a_2, \dots, a_m)$ such that $a_i a_j + n$ is an square, for $1 \le i < j \le m$. The $k$th triangular number is $T_k = \frac{k(k+1)}{2}$ for nonnegative integers $k$. We consider $D(1)$ tuples consisting only of triangular numbers. We prove the nonexistence of any $D(1)$ triangular quadruple and describe an algorithm to generate an infinite family of $D(1)$ triangular triples, which we conjecture contains all $D(1)$ triangular triples. We also consider general $D(n)$ tuples. To aid with computational difficulties, we present an efficient algorithm, using Generalized Pell Equations (GPEs), to determine whether $T_a$ is in a $D(n)$ triangular pair, which runs in $O(a^{1/2})$ time. We then prove that no $D(n)$ triangular pair exists for $n \equiv 2,5 \text{ (mod } 9\text{)}$, and discuss other values of $n$ for which there appear to be no $D(n)$ triangular pairs. We also show that our $D(n)$ equation has solutions in all $\mathbb{Q}_p$, for $p \neq 3$. We then present progress on determining a general criteria on $n$ for which no $D(n)$ triangular pairs exist.
Read the full abstract
Algebraic Geometry
Finding large families of rational curves through Bend-and-Break
We present a new construction that allows us to break off large-degree rational curves from families of higher genus curves. Our construction and results deepen the connection between rational curves and positivity of the anticanonical divisor. Specifically, we show that varieties with large Fujita invariant admit large families of rational curves. We also construct free rational curves on certain singular Fano varieties. As an explicit consequence of our results, we prove that for a general Fano hypersurface of index at least 3, all spaces of genus g curves of sufficiently large degree have the expected dimension.