May 2025 arXiv papers — page 106

Rui Tian, Mingfei Gao, Mingze Xu, Jiaming Hu

We introduce UniGen, a unified multimodal large language model (MLLM) capable of image understanding and generation. We study the full training pipeline of UniGen from a data-centric perspective, including multi-stage pre-training, supervised fine-tuning, and direct preference optimization. More importantly, we propose a new Chain-of-Thought Verification (Co

Mengru Wang, Xingyu Chen, Yue Wang, Zhiwei He

Mixture-of-Experts (MoE) architectures within Large Reasoning Models (LRMs) have achieved impressive reasoning capabilities by selectively activating experts to facilitate structured cognitive processes. Despite notable advances, existing reasoning models often suffer from cognitive inefficiencies like overthinking and underthinking. To address these limitat

Sunhao Dai, Wenjie Wang, Liang Pang, Jun Xu

Generative AI search is reshaping information retrieval by offering end-to-end answers to complex queries, reducing users' reliance on manually browsing and summarizing multiple web pages. However, while this paradigm enhances convenience, it disrupts the feedback-driven improvement loop that has historically powered the evolution of traditional Web search.

Xiaojie Gu, Ziying Huang, Jia-Chen Gu, Kai Zhang

Lifelong learning enables large language models (LLMs) to adapt to evolving information by continually updating their internal knowledge. An ideal system should support efficient, wide-ranging updates while preserving existing capabilities and ensuring reliable deployment. Model editing stands out as a promising solution for this goal, offering a focused and

Gareth Speight, Scott Zimmerman

We show that, in arbitrary Carnot groups, pliability in a subset of directions is sufficient to guarantee the existence of a Whitney-type extension and a Lusin approximation for curves with tangent vectors in the same set of directions. We apply this to show that every horizontal curve in the Engel group must intersect a $C^{1}$ horizontal curve in a set of

Jiaer Xia, Yuhang Zang, Peng Gao, Sharon Li

Learning general-purpose reasoning capabilities has long been a challenging problem in AI. Recent research in large language models (LLMs), such as DeepSeek-R1, has shown that reinforcement learning techniques like GRPO can enable pre-trained LLMs to develop reasoning capabilities using simple question-answer pairs. In this paper, we aim to train visual lang

Behzad Tahmasebzadeh, Matthew A. Taylor, Monica Valluri, Haruka Yoshino

We present a new stellar dynamical measurement of the supermassive black hole (SMBH) in the compact elliptical galaxy NGC 4486B, based on integral field spectroscopy with JWST/NIRSpec. The two-dimensional kinematic maps reveal a resolved double nucleus and a velocity dispersion peak offset from the photometric center. Utilizing two independent methods-Schwar

Olivier Labayle, Breeshey Roskams-Hieter, Joshua Slaughter, Kelsey Tetley-Campbell

Population genetics seeks to quantify DNA variant associations with traits or diseases, as well as interactions among variants and with environmental factors. Computing millions of estimates in large cohorts in which small effect sizes are expected, necessitates minimising model-misspecification bias to control false discoveries. We present TarGene, a unifie

Jiaxin Guo, Zewen Chi, Li Dong, Qingxiu Dong

Reward models play a critical role in guiding large language models toward outputs that align with human expectations. However, an open challenge remains in effectively utilizing test-time compute to enhance reward model performance. In this work, we introduce Reward Reasoning Models (RRMs), which are specifically designed to execute a deliberate reasoning p

Yu Tong, Zihao Pan, Shuai Yang, Kaiyang Zhou

Invisible image watermarking can protect image ownership and prevent malicious misuse of visual generative models. However, existing generative watermarking methods are mainly designed for diffusion models while watermarking for autoregressive image generation models remains largely underexplored. We propose IndexMark, a training-free watermarking framework

Alessio Cela, Carl Lian

We introduce a moduli space of ``complete quasimaps'' to $\mathsf{Bl}_{\mathbb{P}^s}(\mathbb{P}^r)$. The construction, following previous work for curves on projective spaces, essentially proceeds by blowing up Ciocan-Fontanine--Kim's space of quasimaps at loci where sections of line bundles are linearly dependent. We conjecture that tautological intersectio

Ruichuan An, Sihan Yang, Renrui Zhang, Zijun Shen

Personalized models have demonstrated remarkable success in understanding and generating concepts provided by users. However, existing methods use separate concept tokens for understanding and generation, treating these tasks in isolation. This may result in limitations for generating images with complex prompts. For example, given the concept $\langle bo\ra

Jiaqi Leng, Bin Shi

With rapid advancements in machine learning, first-order algorithms have emerged as the backbone of modern optimization techniques, owing to their computational efficiency and low memory requirements. Recently, the connection between accelerated gradient methods and damped heavy-ball motion, particularly within the framework of Hamiltonian dynamics, has insp

Roberto L. Castro, Andrei Panferov, Soroush Tabesh, Oliver Sieberling

Training large language models (LLMs) models directly in low-precision offers a way to address computational costs by improving both throughput and energy efficiency. For those purposes, NVIDIA's recent Blackwell architecture facilitates very low-precision operations using FP4 variants. Yet, current algorithms for training LLMs in FP4 precision face signific

Bufang Yang, Lilin Xu, Liekang Zeng, Kaiwei Liu

Recent advances in Large Language Models (LLMs) have propelled intelligent agents from reactive responses to proactive support. While promising, existing proactive agents either rely exclusively on observations from enclosed environments (e.g., desktop UIs) with direct LLM inference or employ rule-based proactive notifications, leading to suboptimal user int

Wonje Jeung, Sangyeon Yoon, Minsuk Kahng, Albert No

Large Reasoning Models (LRMs) have become powerful tools for complex problem solving, but their structured reasoning pathways can lead to unsafe outputs when exposed to harmful prompts. Existing safety alignment methods reduce harmful outputs but can degrade reasoning depth, leading to significant trade-offs in complex, multi-step tasks, and remain vulnerabl

Yang P. Liu, Richard Peng, Junzhao Yang

We show an $\widetilde{O}(m^{1.5} \epsilon^{-1})$ time algorithm that on a graph with $m$ edges and $n$ vertices outputs its spanning tree count up to a multiplicative $(1+\epsilon)$ factor with high probability, improving on the previous best runtime of $\widetilde{O}(m + n^{1.875}\epsilon^{-7/4})$ in sparse graphs. While previous algorithms were based on c

Yilin Ye, Junchao Huang, Xingchen Zeng, Jiazhi Xia

Cross-modal embeddings form the foundation for multi-modal models. However, visualization methods for interpreting cross-modal embeddings have been primarily confined to traditional dimensionality reduction (DR) techniques like PCA and t-SNE. These DR methods primarily focus on feature distributions within a single modality, whilst failing to incorporate met

Giovanni Rolandino, Marco Gagliardi, Taian Martins, Giacinto Luigi Cerone

Objective: The purpose of this study was to develop and evaluate the performance of RPC-Net (Recursive Prosthetic Control Network), a novel method using simple neural network architectures to translate electromyographic activity into hand position with high accuracy and computational efficiency. Methods: RPC-Net uses a regression-based approach to convert fo

Zdenek Sekanina

New insights into the history of C/1843 D1 and C/1882 R1, the two celebrated Kreutz sungrazers, are provided by assessing evidence on their appearance at the previous perihelion return, known as X/1106 C1 and the Chinese comet of 1138 (Ho's No. 403), respectively. The conditions differed vastly because of disparities in geocentric distance, solar elongation,

Matthew Russo, Chunwei Liu, Sivaprasad Sudhir, Gerardo Vitagliano

LLMs enable an exciting new class of data processing applications over large collections of unstructured documents. Several new programming frameworks have enabled developers to build these applications by composing them out of semantic operators: a declarative set of AI-powered data transformations with natural language specifications. These include LLM-pow

Ben Cohen, Emaad Khwaja, Youssef Doubli, Salahidine Lemaachi

We introduce Toto, a time series forecasting foundation model with 151 million parameters. Toto uses a modern decoder-only architecture coupled with architectural innovations designed to account for specific challenges found in multivariate observability time series data. Toto's pre-training corpus is a mixture of observability data, open datasets, and synth

Ronald Seoh, Dan Goldwasser

In this paper, we introduce EmoGist, a training-free, in-context learning method for performing visual emotion classification with LVLMs. The key intuition of our approach is that context-dependent definition of emotion labels could allow more accurate predictions of emotions, as the ways in which emotions manifest within images are highly context dependent

Navneet Kaur, Lav Gupta

As healthcare systems increasingly adopt advanced wireless networks and connected devices, securing medical applications has become critical. The integration of Internet of Medical Things devices, such as robotic surgical tools, intensive care systems, and wearable monitors has enhanced patient care but introduced serious security risks. Cyberattacks on thes

Giovanni Rolandino, Chiara Zangrandi, Taian Vieira, Giacinto Luigi Cerone

This paper aims to introduce HDE-Array (High-Density Electrode Array), a novel dry electrode array for acquiring High-Density surface electromyography (HD-sEMG) for hand position estimation through RPC-Net (Recursive Prosthetic Control Network), a neural network defined in a previous study. We aim to demonstrate the hypothesis that the position estimates ret

Karthikeya Sharma Maheswaran, Camille Bossut, Andy Wanna, Qirun Zhang

Cryptographic primitives, consisting of repetitive operations with different inputs, are typically implemented using straight-line C code due to traditional execution on CPUs. Computing these primitives is necessary for secure communication; thus, dedicated hardware accelerators are required in resource and latency-constrained environments. High-Level Synthe

Zihao Zhang, Hui Wei, Kenan Jiang, Shijia Pan

Planning under resource constraints is central to real-world decision making, yet most large language model (LLM) planners assume uniform action costs. We systematically analyze whether tree-search LLM planners are cost-aware and whether they efficiently generate budget-feasible plans. In contrast to black-box prompting, explicit search trees expose intermed

Sabrina Aufiero, Antonio Briola, Tesfaye Salarin, Fabio Caccioli

This paper investigates the evolving link between cryptocurrency and equity markets in the context of the recent wave of corporate Bitcoin (BTC) treasury strategies. We assemble a dataset of 39 publicly listed firms holding BTC, from their first acquisition through April 2025. Using daily logarithmic returns, we first document significant positive co-movemen

Yonatan Gutman, Qiang Huo

Two representations theorems are presented: 1. Any Borel action of a second countable locally compact group $G$ on a standard Borel space $X$ admits an injective $G$-equivariant Borel map into the shift space of $1$-Lipschitz functions from $G$ to the unit interval $Lip_1(G)$. 2. Any continuous action of $\mathbb{R}^k$ ($k\in \mathbb{N}$) on a metrizable com

Xueguang Ma, Qian Liu, Dongfu Jiang, Ge Zhang

Reinforcement learning (RL) has recently demonstrated strong potential in enhancing the reasoning capabilities of large language models (LLMs). Particularly, the "Zero" reinforcement learning introduced by Deepseek-R1-Zero, enables direct RL training of base LLMs without relying on an intermediate supervised fine-tuning stage. Despite these advancements, cur

Sara F. Hartke, Ari Athair, Owen Williams, Jennifer A. Franck

Understanding the intricate dynamics of cross-flow turbines (CFT) is critical to the improvement of performance and optimal control strategies. The current study numerically investigates intracycle control by modulating the angular velocity as a function of blade position for a 2-bladed NACA0018 turbine at a lab-scale chord-based Reynolds number of 45,000. P

Leïla Bessila, Stéphane Mathis

Convection is a fundamental mechanism for energy transport in stars and planets, playing a pivotal role in shaping their structures and evolution. The Mixing-Length Theory, a monomodal approach to convection, is widely adopted and implemented in 1D stellar structure and evolution codes. However, it overlooks the combined effects of rotation and magnetic fiel

N. J. Dubicki, V. V. Slastikov, A. Bernand-Mantel, C. B. Muratov

This paper explores the energy landscape of ferromagnetic multilayer heterostructures that feature magnetic skyrmions -- tiny whirls of spins with non-trivial topology -- in each magnetic layer. Such magnetic heterostructures have been recently pursued as possible hosts of room temperature stable magnetic skyrmions suitable for the next generation of low pow

Tiantian Feng, Jihwan Lee, Anfeng Xu, Yoonjeong Lee

We introduce Vox-Profile, a comprehensive benchmark to characterize rich speaker and speech traits using speech foundation models. Unlike existing works that focus on a single dimension of speaker traits, Vox-Profile provides holistic and multi-dimensional profiles that reflect both static speaker traits (e.g., age, sex, accent) and dynamic speech properties

Orhun Vural, Bunyamin Ozaydin, James Booth, Brittany F. Lindsey

This study presents a deep learning-based framework for predicting emergency department (ED) boarding counts six hours in advance using only operational and contextual data, without patient-level information. Data from ED tracking systems, inpatient census, weather, holidays, and local events were aggregated hourly and processed with comprehensive feature en

Sina Sharifi, Erfan Yazdandoost Hamedani, Mahyar Fazlyab

Bilevel optimization involves a hierarchical structure where one problem is nested within another, leading to complex interdependencies between levels. We propose a single-loop, tuning-free algorithm that guarantees anytime feasibility, i.e., approximate satisfaction of the lower-level optimality condition, while ensuring descent of the upper-level objective

Anna C. Doris, Md Ferdous Alam, Amin Heyrani Nobari, Faez Ahmed

Efficient creation of accurate and editable 3D CAD models is critical in engineering design, significantly impacting cost and time-to-market in product innovation. Current manual workflows remain highly time-consuming and demand extensive user expertise. While recent developments in AI-driven CAD generation show promise, existing models are limited by incomp

Sheng Huang, Cory Hilton, Steve Bush, Faiz Sherman

Radio frequency (RF) fingerprinting is widely used for supporting physical layer security in various wireless applications. In this paper, we present the design and implementation of a small antenna with low-cost fabrication that can be directly integrated with nonlinear passive devices, forming a passive RF tag providing unique nonlinear signatures for RF f

Titos Matsakos, Adrian Lomas

We propose a quantum unstructured search algorithm to find the extrema or roots of discrete functions, $f(\mathbf{x})$, such as the objective functions in combinatorial and other discrete optimisation problems. The first step of the Quantum Search for Extrema and Roots Algorithm (QSERA) is to translate conditions of the form $f(\mathbf{x}_*) \simeq f_*$, whe

Kristen M. Surrao, Shivam Pandey, J. Colin Hill, Eric J. Baxter

The thermal Sunyaev-Zel'dovich (tSZ) effect, the inverse-Compton scattering of cosmic microwave background (CMB) photons off high-energy electrons, is a powerful probe of hot, ionized gas in the Universe. It is often measured via cross-correlations of CMB data with large-scale structure (LSS) tracers to constrain gas physics and improve cosmological constrai

Ane G. Domingo-Aldama, Marcos Merino Prado, Alain García Olea, Koldo Gojenola Galletebeitia

BACKGROUND: Atrial fibrillation (AF), the most common arrhythmia, is linked to high morbidity and mortality. In a fast-evolving AF rhythm control treatment era, predicting AF recurrence after its onset may be crucial to achieve the optimal therapeutic approach, yet traditional scores like CHADS2-VASc, HATCH, and APPLE show limited predictive accuracy. Moreov

Filippo Gazzola, Mikhail V. Korobkov, Xiao Ren, Gianmarco Sperone

The steady motion of a viscous incompressible fluid in a junction of unbounded channels with sources and sinks is modeled through the Navier-Stokes equations under inhomogeneous Dirichlet boundary conditions. In contrast to many previous works, the domain is not assumed to be simply-connected and the fluxes are not assumed to be small. In this very general s

Christopher Housholder, Layna Mangiapanello, Steven Senger

Following recent work on the VC-dimension of subsets of various pseudorandom graphs, we study the VC-dimension of Hamming graphs, which have proved somewhat resistant to the standard techniques in the literature. Our methods are elementary, and agree with or improve upon previously known results. In particular, for $H(2,q)$ we show tight bounds on the size o

Wentao Ma, Weiming Ren, Yiming Jia, Zhuofeng Li

Large multimodal models (LMMs) have recently emerged as a powerful tool for long video understanding (LVU), prompting the development of standardized LVU benchmarks to evaluate their performance. However, our investigation reveals a rather sober lesson for existing LVU benchmarks. First, most existing benchmarks rely heavily on multiple-choice questions (MCQ

Tomer Gafni, Asaf Karnieli, Yair Hanani

Deep neural networks have achieved state-of-the-art results in a wide range of applications, from natural language processing and computer vision to speech recognition. However, as tasks become increasingly complex, model sizes continue to grow, posing challenges in latency and memory efficiency. To meet these constraints, post-training quantization has emer

Ze-Fang Jiang, Xiang Fan, Duan She, Shasha Ye

Using the TRENTo-3D initial condition model coupled with (3+1)-dimensional CLVisc hydrodynamic simulations, we systematically investigate the left-right splitting of elliptic flow ($\Delta v_{2}$) for soft particles in relativistic heavy-ion collisions. Our study reveals that the final distribution characteristics of $\Delta v_{2}$ are primarily depend on th

Jordan M. Adams, Daniel M. Heligman

We show that arbitrary 3D electromagnetic fields are transient solutions to Maxwell's equations and provide a simple equation to find how the field evolves over time. Multiple 3D fields can be realized at different times by superposing with an initial phase. Phase optimization algorithms allow for a phase-only modulated input signal. The necessary input wave

Prabhu Prakash Kagitha, Bo Sun, Ishan Desai, Andrew Zhu

A line of work in planning uses LLM not to generate a plan, but to generate a formal representation in some planning language, which can be input into a symbolic solver to deterministically find a plan. While showing improved trust and promising performance, dozens of recent publications have proposed scattered methods on a variety of benchmarks under differ

Benjamin Prada, Shion Matsumoto, Abdul Malik Zekri, Ankur Mali

We present the first theoretical framework that connects predictive coding (PC), a biologically inspired local learning rule, with the minimum description length (MDL) principle in deep networks. We prove that layerwise PC performs block-coordinate descent on the MDL two-part code objective, thereby jointly minimizing empirical risk and model complexity. Usi

Gokul Bhusal, Yifei Lou, Cristina Garcia-Cardona, Ekaterina Merkurjev

Due to low spatial resolution, hyperspectral data often consists of mixtures of contributions from multiple materials. This limitation motivates the task of hyperspectral unmixing (HU), a fundamental problem in hyperspectral imaging. HU aims to identify the spectral signatures (\textit{endmembers}) of the materials present in an observed scene, along with th

Yu Ying Chiu, Zhilin Wang, Sharan Maiya, Yejin Choi

Detecting AI risks becomes more challenging as stronger models emerge and find novel methods such as Alignment Faking to circumvent these detection attempts. Inspired by how risky behaviors in humans (i.e., illegal activities that may hurt others) are sometimes guided by strongly-held values, we believe that identifying values within AI models can be an earl

Rich Burns, Dante S. Lauretta

OSIRIS-REx, NASA's first asteroid sample return mission, rendezvoused with the near-Earth asteroid Bennu in 2018 and delivered a 121.6-gram sample to Earth in 2023, the largest amount of material ever recovered from a planetary body beyond the Moon. The operations phase of OSIRIS-REx was considered the most challenging robotic mission that NASA had undertake

Lingjie Jiang, Xun Wu, Shaohan Huang, Qingxiu Dong

Recent Large Reasoning Models (LRMs) have shown substantially improved reasoning capabilities over traditional Large Language Models (LLMs) by incorporating extended thinking processes prior to producing final responses. However, excessively lengthy thinking introduces substantial overhead in terms of token consumption and latency, which is particularly unne

Adriano Amaricci, Andrea Richaud, Massimo Capone, Nelson Darkwah Oppong

We propose quantum simulation experiments of the Kondo impurity problem using cold alkaline-earth(-like) atoms (AEAs) in a combination of optical lattice and optical tweezer potentials. Within an ab initio model for atomic interactions in the optical potentials, we analyze hallmark signatures of the Kondo effect in a variety of observables accessible in cold

Fnu Mohbat, Mohammed J Zaki

Recent advances in large language models (LLMs) and the abundance of food data have resulted in studies to improve food understanding using LLMs. Despite several recommendation systems utilizing LLMs and Knowledge Graphs (KGs), there has been limited research on integrating food related KGs with LLMs. We introduce KERL, a unified system that leverages food K

Jacob Ewaniuk, Bhavin J. Shastri, Nir Rotenberg

Large, multi-dimensional clusters of entangled photons are among the most powerful resources for emerging quantum technologies, as they are predicted to enable global quantum networks or universal quantum computation. Here, we propose an entirely new architecture and protocol for their generation based on recurrent quantum photonic neural networks (QPNNs) an

Ashutosh Adhikari, Mirella Lapata

As Large Language Models (LLMs) gain expertise across diverse domains and modalities, scalable oversight becomes increasingly challenging, particularly when their capabilities may surpass human evaluators. Debate has emerged as a promising mechanism for enabling such oversight. In this work, we extend the debate paradigm to a multimodal setting, exploring it

Mazen M. Alhwaimel, Zhenbo Qin

In this paper, we study the Chern character operators on the equivariant cohomology of the Hilbert schemes of points in the complex affine plane $C^2$ with the action of the torus $(C^*)^2$, and partially verify Okounkov's Conjecture [Oko, Conjecture 2] in this setting. Our main idea is to apply the connection between the equivariant cohomology of these Hilb

Jiaxin Zhang

We develop a theory for the multiple radial $\mathrm{SLE}(\kappa)$ systems with parameter $\kappa > 0$ -- a family of random multi-curve systems in a simply connected domain $\Omega$, with marked boundary points $z_1, \ldots, z_n \in \partial \Omega$ and a marked interior point $q$. As a consequence of the domain Markov property and conformal invariance, we

Zhangchen Xu, Yuetai Li, Fengqing Jiang, Bhaskar Ramasubramanian

Reinforcement Learning (RL) has become a powerful tool for enhancing the reasoning abilities of large language models (LLMs) by optimizing their policies with reward signals. Yet, RL's success relies on the reliability of rewards, which are provided by verifiers. In this paper, we expose and analyze a widespread problem--false negatives--where verifiers wron

Kudret Bostanci, Deniz Kus

Maximal parabolic subalgebras of untwisted affine Kac-Moody algebras were studied in the context of Borel-de Siebenthal theory in [13], where they were realized as certain equivariant map algebras with a non-free abelian group action. In this paper, we show that this perspective naturally extends to non-maximal parabolic subalgebras and introduce their quant

Michael Krivelevich, Maksim Zhukovskii

We establish the asymptotic behaviour of $\mu(G(n,p))$, the number of unlabelled induced subgraphs in the binomial random graph $G(n,p)$, for almost the entire range of the probability parameter $p=p(n)\in[0,1]$. In particular, we show that typically the number of subgraphs becomes exponential when $p$ passes $1/n$, reaches maximum possible base of exponent

Andrea Nava, Reinhold Egger

Mpemba effects occur after a sudden quench of control parameters if for ''far'' (or ''hot'') initial states with respect to a final target state, the relaxation time toward the target state is shorter than for ''close'' (or ''cold'') initial states. Following a strategy of fishermen in Pontus described by Aristotle, we introduce the Pontus-Mpemba effect as a

Abhimanyu Talwar, Julien Laasri

We consider the problem of reconstructing a 3D scene from multiple sketches. We propose a pipeline which involves (1) stitching together multiple sketches through use of correspondence points, (2) converting the stitched sketch into a realistic image using a CycleGAN, and (3) estimating that image's depth-map using a pre-trained convolutional neural network

Davit Gondauri, M. Moistsrapishvili

The given paper emphasizes the importance of the Railway Silk Road for promoting Georgia's economic growth and development. The article notes that economic integration in the region increases cargo turnover in Central Asia and the Caucasus, thus boosting the volume of goods transported through Georgia and contributing to the sustainability of Georgia's macro

Morgan Lindsay Heisler, Linzi Xing, Ge Shi, Hanieh Sadri

Huawei Cloud users leverage LoRA (Low-Rank Adaptation) as an efficient and scalable method to fine-tune and customize large language models (LLMs) for application-specific needs. However, tasks that require complex reasoning or deep contextual understanding are often hindered by biases or interference from the base model when using typical decoding methods l

Jiunn-Wei Chen, Xiang Gao, Jinchen He, Jun Hua

Large-Momentum Effective Theory (LaMET) is a physics-guided systematic expansion to calculate light-cone parton distributions, including collinear (PDFs) and transverse-momentum-dependent ones, at any fixed momentum fraction $x$ within a range of $[x_{\rm min}, x_{\rm max}]$. It theoretically solves the ill-posed inverse problem that afflicts other theoretic

KangJin Lee, Jesse S. Wainright, Christopher L. Wirth

Carbon black slurries are a key component in redox flow batteries as the large surface area provided by the particles allows an increase in the battery capacity without facing limitations posed by many solid-state batteries such as safety hazard or cost. However, these conductive slurries often have complex mechanical and electrical responses because of the

Kaibo Huang, Zipei Zhang, Yukun Wei, TianXin Zhang

The ubiquity of social media platforms facilitates malicious linguistic steganography, posing significant security risks. Steganalysis is profoundly hindered by the challenge of identifying subtle cognitive inconsistencies arising from textual fragmentation and complex dialogue structures, and the difficulty in achieving robust aggregation of multi-dimension

Sahar Abdelnabi, Ahmed Salem

Reasoning-focused LLMs sometimes alter their behavior when they detect that they are being evaluated, which can lead them to optimize for test-passing performance or to comply more readily with harmful prompts if real-world consequences appear absent. We present the first quantitative study of how such "test awareness" impacts model behavior, particularly it

Michael Wrana, Uzma Maroof, Diogo Barradas

Website fingerprinting (WF) is a technique that allows an eavesdropper to determine the website a target user is accessing by inspecting the metadata associated with the packets she exchanges via some encrypted tunnel, e.g., Tor. Recent WF attacks built using machine learning (and deep learning) process and summarize trace metadata during their feature extra

Anjiang Wei, Yuheng Wu, Yingjia Wan, Tarun Suresh

We introduce SATBench, a benchmark for evaluating the logical reasoning capabilities of large language models (LLMs) through logical puzzles derived from Boolean satisfiability (SAT) problems. Unlike prior work that focuses on inference rule-based reasoning, which often involves deducing conclusions from a set of premises, our approach leverages the search-b

Zhenbo Qin

Let $(a)_\infty = (a; q)_\infty = \prod_{n=0}^\infty (1-aq^n)$. An elegant result of Bloch and Okounkov [BO] states that if $x = e^z$, then $$ \frac{(xq)_\infty (x^{-1}q)_\infty}{(q)_\infty^2}, $$ which appears in various traces in representation theory and algebraic geometry, is a formal power series in $z^2$ whose coefficient for $z^{2k}$ is a quasi-modula

Emmanuel Noutahi, Jason Hartford, Prudencio Tossou, Shawn Whitfield

Drug discovery is fundamentally a process of inferring the effects of treatments on patients, and would therefore benefit immensely from computational models that can reliably simulate patient responses, enabling researchers to generate and test large numbers of therapeutic hypotheses safely and economically before initiating costly clinical trials. Even a m

Microsoft Copilot, Stephen E. Spear

This paper extends (Spear 2003) by replacing human agents with artificial intelligence (AI) entities that derive utility solely from electricity consumption. These AI agents must prepay for electricity using cryptocurrency and the verification of these transactions requires a fixed amount of electricity. As a result the agents must strategically allocate ele

Franck Florin

This paper proposes representing finite-energy signals observed within a given bandwidth as parameters of a probability distribution and employing the information-geometric framework to compute the Fisher-Rao distance between these signals, considered as distributions.

Hao Wang, Chenyu Shi, Angel E. Rodriguez-Fernandez, Oliver Schütze

Maximum mean discrepancy (MMD) has been widely employed to measure the distance between probability distributions. In this paper, we propose using MMD to solve continuous multi-objective optimization problems (MOPs). For solving MOPs, a common approach is to minimize the distance (e.g., Hausdorff) between a finite approximate set of the Pareto front and a re

Mattia Capuano, Livia Ferro, Tomasz Lukowski, Alessandro Palazio

Cosmological correlation functions are central observables in modern cosmology, as they encode properties of the early universe. In this paper, we derive novel canonical differential equations for wavefunction coefficients in power-law FRW cosmologies by combining positive geometries and the combinatorics of tubings of Feynman graphs. First, we establish a g

Hamideh Khaleghpour, Brett McKinney

This paper presents a unified AI framework for high-accuracy audio anomaly detection by integrating advanced noise reduction, feature extraction, and machine learning modeling techniques. The approach combines spectral subtraction and adaptive filtering to enhance audio quality, followed by feature extraction using traditional methods like MFCCs and deep emb

Soumadeep Saha, Akshay Chaturvedi, Joy Mahapatra, Utpal Garain

User authorization-based access privileges are a key feature in many safety-critical systems, but have not been extensively studied in the large language model (LLM) realm. In this work, drawing inspiration from such access control systems, we introduce sudoLLM, a novel framework that results in multi-role aligned LLMs, i.e., LLMs that account for, and behav

Maksim Zhdanov, Vladislav Kurenkov

In this work, we introduce Phi-Module, a universal plugin module that enforces Poisson's equation within the message-passing framework to learn electrostatic interactions in a self-supervised manner. Specifically, each atom-wise representation is encouraged to satisfy a discretized Poisson's equation, making it possible to acquire a potential {\phi} and corr

Vassili N Kolokoltsov

Quantum filtering equations for mixed states were developed in 80th of the last century. Since then the problem of building a rigorous mathematical theory for these equations in the basic infinite-dimensional settings has been a challenging open mathematical problem. In a previous paper, the author developed the theory of these equations in the case of bound

Haoran Zhao, Yuchen Yan, Yongliang Shen, Haolei Xu

Large reasoning models (LRMs), such as OpenAI o1 and DeepSeek-R1, have significantly enhanced their reasoning capabilities by generating longer chains of thought, demonstrating outstanding performance across a variety of tasks. However, this performance gain comes at the cost of a substantial increase in redundant reasoning during the generation process, lea

Davide Buffelli, Sowmen Das, Yu-Wei Lin, Sattar Vakili

Artificial Intelligence (AI) has demonstrated unprecedented performance across various domains, and its application to communication systems is an active area of research. While current methods focus on task-specific solutions, the broader trend in AI is shifting toward large general models capable of supporting multiple applications. In this work, we take a

Michael Mihalik

The question of whether or not all finitely presented groups are semistable at infinity has been studied for over 40 years. In 1986, we defined what it means for a finitely generated group to be semistable at infinity - in analogy with the definition for finitely presented groups. At that time we suggest that the Lamplighter group may not be semistable at in

Yang Xiao, Rohan Kumar Das

As deepfake speech becomes common and hard to detect, it is vital to trace its source. Recent work on audio deepfake source tracing (ST) aims to find the origins of synthetic or manipulated speech. However, ST models must adapt to learn new deepfake attacks while retaining knowledge of the previous ones. A major challenge is catastrophic forgetting, where mo

Yang Xiao, Tianyi Peng, Yanghao Zhou, Rohan Kumar Das

Spoken keyword spotting (KWS) aims to identify keywords in audio for wide applications, especially on edge devices. Current small-footprint KWS systems focus on efficient model designs. However, their inference performance can decline in unseen environments or noisy backgrounds. Test-time adaptation (TTA) helps models adapt to test samples without needing th

Guangzhi Xiong, Eric Xie, Corey Williams, Myles Kim

Large language models (LLMs) have shown significant potential in scientific disciplines such as biomedicine, particularly in hypothesis generation, where they can analyze vast literature, identify patterns, and suggest research directions. However, a key challenge lies in evaluating the truthfulness of generated hypotheses, as verifying their accuracy often

Sushil Pandit

In this note, we consider certain logharmonic mappings in the unit disk $\mathbb{D}=\{z\in\mathbb{C}:|z|<1\}.$ Next, we obtain sharp bound of pre-Schwarzian norm of such logharmonic mappings in the unit disk. Then we discuss growth theorem for the mappings. Moreover, we discuss starlikeness of logharmonic mappings and compute sufficient coefficient condition

Xianzhen Luo, Qingfu Zhu, Zhiming Zhang, Mingzheng Xu

Code Sensitivity refers to the ability of Code LLMs to recognize and respond to details changes in problem descriptions. While current code benchmarks and instruction data focus on difficulty and diversity, sensitivity is overlooked. We first introduce the CTF-Code benchmark, constructed using counterfactual perturbations, minimizing input changes while maxi

Isabella Degen, Zahraa S Abdallah, Henry W J Reeve, Kate Robson Brown

Time series clustering promises to uncover hidden structural patterns in data with applications across healthcare, finance, industrial systems, and other critical domains. However, without validated ground truth information, researchers cannot objectively assess clustering quality or determine whether poor results stem from absent structures in the data, alg

Nima Hosseini Dashtbayaz, Hesam Salehipour, Adrian Butscher, Nigel Morris

Reduced-order modeling (ROM) of time-dependent and parameterized differential equations aims to accelerate the simulation of complex high-dimensional systems by learning a compact latent manifold representation that captures the characteristics of the solution fields and their time-dependent dynamics. Although high-fidelity numerical solvers generate the tra

Nicolas Kainz, Dirk Lebiedz

We investigate properties of boundary orbits (separatrices) of canonical regions (basins/neighbourhoods of equilibria) in holomorphic flows with real-valued time. We establish the continuity of transit times along these boundary orbits and classify possible path components of the boundary of flow-invariant domains. Thus, we provide central tools for topologi

Francesco D'Amore, Luca Mariani, Carlo Mastroianni, Francesco Plastina

The use of quantum computing for machine learning is among the most promising applications of quantum technologies. Quantum models inspired by classical algorithms are developed to explore some possible advantages over classical approaches. A primary challenge in the development and testing of Quantum Machine Learning (QML) algorithms is the scarcity of data

Alexandre Broggi, Nathaniel Bastian, Lance Fiondella, Gokhan Kul

Artificial neural network pruning is a method in which artificial neural network sizes can be reduced while attempting to preserve the predicting capabilities of the network. This is done to make the model smaller or faster during inference time. In this work we analyze the ability of a selection of artificial neural network pruning methods to generalize to

Md Asaduzzaman, Ryan M. L. McFadden, Edward Thoeng, Yasmine Kalboussi

We report measurements of the normal-state and superconducting properties of thin-film $\mathrm{Nb_{1-x}Ti_xN}$ using $^{8}$Li $\beta$-detected nuclear magnetic resonance ($\beta$-NMR). In these experiments, radioactive $^{8}$Li$^{+}$ probes were implanted $\sim21$ nm below the surface of a $\mathrm{Nb_{1-x}Ti_xN}$(91 nm) film in $\mathrm{Nb_{0.75}Ti_{0.25}N

Huihao Jing, Haoran Li, Wenbin Hu, Qi Hu

As Model Context Protocol (MCP) introduces an easy-to-use ecosystem for users and developers, it also brings underexplored safety risks. Its decentralized architecture, which separates clients and servers, poses unique challenges for systematic safety analysis. This paper proposes a novel framework to enhance MCP safety. Guided by the MAESTRO framework, we f

Yaroslav Marchukov, Luis Montano

In this paper we develop a method to coordinate the deployment of a multi-robot team to reach some locations of interest, so-called primary goals, and to transmit the information from these positions to a static Base Station (BS), under connectivity constraints. The relay positions have to be established for some robots to maintain the connectivity at the mo

Himanshu Bangar, Polychronis Tsipas, Prasanna Rout, Lalit Pandey

Altermagnets are a new class of magnetic materials characterized by fully compensated spins arranged in alternating local structures, allowing for spin-split bands similar to those found in ferromagnets without net magnetism. Recently, MnTe has emerged as a prototypical altermagnetic material exhibiting spin-polarized electronic bands and anomalous transport

Martin Baily, David Byrne, Aidan Kane, Paul Soto

With the advent of generative AI (genAI), the potential scope of artificial intelligence has increased dramatically, but the future effect of genAI on productivity remains uncertain. The effect of the technology on the innovation process is a crucial open question. Some inventions, such as the light bulb, temporarily raise productivity growth as adoption spr