Peer Revenue Now accepting positions

Papers

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

All of arXiv

Monday 9 November

Artificial Intelligence

Representational Alignment between Deep Neural Networks and Human Brain in Speech Processing under Audiovisual Noise

Speech comprehension in the human brain is an incremental process that begins with acoustic processing and advances to linguistic processing. While recent studies have revealed that the hierarchy of deep neural networks (DNNs) correlates with the ascending auditory pathway, the exact nature of this DNN-brain alignment remains underexplored. In this study, we investigate how DNN representations align with the brain's acoustic-to-linguistic processing. Specifically, we employed neural encoding models to simulate neural responses to acoustic (i.e., speech and noise envelope) and linguistic features (i.e., word onset and surprisal). By applying representational similarity analysis (RSA), we quantified the similarity between these neural responses and the DNN embeddings generated by a pre-trained automatic speech recognition (ASR) model, both before and after fine-tuning on audiovisual noisy data. Our results demonstrate significant DNN-brain alignment: embeddings from shallow layers exhibit higher similarity to neural responses associated with acoustic features, while those from deeper layers align more closely with neural responses related to linguistic features. Importantly, the fine-tuning process enhances this alignment by improving noise processing in shallower layers and refining linguistic representations in deeper layers. These results suggest that fine-tuned DNN models can naturally develop human-like processing patterns in noisy environments, highlighting a functional alignment between the human brain and DNNs in speech representation.
Read the full abstract

All authors Cheng Luo, Jian Zhu, Zhoujian Sun, Yunyi Qi, Yu Qi, Wenhui Sun, Wenyuan Yu

Materials Science

Ultrafast Liquid Jet X-ray Total Scattering

X-ray free-electron lasers (XFELs) provide the temporal resolution needed to extend total scattering and pair distribution function (PDF) analysis to investigate structural dynamics on femtosecond-to-picosecond timescales. We recently demonstrated that high-quality total scattering data and PDFs can be obtained from individual XFEL pulses; here, we take the next step towards time-resolved studies of structural dynamics in solution. Using the Materials Imaging and Dynamics (MID) instrument at the European XFEL, we obtain total scattering data up to $Q_{\max} = 16.4\,\mathring{\mathrm{A}}^{-1}$ from both conventional capillary samples and, critically, from samples delivered by a liquid jet. Despite the large air and solvent backgrounds inherent to this setup, PDFs suitable for quantitative structural analysis are recovered from liquid dispersions and molecular clusters in solution. We further demonstrate pump-probe total scattering from $\mathrm{CsPbBr}_3$ nanocrystals in a liquid jet, resolving a fluence-dependent photoinduced isotropic lattice strain. Additionally, empirical potential structure refinement (EPSR) of XFEL total scattering data from an aqueous Keggin cluster reveals site-specific hydration environments and enables three-dimensional mapping of the surrounding solvent structure. Together, these advances open the door to tracking not only ultrafast structural changes in solution, but also the evolving solvent environments that govern molecular behaviour and reactivity.
Read the full abstract

All authors Adam F. Sapnik, Philip A. Chater, Dean S. Keeble, Tobias A. Bird, John S. O. Evans, Federica Bertolotti, Antonietta Guagliardi, Norberto Masciocchi, Daniel T. Bowron, Lise J. Støckler, Elodie A. Harbourne, Anders B. Borup, Rebecca S. Silberg, Viktoria A. Kaznelson, Stefanie Punke, Mariia Svyrydenko, Maryna I. Bodnarchuk, Dmitry N. Dirin, Simon C. Boehme, Maksym V. Kovalenko, Bo B. Iversen, Emil S. Bozin, Kirsten M. Ø. Jensen, Gabriele Ansaldi, Karen Appel, Alexander Gierke, Jörg Hallmann, Aliaksandr Leonau, Anders Madsen, Johannes Möller, Jan-Etienne Pudell, Angel Rodriguez-Fernandez, Andreas Schmidt, Roman Shayduk, Cornelius Strohm, Joana Valerio, Ulrike Boesenberg, David A. Keen

Algebraic Geometry

Sharp minima of higher Milnor and Tjurina numbers of isolated singularities

We determine the minimum higher Milnor and Tjurina numbers of isolated complex polynomial singularities with fixed ambient dimension and multiplicity. A Fermat polynomial of multiplicity $m$, perturbed by the $(m+1)$-st power of a linear form, attains both minimum sequences at every order. These values remain the minima at each prescribed total degree $d>m$; the homogeneous case $d=m$ has a separate Tjurina formula. The proof expresses the Tjurina defect through multiplication in the filtered Milnor algebra and uses the monomial strong Lefschetz theorem to attain the resulting rank bound. We also determine the associated graded higher Tjurina algebras of this family and show that they are unchanged by terms of order at least $m+2$.
Read the full abstract

All authors Chuangqiang Hu, Stephen S.-T. Yau, Lishan Yu, Huaiqing Zuo

Software Engineering

ForgeSystem: An Architecture for Demand-Specific Artefact Production with AI Agents

General-purpose software serves many different demands, but this flexibility costs performance and efficiency on any specific demand. Software tailored to a specific demand avoids these losses, yet developing it requires substantial time and effort from human experts. AI agents offer an alternative: they can design, implement and test artefacts tailored to individual demands. We call this production paradigm Forge. Forging for a demand is economically justified when the cumulative benefit of specialisation exceeds the total cost of construction, verification and maintenance. To make Forge a repeatable production process, we propose ForgeSystem, which consists of a control layer, a forge loop and a recursive self-improvement (RSI) loop. The control layer selects which artefact serves each demand and decides which demands to forge, when to forge them and from which starting point; the forge loop iterates between production and verification until a candidate artefact passes verification; and the RSI loop draws on accumulated production records to supply starting points, refine the harness and train models. We implement ForgeSystem for training frameworks, decoding megakernels and individual computational kernels. Forged training engines reach 1.047-1.332x the model-FLOPs utilisation of their reference frameworks, forged megakernels achieve a geometric-mean decoding speedup of 1.21x over SGLang across thirteen configurations, and forged kernels match or exceed expert libraries in eighteen of twenty-two configurations, while satisfying task-specific quality requirements. Controlled comparisons show that explicit quality requirements steer agents towards optimisations that preserve training quality. We also discuss a gradual transition to Forge through service routing, and how demand-specific production may shift value accumulation and trust towards production systems.
Read the full abstract

All authors Qingfeng He, Zhui Zhu, Yaojian Chen, Shangzhan Li, Leshan Li, Tingjian Zhang, Qian Xie, Yuxuan Li, Xu Han, Zhiyuan Liu

Statistical Methodology

Algorithm-aware Statistical Inference in High-dimensional Estimating Equations

Iterative algorithms are integral to many statistical procedures, whereas sampling uncertainty along their iteration is generally not quantified. We develop simultaneous inference for discrete trajectories generated from high-dimensional estimating equations, with iteration-specific population estimands that preserve algorithmic regularization. Our proposed fixed-Jacobian bootstrap propagates perturbations along the observed trajectory without repeatedly running the original nonlinear algorithm. We establish uniform first-order expansions and bootstrap validity over possibly infinite sets of contrasts and iterations under stability, entropy and moment conditions. Simulations in sparse linear regression, paired Fisher discrimination, and temporal-difference learning demonstrate accurate simultaneous coverage. A micro-randomized trial application illustrates inference for a treatment trajectory with participant-level clustering.
Read the full abstract

Robotics

VLA-MOPD: On-Policy Distillation for Multi-Task VLA Policy Consolidation

Physical intelligence requires embodied agents to ground language in observations and execute actions while retaining useful skills. Vision-language-action (VLA) policies offer a shared interface, but task-specific post-training often produces separate specialists. Consolidation is difficult: mixed-task reinforcement learning provides sparse trajectory-level credit, while offline distillation may miss student-visited observations and endpoint distillation does not supervise intermediate action-generation states. We introduce VLA Multi-teacher On-Policy Distillation (VLA-MOPD), a protocol for consolidating flow-matching VLA specialists. After acquiring task or suite experts, a shared student collects robot rollouts and queries the instruction-routed frozen teacher at each visited observation and intermediate ordinary differential equation (ODE) state. Detached per-flow-step transition targets supervise valid action chunks through a masked local matching loss; an optional reference-policy term supports retention on other tasks. Starting from one trajectory per LIBERO task, VLA-MOPD raises mean success-once from 75.4% to 98.9%, compared with 98.15% for endpoint action distillation. It improves mean RoboTwin 2.0 success from 73.0% to 89.3%; on a Piper robot, two consolidation rounds improve average success from 70.0% to 90.0%. In a near-matched Spatial-adaptation study, reference matching limits the Object suite decrease to 3.0 points, versus 13.2 without the reference term.
Read the full abstract

All authors Yongxuan Lv, Zhen Fang, Lichen Man, Hao Li, Pu Li, Hengyi Wang, Langfeng Cao, Lin Chen, Jie Huang, Xiu Su, Shan You, Fei Wang, Feng Zhao

Machine Learning

Learning Evidence-Grounded Reasoning for Context-Aware Time Series Forecasting

LLM-driven time series forecasting has become increasingly promising by leveraging contextual features to generate future values directly or refine forecast priors. However, incorporating contextual features alone does not ensure that models can identify which features matter for forecasting and understand what they suggest about future dynamics. We argue that the fundamental challenge is how to extract predictive evidence from contextual features and when and how to leverage it to guide forecasting decisions. In this paper, we propose CastReasoner, an evidence-grounded reasoning model for context-aware time series forecasting. Specifically, CastReasoner uses a structured chain of thought (CoT) to extract predictive evidence from contextual features, infer the future dynamics suggested by this evidence, and assess these dynamics against the forecast prior to guide adaptive forecasting decisions. To enable this capability, we design a reasoning-centric two-stage post-training framework including supervised fine-tuning (SFT) and reinforcement learning (RL). First, we synthesize reasoning trajectories through controllable multi-teacher guidance and select high-quality ones using a novel student-aware difficulty metric. Second, we design fine-grained multi-objective rewards to further improve reasoning and forecasting. To address heterogeneous learning needs across domains in both stages, we introduce a convergence-guided data balancing strategy. Through this two-stage training, CastReasoner improves both reasoning quality and predictive accuracy. Extensive experiments across multi-domain forecasting benchmarks demonstrate that CastReasoner achieves superior forecasting performance.
Read the full abstract

All authors Xingpeng Gao, Mingyue Cheng, Daoyu Wang, Huajian Zhang, Xiaoyu Tao

Astrophysics of Galaxies

A Comprehensive Analysis of the H$\alpha$ Luminosity Functions at $z\sim 1.3$ and $z\sim 2.0$ in the CANUCS Lensing Fields with JWST NIRISS

We present H$\alpha$ luminosity function (LF) measurements at redshifts $z\sim1.3$ and $z\sim2.0$. We adopt JWST grism data in five lensing cluster fields and their parallel fields from the CANUCS program: ABELL-370, MACS-J0416, MACS-J0417, MACS-J1149, MACS-J1423. Based on emission lines spectroscopically identified in the F115W, F150W and F200W filters, our final sample consists of 762 H$\alpha$ emitters, including 433 at $1.1<z<1.5$ and 329 at $1.8<z<2.2$. Through detailed effective volume and completeness analysis for each source, we construct the H$\alpha$ LF in two redshift bins. Thanks to the sensitivity of NIRISS WFSS and gravitational lensing magnification, our sample reaches intrinsic H$\alpha$ luminosities down to $L_{\rm H\alpha}\sim10^{40.6}~\rm erg~s^{-1}$ at $z\sim1.3$ and $L_{\rm H\alpha}\sim10^{41}~\rm erg~s^{-1}$ at $z\sim2.0$ with completeness larger than 80%. The larger survey volume and coverage of multiple independent sightlines help mitigate cosmic variance. After accounting for cosmic variance, we constrain the faint-end slopes to be $\alpha_{\mathrm{H}\alpha}=-1.51^{+0.06}_{-0.04}$ at $z\sim1.3$ and $-1.42^{+0.09}_{-0.06}$ at $z\sim2.0$, indicating a mild evolution between the two redshift bins. We find a negligible contribution from bright active galactic nuclei in our sample. We estimate integrated cosmic star formation rate densities of $0.103^{+0.008}_{-0.007}~M_{\odot}~\rm yr^{-1}~Mpc^{-3}$ at $z\sim1.3$ and $0.149^{+0.016}_{-0.013}~M_{\odot}~\rm yr^{-1}~Mpc^{-3}$ at $z\sim2.0$. The emission-line samples presented here will enable further detailed studies of galaxy properties including metallicities. The methodology presented here can be readily applicable to other JWST slitless spectroscopic datasets and future wide-field slitless surveys, including those from Euclid, Roman, and the Chinese Space Station Telescope.
Read the full abstract

All authors Qianqiao Zhou, Yuxuan Pang, Xin Wang, Karl Glazebrook, Shengzhe Wang, Hang Zhou, Pengfei Ren, Shamuhawu Hailanhazi, Peter J. Watson, Maruša Bradač, Nicha Leethochawalit, Matthew A. Malkan, Themiya Nanayakkara, Xue-Bing Wu, Hu Zhan

Probability

Large deviations, optimal trajectories and conservation of momenta in generalised Jackson networks

We study properties of optimal trajectories for the quasipotential that governs large deviations of queue lengths in generalised Jackson networks (GJN). The trajectories are shown to satisfy Hamilton equations. The Hamiltonian is identified. Optimal trajectories and associated costs are found for networks with two and three stations. A conservation law for adjoint variables is discovered to the effect that certain components are maintained along an optimal trajectory. A process is proposed for finding optimal trajectories and computing their costs in general GJN.
Read the full abstract

High Energy Physics (Phenomenology)

A Self-Interacting Dark-Matter Route to Supermassive Kerr Black Holes via Super-Eddington Bondi Accretion: Mass-Spin Evolution

Recent JWST observations motivate mechanisms for rapid black-hole growth in the early Universe. Extending our previous Schwarzschild study, we show that critical relativistic Bondi accretion of repulsively self-interacting dark matter remains an efficient super-Eddington channel for Kerr black holes with rotating inflow. Within a stationary equatorial quasi-spherical framework, smooth critical solutions exist over a finite range of inflow angular momentum set by the Kerr marginally bound capture values. Rotation changes the sonic structure and accretion efficiency, while the critical density cancellation and dark-sector/halo matching retain their Schwarzschild form. The resulting rapid growth persists until the finite halo supply becomes limiting. Because the inflow also carries angular momentum, stellar-mass seeds can grow and spin up simultaneously; illustrative population evolutions therefore provide a joint mass–spin route to early supermassive Kerr black holes.
Read the full abstract

Mesoscale and Nanoscale Physics

Tuning correlated states in rhombohedral pentalayer graphene with magnetic encapsulation

Rhombohedral graphene multilayers provide a flexible platform for correlated physics associated with their flat-band electronic structure. While a variety of correlated states appear in these systems as a function of bias and electrostatic doping, promoting specific correlated flavor symmetry breaking states remains a challenge. Here we show that magnetic encapsulation by CrGeTe3 achieves this in rhombohedral pentalayer graphene, by using switchable exchange couplings induced by ferromagnetic substrates. We show that correlations in rhombohedral multilayers at charge neutrality strongly renormalize the exchange proximity splitting,giving rise to collective spin modes whose character is switched by the parallel or antiparallel alignment of the encapsulation. We further show that upon doping of the flat bands, spin triplet f-wave superconducting solutions emerge, whose layer and spin selectivity is controlled by the magnetic encapsulation, hosting in-gap excitations at the edges. Our results establish magnetic alignment as a means to control spin symmetry breaking and unconventional superconductivity in rhombohedral pentalayer graphene.
Read the full abstract

Quantum Physics

Calibration-free quantum thermometry with strain-swept two-level defects

Superconducting qubits operate within a well-defined temperature range, making local temperature monitoring important. Here we show that intrinsic two-level defects in Josephson-junction barriers can act as local thermometers by exploiting the shape of their population response, without requiring calibration of the readout offset or contrast. Marginalizing over these unknown parameters imposes an identifiability requirement absent from single-parameter thermometric bounds: for narrow sweeps, usable information scales as the fourth power of the sweep width. Hence, a narrow sweep becomes locally indistinguishable from a change in offset or contrast. We illustrate this principle with a strain-swept two-level defect in a Josephson junction. For a reported defect with tunneling energy $\Delta_{0}/h=\SI{1.3}{\giga\hertz}$, swept along one branch of its hyperbola to \SI{2.69}{\giga\hertz} (an asymmetry excursion of \SI{2.36}{\giga\hertz}, or \SI{8}{\volt} of piezo voltage), the analysis projects a single-defect fit precision of \SI{0.58}{\percent} at \SI{20}{\milli\kelvin}. The resulting response enables temperature estimation without prior calibration of the readout offset and contrast, while the same sweep overdetermines the Boltzmann form and identifies significant deviations from thermality. These results establish intrinsic junction defects as a resource for calibration-free local thermometry and highlight identifiability as a key design criterion for primary temperature sensing.
Read the full abstract

Quantum Physics

Deploying QKD on the fiber links of the National Quantum Communication Infrastructure in Sweden

Quantum key distribution (QKD) is being deployed across Europe to evaluate its performance and practicality in real-world scenarios. When preparing this deployment, it is important to survey the properties of the fiber links that will be used for the quantum channel, cross-checking them with those of QKD devices, to ensure compatibility and optimal performance. We report here our studies from the Swedish chapter of this endeavor, the National Quantum Communication Infrastructure in Sweden. We focus on procedures to characterize the links and evaluate their suitability for QKD. We relate measurable properties to the key factors that affect the secret key rate using a simplified model that aims at the essence of QKD performance. Finally, we confirm the feasibility of QKD on one of our links by reporting a one-month long successful test.
Read the full abstract

All authors Giulio Foletto, Mira Abu Yahia, Shreya Gouravaram Navalur, Carlos Brotons Montoro, Hilma Karlsson, Dena Karunianto Wibowo, Pascal Lefebvre, Katia Gallo

Geophysics

Probabilistic wavefield decomposition from array covariance

Inter-sensor coherence loss carries information about the wavefield that conventional array processing discards. The array-covariance matrix is modelled using a mixture of energy distributions over slowness with analytical characteristic functions, e.g., skew-normal and angular-harmonic distributions, and inverted using variational Bayesian inference. The posterior supports uncertainty quantification, experimental design, model comparison and hypothesis testing. For a synthetic hydroacoustic T-phase, information is increased by an additional sensor and decreased by noise; the data favour a distributed source over a point source. For an observed thermospheric infrasound return, energy is skewed back towards the source. The method is implemented in the software Topophone2.
Read the full abstract

Strongly Correlated Electrons

Positronium resonance of electric monopoles in quantum spin ice

Quantum spin ice attracts considerable attention due to its remarkable low-energy properties, described by the effective gauge theory, which mimics the quantum electrodynamics (QED). The “effective" QED in quantum spin ice is beyond the “real" QED in our universe in many respects. Above all, the effective QED hosts two kinds of gapped particles, the electric monopole, in addition to the magnetic monopole. Among the two monopoles, the electric monopole is of purely quantum origin, and its character has not been well resolved. In this work, we focus on the Rokhsar Kivelson point of the quantum dimer model, as the low-energy description of the quantum XXZ model on a pyrochlore lattice, and address the excitation spectrum of the electric monopoles, by combining the quantum Monte Carlo simulation with the stochastic analytical continuation. Most strikingly, the electric monopole spectrum exhibits clear double-peak structure, corresponding to the bound states and scattering states of electric monopoles. The lower-energy peak is attributed to the exciton-like bound state of an electric monopole pair, a signal of positronium in quantum spin ice. We found that the flippable loop operator serves as a local physical observable to excite a pair of electric monopoles, implying that monopoles can be detected through lattice or optical response.
Read the full abstract

Algebraic Geometry

The rank of the Fano normal function via the Fermat cubic

We prove that the canonical derivative of the Fano normal function has rank ten at the Fermat cubic threefold. This yields a direct proof of its generic maximal rank, a result also implied by a general theorem of Gao–Zhang. Our approach combines Hain's equivariance principle with an explicit representation-theoretic computation and the nonvanishing of the Griffiths infinitesimal invariant established by Collino–Naranjo–Pirola.
Read the full abstract

Computational Engineering

SimpFlow: Transferable Pretraining for Fluid Dynamics via Simplified Physics

Neural surrogates enable rapid fluid simulation, but learning across diverse three-dimensional (3D) geometries requires costly computational fluid dynamics (CFD) labels. We introduce SimpFlow, a pretraining approach that generates supervision from simplified physics on unlabeled geometry libraries. A source-panel potential-flow solver generates velocity and pressure fields for approximately 9,000 3D geometries, capturing geometry-induced acceleration, stagnation, and suction while omitting viscosity and separated wakes. The solver imports only a surface mesh and reuses precomputed velocity fields across inflow directions, enabling inexpensive label generation. In our benchmark, amortized labeling takes 0.111 s per example using four CPU threads. We pretrain a neural surrogate on these labels and fine-tune it on independent CFD datasets. Across five flow tasks, SimpFlow improves over training from scratch in all 24 task-budget settings. With fine-tuning fixed at 200 epochs, interpolation indicates that SimpFlow requires 35-68% fewer CFD training cases to match the error of scratch models trained at each task's largest data budget. These results demonstrate that simplified physical solutions can provide transferable supervision, offering a practical way to reduce the high-fidelity data requirements of fluid surrogates.
Read the full abstract

Analysis of PDEs

Asymptotic spreading of Fisher-KPP equations in shifting environments with time-periodic limits

In this paper, we study the rightward spreading speed of solutions for Fisher-KPP equations in shifting environments. The linear growth rate has time periodic limits at the two ends of the moving frame. Each limit has a positive average and may be negative during a period. When the growth rate lies between the two limits at each time, we show that the spreading speed depends only on their averages and the asymptotic shift speed. For intermediate shift speeds, the formulas give nonlocal pulling if the right end has the larger average and speed locking if the left end has the larger average. The interior growth profile may be nonmonotone in space and aperiodic in time. We prove the formulas by constructing upper and lower solutions for a stepwise profile and then comparing with shifted stepwise problems.
Read the full abstract

Classical Analysis and ODEs

Translation continuity for the local bilinear circular maximal operator

We prove a positive power translation-continuity estimate for the local bilinear circular maximal operator on the real line. More precisely, there is $\eta>0$ such that the operator applied to a translation difference in either input maps $L^8(\mathbb{R})\times L^8(\mathbb{R})$ to $L^4(\mathbb{R})$ with norm $O(|h|^\eta)$. This answers Question 6.3 in Borges's survey on bilinear spherical averages. The main step is a frequency-localized estimate for arbitrary bounded inputs. A $TT^*$ argument converts it into a linear maximal problem on the plane, associated with pairs of circles. We verify the corrected multi-parameter cinematic curvature condition of Chen, Guo, and Yang by proving an irreducibility and infinitesimal independence statement for the resulting quartic curves. Uniform algebraic sublevel estimates control the degeneracies. Interpolation also gives continuity for all finite exponents $2<p,q<\infty$, with output exponent $pq/(p+q)$.
Read the full abstract

Analysis of PDEs

Resonances of the Kramers-Fokker-Planck Operator

We study the subellipticity of the analytically deformed Kramers-Fokker-Planck operator and give two definitions of resonances, one based on analytic dilation and the other on analytic distortion in the $x$-variables. Due to the failure of global subellipticity of the distorted operator in dimensions $n \geq 2$, the second method can only be used in dimension one. We show that, whenever both methods are applicable, they give rise to the same set of resonances. We also prove that, when $|\na V(x)| = O(\w x^{-\rho})$ with $\rho>1$, positive resonances are embedded eigenvalues and/or scattering resonances whose resonant states belong to weighted $L^2$-spaces and satisfy an outgoing radiation condition.
Read the full abstract

Cryptography and Security

Public, Identifiable, and Effectively Immutable: Ethical Lessons from Research on arXiv

Ethical decision-making is central to safe cybersecurity research. Recent studies, one of which is our own, analyzing the source files of arXiv papers bring this challenge to light - arXiv publicly distributes the source files, to the surprise of many authors. Unique aspects reveal ethical challenges and dilemmas that extend beyond established ethical frameworks, institutional processes, and community best practices. In this work, we examine this case and provide insights for similar research on other public live resources, such as Zenodo or the Wayback Machine. While such empirical studies are particularly realistic, they must account for personally identifiable information, leaked secrets, and other sensitive details that could expose, shame, or otherwise negatively affect individuals, possibly leading to both reputational and practical harm. sDrawing on our collective experience and discussions among researchers, affected authors, and readers, we make three contributions: First, we define the ethical setting and highlight that even intentionally-public data with deliberate real-name attribution can raise ethical concerns when repurposed for empirical analysis. Second, we discuss the ethical issues surrounding the use of arXiv data and derive four higher-level dilemmas that researchers may face when working with similar data. Third, we derive practical guidance for ethically navigating such settings and, in doing so, distill four lessons for research on arXiv and similar public, live datasets: public availability should serve as the starting point of ethical reasoning; research should minimize the amplification it induces on exposed information; disclosure should be calibrated to the harm, confidence, and actionability of a finding; and effective immutability must be accounted for. Together, these contribute to community resources for ethically-sound cybersecurity experiments.
Read the full abstract

All authors Jan Pennekamp, Johannes Lohmöller, Peter Mayer, Arjun Arunasalam

Physics and Society

Modelling socio-economic effects of the pandemic in a metropolitan area

One of the most notorious measures taken by authorities in every country trying to prevent the uncontrollable spread of the Covid 19 pandemic was to impose severe measures to diminish the mobility of people. In recent publications we proposed a mathematical model to address the consequences of these measures and to predict the infections under different scenarios, for various different countries. In this paper we extend the model, called the BTH–SEIRS-model, to deal with the reduction of mobility in an urban area. The chosen region was the densely-populated metropolitan area around Helsinki, the capital of Finland. This choice obeyed to the fact that there are precise data related to mobility from a mobile telephone company, and that this region was completely isolated from the rest of the country for a short period during the pandemic. We also used public transport data from railways and the accommodation occupation rates in the area of interest. We show that the BTH–SEIRS model is capable of simulating the real infection data as well as the mobility-related economic activity, exhibiting the main trends present in the three indicators used for comparison. Our calculations allow us to highlight the difference of the behaviour of the pandemic in a totally different spatial scale, and to discover the hidden factors that govern the behaviour of people who are forced to commute every day from home to work.
Read the full abstract

All authors Jan E. Snellman, Rafael A. Barrio, Cecilia I. Ventura, Tzipe Govezensky, Kimmo K. Kaski, Maarit J. Korpi-Lagg

Multiagent Systems

Agent4Crypto: A Dual-Reflective and Market-Feedback-Aligned LLM-Based Multi-Agent System for Cryptocurrency Trading

Cryptocurrency markets provide a natural testbed for autonomous AI agents because they are digital-native, high-frequency, and machine-speed environments, yet existing methods often struggle with non-stationary market regimes, opaque decisions, and fixed aggregation rules. In this paper, we propose Agent4Crypto, a dual-reflective and market-feedback-aligned multi-agent system. The framework coordinates specialized agents for data collection, analysis, risk assessment, reflection, and trading, synthesizing heterogeneous signals from prices, indicators, news, and charts in an interpretable decision pipeline. Our central design is a trader-centric decision mechanism: three style-specific trader agents, Aggressive, Neutral, and Conservative, generate candidate actions and natural-language rationales from the same market context, while their influence is learned rather than manually fixed. A PPO-based gating policy dynamically allocates trader weights through a dual-reflective feedback process. One reflection stream evaluates post-trade decision quality from realized portfolio outcomes and distills reusable lessons into a Financial ReasoningBank for retrieval in subsequent trader prompts, while the other estimates trader-level counterfactual contribution, indicating which style has recently aligned better with market movements. The PPO state combines Counterfactual Contribution Memory with market features, previous trader weights, and current trader actions, so trader influence adapts across bull, sideways, and bear regimes. Experiments on BTC, ETH, and XRP show that Agent4Crypto improves total return and Sharpe ratio over traditional strategies, machine learning and reinforcement learning baselines, and existing LLM-based trading systems, demonstrating the value of interpretable trader competition and market-feedback-driven adaptation in volatile cryptocurrency markets.
Read the full abstract

All authors Guangyan Gan, Jiahao Li, Jiashun Wang, Lin William Cong

Machine Learning

Odds-Structured Distributional Regression for Interpretable Time-to-Event Prediction

Time-to-event prediction is fundamentally challenged by partial observability and aleatoric uncertainty. In clinical medicine, for instance, the timing of disease progression or acute readmission is contingent on highly dynamic and largely latent health trajectories. Rather than forcing a deterministic point forecast, distributional regression offers a principled means to address this complexity by predicting a probability distribution over time. However, explaining these distributional predictions presents a key obstacle: raw distribution parameters output by the model are abstract mathematical quantities that do not align well with domain concepts. To gain interpretability, we propose a novel paradigm where output parameters are designed to be additive in log-odds—a widely established measure in clinical risk modeling. Our approach, called ODDS-DR, moves beyond technical convenience to deliver clear domain alignment as well as the layer-wise computational structure required for robust, time-dependent feature attribution. Validated across two real-world clinical cohorts, our framework matches state-of-the-art predictive performance while resolving distributional outputs into interpretable risk markers across flexible prognostic horizons.
Read the full abstract

All authors Jakob Versnjak, Manuel Welte, Titus Kuehne, Marcus Kelm, Grégoire Montavon

Materials Science

Defects and Structural Evolution in CrSBr: Machine-Learning Potentials for Atomistic Simulations

The layered van der Waals semiconductor CrSBr combines intrinsic magnetic order with semiconducting and optical properties, making it a promising material for multifunctional electronic and photonic applications. Understanding how defects and structural transformations affect its properties under experimentally relevant conditions, however, remains challenging for first-principles simulations due to the relevant length and time scales. Here, we develop and benchmark machine-learning interatomic potentials (MLIPs) for the Cr-S-Br system, enabling large-scale atomistic simulations of CrSBr with accuracy approaching that of density functional theory (DFT). The charge-aware variant qNEP achieves near-DFT accuracy in predicting lattice constants, formation enthalpies, vacancy energetics, and the anisotropic pressure response of o-CrSBr. Nudged elastic band calculations reveal that pre-existing S vacancies lower the Cr Frenkel-pair migration barrier by up to 1.1 eV, providing an atomistic mechanism for beam-induced Cr redistribution and structural transformations. Displacement threshold calculations further show that Br is the most susceptible species to electron-beam-induced displacement, identifying the relative ease of removing different atomic species. Together, these results establish a validated computational framework for investigating atomic-scale disorder and structural transformations in CrSBr, providing a basis for defect engineering in materials for future magneto-optoelectronic applications.
Read the full abstract

All authors Arsalan Hashemi, Daniil Kruklinskii, Xiaoya Chang, Tapio Ala-Nissila, Mikko Karttunen, Arkady V. Krasheninnikov, Mahdi Ghorbani-Asl

Statistical Methodology

Data-Adaptive Subset-Size Selection for Minimum Covariance Determinant Estimation

Subset size determines the balance between robustness and efficiency in minimum covariance determinant estimation. Conventional choices fix this balance before the extent and structure of contamination are known. Reweighting does not remove the influence of the initial subset size. We propose a data-adaptive rule based on ordered squared Mahalanobis distances. It selects the first window maximum, with window width proportional to the sample size. For a minimum covariance determinant subset, the statistic equals the smallest squared distance among excluded observations. For smooth elliptical distributions with small one-point contamination, we characterize the first population window maximum. Under suitable conditions, it is the last retained fraction at which excluding the contaminating point remains optimal. We derive a first-order expansion for this boundary and establish sample localization along the full integer path. Without contamination, the selected fraction tends to one. A fast depth-based implementation reuses one ordering across candidate sizes. Simulations under Gaussian and heavy-tailed sampling show gains over fixed subset sizes, most consistently for scatter estimation. Real data analyses show that the initial subset size affects which observations are retained after reweighting.
Read the full abstract

High Energy Physics (Experiment)

Beam-beam effects on measurements at electron-positron colliders: LEP neutrino counting and Higgs factories

Beam-beam electromagnetic interactions affect the luminosity, collision energy, polarization, and particle backgrounds relevant to precision measurements at electron-positron colliders. We investigate these effects using Fast $e^{\pm}e^{\pm}$ Beam and Luminosity (FeeBL), a new beam-beam simulation framework, and compare its predictions with published results from LEP and proposed future colliders. We reproduce the analytical field-integration calculations of LEP Bhabha scattering and identify an approximately 2% bias in previously published beam-deflection simulations. Applying the corresponding luminosity correction and a revised uncertainty assessment motivated by the observed agreement among the four LEP experiments yields ($N_\nu = 2.9965 \pm 0.0040$), compared with the previously reported ($2.9963 \pm 0.0074$). The revised uncertainty is conditional on an assumed common rescaling of experiment specific uncertainties and motivates further investigation of the original uncertainty budgets. We identify and correct implementation errors in the modeling of Breit-Wheeler, Bethe-Heitler, and Landau-Lifshitz pair production, with implications for detector backgrounds at Higgs factories. Simulated beam depolarization and luminosity spectra are compared with existing ILC and HALHF predictions. Finally, we investigate collision energy monochromatization for direct Higgs production, obtaining sub-MeV collision energy peak widths in exploratory linear collider configurations at substantial luminosity cost, while finding significantly weaker narrowing for the FCC-ee configurations studied. We derive a rigid Gaussian beam overlap model and estimate that FCC-ee needs roughly 10 m of horizontal only dispersion to reach widths of $\approx1~\mathrm{MeV}$.
Read the full abstract

Materials Science

Multichannel Phonon Interconversion Drives 90 Polarization Rotation for Stable Domain Walls in Hafnium Oxide

Polarization rotation is a fundamental mechanism in ferroelectric materials, yet how such rotation arises microscopically in fluorite ferroelectrics and gives rise to distinct domain walls (DW) remains unresolved. Phonon modes offer a microscopic route to such structural transformations, as geometric orientations can modify their frequencies, dispersion relations, and vibrational eigenstates when their characteristic wavelengths become comparable to relevant physical dimensions. Here, by combining condensed phonon mode decomposition with first-principles calculations, we investigate the origin of 90 polarization rotation in the ferroelectric phase of HfO2. Using the cubic parent phase as reference, we show that polarization rotation originates from multiple symmetry equivalent phonon modes of distinct eigenstates at different wavevectors in the Brillouin zone, whose coupling and cooperative reorientation continuously evolve to generate distinct 90 Pca21 variants and switching pathways. Among the 90 variants, the lowest-energy switching pathway through the Pmn21 exhibits an energy barrier comparable to, but slightly lower than, that of the 180 polarization reversal pathway through the P42/nmc phase, indicating that 90 polarization rotation is an energetically accessible alternative switching route. We further analyze 90 DW and find that frustrated phonon modes emerge at the wall center, where the coexistence and competition of condensed modes promote local reconfiguration of the lattice distortions, contributing to the elevated DW energies. Importantly, experimental observations with STEM imaging provide direct evidence of atomic mode distortions in stable 90 DW, consistent with the theoretically identified low-energy DW. These findings provide the fundamental structural mechanism underlying 90 switching dynamics, competing domain-wall physics, and macroscopic ferroelectric behavior.
Read the full abstract

All authors Fatoye Sawyerr, Hamoon Fahrvandi, Pawan Kumar, Zekun Zhang, Shuning Lv, Yue Yu, Xuan Gao, Kang Jia, Shifeng Wen, Zheng Wen, Li-Min Liu, Xiaoli Fan, Jun Hee Lee, Tengfei Cao

Dynamical Systems

Rates in the weak invariance principle for slowly mixing sequential dynamical systems

We obtain rates of convergence in the weak invariance principle for certain classes of slowly mixing sequential dynamical systems. In particular, for systems satisfying a Functional Correlation Bound with rate $n^{-\gamma} \ (\gamma>1)$, we establish an explicit bound in the Wasserstein-$r$ metric for $1\le r<\gamma+1$ whenever the variance grows at a sufficiently fast polynomial rate. The proof is based on Bernstein's “big block–small block" method. Examples covered by our result include sequential heterochaos baker maps and sequential LSV maps. In particular, for random LSV maps over an ergodic base, we obtain rates in the quenched weak invariance principle.
Read the full abstract

Classical Analysis and ODEs

Asymptotics of matrix orthogonal polynomials with non-scalar varying weights

We study the $2\times 2$ matrix orthogonal polynomials $\mathbf{P}_N$ on the real line with respect to the varying weights $\mathrm{e}^{-N\mathbf{Q}(z)}\mathbf{B}(z)$, where $\mathbf{Q}$ and $\mathbf{B}$ are real symmetric matrix polynomials that commute with a Hermite-type matrix weight, $\mathbf{Q}$ is not scalar and $\mathbf{B}$ is positive definite. We are motivated by the non-Abelian Toda deformation of the Hermite-type weight, in the sense of Deaño, Morey and Román. The weight depends on $N$ through a matrix exponential that is not scalar, so that its two eigenvalues have different exponential rates. The matrix orthogonality reduces to two scalar orthogonality problems on the real line, with a varying weight that has an essential singularity at the origin. Assuming that the associated equilibrium measure is one-cut regular, we obtain the strong asymptotics of $\mathbf{P}_N$ and of $\det\mathbf{P}_N$ by the Deift-Zhou steepest descent method. Our main result is the limiting zero distribution of the entries of $\mathbf{P}_N$. These zeros are complex in general, and they accumulate on a curve where the contributions of the two eigenvalues of the weight have the same size. When this curve crosses the real segment that attracts the zeros of $\det\mathbf{P}_N$, the zeros of the entries display a critical behavior, and a positive fraction of them accumulates on this segment.
Read the full abstract