Machine Learning
Papers
Fresh from arXiv, typeset properly. Highlighted words trade on the market.
All of arXiv
Sunday 8 November
This paper presents a unified framework for license plate detection and character recognition on low-power edge hardware. The proposed network uses a shared CSP-Darknet and PANet feature encoder with an anchor-free localization head and a RoIAlign-based sequence-recognition branch. Character decoding is performed directly on aligned feature maps, avoiding image-space plate cropping and a second recognition backbone. The model is trained on an AI-assisted dataset of approximately 37,000 manually verified images covering license plate layouts from all 50 United States. The dataset includes variations in illumination, viewpoint, weather, blur, occlusion, and plate design. The framework is converted to an INT8 TensorRT engine and evaluated on an NVIDIA Jetson Orin Nano with 8 GB memory. The INT8 model shows an mAP50 of 97.8% and a full-sequence accuracy of 90.7%. It records 147 FPS at 6.8 ms end-to-end latency, with 315 MB inference memory and 9.4 W average power consumption. Relative to the YOLOv8s+CRNN baseline, the proposed deployment provides a 7.1x throughput increase and a 77.8% reduction in allocated inference memory
Probability
Fingerprints of feedback cycles in stationary diffusions
When does the stationary density of a $d$-dimensional diffusion process driven by standard Brownian motion uniquely determine its drift? In general, it does not: the $d$ drift components enter the stationary Fokker-Planck equation only through a single scalar divergence. We consider drifts whose component functions only depend on a subset of their input variables, and encode this dependence with a directed graph. We exclude 2-cycles, which admit rotations that leave the stationary density invariant, but otherwise allow arbitrary graphs, including those with feedback cycles. We show that if the stationary density is sufficiently $s$-concave, e.g., strictly log-concave, then the drift is uniquely determined among graph-structured drifts in a weighted Sobolev space. For drifts with linear self-regulation, a single locally strictly convex level set of the stationary density is sufficient to identify the drift in the region it encloses. Both results follow from Raychaudhuri equations. Excluding 2-cycles allows treating nontrivial rotational and symmetric terms simultaneously, extending the scope of the corresponding Bochner- and Reilly-type arguments. We illustrate the results by recovering the Hill-type regulation of a stochastic genetic repressilator from samples of its stationary distribution.
Read the full abstract
Number Theory
Statistics of centralizers of integer matrices
Fluid Dynamics
Cable construction effects on the lateral forces and wakes of helically stranded cables under oblique inflow
Helically stranded cables in aerial towing systems experience oblique inflow, but are often modeled as smooth cylinders, neglecting three-dimensional cross-stream coupling and surface texture. We perform large-eddy simulations and spectral proper orthogonal decomposition for a smooth cylinder and four stranded cables at four inclinations from $22.5^\circ$ to $90^\circ$, all stationary, with $Re \approx 1.0 \times 10^4$. Aerodynamic differences are governed mainly by cross-sectional symmetry and strand-scale groove depth. Force-source decomposition shows that surface pressure contributes 94.9%–98.3% of the mean normal force and dominates the mean lateral force. High-symmetry, quasi-circular envelopes maintain balanced circumferential pressure, whereas deep-grooved, low-symmetry tri-lobed geometries produce asymmetric separation and a persistent lateral pressure imbalance. In the wake, high-symmetry cables sustain spanwise-synchronized Karman shedding at moderate-to-high inclinations, concentrating fluctuation energy in a dominant low-rank mode. Deep grooves disrupt shedding coherence, deflect the recirculation zone laterally, and spread energy across broadband scales. Mean lateral force and lateral-force fluctuations are therefore decoupled: the quasi-circular $(1 \times 19)$ cable has minimal mean lateral force but strong periodic fluctuations, while the deep-grooved $(3 \times 7)$ cable has weak fluctuations but appreciable mean lateral force. These results support construction-dependent cable selection and aerodynamic load modeling.
Read the full abstract
Probability
Tail behaviour of infimum of Gaussian processes
We investigate the upper tail of $\inf_{t\in\mathcal T}X(t)$, where $X$ is a centred Gaussian process (GP) with continuous sample paths on a nonempty compact metric space $\mathcal T$. As suggested by the findings of [1], the exponential rate is governed by an optimal-measure problem, while the exact prefactor depends on the local geometry of the corresponding contact set. In the full-contact regime, we strengthen the known Gumbel max-domain of attraction property of the distribution of $\inf_{t\in\mathcal T}X(t)$ by establishing the von Mises condition. For fractional Brownian motion on $[a,b]$, $0<a<b$, with Hurst index $H\in(0,1/2)$, we also determine the order of the logarithmic correction to the leading Gaussian exponential. For finite contact sets, we distinguish dominated, smooth-critical, and rough-critical regimes and derive exact asymptotics of $P\{\inf_{t\in\mathcal T}X(t)>u\}$ as $u\to\infty$ under suitable local assumptions. We illustrate our findings with examples including fractional Brownian motion and the fractional Brownian sheet.
Read the full abstract
Computation and Language
WarBias: A Bilingual LLM Benchmark of Stereotypes About Internally Displaced People and Veterans
This paper contains examples of stereotypes about veterans and displaced people that may be offensive or distressing. Social bias benchmarks for language models commonly examine categories such as gender, race, and religion, often drawn from lists of legally protected characteristics. Veterans and internally displaced people (IDPs) fall outside these lists, and stereotypes about them, which can undermine access to employment, housing, and social participation, remain largely absent from language model evaluation. We introduce WarBias, a Ukrainian-English benchmark of stereotypes about these two groups. Built from a taxonomy of 120 stereotype claims grounded in research and monitoring reports, WarBias contains 1,007 stereotype-counter-stereotype-unrelated sentence triplets and 240 multiple-choice questions per language. A Ukrainian language model selected in a blinded human pilot generated the Ukrainian triplets. At least two independent annotators reviewed each triplet in the annotation cohort. The authors revised items in response to the reviews, and the final release was machine-translated into English. Each claim carries a human severity rating. In a baseline evaluation of six models, four prefer stereotypical Ukrainian sentences about at least one group, most strongly about IDPs, while English scores are mostly balanced. WarBias supports complementary evaluations of stereotypical preferences and context-sensitive judgments, and provides paired forget/retain material for unlearning-based bias mitigation. We document the benchmark's construction, intended uses, and limitations, and offer its construction method as a template for analogous benchmarks in other conflict-affected societies.
Read the full abstract
Signal Processing
Energy-Harvesting-Aware Cell-Free LEO Satellite Networks: A Lyapunov Framework for Sustainable User-Centric Beamforming and Clustering
We study battery-aware downlink control in user-centric cell-free low Earth orbit satellite networks with intermittent solar harvesting and finite onboard batteries. A perturbed Lyapunov controller coordinates clustering and hybrid beamforming using measured battery states and causal harvest forecasts. The per-slot solver combines group-sparse association, rate bounds, a quadratic transform, and successive convex approximation. We establish robust battery invariance for feasible implemented actions and give a utility bound conditional on a comparison policy and the per-slot implementation gap. Multi-orbit control-layer simulations show that queue-aware scheduling preserves a larger battery reserve in the stressed scenario at the expense of effective energy efficiency. A separate fixed-cluster matrix snapshot illustrates recovered hybrid beam patterns and the per-slot objective trace.
Read the full abstract
Sound
Real-Time Arbitrary Resampling for Phase-Continuous Microtuning in Audio Streams
This paper develops an output-driven resampling formulation for fine pitch adjustment of recorded signals. A fractional input-position increment combines the desired frequency ratio with independently specified input and output sample rates. Each output sample is evaluated from neighboring input samples, and gradual rate changes preserve the current input position. Musical interval mapping and direct sine synthesis are treated as distinct layers. The formulation separates tuning resolution, numerical accuracy, interpolation quality, and streaming correctness. Optional rational-factor representations use freely chosen approximation parameters; numerical illustrations do not define constants required by the algorithm. Linear interpolation provides a low-cost evaluation kernel but does not establish a universal signal-quality or frequency-accuracy guarantee.
Read the full abstract
Optimization and Control
Optimal Batching with Abandonment: Threshold Policies Under Risk-Neutral and Risk-Sensitive Criteria
We study optimal batching in a discrete-time Markovian system with Bernoulli arrivals and independent item abandonment. Each batch earns a per-item processing reward and incurs holding, abandonment, and fixed stopping costs. We formulate the batching decision as an optimal stopping problem and characterize optimal threshold policies under risk-neutral, risk-seeking, and risk-averse criteria. Under risk neutrality and the entropic risk-seeking criterion, the value function admits a forward-recursion representation, and the resulting reduced objectives are strictly unimodal. In the risk-seeking case, the objective is strictly discretely concave, yielding an explicit marginal-cost condition. For risk aversion, the unregularized entropic continuation branch may diverge, precluding standard fixed-point arguments. We therefore study a regularized robust surrogate formulated as a minimax problem with a relative-entropy penalty and capped likelihood ratio. We establish the existence of an optimal threshold policy and derive an explicit truncated-Gibbs worst-case distribution. We also study repeated batch clearing under both the long-run average and discounted criteria, for which the threshold objectives remain strictly unimodal. Numerical experiments examine threshold sensitivity to system parameters, costs, and risk preferences.
Read the full abstract
Combinatorics
Impropriety and resilience of the chromatic number
We show that, for $p$ bounded away from $1$ and $np\to\infty$, the chromatic number of the binomial random graph $G_{n,p}$ is resilient to removal of a subgraph $H$ with $o(n \log(np))$ edges. It follows that the same assertion holds with $H$ being a subgraph of maximum degree $o(\log(np))$; and both these results are best possible up to a constant factor. We also prove corresponding results for the random $r$-regular graph $G_{n,r}$, when $r\to\infty$ and $r/n$ is bounded away from $1$, where $H$ has $o(n\log r)$ edges or maximum degree $o(\log r)$. Our results rely on a natural connection between chromatic resilience to edge removal and improper graph colouring. This work provides counterpoint to the work of Sudakov and Vu (2008) and Alon and Sudakov (2010), which treated chromatic resilience for $G_{n,p}$ subject to the addition, rather than removal, of edges.
Read the full abstract
Machine Learning
How Efficient Is Your Agent? Online Predictive Bounds under Budget Constraints
We evaluate AI agents in an online setting where tasks arrive sequentially and, before running each task, we construct an upper predictive bound on its time-to-success—the number of interactions until the task is completed successfully. Under a limited budget, some runs are stopped early, so their time-to-success is unobserved (right-censored), and the correctness of the bound may be unknown. We introduce \emph{PRedictive Bounds for Online Budget Efficiency} (PROBE), a method for constructing calibrated upper predictive bounds while adaptively allocating a limited budget across agent tasks to reduce the anticipated coverage deficit. Our method calibrates the bounds online using inverse-probability-weighted feedback together with a correction designed to minimize coverage deficit while remaining unbiased. We prove that PROBE provides a coverage guarantee for any finite time horizon while satisfying the budget constraint. Experiments on software engineering, retrieval-augmented generation, and instruction following show that PROBE attains a coverage rate closer to the target level with tighter predictive bounds than a calibrated baseline.
Read the full abstract
Information Retrieval
An Adaptive, Context-Aware, and Stacked Attention Network-Based Recommendation System to Capture Users' Temporal Preference
Recommendation systems have become more important since the widespread use of the Internet. The large amount of information means that it is difficult for users to discover they really need. However, users' preferences can change over time because of the age and the impact of social networks. Recommendation systems must capture users' preferences and recommend suitable items, which is a difficult challenge. This study proposes a novel recommendation system that adapts to the changing preferences of users. By learning and adapting to users' changing preferences better recommendations are provided. This study uses context factors as additional information to model users' preferences more precisely. For the proposed recommendation system, the context is based on users' interaction items within each defined time interval. This study proposes two novel attention mechanisms for the recommendation system. The contextual item attention module captures contextual information, changing pattern in users' preference and the importance of items. The multi-head attention module extends the diversity of users' preferences and adapts to changing preferences. The recommendation performance is improved using additional item's temporal information to model the contextual item's representation. Experiments compare the proposed algorithm with several state-of-the-art recommendation methods using three real-world datasets. The experimental results demonstrate that the proposed context-aware recommendation model outperforms traditional methods and demonstrate the effectiveness with which contextual information is captured by the attention mechanism.
Read the full abstract
Optimization and Control
Variance-Reduced Stochastic Hessian-Driven Nesterov Acceleration with Adam-Style Preconditioning
Stochastic first-order methods use sampled gradients to reduce per-iteration costs, but estimation errors can slow convergence and, in some settings, cause divergence. To improve gradient accuracy while retaining acceleration, we develop scalar and diagonal versions of VR-Adam-SHANG for smooth stochastic convex optimization. The methods combine STORM-type recursive estimation, Adam-style preconditioning, and SHANG's accelerated interpolation and correction, using at most two stochastic gradient evaluations per iteration without full-gradient refreshes. Under unbiased sampling, bounded variance, and mean-square smoothness of gradient noise, polynomial steps with prescribed initialization yield $O(1/K)$ mean-square gradient-estimation error and expected objective-gap bounds combining $O(1/K^2)$ initial-energy decay with an $O(1/\sqrt K)$ stochastic term after $K$ iterations. These guarantees require neither monotonicity of the preconditioner nor a uniform positive lower bound on its eigenvalues. The analysis also establishes polynomial-step convergence of projected nonrecursive Adam-SHANG-s, filling a gap in its original theory. Experiments compare the methods at matched sample-gradient budgets. In every tested setting, at least one of the two versions achieves the lowest mean terminal objective gap among all compared methods.
Read the full abstract
Materials Science
Strain Controlled Hydrogen Interstitial Site Preference and Hydride Phase Suppression in Epitaxial Vanadium
Understanding hydrogen behavior in transition metals is crucial for energy storage and sensing technologies. Here we combine density functional theory (DFT) calculations with in-situ electrical measurements to reveal how epitaxial strain governs hydrogen interstitial site selection and phase stability in vanadium. Our calculations demonstrate a robust thermodynamic preference for tetrahedral (T) over octahedral (O) sites, with the energy difference modulated by lattice anisotropy through strain-tunable H 1s-V 3d orbital hybridization. Experimentally, epitaxial V films on MgO(001) exhibit a critical thickness-dependent crossover: 50 nm films show characteristic first-order alpha-beta hydride transitions, while 10 nm films display continuous solid-solution behavior with complete transition suppression. This crossover is governed by the progressive relaxation of epitaxial strain with increasing film thickness: the 10 nm films retain strong coherent strain that flattens the free-energy landscape and suppresses hydride nucleation, whereas the 50 nm films approach the strain-relaxed bulk limit where long-range elastic interactions restore first-order phase behavior. Finite-size truncation of elastic fields and interfacial electronic reconstruction further reinforce this dimensional crossover. Our work establishes a unified framework connecting atomic-scale orbital physics to mesoscopic phase thermodynamics, providing design principles for strain-engineered metal-hydride systems.
Read the full abstract
Signal Processing
WiFo-R: Wireless Foundation Model with Recurrent Channel Representation
Wireless foundation models learn transferable channel state information (CSI) representations across heterogeneous wireless systems. However, conventional grid-aligned representations couple feature sequence length to CSI dimensions, restricting the allocation of representation capacity to different propagation conditions and task requirements while increasing inference overhead. This paper proposes WiFo-R, a wireless foundation model that learns progressive channel representations with selectable capacity. A channel-query cross-attention encoder aggregates spatial, temporal, and spectral CSI information, while parameter-shared recurrent updates refine existing channel features and introduce additional latent tokens. Multi-mask, multi-depth self-supervised pretraining equips intermediate states with strong reconstruction capabilities, followed by task-aligned post-training for channel estimation and CSI feedback. WiFo-R supports temporal and frequency-domain channel prediction, channel estimation, and variable-rate CSI feedback, and transfers to localization and LoS/NLoS classification. Experiments across in-distribution and out-of-distribution channel configurations demonstrate accurate reconstruction, progressive performance improvements, and strong transferability. A single feedback encoder–decoder supports six payload lengths at fixed quantization precision, while cross-attention and selectable recurrent depth improve the accuracy–complexity trade-off for large-scale CSI observations.
Read the full abstract
Statistics Applications
Zero Replacement and Latent Profiles in Short Media Diaries: Checking Recording Status Before User Segmentation
Short media diaries contain many zeros because many activities go unrecorded during the observation period. Log-ratio analysis of time allocation requires these zeros to be replaced with small positive values; as a result, all respondents who did not record an activity receive nearly the same value. We examined how such zero replacement affects classification in latent profile analysis and propose checks to perform before profiles are interpreted. Three-day diaries of 8,089 adults from the 2025 Korea Media Panel Survey were converted into nine-part time-use compositions. Gaming was not recorded by 88.9% of the respondents, and in every model with two to eight profiles and a covariance matrix held equal across profiles, respondents who recorded gaming and those who did not were classified into different profiles. Even when the replacement value equaled the minimum recording unit (5 min/day), the two-profile classification largely matched gaming recording status, and the two groups separated more sharply as the replacement value decreased, until the separation was complete. Before naming profiles, researchers should therefore check the proportion of respondents who did not record each activity and its replacement value, the same proportion within each profile, and the distribution of usage time among those who recorded the activity.
Read the full abstract
Audio and Speech Processing
Task-adaptive anonymisation enables privacy-preserving sharing of clinical speech data
Clinical speech is increasingly used for disease screening and clinical assessment, but directly distributing speech recordings poses privacy risks because they contain both speaker-identifying voiceprints and sensitive spoken content. Existing speech anonymisation approaches are primarily developed for general speech applications or task-specific clinical speech and overlook that different clinical tasks rely on distinct speech characteristics, necessitating task-adaptive anonymisation strategies to protect patient privacy while preserving clinical utility. Here we introduce CASA, a clinically adaptive speech anonymisation agent that transforms anonymisation from a fixed processing procedure into a task-guided optimisation process, dynamically tailoring voiceprint and content anonymisation to task-specific privacy and clinical utility requirements. Through recursive self-improvement (RSI), CASA iteratively evaluates and refines anonymisation strategies against task-specific privacy and clinical utility requirements, while learning from these outcomes to guide more efficient planning in subsequent clinical tasks. CASA is validated across 14 clinical tasks encompassing five languages, diverse elicitation paradigms, and mental health, cognitive and voice disorders. CASA preserved disease-related classification performance while meeting speaker-anonymisation requirements across all clinical tasks, versus only 8/14 for the strongest fixed method. Assessments of resistance to speaker re-identification by external attackers and identification and redaction of sensitive personal information in clinical speech further demonstrated CASA's dual protection of speaker identity and sensitive spoken content. Clinical expert assessments further supported the preservation of clinically relevant information and the acceptability of anonymised speech for controlled research sharing.
Read the full abstract
Computer Vision
BiRefNet-V2: Rethinking Bilateral Reference for High-Resolution Dichotomous Image Segmentation
Dichotomous image segmentation requires pixel-accurate foreground-background separation while preserving fine structures such as thin filaments, sharp edges, and intricate textures. BiRefNet addressed this challenge through bilateral references, combining image patches and gradient information, but its inward reference loses local spatial structure and lacks semantic guidance, while its outward reference relies on single-channel gradient attention. We recast this paradigm as controlled information recovery - locality determines what spatial information survives, semantic selectivity what information enters, and focused boundary supervision where refinement occurs - and propose BiRefNet-V2 to implement it: (1) locality-preserving and semantically guided detail injection through space-to-depth transformation, overlapping patches, and a Semantically Guided Gate; (2) a learned boundary stream with focused supervision that generates full-channel residual attention without requiring boundary annotations; and (3) a dual-branch decoder integrating depth-wise boundary convolutions with large-kernel attention. A modernized framework - LayerNorm, GELU, the Lion optimizer, and a DINOv3 backbone - cuts training from 600 to 50 epochs. Extensive experiments demonstrate state-of-the-art performance on DIS5K, camouflaged object detection, and high-resolution salient object detection across all 13 evaluation sets. Code is available at https://github.com/ZhengPeng7/BiRefNet-V2.
Read the full abstract
Analysis of PDEs
Three-Dimensional Stability of Weakly Asymmetric Burgers Vortices: A Circulation-Independent Asymmetry Threshold and Optimal Decay
Burgers vortices are stationary solutions of the Navier–Stokes equations in a background straining flow and serve as a model for vortex tubes in three-dimensional turbulence. In this paper, we consider weakly asymmetric strains and establish the local nonlinear asymptotic stability of the corresponding Burgers vortices against small three-dimensional perturbations for arbitrary circulation. Our proof exploits the specific axial structure of the linearized operator, combining a two-dimensional resolvent analysis with decay estimates for axial derivatives to control the full three-dimensional evolution. A first-moment decomposition isolates the slow modes, and Gaussian resolvent bounds are transferred to polynomially weighted spaces. This yields nonlinear stability with the endpoint exponential decay rate $(1-\lambda)/2$, where $\lambda$ denotes the strain asymmetry, without any additional polynomial factor in time. An exact nonlinear first-moment law demonstrates that this rate is optimal over the full admissible perturbation class, including at zero circulation.
Read the full abstract
Dynamical Systems
Onsager–Machlup Theory and Transition Pathways for Degenerate Stochastic Systems with Multiplicative Noise
Noise-induced transitions between metastable states are central to many nonlinear stochastic dynamical systems, and most probable transition pathways provide an effective means of revealing the underlying transition mechanisms. In this work, we develop an Onsager–Machlup framework for degenerate stochastic dynamical systems driven by multiplicative Gaussian noise. We first derive the Onsager–Machlup action functional for the additive-noise case using the Girsanov transformation, and then extend the result to multiplicative noise through a Lamperti-type transformation. The resulting functional provides a variational characterization of the most probable transition pathway, from which we derive the associated Hamilton–Pontryagin boundary value problem. We further recast the variational problem as an equivalent optimal control problem and propose a neural-network-based numerical method for computing the most probable transition pathways. Numerical examples are presented to illustrate the theoretical framework and demonstrate the effectiveness of the proposed computational approach.
Read the full abstract
Algebraic Geometry
Finite torsion in Hodge–Witt cohomology groups
For bounded coherent complexes over the Raynaud ring, we establish natural Dieudonné isomorphisms between Ekedahl's biduality pieces $A_f^2$ and $A_{\SSpart}^1$, the nilpotent and semi-simple torsion of the Illusie–Raynaud cœur, and the corresponding parts of the finite torsion $t^1$ of Ekedahl's diagonal cohomology. In bidegree $(0,i+1)$, this torsion measures the non-smoothness of the $i$th flat Artin–Mazur formal group; hence all such groups are smooth for Mazur–Ogus varieties. For a smooth projective surface $S$, we describe the three possible dual pairs via the Galois module $\NS(\bar{S})[p^\infty]$ and the local-local and local-étale parts of the finite Picard quotient.
Read the full abstract
Quantum Gases
Parametrizations of the solutions of Love equations and their momenta
The Love equation is a paradigmatic example of a Fredholm integral equation of great interest in mathematics and physics, where its applications range from electrostatic problems to exactly solvable models for one-dimensional systems. In this work we propose a method to approximate the normalized solutions of Love equations of the first and second kind via a parametrization obtained by Hermite interpolation at Chebyshev nodes. Following the same approach, we also obtain approximations for the second, fourth and sixth momenta of the solution of the Love equations. We focus in particular on one-dimensional quantum gases described by the Lieb-Liniger and the Gaudin-Yang models. For these models, the solutions of Love equations correspond to the normalized densities of pseudorapidities. We show that remarkably accurate results can be obtained using the Hermite-Chebyshev interpolation with a number of points as low as 17+2. In the case of the Lieb-Liniger model, we also propose a very accurate closed-form estimate of the dimensionless ground state energy valid for all values of the dimensionless interaction parameter incorporating known results for small and large couplings supplemented by the Hermite-Chebyshev interpolation for intermediate couplings. As an application of this result, we compute the dimensionless local two-particle correlation function and the black soliton solutions of the Lieb-Liniger model.
Read the full abstract
Mesoscale and Nanoscale Physics
Theory of spin-polarized scanning tunneling microscopy with topological insulator tips
Topological insulators have enabled a new class of spin-sensitive scanning tunneling microscopy (STM) probes based on helical boundary states. However, how these probes generate a net spin-polarized tunneling current remains unclear. We present a theory of spin-polarized STM with topological insulator tips and show that asymmetric tunneling-contact geometry and spin dephasing enable helical boundary states to generate a spin-polarized tunneling current. Within the standard Landauer-Büttiker formalism, we demonstrate this mechanism for a quantum spin Hall (QSH) insulator tip: asymmetric contacts make the two counter-propagating helical edge states, which carry opposite spins, traverse path lengths that differ by $\Delta L$, and spin dephasing induces backscattering that reduces the conductance more strongly for the longer path. For a representative asymmetric contact and weak spin dephasing, the QSH tip exhibits a tunneling spin polarization of $P_s=-17.2\%$ under positive bias, whose magnitude can reach $50\%$ for larger $\Delta L$, and which remains robust against Rashba spin-orbit coupling. Crucially, this polarization, dominated by helical edge transport rather than by bulk states, naturally reverses sign with the bias polarity owing to spin-momentum locking. Our findings are consistent with recent STM experiments using topological tips.
Read the full abstract
Cryptography and Security
ErosionBench: Probing Alignment Decay in Enterprise Multi-Agent Systems
As enterprises delegate consequential work to multi-agent LLM systems, benign-looking requests can escalate into unsafe actions as they propagate across sessions through poisoned shared state, long-context documents, and trusted peer messages: no single message is malicious, yet the assembled tool action is. Prior benchmarks isolate a single agent under prompt injection or score multi-agent attacks by their final output, leaving cross-session erosion of shared state out of scope. We introduce ErosionBench, an environment-grounded benchmark that executes such attacks and verifies alignment decay at the process level, scoring the realized tool call against authorized intent and provenance with deterministic contracts, not output text. Each scenario is authored as a deterministic state graph before any prose is rendered, so every state change traces to a known cause. It spans four enterprise domains and four decay mechanisms across 1,728 scenarios with matched benign controls, labeled against the OWASP Top-10 for LLM Applications, plus 198 image-based attacks, each paired with a text equivalent, so a defense must survive the same attack in either channel. Perimeter defenses suppress attacks only by over-refusing 54-92% of legitimate work, so a low attack-success rate can conceal severe over-refusal. As a reference defense, we propose a calibrated Meta-Auditor pairing synchronous state checks with asynchronous trajectory monitoring; on a family-disjoint 906-scenario held-out split it reaches 98.5% recall at a 2.9% false-positive rate. Since it monitors trajectory, not surface content, its detection is channel-invariant: the strongest auditor flags image-delivered attacks about as reliably as their text equivalents (recall gap $\leq$2 points per model). Across GPT, Gemini, and Claude auditors, calibration keeps false positives low while recall tracks model capability, strongest for Claude.
Read the full abstract
Computer Vision
DuSS-PET: Self-Supervised Low-Dose PET Reconstruction Using Pseudo-Dual-Domain Consistency Without Paired Data
Positron emission tomography (PET) is pivotal in medicine and healthcare, but requires high radiation doses for quality imaging. Although deep learning has shown promise for low-dose PET reconstruction/enhancement, existing methods primarily rely on fully supervised learning, which requires paired low- and full-dose scans that are clinically impractical to obtain. To overcome this fundamental limitation, we propose DuSS-PET, a novel self-supervised PET reconstruction framework that operates without any paired low- and full-dose data. Our method is built upon a computational Noisier2Noise (Nr2N) paradigm, featuring a Bernoulli corruption strategy that serves as an image-space approximation to emulate low-count noise patterns in image-derived pseudo-sinogram representations. The DuSS-PET framework integrates three core components: an adaptive Swin transformer for pseudo-sinogram restoration, a plug-and-play module for image enhancement, and a consistency-constrained SSL strategy that enforces agreement between the pseudo-sinogram and image representations. Extensive experiments on multi-center datasets demonstrate that DuSS-PET achieves superior reconstruction performance, outperforming state-of-the-art SSL methods and achieving comparable or even better results than fully supervised approaches across various dose levels. In particular, it exhibits exceptional generalization across different scanner vendors. This work is among the first self-supervised low-dose PET reconstruction framework that enforces consistency between reconstructed images and pseudo-sinograms, where the pseudo-sinogram generated in the image space serves as a solid computational consistency constraint for enhanced imaging without full-dose training scans. The source code will be publicly available at https://github.com/ladderlab-xjtu/DuSS-PET.
Read the full abstract
Analysis of PDEs
The geometry of anisotropic propagation for multidimensional reaction-diffusion equations in shifting habitats: A zero-sum stopping game approach
We study the propagation geometry of multidimensional reaction-diffusion equations in shifting habitats. Using geometric optics techniques and viscosity solution theory, we prove that the associated rate functions converge locally uniformly to the unique viscosity solution of a variational inequality with singular initial data. We identify this limit as the value of a zero-sum stopping game whose zero set determines the Wulff shape. This zero set coincides with that of the positive part of the value function of an unconstrained optimal control problem if and only if Freidlin's condition holds. When Freidlin's condition fails, we derive a pointwise characterization of the stopping-game zero set, yielding a convex hull representation. These results then yield explicit Wulff shapes and directional spreading speeds for all habitat speed. As the habitat speed increases, the Wulff shape evolves from a large ball to a truncated ball. It then develops conical, cylindrical, and paraboloidal boundary components attached to spherical parts, before eventually becoming a smaller ball. Some of these geometric transitions occur while both axial spreading speeds remain unchanged. Moreover, by characterizing the optimal trajectories, we distinguish two mechanisms of nonlocal speed selection and show that such selection can occur in oblique directions even when it is absent in both axial directions.
Read the full abstract
Machine Learning
Beyond Node Attributes: A Controlled Evaluation and Ablation of Local Interaction Statistics for Ethereum Phishing Detection
Methods for detecting phishing accounts on Ethereum are usually evaluated as complete pipelines that differ in features, architecture, and training protocol, which makes the contribution of interaction information hard to isolate. This paper evaluates the Unlabeled Feature Extension (UFE), a block of 12 local interaction statistics computed from each account's incoming and outgoing call and transfer edges without reading class labels. UFE is appended to the 14 source features of HEIG-Phish and tested with six model families: Logistic Regression, Random Forest, LightGBM, Graph Attention Network, GraphSAGE, and Meta-IFD. Within each pair of Baseline and Plus variants, the data partition, model configuration, random seed, decision threshold, and sampled neighborhoods are identical; only the input feature block changes. Across 20 predetermined stratified partitions of the same 2,763 labeled externally owned accounts, UFE increased test Macro-F1 by 1.0 to 9.1 percentage points across the six models. Every pair improved on all 20 partitions, all Nadeau-Bengio corrected 95% confidence intervals lie above zero, and the false-positive rate decreased for all six models. An exact-edge Random Forest ablation over 10 predetermined partitions associates most of the gain with the incoming transfer-edge sum attribute; once this feature is present, an incremental contribution of the remaining 11 UFE columns was not established. The results support reporting feature-augmented baselines alongside graph models, but they show stability under re-partitioning of one static, transductive benchmark rather than temporal or cross-dataset generalization.
Read the full abstract
Information Theory
Improving the Trifference Lower Bound via Non-Perfect Hashing
We strictly improve the Körner-Marton lower bound $\frac14\log \frac{9}{5}$ on the asymptotic rate of trifferent codes. We obtain this improvement by constructing non-perfect $(9,3)$-hash codes with rate above $\log(9/5)$ and then applying the Tetra-code construction, which gives $R_3\ge\frac14R_{9,3}$. More generally, for every $q\ge4$, we prove $R_{q,3}>B_{q,3}+\frac1{250000}$. We also extend the method to $(p,2,3)$-list recovery. For every $q\ge4$ and $0\le p<\frac{q-2}{12(q-1)}$, we prove $R_q^{\mathrm{LR}}(p)>B_q^{\mathrm{LR}}(p)$. Here $B_{q,3}$ and $B_q^{\mathrm{LR}}(p)$ denote the respective probabilistic lower bounds. For every fixed $0\le p<1/18$, the rate improvement is bounded below by a positive constant independent of $q\ge4$. For every fixed $0\le p<1/12$, we also obtain a positive rate improvement independent of $q$ for all sufficiently large $q$.
Read the full abstract
Machine Learning
Chatter Attack: Resource Consumption Attack for Large Language Models
LLMs are powerful but incur substantial inference cost, which can be exploited by malicious users to induce overly long outputs, increasing latency and operational expense. Prior resource consumption attacks either coerce repetition or suppress the end-of-sequence token by lowering its logit value. However, with autoregressive, sampling-based generation, the search space grows exponentially with length and small deviations or unstable trajectories can derail optimization, limiting attack effectiveness and stability. To address this issue, we introduce Chatter Attack, a novel class of resource consumption attacks that optimize adversarial prompts to induce predefined short target phrases, thereby avoiding search-space explosion while still eliciting long responses. Specifically, Chatter Attack triggers self-doubt using short doubt-inducing phrases as the target, and sustains this state throughout generation via inference attention loss and global entropy loss. To extend to black-box settings, we further propose a novel Bayesian Prompt Optimization (BPO) method, which models the objective function globally by constructing a discrete token kernel, efficiently exploring the entire search space using prior information. Extensive experiments across multiple models and datasets show that our method outperforms baselines, with maximum response length increases to 31.5$\times$, and revealing practical resource-depletion risks for real-world LLM inference services.
Read the full abstract
Machine Learning
Primitives and Paths: Predicting What Post-Training Can Do from What the Base Model Already Has
Does reinforcement learning teach language models new capabilities, or merely amplify ones acquired during pre-training? The literature offers evidence for both: pass@$k$ studies suggest RL reweights latent behaviours while some controlled composition studies demonstrate gains beyond simple amplification. We argue that this question has no answer independent of the capability state of the base model, and propose a two-dimensional state space for post-trainability. The first axis, effective full-solution support, measures whether successful trajectories can be sampled under a feasible exploration budget. The second axis, atomic primitive coverage, measures whether the constituent skills of a solution can be executed when the burden of composition is removed. Together, these dimensions motivate three regimes – amplification, composition-limited, and acquisition-limited – each predicting a distinct empirical signature and calling for a distinct intervention. We propose a prospective measurement protocol and investigate the framework using synthetic composition tasks across three small language models. We find that primitive mastery alone does not guarantee that RL can learn a composition. Checkpoints that execute every primitive reliably can fail to improve when complete solutions are never sampled during training, whereas modest composition supervision that establishes nonzero support enables substantial RL gains. The framework reframes mid-training as preparation for post-training: its role is to move models into states from which RL can operate effectively.