Peer Revenue Now accepting positions

Papers

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

All of arXiv

Monday 9 November

Computer Vision

PhysOPSD: Code as Privilege for Physically Plausible Video Generation

While modern video generative models achieve remarkable visual fidelity, they remain noticeably brittle when generating physically plausible dynamics and complex interactions. Targeted post-training is crucial, yet supervised fine-tuning suffers from offline state mismatch, and reinforcement learning is hindered by sparse, ill-defined physical reward functions. On-Policy (Self-)Distillation (OPSD) reconciles on-policy rollouts with dense corrective feedback, shifting the core challenge to designing effective privileged context. However, conventional modalities fall short: text prompts are inherently declarative, whereas real reference videos heavily entangle physical laws with visual appearance without parametric control. To this end, we propose PhysOPSD, introducing executable code as privileged physical context for video generation. Unlike static visual observations, executable code instantiates an editable physical world with intrinsic abstraction and controllability. Powered by PhysOPSD-Engine, we translate real-world physical scenarios into verifiable simulation programs. PhysOPSD then incorporates (i) dual-level physical privilege to bridge the information gap via an intermediate bottleneck, alongside (ii) controlled physical variations under shared rollout noise to expose causal dynamic shifts. Extensive experiments demonstrate that PhysOPSD substantially improves physical plausibility across diverse backbones, outperforming SFT and RL baselines with only $3$K samples, while preserving general video generation quality.
Read the full abstract

All authors Harold Haodong Chen, Rui Peng, Lingteng Qiu, Kejie Qiu, Chao Xu, Zilong Dong

Plasma Physics

Magnetized plasma modulator for polarization-tunable ultrahigh-power single-cycle lasers

Ultrahigh power lasers with tunable polarization and few-cycle duration have long been considered the holy grail of high-power laser science, as they push light-matter interactions into an entirely new physical regime. Yet such pulses remain far beyond the reach of conventional solid-state optics and technologies. Here, we introduce a novel magnetized plasma modulator that boosts linearly polarized lasers from petawatt to exawatt powers while simultaneously enabling tunable polarization, offering a promising approach for high-power laser development. This is achieved in density-tailored plasmas by combining dispersive compression of a chirped picosecond laser pulse with magneto-chiral group-velocity splitting in a few-Tesla magnetic field. It can generate circularly or elliptically polarized pulses with extremely high compression ratios (over 1000) and extremely short durations (down to the single-cycle level) in a compact and efficient manner. These findings demonstrate a feasible pathway toward exawatt-class few-cycle lasers with tunable polarization, opening new avenues for a wide range of cutting-edge research and applications.
Read the full abstract

Robotics

Bridging Task Planning and Execution for Language-Conditioned Bimanual Manipulation

Robots are increasingly expected to execute natural-language manipulation tasks in real-world environments. This is particularly relevant in household assistance and logistics, where clutter requires reasoning about both task objectives and execution feasibility. This challenge becomes even more pronounced for bimanual humanoid platforms, where coordinating two manipulators introduces additional interaction and kinematic constraints. We present a zero-shot vision-language framework for execution-aware bimanual manipulation that jointly reasons about task objectives and physical feasibility. Planning is grounded in a structured scene representation that combines semantic object relations with execution constraints, enabling the selection of obstacle relocations and execution choices while preserving task semantics. Compatible actions are organised into parallel bimanual stages with explicit pre- and post-conditions. A hierarchical validator evaluates these conditions under perceptual uncertainty and uses stage-level execution history to support local recovery or targeted replanning. Experiments on the CENTAURO hybrid wheeled-legged humanoid robot across nine language-conditioned manipulation scenarios show higher task success than a non-execution-aware baseline. In dedicated failure scenarios, hierarchical validation recovers from execution failures with fewer model queries than always replanning. These results demonstrate the benefits of combining execution reasoning directly into task planning. Videos and additional material are available on the Project Webpage: https://advrhumanoids.github.io/bridging-planning-execution/.
Read the full abstract

All authors Emanuele Cuzzocrea, Giuseppina Iannotti, Luca Rossini, Arturo Laurenzi, Nikos Tsagarakis

Instrumentation and Detectors

Frequency tuning capability of self-powered piezoelectric tuned mass dampers

Tuned mass dampers suffer from frequency shift sensitivity. This paper investigates the use of tunable piezoelectric tuned mass dampers (PE-TMD) by studying their behavior with tunable resistive and capacitive loads to adapt their resonant frequency. An analytical model is developed to assess the influence of key dimensionless parameters-electromechanical coupling coefficient, mass ratio, and quality factoron vibration attenuation, frequency tuning capability, and energy harvesting. The study reveals trade-offs between attenuation, tunability, and energy harvesting. It provides design guidelines for autonomous PE-TMDs and identifies key challenges for their practical implementation, including the interest of high coupling coefficients, whose the implication is limited is with growing mass ratio. Notably, it highlights the necessity of high coupling coefficient, the influence of which on frequency tuning is limited as the mass ratio grows.
Read the full abstract

All authors David Gibus, Omar Larhlid, Ahmed Guerine, Adrien Badel

Computation and Language

BDBBQ: A Human-AI Collaborative Benchmark for Bengali Social Bias in LLMs

Despite growing attention to social bias in large language models (LLMs), existing bias benchmarks provide limited coverage of culturally grounded evaluation in Bengali, particularly within Bangladesh's sociocultural context. To address this gap, we develop a bilingual extension of BanglaBBQ comprising 61,312 question-answering instances across 12 social categories in English and Bengali. The benchmark incorporates balanced answer permutations, contrasting question framings, and both ambiguous and disambiguated contexts to systematically examine model performance and stereotype-aligned bias across languages. We validate the dataset through structural checks, 220 statistical tests, and expert annotation of 150 instances by four annotators, achieving strong inter-annotator agreement (Fleiss' ąppa}=0.8349). We evaluate eight LLMs across 490,496 instances and identify five key findings. First, disambiguated contexts yield substantially greater accuracy improvements in Bengali (+36.59 percentage points) than in English (+17.71). Second, all evaluated models exhibit lower accuracy in Bengali, although Mistral-Nemo shows the smallest cross-lingual accuracy gap (5.12 percentage points). Third, stereotype-aligned bias exhibits language and model dependent variation, increasing in Bengali for Gemma 2 9B but decreasing for Mistral-Nemo. Fourth, higher accuracy does not necessarily correspond to lower stereotype-aligned bias. Finally, bias varies considerably across social categories, with Qwen 2.5 7B exhibiting stronger stereotype alignment in Bengali for Politics and Physical Appearance. These findings demonstrate that cross-lingual performance and social bias are distinct dimensions of LLM behavior and underscore the importance of culturally grounded, language-sensitive benchmarks for evaluating fairness beyond English-centric settings.
Read the full abstract

All authors Provat Saha, Ashika Habib Khan, Tahseen Tanha, Hasin Ishrak, Muhammad Iqbal Hossain

Audio and Speech Processing

REAL: Reward Modeling through Evidence-Linked Analysis of Lifelikeness in Generated Speech

As modern text-to-speech systems become increasingly intelligible and natural, evaluation is shifting from detecting gross synthesis failures to identifying subtle departures from lifelike human delivery. Because large-scale listening tests are costly and slow, speech reward models can serve as scalable automatic judges for comparing generated utterances, curating data, and supporting iterative TTS development. Existing speech reward models can predict scores or pairwise preferences and may generate natural-language rationales, but their supervision is typically defined at the utterance or pair level. The connection between local acoustic phenomena and the final preference therefore remains implicit. We introduce REAL, a framework for Reward modeling through Evidence-linked Analysis of Lifelikeness in generated speech. Its central idea is to supervise not only the final pairwise preference but also a structured account of its supporting observations, linking local audio spans to perceptual phenomena, affected dimensions, and the overall judgment. Specifically, REAL-Jury constructs these evidence-linked reports through timestamp alignment, multi-model analysis, and dispute-oriented deliberation. REAL-Judge learns to generate them from paired audio inputs. REAL-Eval assesses preference prediction in two tiers, from salient perceptual differences to subtle distinctions among strong TTS systems. Instantiated with MiMo-Audio, REAL-Judge improves pairwise preference accuracy over its backbone by 16.01 and 11.41 percentage points on the two tiers, respectively, and outperforms the open-source speech reward models SpeechJudge and UniSRM on both tiers.
Read the full abstract

All authors Yaodong Song, Jie Lian, Xusheng Yang, Genliang Zhao, Hongjie Chen, Jian Kang, Zhen-Hua Ling, Jie Li, Yongxiang Li

Computer Vision

PTC-IoU: CPU-Friendly Detection Quality Estimation for Edge Video Analytics

An edge video analytics system must estimate the quality of local detections since there are no ground-truth labels available during runtime. Detectors usually provide confidence scores for their results, but these scores can not reveal every missed object or misplaced bounding box. So a quality estimator is necessary to provide a decision signal indicating which frame has potential detection failures that need to be escalated to a more powerful model in the cloud. Neural network based estimators add workload to the GPU already running detection, while CPU capacity may remain available. We present Predictive Temporal Consistency IoU (PTC-IoU), which moves this workload to CPU by compacting detection boxes histories to support deterministic result scoring. Comparing to neural counterparts, experiments on representative edge hardware, Nvidia Jetson Orin Nano, show that our method reduced per-frame estimator latency from 10.3 ms for neural counterparts to 2.2 ms (4.7x faster), increased pipelined throughput by 13.6%, and cuts estimator energy consumption by up to 95%, while maintaining competitive final detection quality verified on multiple datasets.
Read the full abstract

All authors Wei Geng, Hyerin Kim, Chang Lu, Zhenghao Lu, Joerg Ott

High Energy Physics (Phenomenology)

From oxygen to lead: testing a common QGP description with high-$p_\perp$ probes

The observation of suppression and finite high-$p_\perp$ elliptic anisotropy in O+O and Ne+Ne raises the question of whether a common quark–gluon plasma (QGP) description can simultaneously account for both observables across systems from oxygen to lead. We present predictions using event-by-event bulk evolution constrained by soft-sector observables and coupled to DREENA radiative and collisional energy loss, with the same microscopic parameter set for O+O, Ne+Ne, Xe+Xe, and Pb+Pb. The calculation reproduces the observed ordering of suppression and predicts elliptic anisotropies of comparable magnitude across the four systems in 10–20$\%$ collisions. We predict a modest centrality dependence of $v_2^{\mathrm{SP}}$ and a more pronounced decrease of $v_3^{\mathrm{SP}}$ toward peripheral light-ion collisions, in contrast to the corresponding trends in larger systems. In light ions, we predict that $v_2^{\mathrm{SP}}/(1-R_{AA})$ is nearly independent of momentum at high $p_\perp$. We further predict that the differential ratio $v_2\{4\}/v_2\{2\}$ depends only weakly on momentum at high $p_\perp$, with magnitudes and centrality trends broadly comparable to those of the measured soft-sector ratios, suggesting that the soft-sector fluctuation pattern persists in hard–soft correlations. Predictions for D and B mesons extend the comparison to probes of different flavors. Taken together, these system-size, harmonic, fluctuation, and flavor dependences provide experimental tests of how far a common QGP description extends.
Read the full abstract

Systems and Control

Observer-based parameter estimation in a Darcy model towards dam internal seepage monitoring with validation on real data

This paper addresses internal seepage monitoring in hydraulic infrastructures through the estimation of physical parameters in a distributed model. A physics-based approach relying on Darcy's law is adopted to describe the spatiotemporal evolution of hydraulic head within the structure. To enable real-time applicability, a Proper Orthogonal Decomposition (POD) is employed to derive a reduced-order model of the underlying partial differential equation. An Extended Kalman Filter (EKF) is then designed to perform online estimation of the hydraulic diffusivity, a key parameter related to the structural state. The proposed method is validated directly on real measurement data through a twofold approach. First, the estimated parameter is compared with the value obtained via offline batch optimization performed on the same dataset, showing consistent results. Second, the ability of the resulting model, parametrized by the estimated diffusivity, to reconstruct the observed measurements is assessed, to evaluate its suitability for seepage monitoring.
Read the full abstract

Dynamical Systems

Singularities in leaf closures on complex surfaces

We prove a geometric and dynamical criterion excluding minimal sets contained in the regular locus of a holomorphic foliation on a compact complex surface. In particular, we prove that every Jouanolou foliation of degree at least two has no exceptional minimal set. We also give an explicit lower bound, growing linearly with the degree, for the number of singularities in every regular leaf closure of a Jouanolou foliation. This bound persists under small perturbations, as does the existence of a nonempty open saturated set whose leaves accumulate at all singularities. Under finite pullback unramified over the singular set, the bound is multiplied by the degree of the normalised component over the base leaf. Together with a general persistence theorem, this yields nonempty analytic open sets of foliations on every smooth projective surface with arbitrarily large uniform lower bounds on the number of singularities in their regular leaf closures. On double-plane K3 surfaces, we obtain foliations with ample canonical bundle and explicitly describe and count their singularities.
Read the full abstract

Classical Physics

Data-Driven Conservation Laws for Newtonian Systems: Physics-Informed Neural Networks and Eisenhart-Duval Geometry

We introduce a Physics-Informed Neural Network approach for the data-driven reconstruction of the symmetries and conservation laws of Newtonian dynamical systems, based on the geometrization of the dynamics through the Eisenhart-Duval lift. Within the Eisenhart-Duval lift, the symmetry conditions reduce to the linear Killing or Conformal Killing equations of the extended metric. We combine a Hamiltonian neural network, which reconstructs the potential function from the data, with a random-feature neural network, which solves the symmetry conditions for the collineations of the extended geometry. From the learned vector fields we construct the Lie algebra and, through Noether's theorem, the corresponding conserved quantities, while the derivation of the curvature invariants for the extended geometry provide allow us to perform an independent consistency test. We apply the method to the one-dimensional Ermakov-Pinney equation and to the one-dimensional harmonic oscillator, with and without a quartic perturbation, by using different level of noise in the phase-space data. For the Ermakov-Pinney equation, the three Killing vectors of the flat extended plane, and the corresponding nonlocal conservation laws, are recovered for, while for the harmonic oscillator, the ten conformal Killing vectors which form the $so(3,2)$ Lie algebra are determined for noise levels up to $10\%$. When the quartic term is added, the broken symmetries are identified as approximate symmetries.
Read the full abstract

Statistical Mechanics

Discrete time crystals stabilized by bounded measurement noise

Periodically driven clean interacting systems generically heat up, which destroys discrete time-crystalline order. Here we ask whether continuous weak measurement can stabilize such order. We study a long-range interacting spin chain whose local occupation is continuously monitored. Measurements generally induce a transition from volume-law to area-law entanglement. Yet low entanglement alone is not enough: along typical quantum trajectories, the system still reaches a featureless state with no subharmonic response. In contrast, the most-probable trajectory, obtained by suppressing measurement fluctuations, shows robust period doubling. To connect these limits, we introduce a bounded-noise protocol that restricts measurement fluctuations to a finite window. Narrow windows preserve period doubling. Intermediate windows produce coexisting time-crystalline and non-time-crystalline trajectories. Over time, trajectories leave the time-crystalline group one by one, so the order is long lived but finite. Wide windows destroy the order. Our results establish bounded measurement noise, discarding the large-deviation trajectories, as a route to time-crystalline order in clean systems. They also show that the measurement-induced area-law phase is necessary, but not sufficient, for this order.
Read the full abstract

Materials Science

Quantum Embedding from Molecular to Crystalline Environments at the Cost of a Gas-Phase Calculation

We present a direct-space formulation and implementation of quantum density functional theory (DFT) embedding for molecular active subsystems embedded in frozen environments (hosts) of arbitrary dimensionality, from finite molecular hosts to extended systems. The host is computed once without the embedded molecule (the guest) and is represented only by its electron density fitted in an atom-centered auxiliary Gaussian basis together with its nuclear charges with the periodic electrostatics separated into an explicit near-field and a continuous fast-multipole far-field operator. The method allows the removal of host atoms from this representation to create vacancies or other voids without requiring a supercell creation. Analytic nuclear gradients are also implemented, enabling structural relaxation of the active subsystem. We demonstrate the versatility of the approach through applications to adsorbates on periodic surfaces and molecules embedded in bulk solvent environments. For CO on MgO(001), we show that DFT-in-DFT converges rapidly and reproduces the periodic interaction energy within 0.44 kJ/mol while reducing the cost by a factor of 27. Real-time time-dependent embedded DFT (RT-TDDFET) of methylenecyclopropene in periodic water is showcased to produce a ~0.6 eV solvatochromic blue shift, while strong-field propagation of CO2 on hexagonal boron nitride (hBN) yields substrate-induced effects at the cost of a gas-phase propagation. Coupling the embedding scheme to excited-state correlated wave function theory methods such as CC2 yields a solvent shift within 0.01 eV of supermolecular CC2 references, while reducing the computational cost by over 500 times.
Read the full abstract

Nuclear Theory

Low-lying states of $^{115}$Cd modeled as a valence neutron coupled to $^{114}$Cd

$^{114}$Cd has historically been regarded as an ideal quadrupole vibrator due to its near-canonical level spacings, though this interpretation is disputed due to electromagnetic transition data from the two- and three-phonon states. In this paper we investigate whether, with the extension of including the single-phonon octupole state, the model is still useful in describing the coupling of a neutron to a cadmium-114 core to model the lowest- energy cadmium-115 states, as the triplet states have small to no influence in the calculation. The results thus provide supporting evidence for the tentatively-assigned $J^\pi$ values of the lowest-energy $^{115}$Cd states. A more- rigorous derivation than often used for the vibrational B(E{\lambda}) is used, in conjunction with the $^{115}$Cd structure calculation, to provide a theoretical prediction of the $^{114}$Cd octupole deformation.
Read the full abstract

All authors K. Amos, P. R. Fraser, S. Karataglidis, D. Bhattacharyya, L. Canton, K. Murulane

Accelerator Physics

Physics design of a dual-ring light source toward MHz-rate free-electron laser radiation

Combining the high peak brilliance and short pulse duration of free-electron lasers (FELs) with the high repetition rate of diffraction-limited storage rings (DLSRs) is a promising direction for accelerator light sources. In this paper, we propose a dual-ring light source concept that, at MHz repetition rates, delivers pulses with a pulse durations and peak power intermediate between those of FELs and DLSRs. Such a light source can be realized as an upgrade of an existing DLSR by adding an isochronous ring equipped with transverse gradient undulators (TGUs) and connecting the two rings via bunch compression and decompression systems. We present the conceptual layout and show that bunch compression shortens the electron bunches to ps-scale durations; multi-turn radiation in the isochronous ring raises the repetition rate to 1.36 MHz from a single TGU; and the long TGUs enhance the brilliance, delivering $10^{7}$–$10^{9}$ photons per pulse across the soft and hard X-ray range, an increase of 2–4 orders of magnitude over the original DLSR. The dual-ring light source thus provides a cost-effective route to MHz FEL operation based on existing storage-ring infrastructure, combining the high repetition rate of a DLSR with the short pulses and high brilliance of a linac-driven FEL.
Read the full abstract

All authors Fancong Zeng, Yi Jiao, Weihang Liu, Yu Zhao, Weilun Qin

Number Theory

Linear disjointness and values of Mahler M-functions

We prove that the algebra generated by the values of Mahler M-functions at algebraic points admits a decomposition as the tensor product, over Q, of a countable family of linearly disjoint algebras. Consequently, the study of linear and algebraic relations among such values reduces to the corresponding problem within each of these smaller algebras. Our proof relies on recent advances of the authors on Mahler's method in several variables. As an application, we obtain a complete description of the algebraic relations among the values of the Fredholm series and their derivatives at nonzero algebraic points. This fully resolves an old problem originating in Mahler's work.
Read the full abstract

Dynamical Systems

Orbits along primes in totally minimal systems

In this paper, we show that every totally minimal system admits a point whose orbit along the primes is dense. Moreover, assuming the Dickson-Hardy-Littlewood conjecture, we show that in any totally minimal distal system, the orbit along the primes of every point is dense.
Read the full abstract

All authors Wen Huang, Song Shao, Hui Xu, Xiangdong Ye

Atomic Physics

HFR distorted-wave photoionization of Te III: a preliminary benchmark for large-scale atomic data production to model nebular-phase kilonovae

The shortage of photoionization cross sections for heavy elements remains an obstacle to non-local thermodynamic equilibrium (NLTE) modeling of nebular-phase kilonova ejecta, where hydrogenic prescriptions have often been adopted in the absence of ion-specific calculations. We present a first preliminary benchmark of direct photoionization cross sections for doubly ionized tellurium, obtained using the distorted-wave (DW) approximation within the pseudo-relativistic Hartree-Fock (HFR) method implemented in the Cowan codes, in which a semi-empirical adjustment based on the observed energy levels is performed. For the photoionization from the ground level of Te III to the ground level of Te IV, comparison with Dirac R-matrix data shows encouraging agreement in the magnitude and overall decrease of the non-resonant background in the lower photon-energy region, approximately below 4 Ry (autoionizing resonances are absent from the present direct treatment). These preliminary results support HFR distorted-wave calculations as a promising and computationally tractable approach for large-scale production of photoionization cross sections, with lanthanides and actinides as longer-term targets, in light of their importance to kilonova modeling and of the severe lack of atomic data for these species.
Read the full abstract

All authors Jérôme Deprince, Lucas Maison, Charlie Guillaume, Exaucé Bokamba Motoumba

Artificial Intelligence

AutoBuddy: Evaluating Proactive, Always-On Agents on Task Discovery and Execution in a Simulated Workplace

As LLM agents become more autonomous, they are increasingly expected to run as always-on assistants on a user's behalf, identifying work to be done and completing it without being asked. Yet existing benchmarks usually give the agent its task, whereas in knowledge work the requests sit in the user's workspace and the agent must first recognise which obligations exist. We study proactive task discovery and execution, in which an agent must infer a user's pending tasks from their workspace and complete them in the systems where the work lives. This setting poses two challenges. For the agent, the tasks are buried in a large volume of mail, chat, documents and records. For evaluation, a benchmark needs a simulated organisation whose people, events and documents are mutually consistent, so that tasks arise realistically and their correct outcomes are known. We introduce AutoBuddy-Bench, a simulator that addresses these challenges. It models an organisation's employees, projects, business systems and event history as a temporal graph, renders it into realistic workspace artefacts, derives ground-truth tasks from the events that create obligations, and computes each task's outcome from the world state; execution is judged by the agent's changes to the workspace. To evaluate current LLMs, we build AutoBuddy, a reference agent with interchangeable backbone LLMs: it identifies tasks from the workspace, grounding each in its source records, and executes each in a separate session with write access. Varying only the backbone, we find a systematic shift in where tasks are lost: weaker models fail mainly at discovery, intermediate models identify tasks but often stop short of carrying them out, and the strongest models fail mainly during execution, most often on business-system operations. Discovery precision is only weakly related to execution quality, suggesting the two are distinct capabilities.
Read the full abstract

All authors Shiqi Chen, Daniel Madrigal Diaz, Wei-Ning Chen, Gbola Afonja, Fangkai Yang, Rujia Wang, Chetan Bansal, Victor Rühle, Saravan Rajmohan, Dongge Han

Computation and Language

More Than Just Style: Benchmarking Personality Influence in Large Language Models

In addition to the effects of explicit input prompts and instructions, large language model outputs are influenced by underlying behavioral preferences similar to human personality tendencies. These tendencies can affect how models interpret situations and make decisions. However, their influence is difficult to disentangle from the effects of model capabilities and instructions. Therefore, we introduce SPACE, a benchmark framework combining a controlled evaluation protocol with complementary datasets for personality verification, capability assessment, and hierarchical situational tasks. Controlled comparisons across guidance conditions help distinguish personality influence from measured capability changes and immediate instruction effects. Experimental results demonstrate that although prompting can shift generation style, models' internal personality tendencies are persistent and not readily overridden by opposing instructions. Furthermore, these underlying tendencies have an observable influence on LLM task performance, highlighting the need for greater attention to these characteristics and potential applications. Training code and SPACE benchmark specifications are available at https://github.com/Niklauseik/SPACE-training.
Read the full abstract

All authors Yuhao Liu, Mengyu Wang, Xiao Chen, Lu Liu

Networking

Semantic-Aware Communications in Industrial IoT:What Can Be Tested Today, and What Is Missing?

Classical networks treat every bit as equally important, which wastes time and energy in factories and requires extra traffic engineering for critical traffic. Semantic communications send only what a task needs, but how can they be tested under realistic conditions? This paper examines what can be tested today in 6G industrial IoT, that is, networks of factory machines, robots, and sensors, and what is still missing. It introduces semantic importance, a score attached to each message that states how much the message matters, and uses it in a context-driven design to derive eight testing needs. Existing simulators, emulators, testbeds, codecs, and learning frameworks are assessed against these needs, which reveals five gaps. We propose a toolchain of four components built around the ns-3 simulator for simulated and emulated scenarios.
Read the full abstract

Solar and Stellar Astrophysics

Zeeman-Doppler imaging of early-type stars. Testing Stokes parameter inversions with theoretical magnetohydrodynamic models

Zeeman-Doppler imaging (ZDI) is widely used to reconstruct magnetic fields in early-type stars from time-series spectropolarimetric observations, yet its accuracy has not been systematically validated against physically self-consistent, complex magnetic field models. We aim to assess the reliability and limitations of modern ZDI techniques applied to fossil fields of early-type stars. We generated synthetic spectropolarimetric data from three-dimensional magnetohydrodynamic (MHD) simulations of fossil magnetic fields. The resulting Stokes profiles - individual lines and least-squares deconvolved (LSD) profiles - were inverted with contemporary ZDI codes using Stokes IQUV and IV data. We compared the reconstructed maps with the input MHD surface fields and quantified deviations in field strength, vector orientation, and harmonic energy decomposition. For Stokes IQUV inversions, ZDI recovers field vector components and strength with better than 10% accuracy, while Stokes IV inversions degrade to 15-20% for the most complex geometries. The field inclination is recovered to within 10 deg, with errors increasing to 15-20 deg in the same cases. Full Stokes inversions reproduce poloidal and toroidal energies to within a few percentage points. LSD-based inversions yield accuracy comparable to, and sometimes better than, that achieved with ZDI using individual Fe II lines. Imposing a current-free poloidal constraint, which the input field satisfies, only helps where the unconstrained Stokes IV inversion performs poorly. Cross-talk between magnetic and abundance reconstructions is weak, with spurious inhomogeneities at or below 0.1 dex on average. Modern ZDI robustly reconstructs fossil magnetic fields in early-type stars, including their smaller-scale non-dipolar features, particularly when full Stokes information is available.
Read the full abstract

Computational Finance

Diversification Across Predictors: Heterogeneous Ensemble Learning for Cross-Sectional Return Prediction

Cross-sectional stock return prediction remains challenging due to extremely low signal-to-noise ratios, high-dimensional factor spaces, and non-stationary predictor effectiveness. Individual model classes exhibit complementary failure modes in this regime: neural networks overfit noise due to high variance, while linear models underfit nonlinear structure due to high bias. We propose a heterogeneous ensemble framework that extends the financial principle of diversification to the predictor level. Through a bagging procedure combining linear predictors with nonlinear predictors, each model functions as a distinct specialist whose prediction errors exhibit low cross-covariance, enabling error cancellation upon aggregation. We formalize this mechanism via bias-variance-covariance decomposition, establishing an exact mathematical correspondence between portfolio diversification and predictor aggregation. On CSI300 daily-frequency Alpha158 factor datasets, the proposed ensemble achieves a Rank IC improvement exceeding 13% over standalone neural networks and surpasses tree-based state-of-the-art methods including XGBoost and Double Ensemble, as measured by both predictive accuracy and portfolio-based evaluation metrics.
Read the full abstract

Accelerator Physics

Longitudinal beam dynamics II. Lecture 2: intensity effects and beam instabilities

In this lecture, dealing with intensity effects, the equations of longitudinal motion with induced beam voltage are analyzed. This allows us, first, to describe the single-bunch potential well distortion and the loss of Landau damping. Then the linearized Vlasov equation is used to derive step-by-step the Lebedev equation. This matrix equation is then applied to find the thresholds, growth rates and spectra of an unstable multi-bunch beam. The lecture is concluded by a discussion of possible beam-instability cures.
Read the full abstract

High Energy Physics (Theory)

On a soliton with a non-normalisable zero mode

We consider the holomorphic baby Skyrme model, which we propose as a simple tractable example of a soliton with a non-normalisable zero mode. The $1$-soliton has the same structure as a lump of fixed size. The size modulus of the sigma model becomes massive, while the phase modulus remains massless. The latter is non-normalisable in the $(0+1)$-dimensional sense because of the soliton tail. A suitable way to treat it is to consider the radial $(1+1)$-dimensional effective action on the half-plane. The soliton core sets the boundary condition for the effective model, which we evaluate with the WKB approximation and also numerically. We discuss the analogue of the effective moduli space approximation with a $1+1$ model and a boundary term at a certain cutoff scale. This may serve as a prototype for more complex systems featuring non-normalisable zero modes such as certain kinds of monopoles in $3+1$ dimensions.
Read the full abstract

Cryptography and Security

Say Shibboleth: Fingerprinting the LLMs and Harnesses Behind Agentic Attacks

We study the fingerprinting of agentic cyberattacks. Our objective is to reliably infer information about the attacker's configuration, such as LLM version and harness, from telemetry. To achieve this, we introduce a technique we call shibboleth traps: modular honeypot services designed to elicit discriminative behaviors from agentic attackers. The traps make informative interactions a condition for obtaining resources the agent seeks, eliciting model declarations and characteristic writing without requiring a successful prompt injection. To achieve harness fingerprinting, we introduce general mechanisms to indirectly probe peculiar execution behaviors of harnesses. These probes reveal how a harness masks secrets in tool output, enforces command time limits, and schedules tool calls, using only the effects visible in subsequent requests. We show that combined declarative and behavioral inference distinguishes among 15 candidate models with an average 90% top-1 accuracy, with behavioral classifiers evaluated on held-out harnesses or task families. In a separate experiment across four harnesses, fixed probe rules identify the correct harness with 92% accuracy. When instructed to conceal their identity, agents still leave useful behavioral evidence even as their declarations become uninformative.
Read the full abstract

All authors Dario Pasquini, Reworr, Giuseppe Ateniese, Manish Bhatt, Sergej Epp

Tissues and Organs

Clinical predictions using artificial intelligence in neurosurgery and its subspecialties: a systematic review and meta-analysis

Artificial intelligence models for clinical prediction are proliferating across neurosurgery, but their aggregate discriminative performance, distribution along the care pathway and methodological quality are unclear. We systematically reviewed and meta-analysed these studies in neurosurgery and its subspecialities along with their risk of bias appraisal. We performed a PRISMA search for peer-reviewed original studies using AI to predict a clinical outcome in neurosurgical populations. Metrics were pooled using random effects modelling. Each study was classified as diagnostic, prognostic or forecasting and was assigned an evidence level. Clinical prediction task sub-stratification analysis was performed. Risk-of-bias was assessed using PROBAST. 1063 studies were included. Across 451 studies reporting AUC, pooled discrimination was 0.86, ranging from 0.84 (spine) to 0.87 (neuro-oncology), with extreme heterogeneity. Performance varied by clinical pathway stage, highest for diagnosis, grading and molecular tasks and lowest for post-operative tasks. Only 51 studies were using prospective data with external validation, whereas 619 rested on internal validation alone or no validation. 89 percent studies were at high risk-of-bias, driven by the analysis domain. AI models report high performance across all the neurosurgical subspecialities, concordant across AUC, accuracy, sensitivity and specificity and robust to risk-of-bias stratification. Performance is highest for imaging-anchored diagnostic tasks and lowest at decision-facing post-operative stage. However, the literature is dominated by optimism-prone, internally-validated and poorly calibrated models with extreme heterogeneity. External validation, calibration analysis and reporting guidelines adherence remain prerequisites for clinical adoption.
Read the full abstract

All authors Prajwal Ghimire, Aswin Chari, Jonathan Shapey, Keyoumars Ashkan

Physics and Society

Retweeting patterns reveal partisan asymmetries in information sharing and signal state-level electoral outcomes in the 2020 US presidential election

As social media has become a primary source of political information, concerns are growing that online information consumption may intensify political polarization and reshape democratic processes. Yet how large-scale online information consumption organizes polarized political ecosystems and relates to offline voting remains unclear. Here we reconstruct partisan information ecosystems from more than 500 million tweets generated during the 2020 US presidential election, integrating information diffusion across media outlets, influencers, and social peers. Contrary to the conventional expectation that broader information exposure promotes moderation, we find that retweeting a wider range of political information is associated with stronger political polarization. We further show systematic asymmetries in political information ecosystems: the Republican ecosystem is highly concentrated and ideologically biased, whereas the Democratic ecosystem is more dispersed and primarily organized around relatively moderate left-leaning sources. Finally, we demonstrate that several shared retweeting signatures robustly map onto state-level electoral outcomes despite cross-state heterogeneity. Our findings reveal asymmetric political information ecosystems as a key organizing principle of online political communication and identify collective retweeting patterns as a scalable signal of real-world voting behaviours.
Read the full abstract

All authors Dawei Li, Qinjun Wan, Xin Wang, Dahe Jiang, Junyu Lu, Shitong Wang, Zhelun Zhang, Zhenyu Guan, Shaoting Tang, Feng Fu

Robotics

PCTac: Mechanochromic Photonic-Crystal Tactile Sensor for Dexterous Manipulation

Modern visuotactile sensors primarily sense contact by imaging the deformation of a compliant interface. Using techniques such as photometric stereo, they can recover detailed contact geometry, often represented as surface normals or depth. However, force-related information typically still needs to be inferred indirectly through calibration, geometric reconstruction, or learned mappings. Mechanochromic photonic-crystal (PC) materials offer an alternative physical encoding mechanism: contact loading changes their periodic microstructure and spectral response, directly transforming local stress into a spatially resolved structural-color field. Despite this potential, PC materials have rarely been explored as visuotactile sensors for closed-loop robot learning. We present PCTac, an open-source mechanochromic PC fingertip that uses a dense hue field as its tactile observation. Loading characterization demonstrates a repeatable relationship between the applied normal load and the observed hue response. We further integrate this representation directly into a Diffusion Policy and evaluate it on three real-robot manipulation tasks. Across 90 physical rollouts, PCTac Hue achieves 68 successes (75.6%), compared with 60 (66.7%) for PCTac RGB, 57 (63.3%) for Wedge RGB, and 29 (32.2%) for vision-only. PCTac Hue also achieves the highest observed success rate on each evaluated task, including Perturbed Test-Tube Insertion, where it achieves 16/30 successes versus 13/30 for tactile RGB baselines and 1/30 for vision-only. These results demonstrate the potential of mechanochromic structural color as a compact, load-sensitive tactile representation for robot learning and closed-loop manipulation. Project page: https://bey0ndt1me.github.io/PCTac-page/.
Read the full abstract

All authors Ruilin Zhang, Xuanshuo Liu, Junzhe Dai, Kefei Wu, Xuhao Qin, Ziyuan Jiao, Wanlin Li, Yurui Qu, Chenxi Xiao

Robotics

AutoVLA-D: Learn a Trajectory-Geometric Condition from Reasoning for Autonomous Driving VLA

Autoregressive VLA planning usually predicts continuous trajectories as the final output, but driving instructions often describe actions rather than ego states, which weakens direct conditioning on the final continuous trajectory. We address this gap with TrajCond, which introduces a discrete trajectory-geometric condition for VLA planning. Specifically, we represent TrajCond as a 1D-Brohan differential action trajectory and learn it with a two-stage training framework before final continuous trajectory prediction. With the simultaneous autoregressive generation of discrete and numeric trajectories, we name our method AutoVLA-D (Double). Experiments on NAVSIM navtest, nuScenes, and NeuroNCAP show that discrete trajectories are weaker as direct outputs but more effective as TrajCond, consistently improving final continuous trajectory prediction across closed-loop and open-loop driving benchmarks. Code is available at https://github.com/ShivaTeam/AutoVLA-D.
Read the full abstract

All authors Yuguang Yang, Zhewen Tan, Canyu Chen, Linlin Yang, Cheng Chi, Chunyang Liu, Kehua Sheng, Jinyu Yang, Baochang Zhan, Bo Zhang, Yan Wang, Xianbin Cao