May 2025 arXiv papers — page 100

Shaonwita Pal, Dibyendu Nandy

Over the past century, the Sun's activity -- which exhibits significant variations -- went through a phase known as the Modern Maximum. Notably, the strongest sunspot cycle on record during this period, and indeed since direct sunspot observations began, was cycle 19; this was followed by a significantly weaker cycle 20. Understanding and reconstructing this

Mengqi Lei, Yihong Wu, Siqi Li, Xinhu Zheng

Visual recognition relies on understanding the semantics of image tokens and their complex interactions. Mainstream self-attention methods, while effective at modeling global pair-wise relations, fail to capture high-order associations inherent in real-world scenes and often suffer from redundant computation. Hypergraphs extend conventional graphs by modelin

Felix Hommelsheim

The Forest Augmentation Problem (FAP) asks for a minimum set of additional edges (links) that make a given forest 2-edge-connected while spanning all vertices. A key special case is the Path Augmentation Problem (PAP), where the input forest consists of vertex-disjoint paths. Grandoni, Jabal Ameli, and Traub [STOC'22] recently broke the long-standing 2-appro

Lingyong Zeng, Jie Wang, Hongyu Liu, Longfu Li

Superconductivity has been one of the focal points in medium and high-entropy alloys (MEAs-HEAs) since the first discovery of the HEA superconductor in 2014. Until now, most HEAs' superconducting transition temperature (Tc) has not exceeded 10 K. Here we report the first observation of superconductivity in a bulk medium-entropy nitride (MEN), TiNbTaN3, which

Silvia Cappelletti, Tobia Poppi, Samuele Poppi, Zheng-Xin Yong

Large Language Models (LLMs) are increasingly evaluated on multiple-choice question answering (MCQA) tasks using *first-token probability* (FTP), which selects the answer option whose initial token has the highest likelihood. While efficient, FTP can be fragile: models may assign high probability to unrelated tokens (*misalignment*) or use a valid token mere

Qi'ao Xu, Yan Xing, Jiali Hu, Yunan Jia

Change detection, a critical task in remote sensing and computer vision, aims to identify pixel-level differences between image pairs captured at the same geographic area but different times. It faces numerous challenges such as illumination variation, seasonal changes, background interference, and shooting angles, especially with a large time gap between im

Mikhail Prokofyev

In this paper, we study the property of hereditary completeness of vector systems $\{x_k\}_{k=1}^\infty$ in a Hilbert space. A criterion of hereditary completeness is obtained in terms of projectors on closed linear spans of systems of the form $\{x_k\}_{k \in N}$, $N \subset \mathbb{N}$. Developed technique has been used to prove that mixed systems of a her

Sara Barahona, Anna Silnova, Ladislav Mošner, Junyi Peng

We present a comprehensive analysis of the embedding extractors (frontends) developed by the ABC team for the audio track of NIST SRE 2024. We follow the two scenarios imposed by NIST: using only a provided set of telephone recordings for training (fixed) or adding publicly available data (open condition). Under these constraints, we develop the best possibl

Soumya Sengupta, Manika Singla, Fikret Anli

Chandrasekhar's H(mu)-function forms the foundation of radiative transfer theory for semi-infinite, isotropically scattering atmospheres under external illumination. However, the classical formulation does not account for thermal emission from internal heat sources, which is essential in many astrophysical environments, including hot Jupiters, brown dwarfs,

Arghya Sinha, Bhartendu Kumar, Chirayu D. Athalye, Kunal N. Chaudhury

The use of denoisers for image reconstruction has shown significant potential, especially for the Plug-and-Play (PnP) framework. In PnP, a powerful denoiser is used as an implicit regularizer in proximal algorithms such as ISTA and ADMM. The focus of this work is on the convergence of PnP iterates for linear inverse problems using kernel denoisers. It was sh

Xiaozhou Yang, Zhaozheng Lyu, Xiang Wang, Enna Zhuo

Artificial Kitaev chains (AKCs), formed of quantum dot-superconductor linear arrays, provide a promising platform for hosting Majorana bound states (MBSs) and implementing topological quantum computing. The main challenges along this research direction would include the tuning up of AKCs for hosting MBSs and the readout of the parity of the chains. In this w

Xin Bai, Guanyi Chen, Tingting He, Chenlian Zhou

Emotional Support Conversations (ESC) are crucial for providing empathy, validation, and actionable guidance to individuals in distress. However, existing definitions of the ESC task oversimplify the structure of supportive responses, typically modelling them as single strategy-utterance pairs. Through a detailed corpus analysis of the ESConv dataset, we ide

Haowei Hua, Wanyu Lin

Crystal structures are defined by the periodic arrangement of atoms in 3D space, inherently making them equivariant to SO(3) group. A fundamental requirement for crystal property prediction is that the model's output should remain invariant to arbitrary rotational transformations of the input structure. One promising strategy to achieve this invariance is to

Mohammad Ali S. Afshar, Mohammad Reza Alipour, Saeed Noori Gashti, Jafar Sadeghi

To effectively utilize the AdS/CFT correspondence, a precise set of rules must be established to guide the translation of computed quantities in the gravitational sector into their CFT counterparts, and vice versa. This framework is commonly referred to as the holographic dictionary. The formulation of such dictionaries opens a two-way gateway, allowing rese

Kazuaki Mishima, Antoni Bigata Casademunt, Stavros Petridis, Maja Pantic

Human facial images encode a rich spectrum of information, encompassing both stable identity-related traits and mutable attributes such as pose, expression, and emotion. While recent advances in image generation have enabled high-quality identity-conditional face synthesis, precise control over non-identity attributes remains challenging, and disentangling i

Yuxuan Shu, Vasileios Lampos

Multivariable time series forecasting methods can integrate information from exogenous variables, leading to significant prediction accuracy gains. The transformer architecture has been widely applied in various time series forecasting models due to its ability to capture long-range sequential dependencies. However, a na\"ive application of transformers ofte

Yurun Yuan, Fan Chen, Zeyu Jia, Alexander Rakhlin

Policy-based methods currently dominate reinforcement learning (RL) pipelines for large language model (LLM) reasoning, leaving value-based approaches largely unexplored. We revisit the classical paradigm of Bellman Residual Minimization and introduce Trajectory Bellman Residual Minimization (TBRM), an algorithm that naturally adapts this idea to LLMs, yield

Tingbin Cao, Risto Korhonen, Wenlong Liu

In this article, we focus on studying the differential-difference equation \[ f'(z) = a(z)f(z+1) + R(z, f(z)), \quad R(z, f(z)) = \frac{P(z, f(z))}{Q(z, f(z))}, \] where the two nonzero polynomials \( P(z, f(z)) \) and \( Q(z, f(z)) \) in \( f(z) \), with small meromorphic coefficients, are coprime, and \( a(z) \) is a nonzero small meromorphic function of \

A. Garcia-Lorenzo, M. Albaladejo, S. Gonzlez-Solis, N. Hammoud

This work studies the $\phi \to 3\pi$ decay and the $\phi \to \pi^0 \gamma^\ast$ transition form factor, utilizing the Khuri-Treiman formalism to account for analyticity, crossing, and unitarity. Using once-subtracted dispersion relations, we perform a simultaneous fit to the $\phi \to 3\pi$ Dalitz plot distribution and the $\phi \to \pi^0 \gamma^\ast$ measu

Ji Guo, Xiaolei Wen, Wenbo Jiang, Cheng Huang

With the widespread application of super-resolution (SR) in various fields, researchers have begun to investigate its security. Previous studies have demonstrated that SR models can also be subjected to backdoor attacks through data poisoning, affecting downstream tasks. A backdoor SR model generates an attacker-predefined target image when given a triggered

Daisuke Niizumi, Daiki Takeuchi, Masahiro Yasuda, Binh Thien Nguyen

Recent advancements in foundation models have sparked interest in respiratory audio foundation models. However, the effectiveness of applying conventional pre-training schemes to datasets that are small-sized and lack diversity has not been sufficiently verified. This study aims to explore better pre-training practices for respiratory sounds by comparing num

Yiwen Song, Qianyue Hao, Qingmin Liao, Jian Yuan

Model ensemble is a useful approach in reinforcement learning (RL) for training effective agents. Despite wide success of RL, training effective agents remains difficult due to the multitude of factors requiring careful tuning, such as algorithm selection, hyperparameter settings, and even random seed choices, all of which can significantly influence an agen

Xin Gao, Qinjian Lou, Yi-Nan Wang

We investigate the quantum tunneling process through a topology transition near a conifold singularity, in the setup of IIB CY3 orientifold compactification. We propose a novel method to do moduli stabilization in an extended moduli space, parametrized by both the geometric moduli and the light D3-brane wrapping modes arisen from the brane quantization. Assu

Seongmin Park, Hyungmin Kim, Sangwoo Kim, Wonseok Jeon

Deep neural network (DNN)-based policy models, such as vision-language-action (VLA) models, excel at automating complex decision-making from multi-modal inputs. However, scaling these models greatly increases computational overhead, complicating deployment in resource-constrained settings like robot manipulation and autonomous driving. To address this, we pr

Johannes Kaiser, Kristian Schwethelm, Daniel Rueckert, Georgios Kaissis

Accurately estimating the informativeness of individual samples in a dataset is an important objective in deep learning, as it can guide sample selection, which can improve model efficiency and accuracy by removing redundant or potentially harmful samples. We propose Laplace Sample Information (LSI) measure of sample informativeness grounded in information t

Jan Simon Haase, Alexander Fieguth, Igor Bröckel, Janina Hamann

We present a novel concept for a compact and robust crossed-beam optical dipole trap (cODT) based on a single lens, designed for the efficient generation of Bose-Einstein condensates (BECs) under dynamic conditions. The system employs two independent two-dimensional acousto-optical deflectors (AODs) in combination with a single high-numerical-aperture lens t

Masakatsu Kenmoku

The magnetic monopole is one of the important problems in the early stage of universe as well as observations and experiments on Earth. We study the existence or non-existence of the Dirac and the 't Hooft-Polyakov magnetic monopole theories using the pre-potential $\boldsymbol{C}$, which is defined to derive the vector potential by the curl operation as $\b

Xin Chen, Kun Yin

In this paper we will study homogenization of for stable-like process with divergence-free drift in ergodic environments. In particular, neither the drift nor the stream function are required to be bounded.

Maodong Li, Longyin Zhang, Fang Kong

Multi-hop question generation (MQG) aims to generate questions that require synthesizing multiple information snippets from documents to derive target answers. The primary challenge lies in effectively pinpointing crucial information snippets related to question-answer (QA) pairs, typically relying on keywords. However, existing works fail to fully utilize t

Sonia Mazelet, Rémi Flamary, Bertrand Thirion

Optimal transport between graphs, based on Gromov-Wasserstein and other extensions, is a powerful tool for comparing and aligning graph structures. However, solving the associated non-convex optimization problems is computationally expensive, which limits the scalability of these methods to large graphs. In this work, we present Unbalanced Learning of Optima

Kangan Qian, Sicong Jiang, Yang Zhong, Ziang Luo

Vision-Language Models (VLMs) show promise for autonomous driving, yet their struggle with hallucinations, inefficient reasoning, and limited real-world validation hinders accurate perception and robust step-by-step reasoning. To overcome this, we introduce \textbf{AgentThink}, a pioneering unified framework that integrates Chain-of-Thought (CoT) reasoning w

Xintong Wang, Yixiao Liu, Jingheng Pan, Liang Ding

Detoxifying offensive language while preserving the speaker's original intent is a challenging yet critical goal for improving the quality of online interactions. Although large language models (LLMs) show promise in rewriting toxic content, they often default to overly polite rewrites, distorting the emotional tone and communicative intent. This problem is

Perukrishnen Vytelingum, Rory Baggott, Namid Stillman, Jianfei Zhang

In this paper, we describe a novel agent-based approach for modelling the transaction cost of buying or selling an asset in financial markets, e.g., to liquidate a large position as a result of a margin call to meet financial obligations. The simple act of buying or selling in the market causes a price impact and there is a cost described as liquidity risk.

Lie Fu, Zhiyuan Li, Teppei Takamatsu, Haitao Zou

We investigate in this paper the so-called pointed Shafarevich problem for families of primitive symplectic varieties. More precisely, for any fixed pointed curve $(B, 0)$ and any fixed primitive symplectic variety $X$, among all locally trivial families of $\mathbb{Q}$-factorial and terminal primitive symplectic varieties over $B$ whose fiber over $0$ is is

Xu yan, Zhaohui Wang, Rong Wei, Jingbo Yu

We propose R3GS, a robust reconstruction and relocalization framework tailored for unconstrained datasets. Our method uses a hybrid representation during training. Each anchor combines a global feature from a convolutional neural network (CNN) with a local feature encoded by the multiresolution hash grids [2]. Subsequently, several shallow multi-layer percep

Qianyue Hao, Yiwen Song, Qingmin Liao, Jian Yuan

Policy exploration is critical in reinforcement learning (RL), where existing approaches include greedy, Gaussian process, etc. However, these approaches utilize preset stochastic processes and are indiscriminately applied in all kinds of RL tasks without considering task-specific features that influence policy exploration. Moreover, during RL training, the

Debraj Mahata, Ananthakrishnan Srinivasana

Three sets of Co-Ni-Ga alloy nanoparticles have been synthesized by a template-free chemical route. Structural, morphological, shape memory, and magnetic properties of room temperature martensite (M) phase, dual (M + secondary $\gamma$) phase and austenite (A) phase Co-Ni-Ga nanoparticles are reported. Temperature-dependent XRD analysis revealed that $Co_{36

Joonho Yang, Seunghyun Yoon, Hwan Chang, Byeongjeong Kim

Large Language Models (LLMs) have significantly advanced text generation capabilities, including tasks like summarization, often producing coherent and fluent outputs. However, faithfulness to source material remains a significant challenge due to the generation of hallucinations. While extensive research focuses on detecting and reducing these inaccuracies,

Yingtong Chen, Fei Wu, Dayong Zhang, Qiang Ji

The impact of international tourism on biodiversity risks has received considerable attention, yet quantitative research in this field remains relatively limited. This study constructs a biodiversity risk index for 155 countries and regions spanning the years 2001 to 2019, analysing how international tourism influences biodiversity risks in destination count

Michał Pilipczuk

We prove that for every $d\in \mathbb{N}$ and a graph class of bounded expansion $\mathscr{C}$, there exists some $c\in \mathbb{N}$ so that every graph from $\mathscr{C}$ admits a proper coloring with at most $c$ colors satisfying the following condition: in every ball of radius $d$, every color appears either zero times or an odd number of times. For $d=1$,

Yuchen Li, Chaoran Feng, Zhenyu Tang, Kaiyuan Deng

We introduce GS2E (Gaussian Splatting to Event), a large-scale synthetic event dataset for high-fidelity event vision tasks, captured from real-world sparse multi-view RGB images. Existing event datasets are often synthesized from dense RGB videos, which typically lack viewpoint diversity and geometric consistency, or depend on expensive, difficult-to-scale

Shital H. Joshi, Ekta Shah

We define and study stronger forms of Devaney chaos and name it as $\mathscr{F}-$Devaney chaos, where $\mathscr{F}$ is a family of subsets of $\mathbb{N}$. Examples of maps which is $\mathscr{F}_t-$Devaney chaotic but not $\mathscr{F}_{cf}-$Devaney chaotic, $\mathscr{F}_s-$Devaney chaotic but neither $\mathscr{F}_t-$Devaney chaotic nor $\mathscr{F}_{cf}-$Dev

Fengting Zhang, Boxu Liang, Qinghao Liu, Min Liu

Mesh reconstruction is a cornerstone process across various applications, including in-silico trials, digital twins, surgical planning, and navigation. Recent advancements in deep learning have notably enhanced mesh reconstruction speeds. Yet, traditional methods predominantly rely on deforming a standardised template mesh for individual subjects, which over

Tao Sun, Enhao Pan, Zhengkai Yang, Kaixin Sui

Academic posters are vital for scholarly communication, yet their manual creation is time-consuming. However, automated academic poster generation faces significant challenges in preserving intricate scientific details and achieving effective visual-textual integration. Existing approaches often struggle with semantic richness and structural nuances, and lac

Bilgesu Arif Bilgin, Olof Hallqvist Elias, Michael Selby, Phillip Stanley-Marbell

This article studies a general divide-and-conquer algorithm for approximating continuous one-dimensional probability distributions with finite mean. The article presents a numerical study that compares pre-existing approximation schemes with a special focus on the stability of the discrete approximations when they undergo arithmetic operations. The main resu

Yanzhi Tian, Zeming Liu, Zhengyang Liu, Yuhang Guo

In-Image Machine Translation (IIMT) aims to translate texts within images from one language to another. Previous research on IIMT was primarily conducted on simplified scenarios such as images of one-line text with black font in white backgrounds, which is far from reality and impractical for applications in the real world. To make IIMT research practically

Ian George, Marco Tomamichel

The maximal correlation coefficient measures the linear correlation in a bipartite distribution and contraction coefficients measure how much information is lost under a noisy channel. Remarkably, Raginsky established a close relation between these two concepts by showing that the $\chi^2$ contraction coefficient equals the maximal correlation coefficient of

Arkadiusz Kobus, Xinwei Li, Mariusz Gajda, Li You

We propose a multisetting protocol for the detection of two-body Bell correlations, and apply it to spin-nematic squeezed states realized in $f$ pairs of SU(2) subsystems within spin-$f$ atomic Bose-Einstein condensates. Experimental data for $f=1$, alongside with numerical simulations using the truncated Wigner method for $f=1,\,2,\,3$, demonstrate the effe

Zixiang Ren, Siyao Zhang, Ling Qiu, Derrick Wing Kwan Ng

This paper investigates robust secure communications in a near-field integrated sensing, communication, and powering (ISCAP) system, in which the base station (BS) is equipped with an extremely large-scale antenna array (ELAA). In this system, the BS transmits confidential messages to a single legitimate communication user (CU), simultaneously providing wire

Fernando Ballesta-Yagüe

An $A_{\infty}$-weight on a Lipschitz curve $\Lambda$ in the plane can be extended analytically to the graph Lipschitz domain $\Omega$ above it. This problem was studied by C. Kenig [Ken80], who introduced the class $AE$ of well-behaved analytic extensions. Later, he and D. Jerison [JK82] added a Smirnov-type condition to the definition of this class. In thi

Hyungjoo Chae, Sunghwan Kim, Junhee Cho, Seungone Kim

Web navigation is a unique domain that can automate many repetitive real-life tasks and is challenging as it requires long-horizon sequential decision making beyond typical multimodal large language model (MLLM) tasks. Yet, specialized reward models for web navigation that can be utilized during both training and test-time have been absent until now. Despite

Rongzhi Zhu, Yi Liu, Jiancheng Wang, Xiangyu Liu

Large reasoning models (LRMs) have achieved remarkable success on complex tasks, yet their tendency to "overthink" leads to inefficiencies. Although "save-thinking" prompts are intended to mitigate this issue, we find that LRMs still frequently enter the "Still-thinking" mode instead of the expected "No-thinking" mode, especia

Seokjun Lee, Seung-Hyun Kong

Oversteer, wherein a vehicle's rear tires lose traction and induce unintentional excessive yaw, poses critical safety challenges. Failing to control oversteer often leads to severe traffic accidents. Although recent autonomous driving efforts have attempted to handle oversteer through stabilizing maneuvers, the majority rely on expert-defined trajectories or

Cécile Rousseau, Tobia Boschi, Giandomenico Cornacchia, Dhaval Salwala

SDForger is a flexible and efficient framework for generating high-quality multivariate time series using LLMs. Leveraging a compact data representation, SDForger provides synthetic time series generation from a few samples and low-computation fine-tuning of any autoregressive LLM. Specifically, the framework transforms univariate and multivariate signals in

Xin Shu, Shuai Wang, Ang Li

Probabilities of causation (PoCs) are fundamental quantities for counterfactual analysis and personalized decision making. However, existing analytical results are largely confined to binary settings. This paper extends PoCs to multi-valued treatments and outcomes by deriving closed form bounds for a representative family of discrete PoCs within Structural C

Chengkang Xu, Fulin Chen, Shaobin Tan

In this paper, we study non-weight modules over gap-$p$ Virasoro algebras, including Whittaker modules, $\mathcal{U}(\mathbb{C} L_0)$-free modules and their tensor products. We establish necessary and sufficient conditions for universal Whittaker modules to be irreducible and study the structure of irreducible Whittaker modules. The $\mathcal{U}(\mathbb{C} L

Eduarda Caldeira, Jan Niklas Kolf, Naser Damer, Fadi Boutros

Face recognition models have made substantial progress due to advances in deep learning and the availability of large-scale datasets. However, reliance on massive annotated datasets introduces challenges related to training computational cost and data storage, as well as potential privacy concerns regarding managing large face datasets. This paper presents D

Xiaotian Zhao, Zixuan Li, Yichen Cai, Xinfei Guo

The increasing number of rectilinear floorplans in modern chip designs presents significant challenges for traditional macro placers due to the additional complexity introduced by blocked corners. Particularly, the widely adopted wirelength model Half-Perimeter Wirelength (HPWL) struggles to accurately handle rectilinear boundaries, highlighting the need for

Chenyu Zheng, Xinyu Zhang, Rongzhen Wang, Wei Huang

Diffusion Transformers have emerged as the foundation for vision generative models, but their scalability is limited by the high cost of hyperparameter (HP) tuning at large scales. Recently, Maximal Update Parametrization ($\mu$P) was proposed for vanilla Transformers, which enables stable HP transfer from small to large language models, and dramatically red

Zhenyu Ning, Guangda Liu, Qihao Jin, Chengwei Li

Recent developments in Video Large Language Models (Video LLMs) have enabled models to process hour-long videos and exhibit exceptional performance. Nonetheless, the Key-Value (KV) cache expands linearly over time, leading to substantial memory overhead and response delay--critical challenges in various real-world online applications, such as Deepseek servic

Stella Civelli, Dario Cellini, Enrico Forestieri, Marco Secondini

Fiber nonlinearity represents a critical challenge to the capacity enhancement of modern optical communication systems. In recent years, significant research efforts have focused on mitigating its impact through two complementary approaches. On the one hand, researchers have investigated practical digital signal processing (DSP) techniques to mitigate or com

Wenmin Li, Shunsuke Sakai, Tatsuhito Hasegawa

Deploying machine learning models in resource-constrained environments, such as edge devices or rapid prototyping scenarios, increasingly demands distillation of large datasets into significantly smaller yet informative synthetic datasets. Current dataset distillation techniques, particularly Trajectory Matching methods, optimize synthetic data so that the m

Xiangkai Chen, Yuhong Li, Xiaofei Zhu, Yun-Long Tang

Oxides have long been regarded as intrinsically brittle due to their strong, directional ionic or covalent bonds, in stark contrast to the ductile behavior of metals, where delocalized electron sharing enables plasticity through facile dislocation glide. Here, we challenge this paradigm by demonstrating that typical oxides, such as SrTiO3 and MgO, can exhibi

Zihao Pan, Yu Tong, Weibin Wu, Jingyi Wang

Adversarial attacks aim to generate malicious inputs that mislead deep models, but beyond causing model failure, they cannot provide certain interpretable information such as ``\textit{What content in inputs make models more likely to fail?}'' However, this information is crucial for researchers to specifically improve model robustness. Recent research sugge

Ronald Quirchmayr, Alden Waters

We prove dispersive estimates for the wave equation in the exterior of a torus. Because no separation of variables into a basis of eigenfunctions and eigenvalues exists for the time harmonic problem, we introduce a related approximate operator for the Dirichlet Laplacian in the exterior of a torus. The approximate operator coincides with the Schr\"odinger op

Om Khangaonkar, Hamed Pirsiavash

By pretraining to synthesize coherent images from perturbed inputs, generative models inherently learn to understand object boundaries and scene compositions. How can we repurpose these generative representations for general-purpose perceptual organization? We finetune Stable Diffusion and MAE (encoder+decoder) for category-agnostic instance segmentation usi

Yuwei Hu, Jun Zheng, Leandro S. Tavares

Let $N>2$, $p\in \left(\frac{2N}{N+2},+\infty\right)$, and $\Omega$ be an open bounded domain in $\mathbb{R}^N$. We consider the minimum problem $$ \mathcal{J} (u) := \displaystyle\int_{\Omega } \left(\frac{1}{p}| \nabla u| ^p+\lambda_1\left(1-(u^+)^2\right)^2+\lambda_2u^+\right)\text{d}x\rightarrow \text{min} $$ over a certain class $\mathcal{K}$, where $\l

Jiatao Li, Mao Ye, Cheng Peng, Xunjian Yin

Existing AI-generated text detection methods heavily depend on large annotated datasets and external threshold tuning, restricting interpretability, adaptability, and zero-shot effectiveness. To address these limitations, we propose AGENT-X, a zero-shot multi-agent framework informed by classical rhetoric and systemic functional linguistics. Specifically, we

Nicolas Bouchot

The random interlacements $\mathscr{I}(u)$ at level $u$ has been introduced by Sznitman, as a Poissonian collection of independent simple random walk trajectories on $\mathbb{Z}^d$, $d\geq 3$, with intensity $u>0$. Since then, several works investigated the properties of the random interlacements intersected with large sets of~$\mathbb{Z}^d$. In this paper,

Hyunseok Lee, Jeonghoon Kim, Beomjun Kim, Jihoon Tack

Recent advances in Multimodal Large Language Models (MLLMs) have enabled autonomous agents to interact with computers via Graphical User Interfaces (GUIs), where accurately localizing the coordinates of interface elements (e.g., buttons) is often required for fine-grained actions. However, this remains significantly challenging, leading prior works to rely o

Josnei Novacoski

For a valued field $(K,v)$, with a fixed extension of $v$ to the algebraic closure $\overline K$ of $K$, and an element $\theta\in\overline K$, we are interested in the possible values of $\theta-\theta'$ where $\theta'$ runs through all the $K$-conjugates of $\theta$. The study of these values is a classic problem in number theory and ramification theory. H

Weixiang Zhao, Jiahe Guo, Yang Deng, Tongtong Wu

Multilingual reasoning remains a significant challenge for large language models (LLMs), with performance disproportionately favoring high-resource languages. Drawing inspiration from cognitive neuroscience, which suggests that human reasoning functions largely independently of language processing, we hypothesize that LLMs similarly encode reasoning and lang

Tatyana Shmykova, Leila Khaertdinova, Ilya Pershin

Accurate segmentation of anatomical structures in volumetric medical images is crucial for clinical applications, including disease monitoring and cancer treatment planning. Contemporary interactive segmentation models, such as Segment Anything Model 2 (SAM-2) and its medical variant (MedSAM-2), rely on manually provided prompts like bounding boxes and mouse

Yuansheng Gao, Peng Gao, Han Bao, Bin Li

Mental manipulation on social media poses a covert yet serious threat to individuals' psychological well-being and the integrity of online interactions. Detecting such behavior is challenging due to the difficult-to-annotate training data, its highly covert and multi-turn nature, and the lack of real-world datasets. To address these challenges, we propose Me

Kyungguen Byun, Jason Filos, Erik Visser, Sunkuk Moon

We propose a speech enhancement system that combines speaker-agnostic speech restoration with voice conversion (VC) to obtain a studio-level quality speech signal. While voice conversion models are typically used to change speaker characteristics, they can also serve as a means of speech restoration when the target speaker is the same as the source speaker.

Théo Leblanc

We prove that the number of points of a stationary linear Hawkes process lying in any bounded subset of the real line has exponential moments, without any other assumption than the one needed for existence of such stationary process, namely the spectral radius of the matrix of L1 norms of interaction functions is smaller than one. The proof relies on a mass

Zhengyi Li, Yue Guan, Kang Yang, Yu Feng

The wide deployment of the generative pre-trained transformer (GPT) has raised privacy concerns for both clients and servers. While cryptographic primitives can be employed for secure GPT inference to protect the privacy of both parties, they introduce considerable performance overhead.To accelerate secure inference, this study proposes a public decoding and

Idriss Malek, Aya Laajil, Abhijith Sharma, Eric Moulines

Although Generative Flow Networks (GFlowNets) are designed to capture multiple modes of a reward function, they often suffer from mode collapse in practice, getting trapped in early-discovered modes and requiring prolonged training to find diverse solutions. Existing exploration techniques often rely on heuristic novelty signals. We propose Loss-Guided GFlow

Suping Xu, Lin Shang, Keyu Liu, Hengrong Ju

Fuzzy rough feature selection (FRFS) is an effective means of addressing the curse of dimensionality in high-dimensional data. By removing redundant and irrelevant features, FRFS helps mitigate classifier overfitting, enhance generalization performance, and lessen computational overhead. However, most existing FRFS algorithms primarily focus on reducing unce

Yerin Hwang, Dongryeol Lee, Kyungmin Min, Taegwan Kang

Recently, large vision-language models (LVLMs) have emerged as the preferred tools for judging text-image alignment, yet their robustness along the visual modality remains underexplored. This work is the first study to address a key research question: Can adversarial visual manipulations systematically fool LVLM judges into assigning unfairly inflated scores

Andre Dourson, Kylie Taylor, Xiaoli Qiao, Michael Fitzke

Self-supervised learning has emerged as a powerful paradigm for training deep neural networks, particularly in medical imaging where labeled data is scarce. While current approaches typically rely on synthetic augmentations of single images, we propose VET-DINO, a framework that leverages a unique characteristic of medical imaging: the availability of multip

Feng-Ji Chen, Chao-Kai Wen, De-Ming Chian

While numerous experimental studies have demonstrated the feasibility of reconfigurable intelligent surface (RIS) technology, most have primarily focused on extending coverage. In contrast, this paper presents an experimental evaluation of multiple active RISs deployed in a 5G multiple-input multiple-output (MIMO) commercial network, emphasizing enhancements

Xiaoling Zhou, Wei Ye, Rui Xie, Shikun Zhang

Deep learning has seen widespread success in various domains such as science, industry, and society. However, it is acknowledged that certain approaches suffer from non-robustness, relying on spurious correlations for predictions. Addressing these limitations is of paramount importance, necessitating the development of methods that can disentangle spurious c

Zihao Jiang, Ben Liu, Miao Peng, Wenjie Xu

While large language models (LLMs) show great potential in temporal reasoning, most existing work focuses heavily on enhancing performance, often neglecting the explainable reasoning processes underlying the results. To address this gap, we introduce a comprehensive benchmark covering a wide range of temporal granularities, designed to systematically evaluat

Mohamad Mestoukirdi, Mourad Khanfouci

This work proposes a new algorithm to mitigate model generalization loss in Vertical Federated Learning (VFL) operating under client reliability constraints within 5G Core Networks (CNs). Recently studied and endorsed by 3GPP, VFL enables collaborative and load-balanced model training and inference across the CN. However, the performance of VFL significantly

Yu Wang

We introduce a direct estimation framework for reconstructing multiple density matrix elements of an unknown quantum state using classical shadow tomography. Traditional direct measurement protocols (DMPs), while effective for individual elements, suffer from poor scalability due to post-selection losses and the need for element-specific measurement configur

Zhiyuan Wei, Jing Sun, Zijian Zhang, Zhe Hou

Large Language Models (LLMs) have shown great promise in code analysis and auditing; however, they still struggle with hallucinations and limited context-aware reasoning. We introduce SmartAuditFlow, a novel Plan-Execute framework that enhances smart contract security analysis through dynamic audit planning and structured execution. Unlike conventional LLM-b

Emma Collins, Myungseo wong, Kim Yun, Finn Kingston

Despite progress in geometry-aware domain adaptation, current methods such as GAMA still suffer from two unresolved issues: (1) insufficient disentanglement of task-relevant and task-irrelevant manifold dimensions, and (2) rigid perturbation schemes that ignore per-class alignment asymmetries. To address this, we propose GAMA++, a novel framework that introd

Yassir Fathullah, Mark J. F. Gales

This paper explores generalised probabilistic modelling and uncertainty estimation in comparative LLM-as-a-judge frameworks. We show that existing Product-of-Experts methods are specific cases of a broader framework, enabling diverse modelling options. Furthermore, we propose improved uncertainty estimates for individual comparisons, enabling more efficient

Peter Súkeník, Christoph H. Lampert, Marco Mondelli

The empirical emergence of neural collapse -- a surprising symmetry in the feature representations of the training data in the penultimate layer of deep neural networks -- has spurred a line of theoretical research aimed at its understanding. However, existing work focuses on data-agnostic models or, when data structure is taken into account, it remains limi

W. F. Holmes-Hewett, J. D. Miller, H. G. Ahmad, S. Granville

Driven by the pursuit of high-performance electronic devices, research into novel materials with properties appropriate for cryogenic applications has unveiled the exceptional properties of the rare-earth nitride series of intrinsic ferromagnetic semiconductors. Here we report on the field focusing on developments, since the most recent comprehensive review

Milou Temmink, Andrew D. Sellek, Danny Gasman, Ewine F. van Dishoeck

Millimetre-compact dust disks are thought to have efficient radial drift of icy dust pebbles, which has been hypothesised to produce an enhanced cold ($T<$400 K) H$_2$O reservoir in their inner disks. Mid-infrared spectral surveys, now with the James Webb Space Telescope (JWST), pave the way to explore this hypothesis. In this work, we test this theory for 8

Koustav Banerjee, Kathrin Bringmann, Ben Kane

We derive an asymptotic expansion with effective error bound for $u(n)$, counting the number of unimodal sequences of size $n$. We prove that $u(n)$ satisfies the higher order Tur\'{a}n inequalities for $n\geq33$ and that certain second $j$-shifted difference of $u(n)$ are positive.

Yifan Liu, Wuyang Li, Weihao Yu, Chenxin Li

Computed Tomography serves as an indispensable tool in clinical workflows, providing non-invasive visualization of internal anatomical structures. Existing CT reconstruction works are limited to small-capacity model architecture and inflexible volume representation. In this work, we present X-GRM (X-ray Gaussian Reconstruction Model), a large feedforward mod

Saqib Qamar, Mohd Fazil, Parvez Ahmad, Shakir Khan

Medical image segmentation plays an important role in various clinical applications; however, existing deep learning models face trade-offs between efficiency and accuracy. Convolutional Neural Networks (CNNs) capture local details well but miss the global context, whereas transformers handle the global context but at a high computational cost. Recently, Sta

Yuxuan Du, Zhendong Wang, Yuhao Luo, Caiyong Piao

The rapid emergence of multimodal deepfakes (visual and auditory content are manipulated in concert) undermines the reliability of existing detectors that rely solely on modality-specific artifacts or cross-modal inconsistencies. In this work, we first demonstrate that modality-specific forensic traces (e.g., face-swap artifacts or spectral distortions) and

Ting Huang, Zeyu Zhang, Ruicheng Zhang, Yang Zhao

3D scene understanding plays a fundamental role in vision applications such as robotics, autonomous driving, and augmented reality. However, advancing learning-based 3D scene understanding remains challenging due to two key limitations: (1) the large scale and complexity of 3D scenes lead to higher computational costs and slower training compared to 2D count

Kabir V. Dabholkar, Omri Barak

Many natural systems, including neural circuits involved in decision making, are modeled as high-dimensional dynamical systems with multiple stable states. While existing analytical tools primarily describe behavior near stable equilibria, characterizing separatrices--the manifolds that delineate boundaries between different basins of attraction--remains cha

Thilo Baumann

We show that Kuznetsov--Shinder's notion of deformation absorption of singularities leads to a new approach for studying the bounded derived category of a hereditary order on a curve. The starting point is a hereditary order which can be interpreted as a smoothing of the finite-dimensional algebra obtained from the restriction to a ramified point. We constru

Qihan Wang, Shidong Pan, Tal Linzen, Emily Black

Large Language Models (LLMs) are known to lack cultural representation and overall diversity in their generations, from expressing opinions to answering factual questions. To mitigate this problem, we propose multilingual prompting: a prompting method which generates several variations of a base prompt with added cultural and linguistic cues from several cul