May 2025 arXiv papers — page 17

Giulio Malavolta, Tamer Mour

A proof of quantumness (PoQ) allows a classical verifier to efficiently test if a quantum machine is performing a computation that is infeasible for any classical machine. In this work, we propose a new approach for constructing PoQ protocols where soundness holds unconditionally assuming a bound on the memory of the prover, but otherwise no restrictions on

Yichen Feng, Zhangchen Xu, Fengqing Jiang, Yuetai Li

Vision language models (VLMs) are expected to perform effective multimodal reasoning and make logically coherent decisions, which is critical to tasks such as diagram understanding and spatial problem solving. However, current VLM reasoning lacks large-scale and well-structured training datasets. To bridge this gap, we propose VisualSphinx, a first-of-its-ki

Luiz Filipi Anderson de Sousa Moura, Carlos Becker Westphall

Quantum computers impose an immense threat to system security. As a countermeasure, new cryptographic classes have been created to prevent these attacks. Technologies such as post-quantum cryptography and quantum cryptography. Quantum cryptography uses the principle of quantum physics to produce theoretically unbreakable security. This tertiary review select

Davi Lopes Medeiros, Euripedes Carvalho da Silva, Emanoel Souza

The abnormal surfaces called snakes and circular snakes, defined in \cite{GabrielovSouza}, are special types of surface germs capturing the outer Lipschitz phenomena relevant to the outer classification problem. We provide algorithms to obtain a minimal pancake decomposition, i.e., where the number of pancakes is minimal, for snakes and circular snakes. We c

A. Ustrzycka, H. Mousavi, F. J. Dominguez-Gutierrez, S. Stupkiewicz

Neutron irradiation in structural alloys promotes defect clustering, which suppresses plasticity and triggers a ductile-to-brittle transition (DBT), a key degradation mechanism limiting fracture resistance in nuclear materials. This study investigates the fracture mechanisms underlying this transition in irradiated Fe-Ni-Cr alloys. Using Molecular Dynamics s

Digvijay Singh, Rahul Shukla, Karunesh Kumar Singh

In this work, wavelet-based filtering operators are constructed by introducing a basic function $D(t_1, t_2, t_3)$ using a general wavelet transform. The cardinal orthogonal scaling functions (COSF) provide an idea to derive the standard sampling theorem in multiresolution spaces which motivates us to study wavelet approximation analysis. With the help of mo

Asaf Goren, Natalie Lang, Nir Shlezinger, Alejandro Cohen

Federated learning (FL) enables collaborative model training across distributed edge devices while preserving data privacy, and typically operates in a round-based synchronous manner. However, synchronous FL suffers from latency bottlenecks due to device heterogeneity, where slower clients (stragglers) delay or degrade global updates. Prior solutions, such a

Michael A. Högele, Torsten Wetzel

Let $L = (L(t))_{t\geq 0}$ be a multivariate L\'evy process with L\'evy measure $\nu(dy) = \exp(-f(|y|)) dy$ for a smoothly regularly varying function $f$ of index $\alpha>1$. The process $L$ is renormalized as $X^\varepsilon(t) = \varepsilon L(r_\varepsilon t)$, $t\in [0, T]$, for a scaling parameter $r_\varepsilon= o(\varepsilon^{-1})$, as $\varepsilon \to

William Merrill, Shane Arora, Dirk Groeneveld, Hannaneh Hajishirzi

The right batch size is important when training language models at scale: a large batch size is necessary for fast training, but a batch size that is too large will harm token efficiency. To navigate this tradeoff, McCandlish et al. (2018) suggest that a critical batch size (CBS), below which training will not substantially degrade loss, can be estimated bas

Yuyang Tian, Desen Sun, Yi Ding, Sihang Liu

As large language models (LLMs) become widely used, their environmental impact, especially carbon emission, has attracted more attention. Prior studies focus on compute-related carbon emissions. In this paper, we find that storage is another key contributor. LLM caching, which saves and reuses KV caches for repeated context, reduces operational carbon by avo

Otman Benchekroun, Eitan Grinspun, Maurizio Chiaramonte, Philip Allen Etter

Designing subspaces for Reduced Order Modeling (ROM) is crucial for accelerating finite element simulations in graphics and engineering. Unfortunately, it's not always clear which subspace is optimal for arbitrary dynamic simulation. We propose to construct simulation subspaces from force distributions, allowing us to tailor such subspaces to common scene in

Stephan Rabanser, Ali Shahin Shamsabadi, Olive Franzese, Xiao Wang

Cautious predictions -- where a machine learning model abstains when uncertain -- are crucial for limiting harmful errors in safety-critical applications. In this work, we identify a novel threat: a dishonest institution can exploit these mechanisms to discriminate or unjustly deny services under the guise of uncertainty. We demonstrate the practicality of t

Anders Aamand, Justin Y. Chen, Siddharth Gollapudi, Sandeep Silwal

We design improved approximation algorithms for NP-hard graph problems by incorporating predictions (e.g., learned from past data). Our prediction model builds upon and extends the $\varepsilon$-prediction framework by Cohen-Addad, d'Orsi, Gupta, Lee, and Panigrahi (NeurIPS 2024). We consider an edge-based version of this model, where each edge provides two

Shriphani Palakodety

Many identity systems assign a single, static identifier to an individual for life, reused across domains like healthcare, finance, and education. These Universal Lifelong Identifiers (ULIs) underpin critical workflows but now pose systemic privacy risks. We take the position that ULIs are fundamentally incompatible with the AI era and must be phased out. We

Jiayi Tian, Ryan Solgi, Jinming Lu, Yifan Yang

Large Language Models (LLMs) have enabled remarkable progress in natural language processing, yet their high computational and memory demands pose challenges for deployment in resource-constrained environments. Although recent low-rank decomposition methods offer a promising path for structural compression, they often suffer from accuracy degradation, expens

Candida Moffa, Daniele Francescone, Alessandro Curcio, Anna Candida Felici

Terahertz radiation enables non destructive, depthresolved analysis of layered artworks. This study demonstrates THz multispectral imaging ability to reveal concealed text beneath mockup of pictorial layers, reconstructing hidden narratives at varying depths through frequency domain analysis.Simultaneously, it maps pigment composition, providing valuable che

Jonas Elsborg, Tejs Vegge, Arghya Bhowmik

Reliably monitoring and recognizing maritime vessels based on acoustic signatures is complicated by the variability of different recording scenarios. A robust classification framework must be able to generalize across diverse acoustic environments and variable source-sensor distances. To this end, we present a deep learning model with robust performance acro

Tao E. Li

Understanding condensed-phase polariton experiments requires accurately accounting for both realistic cavity geometries and the interplay between polaritons and material dark modes arising from microscopic molecular interactions. The finite-difference time-domain (FDTD) approach numerically propagates classical Maxwell's equations in the time domain, offerin

Aurosweta Mahapatra, Ismail Rasim Ulgen, Abinay Reddy Naini, Carlos Busso

Traditional anti-spoofing focuses on models and datasets built on synthetic speech with mostly neutral state, neglecting diverse emotional variations. As a result, their robustness against high-quality, emotionally expressive synthetic speech is uncertain. We address this by introducing EmoSpoof-TTS, a corpus of emotional text-to-speech samples. Our analysis

Rafi Hassan Chowdhury, Sabbir Ahmed

Ensuring food safety is critical due to its profound impact on public health, economic stability, and global supply chains. Cultivation of Mango, a major agricultural product in several South Asian countries, faces high financial losses due to different diseases, affecting various aspects of the entire supply chain. While deep learning-based methods have bee

Henry Conklin

Despite the remarkable success of large large-scale neural networks, we still lack unified notation for thinking about and describing their representational spaces. We lack methods to reliably describe how their representations are structured, how that structure emerges over training, and what kinds of structures are desirable. This thesis introduces quantit

Prathamesh Joshi, Wanting Niu, Chad Hanna, Rachael Huxford

Searches for gravitational waves from compact binary coalescences employ a process called matched filtering, in which gravitational wave strain data is cross-correlated against a bank of waveform templates. Data from every observing run of the LIGO, Virgo, and KAGRA collaboration is typically analyzed in this way twice, first in a low-latency mode in which g

V. Marels, V. Mesa, M. Jaque Arancibia, S. Alonso

Bars are considered an efficient mechanism for transporting gas toward the central regions of galaxies, potentially enhancing nuclear activity. However, the extent to which bars influence active galactic nuclei (AGNs), and whether their efficiency varies with environment, remain open questions. In this study, we aim to quantify the role of bars in triggering

Alexander Duncan, Shreya Sharma

The Amitsur subgroup of a variety with a group action measures the failure of the action to lift to the total spaces of its line bundles. We introduce the "numerical Amitsur group," which is an approximation of the ordinary Amitsur subgroup that can be computed using only the Euler-Poincar\'e characteristic on the Picard group. As an application, we find a u

Candida Moffa, Alessandro Curcio, Camilla Merola, Vittoria Maria Orsini

In this work, we propose a prototype set-up exploiting terahertz time-domain spectroscopy (THz-TDS) to investigate gaseous compounds. The system is portable and allows to perform remote measurements. We used the prototype to characterise for the first time in literature over a broad THz range, pure dichloromethane and chloroform, two pollutants known as very

CUORE Collaboration, D. Q. Adams, C. Alduino, K. Alfonso

We present the analysis techniques developed to explore the keV-scale energy region of the CUORE experiment, based on more than 2 tonne yr of data collected over 5 years. By prioritizing a stricter selection over a larger exposure, we are able to optimize data selection for thresholds at 10 keV and 3 keV with 691 kg yr and 11 kg yr of data, respectively. We

Dylan Zapzalka, Trenton Chang, Lindsay Warrenburg, Sae-Hwan Park

In settings where ML models are used to inform the allocation of resources, agents affected by the allocation decisions might have an incentive to strategically change their features to secure better outcomes. While prior work has studied strategic responses broadly, disentangling misreporting from genuine modification remains a fundamental challenge. In thi

Melika Sepidband, Hamed Taherkhani, Song Wang, Hadi Hemmati

Automatic code generation has gained significant momentum with the advent of Large Language Models (LLMs) such as GPT-4. Although many studies focus on improving the effectiveness of LLMs for code generation, very limited work tries to understand the generated code's characteristics and leverage that to improve failed cases. In this paper, as the most straig

Adriano Fragomeni, Dima Damen, Michael Wray

Text-to-Video (T2V) retrieval aims to identify the most relevant item from a gallery of videos based on a user's text query. Traditional methods rely solely on aligning video and text modalities to compute the similarity and retrieve relevant items. However, recent advancements emphasise incorporating auxiliary information extracted from video and text modal

B. Manjarrez-Montañez, R. A. Méndez-Sánchez, Y. Betancur-Ocampo, A. Martínez-Argüello

A methodology for constructing a quasi-one-dimensional coupled-resonator phononic metamaterial is presented. This is achieved through the design of artificial phononic analogs of two molecular structures: trans-polyacetylene and trans-polyaminoborane. The band structure of trans-polyacetylene is analyzed in relation to the Su-Schrieffer-Heeger (SSH) model, w

Boyuan Chen, Donghai Hong, Jiaming Ji, Jiacheng Zheng

As multimodal large models (MLLMs) continue to advance across challenging tasks, a key question emerges: What essential capabilities are still missing? A critical aspect of human learning is continuous interaction with the environment -- not limited to language, but also involving multimodal understanding and generation. To move closer to human-level intelli

Xiang Meng, Mehdi Makni, Rahul Mazumder

Network pruning reduces the computational requirements of large neural networks, with N:M sparsity -- retaining only N out of every M consecutive weights -- offering a compelling balance between compressed model quality and hardware acceleration. However, N:M sparsity only accelerates forward-pass computations, as N:M patterns are not preserved during matrix

Shai M. Chester, Rishi Mouland, Jesse van Muiden

Extremal cubic couplings in AdS relate bulk fields such that $\Delta_i+\Delta_j=\Delta_k$. Such couplings lead to divergent 3-point Witten diagrams, and do not occur in theories with maximal supersymmetry. We consider the simplest theories where such coupling are non-zero, which is type IIB string theory with $N$ D3 branes probing various configurations of s

Samuel Müller, Arik Reuter, Noah Hollmann, David Rügamer

Training neural networks on randomly generated artificial datasets yields Bayesian models that capture the prior defined by the dataset-generating distribution. Prior-data Fitted Networks (PFNs) are a class of methods designed to leverage this insight. In an era of rapidly increasing computational resources for pre-training and a near stagnation in the gener

Yuanzhe Liu, Ryan Deng, Tim Kaler, Xuhao Chen

Recent studies show that LLMs possess different skills and specialize in different tasks. In fact, we observe that their varied performance occur in several levels of granularity. For example, in the code optimization task, code LLMs excel at different optimization categories and no one dominates others. This observation prompts the question of how one lever

Sriram Balasubramanian, Samyadeep Basu, Soheil Feizi

Chain-of-thought (CoT) reasoning enhances performance of large language models, but questions remain about whether these reasoning traces faithfully reflect the internal processes of the model. We present the first comprehensive study of CoT faithfulness in large vision-language models (LVLMs), investigating how both text-based and previously unexplored imag

Thushara Manjari Naduvilakandy, Hyeju Jang, Mohammad Al Hasan

Causality detection and mining are important tasks in information retrieval due to their enormous use in information extraction, and knowledge graph construction. To solve these tasks, in existing literature there exist several solutions -- both unsupervised and supervised. However, the unsupervised methods suffer from poor performance and they often require

Sho Shibata, Andre Izidoro

The size distribution of planets with radii between 1 and $4 R_\oplus$ peaks near 1.4 and $2.2R_\oplus$, with a dip around $1.8 R_\oplus$ -- the so-called "radius valley." Recent statistical analyses suggest that planets within this valley ($1.5 < R < 2R_\oplus$) tend to have slightly higher orbital eccentricities than those outside it. The origin of this dy

Gaurav Sarkar, Jay Gala, Subarna Tripathi

The design of activation functions remains a pivotal component in optimizing deep neural networks. While prevailing choices like Swish and GELU demonstrate considerable efficacy, they often exhibit domain-specific optima. This work introduces SG-Blend, a novel activation function that blends our proposed SSwish, a first-order symmetric variant of Swish and t

An Vo, Khai-Nguyen Nguyen, Mohammad Reza Taesiri, Vy Tuong Dang

Large language models (LLMs) memorize a vast amount of prior knowledge from the Internet that helps them on downstream tasks but also may notoriously sway their outputs towards wrong or biased answers. In this work, we test how the knowledge about popular subjects hurt the accuracy of vision language models (VLMs) on standard, objective visual tasks of count

Xiantao Fan, Xinyang Liu, Meng Wang, Jian-Xun Wang

Turbulent flows and fluid-structure interactions (FSI) are ubiquitous in scientific and engineering applications, but their accurate and efficient simulation remains a major challenge due to strong nonlinearities, multiscale interactions, and high computational demands. Traditional CFD solvers, though effective, struggle with scalability and adaptability for

Malithi Wanniarachchi Kankanamge, Nick McKenna, Santiago Carmona, Syed Mhamudul Hasan

The ChatGPT Windows application offers better user interaction in the Windows operating system (OS) by enhancing productivity and streamlining the workflow of ChatGPT's utilization. However, there are potential misuses associated with this application that require rigorous forensic analysis. This study presents a holistic forensic analysis of the ChatGPT Win

Adler Marques, Luciane Quoos

Since Massey introduced linear complementary dual (LCD) codes in 1992 and Bhasin et al. later formalized linear complementary pairs (LCPs) of codes - structures with important cryptographic applications - these code families have attracted significant interest. We construct infinite sequences $(C_i)_{i \geq 1}$ of LCD codes and of LCPs $(C', D')_{i \geq 1}$

Keefer Rowan

We construct a smooth velocity field $u$ on $\mathbb{R}_+ \times \mathbb{T}^3$ that exhibits kinematic dynamo action, causing exponential growth in solutions to the magnetohydrodynamic induction equation, with a rate that is uniform in diffusivity, for suitable sequences of diffusivity $\kappa_j \to 0.$ We call this a subsequentially fast dynamo, giving dyna

Natasha Cowley, Sarah Woolner, Oliver E. Jensen

We use a three-dimensional formulation of the cell vertex model to describe the mechanical properties of a confluent planar monolayer of prismatic cells. Treating cell height as a degree of freedom, we reduce the model to a two-dimensional form. We show how bulk effects, associated with cell volume and total surface area, lead to coupling between energy vari

Thiago Bomfim, Victor Carneiro

It is known that all uniformly expanding dynamics $f: M \rightarrow M$ have no phase transition with respect to a Hölder continuous potential $ϕ: M \rightarrow \mathbb{R}$, in other words, the topological pressure function $\mathbb{R} \ni t \mapsto P_{top}(f , tϕ)$ is analytical. Moreover, the associated transfer operator $\mathcal{L}_{f , tϕ}$, acting on th

Galen Pogoncheff, Michael Beyeler

Human-aligned deep learning models exhibit behaviors consistent with human values, such as robustness, fairness, and honesty. Transferring these behavioral properties to models trained on different tasks or data distributions remains challenging: aligned behavior is easily forgotten during fine-tuning, and collecting task-specific data that preserves this be

Wendong Xu, Jing Xiong, Chenyang Zhao, Qiujiang Chen

We present SwingArena, a competitive evaluation framework for Large Language Models (LLMs) that closely mirrors real-world software development workflows. Unlike traditional static benchmarks, SwingArena models the collaborative process of software iteration by pairing LLMs as submitters, who generate patches, and reviewers, who create test cases and verify

Daniel Wurgaft, Ben Prystawski, Kanishk Gandhi, Cedegao E. Zhang

The think-aloud method, where participants voice their thoughts as they solve a task, is a valuable source of rich data about human reasoning processes. Yet, it has declined in popularity in contemporary cognitive science, largely because labor-intensive transcription and annotation preclude large sample sizes. Here, we develop methods to automate the transc

Naomi Omeonga wa Kayembe

This study introduces a novel conceptual framework distinguishing problem-seeking from problem-solving to clarify the unique features of human intelligence in contrast to AI. Problem-seeking refers to the embodied, emotionally grounded process by which humans identify and set goals, while problem-solving denotes the execution of strategies aimed at achieving

A. Albert, S. Alves, M. André, M. Ardid

This study presents a novel search for magnetic monopoles using data collected over a 14 year period (2008-2022) by the ANTARES neutrino telescope. The interaction of magnetic monopoles with matter was modeled according to Kazama, Yang, and Goldhaber cross-section. Upper limits on the flux of magnetic monopoles are obtained for velocities both above and belo

Christof Schmidhuber

The critical dynamics of conformal field theories on random surfaces is investigated beyond the previously studied dynamics of the overall area and the genus. It is found that the evolution of the order parameter in physical time performs a generalization of the multifractal random walk. Accordingly, the higher moments of time variations of the order paramet

Songtao Feng, Jie Fu

Reinforcement learning from human feedback (RLHF) has achieved great empirical success in aligning large language models (LLMs) with human preference, and it is of great importance to study the statistical efficiency of RLHF algorithms from a theoretical perspective. In this work, we consider the online RLHF setting where the preference data is revealed duri

Xuweiyi Chen, Wentao Zhou, Aruni RoyChowdhury, Zezhou Cheng

While massively scaling both data and models have become central in NLP and 2D vision, their benefits for 3D point cloud understanding remain limited. We study the initial step of scaling 3D point cloud understanding under a realistic regime: large-scale multi-dataset joint training for 3D semantic segmentation, with no dataset labels available at training o

Sihyung Park, Leonard A. Stefanski

$\ell_p$-norm penalization, notably the Lasso, has become a standard technique, extending shrinkage regression to subset selection. Despite aiming for oracle properties and consistent estimation, existing Lasso-derived methods still rely on shrinkage toward a null model, necessitating careful tuning parameter selection and yielding monotone variable selectio

Feiteng Fang, Dingwei Chen, Xiang Huang, Ting-En Lin

Currently, most reinforcement learning tasks focus on domains like mathematics and programming, where verification is relatively straightforward. However, in subjective tasks such as role-playing, alignment techniques struggle to make progress, primarily because subjective reward modeling using the Bradley-Terry model faces significant challenges when dealin

David Ma, Huaqing Yuan, Xingjian Wang, Qianbo Zang

Although long-video understanding demands that models capture hierarchical temporal information -- from clip (seconds) and shot (tens of seconds) to event (minutes) and story (hours) -- existing benchmarks either neglect this multi-scale design or scatter scale-specific questions across different videos, preventing direct comparison of model performance acro

Víctor J. Martínez-Lahuerta, Jan-Niclas Kirsten-Siemß, Klemens Hammerer, Naceur Gaaloul

Bragg Diffraction of matter waves is an established technique used in the most accurate quantum sensors. It is also the method of choice to operate large-momentum-transfer, high-sensitivity atom interferometers. It suffers, however, from an intrinsic multi-path character. Optimal control theory (OCT) has recently led to an improved robustness of atom interfe

Theodoros Depastas, Aldo Bonasera, Joe Natowitz

Mid-weight $\alpha$-conjugate nuclei are predicted to possess exotic toroid-like resonances with high angular momenta. The search for these states in $^{28}$Si$^*$ is the main point of two published experimental investigations of the peripheral $^{28}$Si + $^{12}$C reaction by Cao and collaborators and by Hannaman and collaborators. In this work, we develop

Ashfaq Ahmad, Kamran Ahmad, Shoaib Ahmad Khan

A feasibility study for the Standard Model Higgs boson produced in association with a single top-quark or a top-quark pair and a photon (\tHa\, and \ttHa) is presented, using simulated pp collision data corresponding to an integrated luminosity of 350 fb$^{-1}$ at $\sqrt{s}=13.6$ TeV. This study was conducted using simulated data generated with MadGraph. The

Zihao Zuo, Gabriele Cugno, Joseph Michail, Elena Gallo

We present new observations of the black hole X-ray binary A0620-00 using the Mid-Infrared Instrument on the James Webb Space Telescope, during a state where the X-ray luminosity is 9 orders of magnitude below Eddington, and coordinated with radio, near-infrared and optical observations. The goal is to understand the nature of the excess mid-infrared (MIR) e

Neehar Kondapaneni, Oisin Mac Aodha, Pietro Perona

We propose a method for discovering and visualizing the differences between two learned representations, enabling more direct and interpretable model comparisons. We validate our method, which we call Representational Differences Explanations (RDX), by using it to compare models with known conceptual differences and demonstrate that it recovers meaningful di

Charles Bricout, Samira Ebrahimi Kahou, Sylvain Bouix

Motion-related artifacts are inevitable in Magnetic Resonance Imaging (MRI) and can bias automated neuroanatomical metrics such as cortical thickness. These biases can interfere with statistical analysis which is a major concern as motion has been shown to be more prominent in certain populations such as children or individuals with ADHD. Manual review canno

Patrick Lierle, Carl Schmidt

The spatial distribution and linewidth of Mercury's sodium and potassium exosphere were observed using a combination of long-slit and high-resolution point spectroscopy. Effective temperatures were estimated from emission line profiles by forward modeling their Doppler broadening. These serve as an energy metric for collisionless gas that is inherently nonth

Yuxin Wang, Botao Yu, Ivory Yang, Saeed Hassanpour

Large Language Models are widely used for content moderation but often present certain over-sensitivity, leading to misclassification of benign content and rejecting safe user commands. While previous research attributes this issue primarily to the presence of explicit offensive triggers, we statistically reveal a deeper connection beyond token level: When b

Gustavo Sutter Pessurno de Carvalho, Mohammed Abdulrahman, Hao Wang, Sriram Ganapathi Subramanian

The optimization of expensive black-box functions is ubiquitous in science and engineering. A common solution to this problem is Bayesian optimization (BO), which is generally comprised of two components: (i) a surrogate model and (ii) an acquisition function, which generally require expensive re-training and optimization steps at each iteration, respectivel

Pavel Tikhonov, Ivan Oseledets, Elena Tutubalina

In-context learning (ICL) enables Large Language Models (LLMs) to adapt to new tasks using few examples, with task vectors - specific hidden state activations - hypothesized to encode task information. Existing studies are limited by small-scale benchmarks, restricting comprehensive analysis. We introduce QuiteAFew, a novel dataset of 3,096 diverse few-shot

David A. Zarate-Herrada, Isaías Vallejo-Fabila, Lea F. Santos, E. Jonathan Torres-Herrera

Fractal dimensions are tools for probing the structure of quantum states and identifying whether they are localized or delocalized in a given basis. These quantities are commonly extracted through finite-size scaling, which limits the analysis to relatively small system sizes. In this work, we demonstrate that the correlation fractal dimension $D_2$ can be d

Pureum Noh, Hyunggeun Lee, Myung Joon Han, Eun-Gook Moon

The $KJ\Gamma\Gamma'$ spin model-originally derived for an ideal $P\bar{3}1m$ symmetric geometry-has long served as a central framework for understanding candidate Kitaev materials. In realistic crystals, however, this ideal geometry is seldom realized, either at low temperatures or under external perturbations, limiting the model's quantitative applicabilit

Winstead Zhu, Ann Clifton, Azin Ghazimatin, Edgar Tanaka

Discovering and evaluating long-form talk content such as videos and podcasts poses a significant challenge for users, as it requires a considerable time investment. Previews offer a practical solution by providing concise snippets that showcase key moments of the content, enabling users to make more informed and confident choices. We propose an LLM-based ap

Amirhossein Alimohammadi, Aryan Mikaeili, Sauradip Nag, Negar Hassanpour

Image editing is an important task in computer graphics, vision, and VFX, with recent diffusion-based methods achieving fast and high-quality results. However, edits requiring significant structural changes, such as non-rigid deformations, object modifications, or content generation, remain challenging. Existing few step editing approaches produce artifacts

Yu-Jen Chiu, Eric M. Weiner, Ahmad K. Omar

The linear laws of transport phenomena are central in our description of irreversible processes in systems across the physical sciences. Linear irreversible thermodynamics allows for the identification of the underlying forces driving transport and the structure of the relevant transport coefficients for systems that are locally in equilibrium. Increasingly,

Shashank Dattathri, Frank C. van den Bosch, Martin D. Weinberg, Uddipan Banik

We explore the stability of isotropic, spherical, self-gravitating systems with a double-power law density profile. Systems with rapid transitions between the inner and outer slopes are shown to have an inflection in their isotropic distribution function (DF), where ${\rm d} f/{\rm d} E > 0$, thereby violating Antonov's stability criterion. Using high-resolu

Jun-Young Lee, Francisco Villaescusa-Navarro

The standard cosmological model with cold dark matter posits a hierarchical formation of structures. We introduce topological neural networks (TNNs), implemented as message-passing neural networks on higher-order structures, to effectively capture the topological information inherent in these hierarchies that traditional graph neural networks (GNNs) fail to

Elizaveta Safonova, Aleksey Lunkin, Mikhail Feigel&#39; man

We studied global density-of-states correlation function $R(ω)$ for Lévy-Rosenzweig-Porter random matrix ensemble in the non-ergodic extended phase. Using an extension of Efetov&#39;s supersymmetry approach we calculated $R(ω)$ exactly in all relevant ranges of $ω$. At relatively low $ω\leq Γ$\, (with $Γ\gg Δ$ being the effective miniband width) we found GUE

Jeonghoon Lim, Stanley A. Baronett, Jacob B. Simon, Chao-Chin Yang

The streaming instability (SI), driven by aerodynamic coupling between solids and the gas under a global radial pressure gradient, concentrates solids and facilitates planetesimal formation. Unstratified simulations are commonly used to study the SI, based on the assumption that they approximate conditions near the disk midplane. However, it remains unclear

David Fernández-Fernández, Johannes C. Bayer, Rolf J. Haug, Gloria Platero

We theoretically investigate long-range coherent charge transport in linear quadruple quantum dot (QQD) arrays under reduced symmetry configurations. Employing a master equation approach, we identify precise resonant conditions that enable minimal occupation of intermediate dots, thereby facilitating long-range transfer between distant sites. Our results hig

Lasha Berezhiani, Giordano Cintia, Valerio De Luca, Justin Khoury

The superfluid dark matter model offers an elegant solution to reconcile discrepancies between the predictions of the cold dark matter paradigm and observations on galactic scales. In this scenario, dark matter is composed of ultralight bosons with self-interactions that can undergo a superfluid phase transition in galactic environments. In this review, we e

Henry Whitehead, Connar Rowan, Bence Kocsis

We investigate the evolution of black holes on orbits with small inclinations ($i < 2^\circ$) to the gaseous discs of active galactic nuclei. We perform 3D adiabatic hydrodynamic simulations within a shearing frame, studying the damping of inclination by black hole-gas gravitation. We find that for objects with $i<3H_0R_0^{-1}$, where $H_0R_0^{-1}$ is the di

Philipp Grete, Evan Scannapieco, Marcus Brüggen, Liubin Pan

Supersonic turbulence plays a critical role in shaping astrophysical systems, from molecular clouds to the circumgalactic medium. Key properties of this turbulence include the Mach number, driving scale, and nature of the driving mechanism, which can be solenoidal (divergence-free), compressive (curl-free), or a mix of the two. A less studied property is the

Tyler Holland-Ashford, Patrick Slane, Brian Williams

In this work, as a follow-up to our similar analysis of Kepler's supernova remnant (SNR), we estimate total mass ratios of various ejecta elements in Tycho's SNR using Suzaku X-ray data. In our spectral analysis, we account for uncertainties arising from Suzaku's effective area calibration (5%-15%) and from the unknown filling factors of the various plasma c

Patricio Colazo, Nelson Padilla, Federico Stasyszyn

This paper explores the impact of primordial black holes (PBHs) on the abundance of low-mass haloes and subhaloes in the dark and low-stellar-mass regime, and examines how these effects can be measured through fluctuations in strong lensing and brightness fluctuations in clusters of galaxies, providing potential ways to constrain the fraction of dark matter

Emanuele Berti, Vitor Cardoso, Gregorio Carullo, Jahed Abedi

The &#34;ringdown&#34; radiation emitted by oscillating black holes has great scientific potential. By carefully predicting the frequencies and amplitudes of black hole quasinormal modes and comparing them with gravitational-wave data from compact binary mergers we can advance our understanding of the two-body problem in general relativity, verify the predic

Will Barker, Carlo Marzo, Alessandro Santoni

Like general relativity, metric-affine gravity should be a viable effective quantum theory, otherwise it is a mathematical curiosity without physical application. Assuming a perturbative quantum field theory, the universal, flat limit of metric-affine gravity offers a good foundation for model-building only when symmetry constraints are themselves sufficient

Kim V. Berghaus, Yufeng Du, Vincent S. H. Lee, Anirudh Prabhu

The upcoming Deep Synoptic Array 2000 (DSA-2000) will map the radio sky at $0.7-2$ GHz ($2.9 - 8.3 \, \mu$eV) with unprecedented sensitivity. This will enable searches for dark matter and other physics beyond the Standard Model, of which we study four cases: axions, dark photons, dark matter subhalos and neutrino masses. We forecast DSA-2000's potential to d

Panos Betzios

We describe the duality between the gravitating $c=1$ (compact) Sine-Gordon model and a normal matrix model. From a two-dimensional quantum gravity perspective and due to the periodic nature of the potential, this model admits both anti-de Sitter and de-Sitter saddles, similarly to simpler models of Sine-Dilaton gravity, as well as more complicated interpola

Sylvain Lacroix, Nat Levine, Anders Wallberg

Large families of integrable 2d sigma-models have been constructed at the classical level, partly motivated by the utility of integrability on the string worldsheet. It is natural to ask whether these theories are renormalisable at the quantum level, and whether they define quantum integrable field theories. By considering examples, a folk theorem has emerge

Yubo Su, Connar Rowan, Mor Rozner

Active galactic nuclei (AGNs) consist of a central supermassive black hole (SMBH) embedded in a region with both high gas and stellar densities: the gas is present as a thin accretion disc that fuels the central SMBH, while the stars form a dense, roughly isotropic nuclear star cluster. The binaries present in such a cluster could be considered naturally as

Aretaios Lalakos, Alexander Tchekhovskoy, Elias R. Most, Bart Ripperda

Accretion onto supermassive black holes (BHs) can launch relativistic jets that inject energy and momentum into their surroundings. Understanding how such feedback shapes large-scale accretion is key to bridging observations from galactic scales (e.g., the Bondi radius, $r_{\rm B}$) down to event horizon scales ($r_{\rm g}$), spanning 5-6 orders of magnitude

Jonathan J. Heckman, Max Hübner, Chitraang Murdia

The global symmetry data of a $D$-dimensional absolute quantum field theory can sometimes be packaged in terms of a $(D+1)$-dimensional bulk system obtained by extending along an interval, with a relative QFT$_D$ at one end and suitable gapped / free boundary conditions at the other end. The partition function of the QFT$_D$ can then be interpreted as a wave

Yao Xiao, Qiqian Fu, Heyi Tao, Yuqun Wu

Image-text models excel at image-level tasks but struggle with detailed visual understanding. While these models provide strong visual-language alignment, segmentation models like SAM2 offer precise spatial boundaries for objects. To this end, we propose TextRegion, a simple, effective, and training-free framework that combines the strengths of image-text mo

Mostafa Najafi, Ali Morassaei

Let $ f:(0,\infty)\rightarrow \Bbb{R} $ be a completely monotonic function. In this paper, we present some properties of this functions and several new classes of completely monotonic functions. We also give some special functions such that its have completely monotonic condition.

Yunze Man, De-An Huang, Guilin Liu, Shiwei Sheng

Recent advances in multimodal large language models (MLLMs) have demonstrated remarkable capabilities in vision-language tasks, yet they often struggle with vision-centric scenarios where precise visual focus is needed for accurate reasoning. In this paper, we introduce Argus to address these limitations with a new visual attention grounding mechanism. Our a

Wentao Zhang, Woojeong Kim, Yuntian Deng

Conversational agents powered by large language models (LLMs) are rapidly becoming integral to our daily interactions, generating unprecedented amounts of conversational data. Such datasets offer a powerful lens into societal interests, trending topics, and collective concerns. Yet, existing approaches typically treat these interactions as independent and mi

Aneeshan Sain, Subhajit Maity, Pinaki Nath Chowdhury, Subhadeep Koley

As sketch research has collectively matured over time, its adaptation for at-mass commercialisation emerges on the immediate horizon. Despite an already mature research endeavour for photos, there is no research on the efficient inference specifically designed for sketch data. In this paper, we first demonstrate existing state-of-the-art efficient light-weig

Chenyu Yang, Shiqian Su, Shi Liu, Xuan Dong

The rapid advancement of large Vision-Language Models (VLMs) has propelled the development of pure-vision-based GUI Agents, capable of perceiving and operating Graphical User Interfaces (GUI) to autonomously fulfill user instructions. However, existing approaches usually adopt an offline learning framework, which faces two core limitations: (1) heavy relianc

Yunjae Won, Hyunji Lee, Hyeonbin Hwang, Minjoon Seo

Direct Preference Optimization (DPO) has been widely used for aligning language models with human preferences in a supervised manner. However, several key questions remain unresolved: the rationale behind its log-ratio reward, how the statistical structure of preference datasets shapes its training dynamics, and how those dynamics impact downstream capabilit

Amber Yijia Zheng, Cedar Site Bai, Brian Bullins, Raymond A. Yeh

Model immunization aims to pre-train models that are difficult to fine-tune on harmful tasks while retaining their utility on other non-harmful tasks. Though prior work has shown empirical evidence for immunizing text-to-image models, the key understanding of when immunization is possible and a precise definition of an immunized model remain unclear. In this

Heekyung Lee, Jiaxin Ge, Tsung-Han Wu, Minwoo Kang

Rebus puzzles, visual riddles that encode language through imagery, spatial arrangement, and symbolic substitution, pose a unique challenge to current vision-language models (VLMs). Unlike traditional image captioning or question answering tasks, rebus solving requires multi-modal abstraction, symbolic reasoning, and a grasp of cultural, phonetic and linguis

Yusuf Dalva, Hidir Yesiltepe, Pinar Yanardag

We introduce LoRAShop, the first framework for multi-concept image editing with LoRA models. LoRAShop builds on a key observation about the feature interaction patterns inside Flux-style diffusion transformers: concept-specific transformer features activate spatially coherent regions early in the denoising process. We harness this observation to derive a dis