Peer Revenue Now accepting positions

Papers

Fresh from arXiv, typeset properly. Highlighted words trade on the market.

All of arXiv

Monday 9 November

Accelerator Physics

Longitudinal beam dynamics II. Lecture 1: single and double rf systems

In the first part of this lecture, the main characteristics of synchrotron motion, namely, synchronous particle and phase, bucket area and emittance, are introduced. The equations of longitudinal particle motion in a single rf system of a proton synchrotron are rederived and the phase equation is analyzed for small and large amplitude particles. The particle motion below and above transition energy is also considered. The main applications and operation modes of a double rf system are discussed in the second part of the lecture. The intensity effects and beam instabilities are the subject of the following lecture.
Read the full abstract

Algebraic Geometry

Lipschitz inner rates of normal surface singularities

In this article, we connect the local Lipschitz geometry of a normal surface singularity with a rational homology sphere link to the analytic semigroup $S_{an}(\tilde{X})$ associated with a convenient resolution. Under certain restriction, we provide an explicit computational characterization of the Lipschitz rate $q_v$ associated with an irreducible exceptional curve $E_v$ in terms of the semigroup. We also show that removing the restriction we get examples when $q_v$ cannot be determined from the semigroup. Interestingly enough, this phenomenon already happens for rational singularities. Indeed, we exemplify that the inner rates of rational singularities are not topological.
Read the full abstract

Artificial Intelligence

Reconstructing hourly traffic on every street of a capital from a sparse counting network: a graph neural network applied to Paris

Micro-scale air quality models require hourly traffic data on every road segment, whereas permanent counters cover only a small part of the network. A graph neural network (GNN) was developed to reconstruct hourly traffic flows on the 32,163 road segments of Paris intra-muros from the 1,573 segments equipped with a counter (4.9%). The model adapts the Graph Neural Network for Urban Interpolation (GNNUI) to a real, sparse counting network. Flows are decoded on the edges of the junction graph (one edge per segment, both directions summed), a traffic climatology serves as an offset in a zero-inflated negative-binomial likelihood, and training proceeds by masking counters. Eight-fold spatial cross-validation, with counters sharing a published series merged before splitting, distinguishes interpolation at a withheld counter in an instrumented district (R${}^2$ = 0.89, error of a quarter of the mean flow) from extrapolation to a district without counters (R${}^2$ = 0.74, a third of the mean flow). These scores hold on main axes where counters are located, while relative errors are about 50% on the few counted local streets. Ablations show that the climatology offset tends to help extrapolation. Road attributes, time cycles and the live state of the counting network matter most, while weather inputs have no significant effect. The masking rate is the only inherited hyperparameter with a clear effect. When evaluated on January–August 2026 (outside the training period), the frozen model stayed around the error band of previous years.
Read the full abstract

Robotics

Safety-Constrained Intelligent Control of a 2-DOF Robotic Manipulator Using Control Barrier Functions

Control barrier function (CBF) filters depend on dynamic models that can be inaccurate when their parameters and external forces are uncertain. This paper investigates a radial basis function (RBF) neural network compensator whose output is shared by a Cartesian tracking controller and a joint-space CBF filter. An auxiliary state driven by the difference between requested and applied torques corrects the learning signal, reducing the direct influence of filter intervention on adaptation. A two-link manipulator is evaluated through constrained regulation, ablations, matched uncertainty and disturbance trials, and operating-range and timing studies. The results demonstrate that the proposed controller improves tracking while maintaining joint, velocity, and torque limits under parameter uncertainty, and that both the shared compensation and the auxiliary correction contribute to this improvement.
Read the full abstract

Artificial Intelligence

Detecting Shadow AI: Unsanctioned LLM Agents on Enterprise Endpoints: A Taxonomy, an Open Detector and a Controlled Evaluation

Shadow IT, the use of technology that the IT organisation has not approved, is a long-standing governance problem. We argue that autonomous AI agents can make it grow much faster: agents make building tools nearly free, act with their user's privileges, and leave behind scripts, automations and MCP servers that are themselves unsanctioned. A simple compounding model shows that, under its assumptions, shadow IT grows exponentially once agentic artifacts reproduce faster than the organisation retires them; retiring them starts with detecting them. We then ask whether unsanctioned agents can be detected on Windows endpoints. We propose a taxonomy of nine classes of agents and six layers of endpoint footprint, and argue that rules based only on the structure of processes and connections cannot separate an agent from its behavioural twin, the same program talking to a non-AI service. We present Agent Detector, an open-source Windows detector that anchors behaviour to the identity of the remote service and correlates it with configuration, installation and execution history, using a catalog of 39 tools and 184 LLM endpoint entries; it informs users through a notice they must acknowledge instead of blocking. On a controlled bench of 12 synthetic agents, 3 AI components and 9 benign workloads, repeated 5 times on one workstation, two behavioural rules flagged every behavioural twin, while Agent Detector flagged none and detected every agent in the majority of runs. Its limits are temporal: a connection to a provider behind a shared content-delivery network is attributable only while the DNS cache holds the provider's name, one to five minutes, and an intermittent agent was seen by about one scan in three. A scan costs about 154 ms. The bench is too small to rank detectors; it maps what each kind of evidence can see: unsanctioned agents can be partly detected, and not by behaviour alone.
Read the full abstract

Functional Analysis

Vector metrics with values in pre-Riesz spaces I: transfer of order convergence along vector lattice covers

Vector metric spaces in the sense of Çevik and Altun take values in a vector lattice, but their axioms use only the order of the value space. We develop the theory for vector metrics with values in an Archimedean pre-Riesz space $E$, i.e.\ an ordered vector space admitting a bipositive order dense embedding $i$ into a vector lattice $Y$. We ask when an $E$-valued vector metric $d$ and the vector-lattice-valued metric $i\circ d$ have the same convergent and Cauchy nets and sequences. Relatively uniform convergence transfers along every cover. For nets, $\mathrm{o}_3$-convergence transfers if and only if the cover has an intrinsic domination property. This property holds for pervasive covers, by a result of Kalauch and van Gaans, for uniformly dense covers and in finite dimensions; it fails for an example of Kalauch and van Gaans. Sequential convergence transfers under a countable domination property implied by uniform density, but not under pervasiveness. For uncountable $\Gamma$, the vector lattice $\mathbb{R}\mathbf 1\oplus c_{00}(\Gamma)$ is pervasive in its Dedekind completion $\ell^\infty(\Gamma)$ without the transfer, while for countable $\Gamma$ the transfer holds; the difference is explained by order separability. In finite dimensions, with a closed generating cone, all these convergences coincide with norm convergence in every cover. For $C^1[0,1]$-valued integral metrics, positivity of the kernel on the diagonal forces all order convergences to coincide with $L^1$-convergence. For the Volterra kernel $(t-s)_+$, by contrast, $L^1$-convergence, relatively uniform convergence and order convergence are pairwise different, and we describe each explicitly.
Read the full abstract

Sound

AudioCodex: Streaming Autoencoders for Live Audio

We introduce AudioCodex, a fully causal autoencoder for live generative audio applications that compresses stereo audio into compact continuous latent sequences at ~10 Hz. It combines a Mean Flow magnitude autoencoder with latent hypersphere projection, bounded-noise augmentation, representation alignment, and adversarial phase recovery to learn representations that support both efficient downstream generation and high-quality reconstruction. Through extensive ablations, we show that AudioCodex outperforms the state-of-the-art in terms of generation-reconstruction trade-off, whilst remaining strictly causal in both encoding and decoding. This work paves the way towards truly real-time interactive and streaming generative audio applications.
Read the full abstract

All authors Marco Pasini, Javier Nistal, Cyran Aouameur, Stefan Lattner, George Fazekas

Artificial Intelligence

TANSAR: Editable and Attributable Correlation Matrix Memories under Correlated Keys, with Rule-Constrained Multi-Hop Knowledge Editing

Correlation Matrix Memories (CMMs) store associations as sums of rank-one outer products, providing closed-form addressability, exact deletion, and linear attribution. While classical analyses assume random, orthogonal keys, semantic encoders produce correlated representations. We analyse CMMs under correlated keys using frame-theoretic coherence: (i) capacity bounds yielding an encoder-independent limit $n \le d(1+\varepsilon^2)$ for mean squared crosstalk $\varepsilon^2$; (ii) an exact characterisation of the collateral effect of deleting an association, guaranteeing that deletion never invalidates certified associations; and (iii) a geometric transparency bound demonstrating why attribution requires candidate sets of centred keys. On this basis, we introduce Topology-Aware Neuro-Symbolic Associative Retrieval (TANSAR). TANSAR traverses a directed acyclic association graph, prunes invalid paths via sound ground Horn constraints, and injects a representation derived strictly from the verified path into the residual stream of a frozen language model, guarded by a Clopper–Pearson selective-risk gate. Controlled experiments confirm all theoretical bounds. In multi-hop knowledge editing, Horn pruning reduces answer contradictions from 25.1% to 0.5%, whereas Hopfield retrieval incurs 24% contradictions when stale associations remain stored (and 0% if stale associations are removed). An end-to-end edit transaction requires 13.6 ms on a standard CPU and 2.37 ms on GPU. The symbolic component matches gold labels across all 600 ProofWriter theories and achieves 94.0% agreement on FOLIO, demonstrating sound and efficient multi-hop associative retrieval with exact unlearning.
Read the full abstract

All authors Abdulhady Abas Abdulla, Jiyan Salim Mahmud, Fatemeh Daneshfar, Hussein Mohammed Ali, Mourad Oussalah, Muhammad Bilal

High Energy Physics (Phenomenology)

Magnetic moments of the $\Sigma_c$ and $\Sigma_b$ baryons from external-field QCD sum rules

We compute the magnetic moments of the $\Sigma_c$ and $\Sigma_b$ isotriplets with external-field QCD sum rules. The heavy-quark propagator is kept with its full mass dependence and with its couplings to the photon, to soft gluons and to both simultaneously, and the magnetic sum rule is normalized by the mass sum rule of the same interpolating field, which removes the residue and gives the baryon masses within a few percent. The quark-condensate magnetic susceptibility $\chi$, which otherwise dominates the uncertainty, is fixed by the proton magnetic moment computed with the same interpolating field; the neutron moment then follows to about $5\%$. Uncertainties from the QCD inputs are evaluated with a Monte Carlo analysis, and those from the interpolating field, the Borel window, the scale of the heavy-quark mass, the calibration and the normalization are added. With $\overline{\rm MS}$ heavy-quark masses we obtain $\mu_{\Sigma_c^{++}}=2.37\pm0.65$, $\mu_{\Sigma_c^{+}}=0.50\pm0.15$, $\mu_{\Sigma_c^{0}}=-1.39\pm0.37$, $\mu_{\Sigma_b^{+}}=2.16\pm0.65$, $\mu_{\Sigma_b^{0}}=0.55\pm0.16$ and $\mu_{\Sigma_b^{-}}=-1.06\pm0.32$ in nuclear magnetons; at this order the results depend sizably on the heavy-quark mass scheme. The moments are carried by the light diquark; the heavy-quark contribution, whose separation depends on the interpolating field, is $-0.13\pm0.02\,\mu_N$ for $\Sigma_c$ and $+0.013\pm0.003\,\mu_N$ for $\Sigma_b$.
Read the full abstract

Cosmology and Nongalactic Astrophysics

Ultralight scalar field domain walls and Galileo GNSS atomic clocks

In this article, we study the impact of an Earth-sized ultralight scalar field domain wall (DW) on the European Galileo Global Navigation Satellite System constellation and the feasibility of a related detection. In particular, a set of statistical procedures is developed and implemented to characterize the statistical properties of the background noise for a subnetwork of $N$ Galileo satellite clocks, including a false-alarm rate estimation and a detection efficiency analysis. These methods allow to implement a coincidence analysis and establish a preliminary statistical framework for a robust and controlled search for domain-wall crossings in clock data, and for deriving constraints on the associated fundamental particle physics parameters. The activity is addressed from the point of view of the scientific objectives of the Galileo for Science (G4S$\_2.0$) project.
Read the full abstract

All authors Alessandro Di Marco, Feliciana Sapio, Massimo Visco, David Lucchesi, Marco Cinelli, Emiliano Fiorenza, Carlo Lefevre, Pasqualino Loffredo, Marco Lucente, Carmelo Magnafico, Roberto Peron, Francesco Santoli, Francesco Vespe

Materials Science

Kagome Flat Bands from Self-Assembled Water on Non-van der Waals 2D Materials

The properties of two-dimensional (2D) materials are governed not only by their intrinsic electronic structure but also in particular by their surface chemistry and interactions with the environment. The surfaces of non-van der Waals (non-vdW) 2D materials are intrinsically reactive and may require chemical stabilization under realistic experimental conditions. Here, we perform a systematic first-principles study of water passivation of non-vdW 2D materials and demonstrate that molecular water adsorption is energetically and dynamically favored for a wide range of systems. Rather than dissociating, water molecules frequently remain intact and form adsorbed ordered surface overlayers. These overlayers can adopt a kagome lattice geometry, leading to the emergence of flat electronic bands directly at the valence-band edge. Analysis of the electronic structure reveals a strong intermolecular coupling between neighboring water lone pairs within this kagome topology, while the locked molecular orientations simultaneously produce an electrostatically frustrated surface potential. These findings establish water not only as a stabilizing agent but also as an active element for electronic structure engineering of non-vdW 2D surfaces.
Read the full abstract

All authors Tom Barnowsky, Madeleine Christie, Anastasiia Nihei, Rico Friedrich

Distributed Computing

Adapting Multi-Stage Distributed Real-Time Systems Design to Vehicular Edge Computing

Vehicular Edge Computing (VEC) aims to support computation-intensive applications in intelligent transportation systems while meeting stringent latency requirements. When processed remotely, applications require access to different resource types in a specific sequence, forming a multi-stage processing pipeline. This structure closely resembles multi-stage distributed applications executing across heterogeneous resources, making established multi-stage distributed real-time scheduling techniques well suited for modeling VEC systems. The resulting model accommodates diverse resource configurations and access patterns. Building on this model, we propose a joint mapping and scheduling approach that accounts for vehicle mobility and real-time constraints. The proposed heuristic extends optimal priority assignment for multi-stage systems to the VEC setting, to construct feasible schedules while minimizing local-device energy consumption and ensuring that offloaded jobs meet their deadlines. Extensive simulations show that the proposed approach consistently outperforms both a baseline based on the same delay-composition framework and a deadline decomposition-based approach.
Read the full abstract

Instrumentation and Detectors

VNA-Only Calibration of Waveguide Thermistor Mounts with Self-Consistent Determination of the Thermistor-Port Impedance

A method is presented for determining the calibration factor and the effective efficiency of waveguide thermistor mounts from vector-network-analyzer reflection data alone, without a microcalorimeter, extending the three reflection-coefficient method of the coaxial case. In waveguide the thermistor is coupled to the line through a matching structure and the reference impedance of the thermistor port, $Z_0$, is not known a priori; it is inferred through a power-consistency relation between pairs of states, under the convention that it is real. The method reproduces the factory certificates (U = 2 %) of four K486A mounts within 1.0 % rms and the PTB-traceable values (U = 0.85 to 0.98 %) of a fifth within 0.34 % rms; its expanded measurement uncertainty is 0.5 to 1.16 % (k = 2) for the best specimen and below 2 % for the five. A mount with its waveguide-to-coaxial adapter is calibrated directly, without characterizing the adapter, from the model factor of the mount and the geometric factor measured through the adapter. The scattering matrix implied by the solution is not passive; a minimum-norm correction of the thermistor impedances along the two directions invisible to the calibration factor - the ratios of the efficiencies of the three states - brings it to the floor of passivity, where the efficiency depends on the load only through its mismatch factor, leaves the result unchanged, and gives the sensitivity of the efficiency to the bias resistance in closed form: below $4\times10^{-5}$ $\Omega^{-1}$ in band mean, proportional to the mismatch of the operating state and to the loss of the mount.
Read the full abstract

Statistical Methodology

Bayesian precision factor models for high-dimensional macroeconomic time series

Series in a macroeconomic panel are almost all correlated with one another, which makes correlation a poor guide to which pairs are directly linked. For example, Treasury rates at different maturities are all highly correlated, yet a short and a long rate may be related only through the rates at maturities between them; similarly, sectors of the economy such as output, employment, prices, and interest rates may be related directly or only through other series. Both are questions about the inverse covariance matrix. Recovering it from quarterly data is difficult: contemporaneous dependence is entangled with temporal dynamics, the innovations remain densely dependent once those dynamics are removed, and the number of series is comparable to the number of quarters observed. We propose VAPOR (Vector Autoregressive Precision factOR), a Bayesian model that estimates vector autoregressive dynamics jointly with a factor-structured innovation precision matrix, inferring the partial correlations directly from the fitted precision rather than imposing a zero pattern. Applied to 100 quarterly FRED-QD series, VAPOR learns from the data that each Treasury rate is directly related only to the rates at adjacent maturities, places all 26 selected dependencies within a single FRED-QD group, and attains the lowest held-out prediction error among the estimators compared.
Read the full abstract

Cryptography and Security

An Agile Homomorphic Encryption and ML-KEM Scheme for Privacy-Preserving Health Data Processing

The storage and processing of health data raise serious privacy concerns; therefore, the data must be encrypted. Fully Homomorphic Encryption (FHE) allows computations on encrypted data, which makes it a very promising technology in this context. However, although the efficiency of FHE algorithms has improved in recent years, they remain very complex. The security of several prominent FHE schemes is based on variants of the Learning With Errors (LWE) problem, which is also the basis of Module-Lattice-Based Key-Encapsulation Mechanism (ML-KEM), the new post-quantum standard for asymmetric encryption. In this paper, we propose an agile cryptographic scheme -combining ML-KEM and BGV -that allows efficient implementations in an embedded use case. We describe the parameter selection for the proposed scheme and show its feasibility with a proof-of-concept implementation.
Read the full abstract

Numerical Analysis

Reconstructing Flash-Evoked Potentials in the Human Brain and Tracking Cortical Activity Using Spatiotemporal Filtering: Numerical Simulation Study

Electroencephalographic source reconstruction of visual evoked activity is challenging because distributed, temporally evolving, and overlapping cortical sources generate scalp potentials. We present a controlled simulation framework for flash visual evoked potentials (FVEPs) based on a coupled Jansen–Rit neural mass network and a realistic finite-element EEG forward model. The framework generates a sequential activation pattern across early visual, parietal, and frontal regions and enables direct comparison of reconstructed activity with known ground truth. Standardized Kalman filtering (SKF), change-rate SKF (CrSKF), conventional Kalman filtering (KF), and linearly constrained minimum variance (LCMV) beamforming are compared using parcel-wise time courses, temporal lags, and assessments of spatial focality and Earth Mover's Distance. In the inverse comparison, SKF and CrSKF preserve the principal temporal progression more consistently than KF and LCMV, while CrSKF produces smoother temporal estimates. The results also demonstrate parcel-wise leakage and reduced separation of closely spaced components, emphasizing the need to interpret temporal ordering together with spatial reconstruction quality. The proposed framework provides a reproducible test environment for evaluating dynamic EEG inverse methods under controlled noise conditions.
Read the full abstract

Functional Analysis

Completions of vector metric spaces and the diagonal property of order convergence

A vector metric space carries a distance with values in an ordered vector space $E$; a sequence converges if its distances to the limit are dominated by a sequence decreasing to zero in $E$. We show that, for Archimedean $E$, every $E$-valued vector metric space has a completion if and only if order Cauchy sequences in $E$ converge and $E$ has the diagonal property of order convergence (SDP): for every double sequence $(a^k_n)$ with $a^k_n\downarrow_n0$ there are indices $n_k$ and $e_k\downarrow 0$ with $a^k_{n_k}\le e_k$. In vector lattices this is the classical diagonal property of Kantorovich, equivalently the strong Egoroff property. We show that its standard characterisations remain valid in Archimedean ordered vector spaces, and that it holds if and only if convergence in every $E$-valued vector metric space satisfies the iterated limit axiom. For an ordered Banach space with a closed normal cone, all completions exist if and only if the norm is $\sigma$-order continuous; if the cone has an interior point, all vector metrics are then scalar. By a theorem of Veksler and of Hager and van Mill, for compact Hausdorff $K$ either $K$ is an almost $P$-space and $C(K)$-valued vector metrics are scalar, or there are countable $C(K)$-valued vector metric spaces without completion. A product of ordered vector spaces with (SDP) and nonzero cones has (SDP) if and only if there are fewer than $\mathfrak b$ factors, extending a result of Hager and van Mill for $\mathbb{R}^A$. We also prove permanence results for direct summands and uniformly dense vector lattice covers of pre-Riesz spaces.
Read the full abstract

Computer Vision

TRIAD: Multimodal Retrieval across Abstraction, Constraints, and State

Multimodal retrieval increasingly supports agentic systems, where user requests can involve cross-level relations, combine multiple constraints, and evolve across dialogue turns. In these settings, a highly similar result may still violate the user's requirements, while retrieved content may directly inform downstream actions without human screening. We introduce TRIAD, a multimodal retrieval benchmark for evaluating how well retrieval systems satisfy these complex, evolving requests. TRIAD comprises three tracks: abstraction, testing bidirectional retrieval between visual instances and concepts at different hierarchical levels; constraints, testing the joint satisfaction of positive and negative conditions; and state, testing the integration of requirements across dialogue turns and the updating of earlier conditions when users revise their requests. The benchmark contains six subsets, each with 1,000 queries and a fixed pool of 1,000 candidates per query, evaluated using a common multi-positive mean average precision (mAP) protocol. Simpler reference tasks and hard negatives grouped by the requirements they violate help characterize specific retrieval difficulties. We evaluate 20 configurations of 17 systems and find that no system leads across all six subsets. Performance declines with hierarchical distance, and negative-only queries generally underperform positive-only queries with the same number of constraints. Across paired tasks expressing the same final intent, multi-turn retrieval yields lower mAP in 39 of 40 configuration-domain comparisons. The benefits of explicit reasoning also vary across models and tasks. These findings highlight the gap between retrieving similar content and satisfying a complete request.
Read the full abstract

All authors Xiangyang Wu, Chen Ju, Xu Chen, Likang Peng, Jinsong Lan, Haihong Tang, Yang Hu

High Energy Physics (Phenomenology)

Naturalness in the Past from a Time-Dependent Supersymmetry Scale

We consider the possibility that electroweak naturalness may be epoch dependent. In no-scale supergravity, a rolling quintessence modulus can control the supersymmetry-breaking scale, allowing superpartners to be much lighter in the early universe while remaining heavy today. Fifth-force and varying-constant effects can be suppressed via correlated modular weights in the visible-sector Kähler metric and modulus-dependent superpotential Yukawa couplings, while radiative stability of the quintessence potential requires additional tuning of the field-dependent vacuum energy. The resulting effective theory has distinctive implications for early-universe phase transitions, relic abundances, and gravitational waves.
Read the full abstract

Distributed Computing

Release Sooner, Expose Less: Offloading MoE Dispatch in NCCL-EP under Dual-Batch Overlap

Dual-batch overlap (DBO) hides mixture-of-experts (MoE) communication behind another microbatch's computation, but only after dispatch send releases the GPU. The time dispatch send holds the GPU therefore stays exposed, while network progress after release can be hidden. We evaluate GPUDirect Async - Kernel Initiated (GDA-KI) Free Flow, an NCCL-EP research configuration that leaves WQE construction on the GPU but moves send progression to a host thread. On a GB300 NVL72 system, GDA-KI Free Flow reduces dispatch-send time relative to default GDA-KI at every operating point in the evaluated parameter sweep. An experiment that inserts a GPU delay between dispatch send and receive shows that offload increases the in-flight interval. Depending on the load, 32-96 microseconds of computation or other independent work between send and receive hides the deferred progress, leaving less exposed time than default GDA-KI. In disaggregated vLLM serving with DBO, GDA-KI Free Flow lowers median TPOT by 15-25% relative to NCCL-EP default GDA-KI and by 16-42% relative to DeepEP v1.
Read the full abstract

All authors Hsin-Chun Yin, Khaled Hamidouche, Ian Ziemba, Arjun Kashyap, Pak Markthub, Subhadeep Bhattacharya, Ke Wen, Assaf Ravid, Manjunath Gorentla Venkata, Gil Bloch

Statistical Computation

Law-Invariant Operators for Stochastic Mechanics

A central objective in stochastic mechanics is to separate deterministic physical behavior from the probability laws governing uncertain inputs. While classical spectral approaches achieve this separation in linear settings, nonlinear responses introduce interactions beyond conventional second-order variability response functions (VRF). Inspired by the spirit of VRFs, this work establishes that, for a fixed physical system, uncertainty propagation is described by a law-invariant operator mapping input realizations to responses, governed by the underlying physics and the deterministic parameters of the system. Different probability laws affect the input sampling and resulting response statistics, but not the mapping operator. We establish this law invariance under complete, invertible coordinate transformations and develop complementary, Fourier, fixed-reference polynomial-chaos (PCE) and finite element method (FEM) transfer formulations. Fourier captures nonlinear cross-frequency interactions and recovers the classical VRF in the appropriate linear limit, PCE describes coupling among stochastic modes, while FEM expresses the deterministic inverse-stiffness action. Neural approximations of these deterministic operators are investigated through analytical and finite-element benchmarks, examining cross-law generalization, the influence of training distributions and stochastic excitation, and the prediction of response fields and statistics. The results demonstrate the potential for reusing learned physical operators across different stochastic laws without retraining, while highlighting the importance of adequate training coverage and the challenges of accurately estimating low-probability events.
Read the full abstract

Combinatorics

Avoiding short progressions in Euclidean Ramsey theory II

Recently, Currier, Moore, and Yip developed a general framework for constructing colorings avoiding prescribed monochromatic arithmetic progressions in Euclidean Ramsey theory. Writing $\ell_m$ for an $m$-term arithmetic progression with unit spacing, they proved $\mathbb E^n\not\to(\ell_3,\ell_{20})$ and $\mathbb E^n\not\to(\ell_3,\alpha\ell_{6889})$ for every $\alpha>0$. We refine and simplify their framework, proving $\mathbb E^n\not\to(\ell_3,\alpha\ell_{22})$ for every $\alpha>0$. We further establish $\mathbb E^n\not\to(\ell_3,\alpha\ell_{21})$ except possibly for four values of $\alpha^2$, and $\mathbb E^n\not\to(\ell_3,\alpha\ell_{17})$ for all but finitely many $\alpha>0$. Our approach also recovers several earlier results and avoids some technical arguments used in previous works.
Read the full abstract

Statistics Applications

Numerical and analytical strategies to propagate correlated uncertainties through the discrete Fourier transform

The Fourier transform underpins a range of comparisons between atomistic simulations and scattering measurements, yet uncertainty propagation through it is seldom discussed, and correlations between data points are usually ignored. Here, we give a pedagogical account of uncertainty propagation through the discrete Fourier transform, including the full covariance matrix. Two routes are described: a numerical scheme that samples from the multivariate normal distribution describing the data, and an analytical scheme that treats the transform as a change of basis. We discuss the scaling behaviour of these methods and demonstrate their utility on problems that link simulation to scattering: the Faber–Ziman partial structure factors of liquid water, and the dynamic structure factor obtained from a rotational autocorrelation function.
Read the full abstract

Combinatorics

Induced packing treewidth III. Branch decompositions

The notion of induced packing treewidth aims to unify classes defined by forbidden induced subgraphs or induced minors with classes defined by the existence of certain structured tree decompositions. In this paper, we introduce induced packing branchwidth, a branch-decomposition analogue of induced packing treewidth. For a graph $H$, the induced $H$-packing branchwidth of a graph $G$, denoted by $\branchpi_H(G)$, is the minimum, over all branch decompositions of $G$, of the largest number of pairwise anticomplete induced copies of $H$ that cross one displayed cut. For $H=P_2$, it equals sim-width. We prove several structural properties of graphs of bounded $\branchpi_H$, for various choices of $H$, including $\chi$-boundedness, polynomial degree-boundedness, and metric properties.
Read the full abstract

Group Theory

Gorenstein modules over Thompson's group

Let $k$ be any commutative ring of characteristic $0$. We construct an example of a countably presented Gorenstein projective module over the algebra of Thompson's group $F$ with coefficients in $k$, that is not Benson cofibrant. In particular, we provide a negative answer to a conjecture proposed by Dembegioti and Talelli. The construction uses a general method that produces totally acyclic complexes of flat modules and is based on the computation of the (co-)cohomology of $F$ by Brown and Geoghegan.
Read the full abstract

Artificial Intelligence

DForkTree: Accelerating Speculative Decoding with Fork-Aware Block Diffusion Draft Trees

Speculative decoding speeds up language models by having a lightweight drafter propose future tokens that the target model verifies in parallel. Block diffusion drafters such as DFlash predict an entire block in one forward pass, and DDTree builds draft trees directly from their per-position distributions. Each distribution implicitly assumes that the drafter's own greedy continuation is accepted, whereas a draft tree improves over a chain only on branches that leave this greedy path. Once a path forks, the drafter's top-1 accuracy drops from 89% to 18% and its top candidates often miss the true token. Re-running the drafter from the fork token largely restores its accuracy, showing that the errors stem from conditioning on an outdated prefix rather than from limited model capacity. In this paper, we introduce DForkTree, a training-free method that re-anchors each branch on its own tokens. It adjusts the drafter's scores with pointwise mutual information from target-generated n-grams, weighted more heavily after a fork, and a layer-wise GPU search recovers the exact optimal tree under these scores. On Qwen3.5-4B and 9B, DForkTree outperforms all baselines on every benchmark, by up to 41.5% over DFlash and 15.1% over DDTree, with average speedups of up to 4.75x over autoregressive decoding.
Read the full abstract

All authors Zhicheng Zheng, Cheng Zhang, Chenkai Tang, Tianxiang Zhao

Combinatorics

Edge-disjoint Hamilton cycles in the McKay–Miller–Širáň graphs $H_q$ for every prime power $q\geq 3$

Motivated by Hamilton cycle problems and applications to collective communication, we study edge-disjoint Hamilton cycles in the McKay–Miller–Širáň (MMS) graphs. The MMS graph $H_q$ (indexed by a prime power $q$) is a highly symmetric graph arising from finite fields. For every prime power $q\geq 3$, we prove that the MMS graph $H_q$ contains $\big\lceil\frac{q-1}{2}\big\rceil$ edge-disjoint Hamilton cycles. This resolves a problem of Li, Wang and Zheng, who established the result for prime $q$ and for $q=4$. Our construction uses graph lifts over finite fields and joins the components of lifted Hamilton cycles by adding edges within vertex fibres. The resulting cycles can be constructed using $O(|E(H_q)|)$ finite field operations, comparisons, and elementary list operations. As an application, these cycles provide edge-disjoint spanning rings for parallel ring-all-reduce algorithms on MMS networks.
Read the full abstract

Optics

Boosting bottom-up milling of glass with GHz laser bursts

Burst-mode laser processing has gained increasing attention in material machining due to its potential to enhance processing efficiency and quality. However, most studies have focused on conventional top-down machining strategies, while investigations of bottom-up milling (BUM) remain limited. In this work, we optimize processing parameters for high-efficiency glass milling using a bottom-up approach with GHz bursts. The results show that the highest BUM efficiencies are achieved when employing high pulse counts ($\geq$ 20). To the best of our knowledge, we report the highest glass milling efficiency, reaching more than 250 $\mu$m$^3$/$\mu$J for both soda-lime glass and fused silica. Furthermore, increasing the burst repetition rate while maintaining a parameter set associated with high milling efficiency enables a material removal rate larger than 8 mm$^3$/s.
Read the full abstract

All authors Miglė Mackevičiūtė, Juozas Dudutis, Aleksandras Kondratas, Paulius Gečys

Computer Vision

QCC-WM: Linking Macro-State Dynamics and Visual Generation via Quotient Causal Construction

A key challenge in action-conditioned world modeling is to construct macro-states from high-dimensional observations that support recursive prediction under actions, while keeping their predicted evolution consistent with visual changes in generated observations. We propose QCC-WM, a visual world model based on Quotient-based Causal Construction. QCC-WM uses action responses to guide macro-variable construction for recursive action-conditioned prediction. A shared reference quotient measurement constrains the consistency between predicted states and observed changes, allowing macro-level predictions to provide structured constraints for future visual generation. Using translational and rotational actions and their predicted responses, the model aggregates local visual features into macro-nodes with spatial correspondence through causal abstraction, and recursively predicts node content and motion under a shared reference measurement. Macro-level predictions guide future observation generation, while predicted motion transports historical appearance information. Node-level errors between generated observations and macro-level targets are then mapped back to the corresponding spatial regions for correction, improving consistency between visual generation and macro-level predictions. On the US-VLA dataset, QCC-WM achieves an ROI-MSE of 0.01861 for four-step prediction. Compared with the best baseline at each prediction horizon, it reduces LPIPS by 48.59%, 41.68%, and 30.23% for four-step prediction and eight- and sixteen-step recursive extrapolation, respectively. Ablation studies and mechanism analyses show that motion-guided reference transport mainly improves pixel reconstruction, while node-level error feedback further improves perceptual quality and consistency between generated observations and macro-level predictions.
Read the full abstract

All authors Zhizhong Fu, Wei Zhou, Zhaoyang Jiang, Yudong Lin, Songlun Li, Zhi Li, Feifan Chen, Gaoming Li, Xiaorong Ding, Yifan Chen

Statistical Methodology

TRACE: Adaptive Credibility Reweighting for Robust Bayesian Model Selection with Source Summary Statistics

This study focuses on regression estimation and variable selection based on target data, while incorporating information from coefficient estimates obtained in related studies. A weighted combination of these estimates may agree with some target effects but differ from others. We propose Transfer Regression with Adaptive Credibility Estimation (TRACE), a Bayesian method requiring only fitted coefficients and sample sizes from source studies. TRACE learns simplex-constrained source weights shared across variables. Separate latent indicators determine target inclusion and whether each coefficient is guided by the weighted source estimate. We develop an approximate collapsed variational expectation-maximization algorithm with robust source-weight initialization. Under suitable conditions, consistency of variable selection, transfer decisions, and source-weight estimation is established. These results hold at every prespecified fixed iteration in fixed dimension and at global mean-field optima of the exact collapsed variational objective as the dimension grows. Simulations show that TRACE achieves competitive variable-selection performance under different patterns of source heterogeneity. An application to data from the Genotype-Tissue Expression project illustrates its use in cross-tissue gene-expression prediction.
Read the full abstract

All authors Pengfei Chen, Yuehan Yang, Lucy Klinger, Chong You