Machine Learning
Papers
Fresh from arXiv, typeset properly. Highlighted words trade on the market.
All of arXiv
Tuesday 15 September
High-capacity associative memories based on Kernel Logistic Regression (KLR) exhibit a "Ridge of Optimization" characterized by extreme stability and a highly skewed weight spectrum. However, the dynamical process by which learning converges to this critical regime has remained unclear. This paper provides a geometric analysis of the learning trajectories on the statistical manifold of a KLR-trained Hopfield network. By comparing the paths of Gradient Descent (GD) and Natural Gradient Descent (NGD), we elucidate the mechanisms governing the optimization process. Our analysis reveals that learning on the Ridge proceeds in two distinct phases. We show that the extreme curvature of the Ridge causes standard GD to follow a highly oscillatory, non-geodesic path. In stark contrast, NGD explicitly corrects for this geometry, following the ideal geodesic path and completely overcoming the instabilities faced by GD. We demonstrate experimentally that NGD not only converges significantly faster but also achieves a solution with superior generalization performance. These results establish that the highly structured geometry of the Ridge is optimally suited for information-geometric optimization, providing a new perspective on the interplay between learning dynamics and emergent representation geometry.
Machine Learning
SOTER: A Generative Time-Series Foundation Model for Wearable Human Physiological Signals
Time-series foundation models have demonstrated strong cross-domain transfer, yet their common architectural assumptions remain poorly aligned with wearable physiological signals, which are multichannel, irregularly sampled, noisy, and governed by coupled continuous-time dynamics spanning distinct spectral scales. We present SOTER, a generative foundation model for wearable physiological time series that unifies cross-channel coupling, spectrum-guided expert specialization, and continuous-time latent evolution within a single pre-training framework. SOTER combines a spatial feature-aware backbone that models inter-signal dependencies, a power spectral density (PSD)-guided mixture-of-experts layer that routes representations to experts associated with fixed spectral bands through an inspectable, non-learned rule, and a neural controlled differential equation decoder that supports prediction and imputation at arbitrary timestamps. We pre-train SOTER on 226 billion time points from five public physiological datasets and evaluate the same pre-trained model across out-of-distribution zero-shot forecasting, frozen-encoder linear-probe classification, and continuous-time imputation on wearable benchmarks. SOTER achieves the best RMSE on 4 of 6 datasets and the best MAE on 5 of 6 in zero-shot forecasting, the highest average Macro-AUROC in classification, and the lowest imputation error on all six datasets at 75% missingness. It further remains robust to additive acquisition noise, matching or surpassing baselines evaluated on clean inputs even under the strongest corruption. These results indicate that domain-specialized foundation models for wearable physiology benefit from jointly modeling channel structure, spectral scale, and continuous-time dynamics.
Read the full abstract
Machine Learning (Statistics)
On the disintegration of the stochastic majority vote: From PAC-Bayesian bounds to a self-bounding algorithm
Weighted majority votes are central to many successful ensemble methods. PAC-Bayesian theory provides tight generalization guarantees for such models by analyzing the expected risk of stochastic classifiers, while analyzing the risk of deterministic majority votes relies on surrogate bounds. To avoid these surrogates, Zantedeschi et al. ( 2021) introduced guarantees for stochastic majority votes, but the resulting models remain randomized. In this paper, we propose a derandomization framework for stochastic majority votes. To do so, we apply recent advances in disintegrated PAC-Bayesian theory directly to the space of majority vote weight vectors, transforming stochastic guarantees into certificates for a single deterministic majority vote. We derive two families of high-probability generalization bounds, covering both data-independent and data-dependent constructions of the ensemble, which naturally lead to a self-bounding learning algorithm optimizing deterministic majority vote guarantees.
Read the full abstract
Materials Science
ZStar: A unified toolkit for polarization, Born effective charges, dielectric and piezoelectric responses, and infrared and Raman spectra
Born effective charge (BEC) tensors link atomic displacements to polarization, thereby connecting lattice vibrations with dielectric response and infrared activity. We present ZStar, an open-source Python toolkit that automates calculations of polarization, BEC, phonons, static and frequency-dependent dielectric response, and infrared and Raman spectra. ZStar supports bulk crystals, two-dimensional materials, one-dimensional structures, and molecules through an integrated workflow for input preparation, execution, and response analysis. Berry-phase polarization and real-space dipoles describe responses along periodic and open directions, respectively. Its unified, symmetry-adapted finite-displacement framework jointly determines BEC and zone-center force constants from the same calculations, improving computational efficiency while preserving numerical accuracy. Strain-response calculations further provide proper piezoelectric coefficients and their internal-relaxation contributions for bulk crystals. In the ABACUS+PYATB route, electronic Hamiltonian data are reused to obtain Raman response without additional self-consistent calculations. Native response implementations from VASP and CP2K are also integrated. Matched infrared/Raman benchmarks against the implemented separate workflows achieve speedups of up to 8.4. An accompanying agent skill provides an agent-friendly interface for workflow integration. ZStar therefore offers an efficient and automated route for exploring polarization, BEC, dielectric and piezoelectric responses, and vibrational spectra across different material dimensionalities.
Read the full abstract
Analysis of PDEs
Formalization of Landau damping in the Vlasov–Poisson equations in Lean
We present a formalization in Lean 4 of Mouhot and Villani's theorem on nonlinear Landau damping in $\mathbb T^d$ in all Gevrey regularity indices $s > 1/3$ for small backgrounds.
Computation and Language
Smarter by the Moment: Environment-Driven Dynamic Policies for Continual LLM Improvement
Large Language Models (LLMs) have achieved remarkable progress across diverse domains, but continual adaptation to evolving tasks and environments remains a key challenge. Existing memory-augmented approaches retrieve individual past examples as direct references, but do not explicitly synthesize actionable strategies from them, causing the same types of errors to recur. We propose Dynamic Retrieval-based Policy Generation (DRPG), a framework that integrates memory-based retrieval with a dynamic policy generator, leveraging historical data and environment feedback to produce task-specific policies for continual LLM improvement. We evaluate DRPG across six benchmarks spanning text-to-SQL, question answering, medical diagnosis, and Python programming, using seven LLMs from both proprietary and open-weight families. DRPG outperforms strong baselines across most datasets and models. Further analysis demonstrates that DRPG's policy generation is robust to retrieval strategy, operates effectively without prior policy continuity, and can leverage smaller or cross-family models as cost-efficient policy generators. We also find that the benefit of policy-level guidance depends on task characteristics, offering practical insights into when and under what conditions this mechanism is most effective.
Read the full abstract
Materials Science
Fractional-Quantum Ferroelectrics: A Route to High-Mobility Ferroelectric Semiconductors
Ferroelectric semiconductors are promising for multifunctional electronics, yet their typically low carrier mobilities remain a major limitation. Using first-principles phonon-limited transport calculations for monolayer In$_2$Se$_3$, we show that this limitation depends critically on the microscopic origin of ferroelectricity. In displacive $\beta'$-In$_2$Se$_3$, low-frequency ferroelectric modes dominate carrier scattering, with additional contributions from longitudinal-optical (LO) phonons, limiting the room-temperature electron mobility to a few cm$^2$/(V s). By contrast, in fractional-quantum ferroelectric $\alpha$-In$_2$Se$_3$, ferroelectric-mode scattering is absent because polarization arises from discrete lattice-scale atomic displacements rather than soft-mode condensation. Transport is therefore dominated by LO phonons, yielding a room-temperature mobility above 70 cm$^2$/(V s). Carrier doping further screens long-range electron-LO-phonon interactions and raises the mobility beyond 300 cm$^2$/(V s) at experimentally accessible densities. These results establish that fractional-quantum ferroelectricity can decouple robust polarization from strong intrinsic carrier scattering, offering a route toward high-mobility ferroelectric semiconductors.
Read the full abstract
Category Theory
On Models of the Planar Lambda Calculus
We construct an adjunction relating two approaches to modelling the planar lambda calculus: semi-closed operads and planar lambda-models. We use this to obtain a planar version of Scott's representation theorem.
Computer Vision
TEDi: Temporal Memory-Enhanced and Denoising Transformer for Surgical Instrument Segmentation
Query-based segmentation methods have shown promising potential for surgical instrument segmentation and recognition, which is essential for scene understanding and downstream tasks in computer assisted surgery. However, most existing approaches predominantly rely on per-frame predictions and overlook cross-frame temporal priors as well as temporal-consistency constraints. This limitation often leads to unstable query representations and suboptimal category recognition. In this paper, we propose TEDi, a Temporal memory-Enhanced and Denoising transformer for surgical instrument segmentation that addresses these is sues through Memory Search Enhancement and Temporal Consistency Denoising. The former introduces a query-level memory bank and a memory search enhancement encoder to retrieve discriminative representations from historical frames, enriching current-frame features. The latter constructs a temporally consistent reference as a cross-frame semantic anchor to suppress temporally unstable predictions and promote semantic coherence across frames. Extensive experiments on two benchmark datasets, EndoVis 2017 and EndoVis 2018, demonstrate that TEDi consistently outperforms state-of-the-art methods, highlighting its potential to further advance computer-assisted surgery. Our code is available at github.com/argon-xixi/TEDi.
Read the full abstract
Machine Learning (Statistics)
Time-warping estimation via stationarity-based learning of the de-warped signal
Time-warping estimation is a fundamental problem in signal processing with applications in bioacoustics, radar, and biomedical analysis. This paper introduces a Time-Warping Estimation Trainable (TWET) model for estimating timewarping functions from a single observation. The proposed approach formulates time-warping estimation as a stationarization problem in the wavelet domain and leverages a hierarchical dilated convolutional architecture to estimate the time-warping functions. A differentiable stationarity criterion is introduced for end-to-end optimization. TWET is compared with existing approaches. Experimental results show improved deformation reconstruction accuracy together with significantly reduced computation time, making the framework compatible with low-latency applications.
Read the full abstract
Artificial Intelligence
Layers, Sinks, and Scaling: Adaptive Evidence Selection for Multimodal Large Language Models
Multimodal large language models (MLLMs) can answer knowledge-intensive visual questions by combining visual evidence from images with facts retrieved from external sources. However, MLLMs may overlook relevant evidence in both modalities, attending weakly to the textual sentences or visual regions needed for the correct answer. Recent efforts address this by highlighting retrieved text and marking visual regions before generation, but apply a fixed, one-shot policy that cannot adapt to three sources of variation: whether highlighting is necessary, how much evidence different examples require, and when different textual evidence becomes relevant as the answer unfolds. We introduce Adaptive Relevance-guided Evidence Allocation (AREA), a training-free inference-time method that formulates evidence highlighting as adaptive allocation. AREA generates a single probe token to read visual and textual relevance from fixed backbone layers, then makes three decisions: i) whether to intervene (controlled by natural attention coverage and visual sink contamination), ii) how much evidence to expose (determined by relevance entropy), and iii) when to refresh text during generation (triggered by causal context-attention peaks). Across four KB-VQA and seven standard multimodal benchmarks with nine frozen MLLM checkpoints, establishes the best performance among training-free highlighting methods.
Read the full abstract
Solar and Stellar Astrophysics
The study of AF And. I. Hydrogen abundance constraints and dynamically-consistent wind model of the hot state
This paper starts a series of works devoted to an in-depth study of luminous blue variable AF And in the M31 galaxy. Here we present the results of non-LTE modelling of the star in a hot state with CMFGEN code. In addition to the $\beta$-velocity law, widely used in the analysis of stellar atmospheres, we used a velocity distribution obtained from the balance of radial forces in the wind, including centrifugal forces for several models. Such a dynamically consistent approach reduces the parameter space during modelling and, in addition, allows us to estimate the current stellar mass whose inner hydrostatic layers are obscured by a dense wind that contaminates the absorption profiles. Based on our modelling we estimate the parameters of AF And as follows: luminosity $L \approx 6.5\times10^5~\text{L}_\odot$, mass loss rate $\dot{M}$ $\approx 3\times10^{-5}~\text{M}_\odot~\text{yr}^{-1}$, terminal velocity $v_\infty \approx 300~\text{km s}^{-1}$ and effective temperature $T_{\rm eff} \approx 17~\text{kK}$. modelling of the hot state of AF And is of particular interest due to the possibility to place constraints on the ratio of hydrogen and helium abundances for further use in constructing models of the cold state of this star. We conclude that the current hydrogen mass fraction in the wind lies in the range from $20$ to $30$ per cent. Self-consistent modelling gives the current mass $M_*$ estimation in the range $15.6 - 17.7~\text{M}_\odot$. By employing evolutionary tracks, we were able to estimate the initial stellar mass as $M_{*,~init} \approx 50~\text{M}_\odot$.
Read the full abstract
Human-Computer Interaction
Available but Unclaimed: An Empirical Study of Human-AI Synergy
People increasingly reason with large language models (LLMs), yet complementary capabilities do not guarantee outperforming both components. In a between-subjects study, participants (N=535) solved a 40-item battery of matrix reasoning, mental rotation, syllogisms, and letter-string analogies, unaided or with GPT-5.6-Luna, Claude Opus 4.8, Gemini 3.6 Flash, or Kimi K3. Each assisted trial required consultation with the model. Each model answered every item alone 100 times under matched elicitation. The assisted-unaided accuracy difference increased with item-level LLM competence. Deference varied across tasks and increased with competence within tasks. Post-advice confidence distinguished correct from incorrect answers less strongly than unaided confidence. In a reference comparison, about half the increase in LLM accuracy carried through to assisted accuracy. How much of that accuracy gain reached participants differed across the models. These findings motivate evaluating LLMs in interaction with humans and designing support for selective deference that preserves independent reasoning.
Read the full abstract
Differential Geometry
On the Minkowski deficit for surfaces of revolution
By studying the classical Minkowski's inequality, $4\pi A\leq M^{2}$, relating area and mean curvature, in the particular case of convex surfaces of revolution, one is led to a functional inequality that holds for general functions. We state and prove this inequality obtaining a new proof of Minkowski's inequality in this particular case and we establish a sharp inequality between the Minkowski deficit of the surface and the isoperimetric deficit of any meridian curve.
Read the full abstract
Signal Processing
Covariance-Weighted Spectral Delay Fusion With a One-Dimensional Affine Model for High-Precision Distributed Optical-Fiber Sensing
Periodic disturbances can produce ambiguous delay estimates, limiting reliable high-precision localization in distributed optical-fiber sensing. We develop spectral delay fusion for a sensing system using a dual-wavelength bidirectional Mach-Zehnder interferometer, with four phase traces recovered by heterodyne detection and digital demodulation. With calibrated propagation parameters and timing offsets fixed, the six pairwise delay predictions form a one-dimensional affine line segment parameterized by the position of a single dominant disturbance, with sensitivities determined by propagation direction and chromatic dispersion. A generalized least-squares estimator combines unwrapped delays from robust cross-spectral phase slopes with wrapped delays from polarity-invariant phase alignment to jointly estimate position and integer ambiguities under the proposed model, using an effective joint covariance to account for shared-channel and cross-representation dependence. Experiments use a 131.335-km sensing fiber at 1530 and 1550 nm, with periodic phase perturbations applied at five nominal positions from 25 to 125 km. Across the reported groups of 20 records, the proposed method yields sample standard deviations of 1.007-1.685 m at a drive voltage of 500 mV and 0.449-1.324 m at 1 V. The ratio of the smallest single-pair sample standard deviation to that of the proposed method ranges from 2.57 to 19.56 at 500 mV and from 2.12 to 2900 at 1 V. The upper ratio reflects unstable single-pair phase-slope delay estimates for periodic disturbances in the 1-V, nominal 50-km group, where the proposed covariance-weighted fusion maintains meter-scale localization repeatability.
Read the full abstract
High Energy Physics (Phenomenology)
Glueballs: hadrons without quarks
Glueballs are colour-singlet bound states built from gluons alone. They are an unavoidable consequence of the non-Abelian structure of Quantum Chromodynamics (QCD), and, in the pure Yang–Mills limit, they are the only physical excitations of the theory. The present article reviews what is known about them. After establishing which spin, parity and charge-conjugation quantum numbers two- and three-gluon states can carry, the pure-gauge spectrum is surveyed. Each of the main theoretical approaches and their respective conclusions are briefly presented. They include lattice QCD, constituent-gluon and Coulomb-gauge models, functional Dyson–Schwinger and Bethe–Salpeter equations, holographic models, and QCD sum rules. Particular attention is paid to the scale-setting ambiguity that limits how precisely a quenched glueball mass can be converted into physical units. The discussion then turns to full QCD. After discussing the meaning of a glueball assignment in this context, unquenching effects and questions related to glueball–$q\bar{q}$ mixing are addressed. The large-$N_c$ counting that underpins the mixing picture, as well as mass-matrix and effective-Lagrangian treatments of mixing are presented. The selection rules and dynamical mechanisms that shape glueball decays are also discussed. The gluon-rich production mechanisms used experimentally are finally reviewed. The candidates are assessed sector by sector: $f_0(1370)/f_0(1500)/f_0(1710)$ and the competing interpretations of the scalar sector, $\eta(1405)/\eta(1475)$ and $X(2370)$ in the pseudoscalar sector, the crowded tensor region, and the essentially unexplored $C=-1$ sector and its connection to the Odderon. Outlooks on the programmes that could help in validating some of these candidates or identifying others are reviewed as a conclusion.
Read the full abstract
Materials Science
SpectraFit XPS: a free, browser-based application for X-ray photoelectron spectroscopy peak fitting and quantification
SpectraFit XPS is a free web application for the analysis of X-ray photoelectron spectroscopy (XPS) data that runs entirely in the browser, with no installation and no server-side processing: spectra never leave the user's computer. It combines an interactive Levenberg–Marquardt fitting engine with true Voigt, Doniach–Šunjić, and summed Gaussian–Lorentzian line shapes; Shirley (static and dynamic), linear, and Tougaard (two- and four-parameter) backgrounds; spin–orbit doublet generation; and expression-based inter-peak constraints. Quantitative analysis supports ten relative-sensitivity-factor sets, several kinetic-energy correction schemes including TPP-2M inelastic mean free paths, angular correction, and atomic and weight percentages, with the underlying database and algorithms adapted from the open-source KherveFitting project. The application reads and writes ISO~14976 (VAMAS) files—including multi-region and depth-profile data and load-time transmission-function correction—and imports common vendor and spreadsheet formats. A publication plotter exports journal-ready SVG/PNG figures. The quantification pipeline reproduces KherveFitting reference values exactly in a 30-case regression suite, part of an automated test set of 100 tests. The application, an illustrated bilingual user guide, and a synthetic demonstration dataset are freely available at https://spectrafit-xps.web.app/.
Read the full abstract
Machine Learning
Noise2Noise Revisited: Training Pair Distributions Dominate Loss Choice in Self-Supervised Denoising
Noise2Noise (N2N) trains denoisers on pairs of independently corrupted observations, eliminating clean references. We stress-test two natural conjectures about why the L1 loss outperforms L2 here. First, the hypothesis that the L1 loss confers robustness via parameter sparsity confuses the loss with Lasso regularization: an explicit Lasso penalty produces the predicted sparsity yet fails to reproduce L1's cross-noise behavior, while L1- and L2-trained weight distributions are indistinguishable. Second, the population optima of the two losses coincide exactly for symmetric signal posteriors and nearly so for concentrated ones. Measured differences are therefore dominated by optimization dynamics (bounded-influence gradients), which we probe with gradient statistics and contaminated-target training. On Kodak24 with five synthetic noise families, the L1 loss holds a statistically significant edge over L2, below 1 dB PSNR, holding across three seeds on 13 of the 14 noise columns. On real camera noise the loss is not the decisive variable in distribution: on official SIDD validation blocks, synthetic-Gaussian-trained N2N models gain only 0.8 to 3.7 dB over the noisy input regardless of loss, while retraining on SIDD's own noisy pairs, never reading ground truth, gains 9.4 to 11.0 dB, far ahead of BM3D. All metrics are on raw network outputs, and the study makes no leaderboard claim. The training pair distribution, not the loss, carries the inductive bias. That design rule applies wherever clean references are unobtainable, from microscopy to industrial inspection sensors.
Read the full abstract
Distributed Computing
Many Processors, Still One Computer: The Nested Parallel von Neumann Architecture and Nested BSP
The contest in large-scale AI computing is no longer whether one processor can be made stronger, but whether a million of them can still behave as a single computer. This paper argues that two extensions are required. First, classic BSP, nested recursively, becomes a plan that holds at every scale: each superstep consists of parallel work, then exchange and aggregation, then a barrier, and at every layer the participating units are peers. Second, the von Neumann single-machine architecture, extended past its master–slave habit, becomes the Nested Parallel von Neumann Architecture: a nest of peer-equal layers, from chip package to autonomous zone, joined end to end by one memory-semantic interconnect, the Unified Bus. The software nest and the hardware nest correspond layer by layer, and the $\tau$ Scaling law folds time at every layer. We examine which workloads the nested structure serves, AI training above all but much of classic HPC as well, and describe the hardware decisions that make the nesting physical. Many processors, still one computer.
Read the full abstract
Robotics
Optimal Excitation Trajectories for System Identification of Underwater Vehicles
In this work, we propose a structured methodology for the system identification of underwater vehicles through the design of optimal excitation trajectories. To this end, the trajectories are parameterized using Bezier curves, which ensure smooth and differentiable motion profiles while facilitating the enforcement of constraints through appropriate manipulation of the control points. An optimization problem is formulated to determine a dynamically feasible excitation trajectory that respects safety limits and maximizes the quality of the collected data, thereby enabling reliable estimation of the vehicle's dynamic parameters using least squares. The proposed methodology is experimentally validated in a laboratory water tank, where the dynamic parameters, identified from the optimized trajectory, are evaluated by predicting the vehicle's velocity through forward simulation on previously unseen trajectories.
Read the full abstract
Computer Vision
PSMP-CLIP: Patch-Prompt SAM and Multi-Semantic Prompting for CLIP-Based Zero-Shot Anomaly Detection
Zero-shot anomaly detection aims to localize anomalies without target-domain samples. Existing CLIP-based methods suffer from coarse anomaly maps and limited semantic prompts. We propose PSMP-CLIP, integrating patch-prompt SAM2 segmentation (PPSS) and multi-semantic guided prompt regularization (MSGPR). PPSS samples prompts directly from intermediate patch features, avoiding threshold drift and guiding SAM2 to produce precise masks. MSGPR uses multiple learnable prompts constrained by semantic anchors to preserve generalization. Experiments on 14 datasets show highly competitive performance, achieving the best pixel-level AUROC on MVTec AD, BTAD, DTD-Synthetic, CVC-ClinicDB, TN3K, Endo, and Kvasir.
Read the full abstract
Computers and Society
AI literacy over tool design: a mixed-methods study of scaffolded versus unrestricted generative AI in programming education
Generative AI has become a routine resource in programming education, and most institutional responses to it are attempts at control, either by restricting access or by offering students a controlled version of the technology. This paper reports a seven-week mixed-methods pilot study in a master's-level data analytics course, in which 33 students were randomly assigned either to a scaffolded AI Study Coach embedded in the notebook-based laboratory sessions or to unrestricted use of AI tools of their own choosing. The Coach offered stepwise hints, did not generate code, limited the number of hints per session, and required a short reflection at the end of each session. The design assumed, in line with scaffolding theory and recent experimental evidence, that guided and limited support would build confidence and reduce over-reliance, and that the scaffolded group would learn more. Assignment performance did not differ between the conditions. Students in the Coach condition reported higher confidence but managed the hint budget poorly, while students in the unrestricted condition were satisfied with their tools and uneasy about how much they depended on them. In interviews, students in both conditions identified awareness of their own reliance on AI as the most valuable outcome of the course. Students who had formulated their own rules for when to use AI performed better in both conditions, and those with the best understanding of how the models work, in every case self-taught, used the tools most deliberately and achieved the highest scores. The design of the tool mattered less than the students' capacity to govern their own use of it, a capacity that is at present acquired by chance. The paper argues that the appropriate response is structural: assessment that grades the reasoning behind AI-assisted work, and AI literacy taught explicitly as a core skill.
Read the full abstract
Mesoscale and Nanoscale Physics
Diffusive molecules share the 1/3 shot noise suppression of quantum conductors
Electrical noise measurement, especially shot noise, which arises from the discreteness of charge, is an indispensable tool in quantum transport for probing the nature of charge carriers beyond simple conductance measurements. Diffusive conductors share a universal shot noise reduction of 1/3, a result obtained independently from quantum scattering theory, semi-classical kinetics, or by classical exclusion processes8. Until now, however, experimental tests of this universality have been limited to cryogenic conditions, leaving its origin incompletely understood. Here we show that ferrocene redox cycling in a microfluidic gap is a room temperature realization of this universality. We derive the full counting statistics of diffusing single-electron molecular shuttles and verify the predicted current noise experimentally, showing that the universal 1/3 shot noise suppression is recovered in the diffusion-limited regime. Our result identifies diffusion and sequential charge transfer as sufficient ingredients for this universal noise reduction, rather than quantum coherence, fermionic statistics or cryogenic conditions. We anticipate that this study will establish electrochemical microfluidics as a room-temperature platform for mesoscopic counting statistics and bring noise-based probes to molecular transport and reaction kinetics. Furthermore, this liquid-based quantum-inspired study will provide a novel insight on the yet to be understood links between quantum and biology.
Read the full abstract
High Energy Astrophysical Phenomena
Population synthesis of low-mass binaries with wind-accreting neutron stars
We perform population synthesis modeling of neutron star binary systems containing a low-mass main-sequence companion underfilling its Roche lobe. About $10^5$ such systems are expected to exist in the Milky Way. In the majority of such systems, NSs are expected to start accreting the stellar wind matter within a few Gyr, resulting in the appearance of a dim X-ray source. The exact value of this fraction mainly depends on the efficiency of the propeller spin-down, properties of the stellar wind, and common envelope efficiency. Mostly, NSs in such systems are formed in electron capture supernovae. Thus, they obtain a low natal kick. Due to this, the center-of-mass velocities of such binaries are low, too. Only in the case of high common envelope efficiency ($\alpha_\mathrm{CE}\gtrsim 3$), a few tens of long-lived high-velocity binaries can be produced in core-collapse supernovae.
Read the full abstract
Quantum Physics
Initial-Slip Dynamics Enables the Quantum Mpemba Effect
We investigate the quantum Mpemba effect, whereby a system initially farther from equilibrium relaxes faster than one initially closer to equilibrium, emerging in a quantum harmonic oscillator linearly coupled to a bosonic environment. Starting from factorized thermal states at different temperatures, we derive exact analytical criteria for the occurrence of the quantum Mpemba effect valid also beyond the typically considered weak-coupling and Markovian approximations. Most importantly, we show that the effect is enabled by the initial-slip dynamics generated by the sudden switch-on of the system-bath interaction. Neglecting this contribution completely suppresses the anomalous relaxation, identifying transient system-bath correlations as the key ingredient underlying the quantum Mpemba effect.
Read the full abstract
Materials Science
Topological spin textures in 2D altermagnetic chromium chalcogenides: Interplay between magnetic frustration and Dzyaloshinskii-Moriya interaction
Altermagnetism has recently emerged as a distinct magnetic paradigm, yet the exploration of topological spin textures within two-dimensional (2D) altermagnets remains limited. Using first-principles calculations and atomistic spin simulations, we investigate monolayer $\mathrm{Cr_2X_2}$ and Janus monolayer $\mathrm{Cr_2XY}$ altermagnets ($\mathrm{X, Y}=\mathrm{O, S, Se, Te}$, and $\mathrm{X \neq Y}$). Both systems exhibit strong intrinsic exchange frustration arising from competition between nearest-neighbor and further-neighbor exchange interactions. In Janus $\mathrm{Cr_2XY}$, breaking the out-of-plane mirror symmetry additionally activates a substantial in-plane Dzyaloshinskii-Moriya interaction (DMI), mediated by the strong spin-orbit coupling of heavy chalcogen ligands. The interplay between exchange frustration and DMI gives rise to distinct frustration- and DMI-dominated regimes, while magnetic anisotropy determines the preferred spin orientation, leading to a diverse range of non-collinear states. Systematic phase-space scans and geodesic nudged elastic band (GNEB) calculations reveal two distinct mechanisms governing this interplay: a synergistic effect, in which frustration lowers the critical DMI for the onset of non-collinear states, and a competitive effect, in which frustration favors high-topological-charge states while DMI favors low-charge states. Our results establish a general framework for understanding the interplay between frustration and DMI and guiding the design of topological spin textures in two-dimensional altermagnets.
Read the full abstract
Artificial Intelligence
Integrating the Analytic Hierarchy Process with Large Language Models for Transparent Multi-Criteria Decision-Making
LLMs are increasingly employed in a wide range of decision-making tasks. However, the opacity of their internal reasoning makes it difficult to validate or interpret their outputs, and the need for interpretability becomes especially critical in high-stakes settings. This study examines the decision-making capabilities of LLMs through the Analytic Hierarchy Process (AHP), a classical and widely used multicriteria decision-making framework. We construct a new annotated benchmark based on AHP and propose the first end-to-end approach that enables LLMs to perform the complete AHP workflow. Experiments in real-world decision problems in the legal and higher-education ranking domains show that our method significantly improves alignment with expert judgments.
Read the full abstract
Computer Vision
Unifying Semantic Priors and High-Frequency Traces: Enhancing V-JEPA with Mixture-of-Experts for Robust Synthetic Image Forensics
The unchecked proliferation of manipulated images on social media platforms has increased the spread of misinformation, posing a severe threat to public trust and information integrity. Modern deepfake detectors typically rely on Vision Transformers (ViTs) to capture the low-level inconsistencies that characterize fully synthetic or locally tampered images. However, the global understanding of such foundation models is not enough to discriminate alone between real and fake multimedia content, especially in challenging scenarios where images are compressed or transmitted through social media. In this paper we pioneer the application of Joint-Embedding Predictive Architecture (JEPA) models to deepfake detection, taking advantage of the generalized representation of visual reality that such World Models have exhibited. We hypothesize, and empirically demonstrate, that the intrinsic world understanding of JEPA models can be used as a strong prior for a deepfake detector. To fully exploit JEPA capabilities, we propose MoE-JEPA, a dual-stream architecture for deepfake detection. By enhancing a V-JEPA 2 backbone with a Residual Mixture-of-Experts (MoE) mechanism, along with a noise stream branch, our model dynamically internalizes forensic knowledge. Furthermore, a Gated Attention Multiple Instance Learning (MIL) module is employed to ensure precise spatial semantic understanding. Evaluated on the SID-Set benchmark, comprising 300K AI-generated, tampered and authentic images, MoE-JEPA establishes a new state-of-the-art with an accuracy of 95.54%, successfully outperforming vastly larger models.
Read the full abstract
Computation and Language
Benchmarking Factual Robustness of LLMs via Multi-conversation Persuasion
As Large Language Models (LLMs) increasingly serve as primary knowledge retrieval interfaces, their robustness against persuasion attacks—attempts to inject misinformation or enforce counterfactuals—has become a critical safety concern. Existing red-teaming frameworks typically evaluate models in multi-turn dialogues where the target model retains full conversation history. We identify a critical flaw in this setting termed “Refusal Inertia”: a model's initial refusal often propagates through subsequent turns largely to maintain contextual consistency, thereby masking its true vulnerability to sophisticated, isolated persuasion attempts. To rigorously evaluate the “cold-start” defense capabilities of SOTA models, we introduce the SAST-IR (Stateful Attacker, Stateless Target - Iterative Refinement) framework. By enforcing a memory wipe on the target while retaining the attacker's history, we simulate a worst-case adversarial setting using multi-turn (stateless) iterations. Leveraging CP-Agent (Cognitive Persuasion Agent), an enhanced diagnosis-guided agent, our experiments on the custom CounterFact-Strict dataset ($N=50$) yield alarming results: simple, diverse attack strategies achieved a staggering 96% success rate, exposing severe brittleness in memory-less defense. Furthermore, we reveal a “Complexity Paradox”: while complex, iteratively refined attacks are effective, they often trigger defensive compliance, whereas simple strategies achieve a higher rate of genuine persuasion (84.7%). Our code and dataset are available at GitHub, https://github.com/cza1006/llm-persuasion-defense.
Read the full abstract
Combinatorics
Exact Labelled Layer-Order Classification for Period-2 Mountain–Valley Assignments on 2 x 2k Maps
We classify the labelled final layer orders of ordinary orthogonal 2 x 2k maps whose mountain-valley assignments are periodic with period two in each crease family. For k >= 2, local flat-foldability leaves 16 of the 64 formal period words. Eight pleat words have a unique labelled state. The remaining eight words reduce, under labelled symmetries, to two constant-row families described by row pivots. For the canonical word MMMMVV, an ordered pivot pair (a,b) is realizable exactly when a = 1, a = m, or a = b, where m = 2k - 1. For MVMMMM, realizability is equivalent to BR(a) = BR(b). In both families each compatible ordered pivot pair determines a unique labelled state. The resulting state counts are 0, 1, 6k - 5, and 4k - 3. The case k = 1 is treated separately because two formal period bits have no physical crease carriers.