March 2025 arXiv papers — page 220

Jamie Wynn, Zawar Qureshi, Jakub Powierza, Jamie Watson

Exploring real-world spaces using novel-view synthesis is fun, and reimagining those worlds in a different style adds another layer of excitement. Stylized worlds can also be used for downstream tasks where there is limited training data and a need to expand a model's training distribution. Most current novel-view synthesis stylization techniques lack the ab

Patricia Mayer, Florian Joseph Baader, David Yang Shu, Ludger Leenders

The chemical industry's transition to net-zero greenhouse gas (GHG) emissions is particularly challenging due to the carbon inherently contained in chemical products, eventually released to the environment. Fossil feedstock-based production can be replaced by electrified chemical production, combining carbon capture and utilization (CCU) with electrolysis-ba

Faraz Faruqi, Maxine Perroni-Scharf, Jaskaran Singh Walia, Yunyi Zhu

Recent work in Generative AI enables the stylization of 3D models based on image prompts. However, these methods do not incorporate tactile information, leading to designs that lack the expected tactile properties. We present TactStyle, a system that allows creators to stylize 3D models with images while incorporating the expected tactile properties. TactSty

Alexander Zlotnik

The compact fourth-order finite-difference scheme for solving the 1d wave equation is studied. New error bounds of the fractional order $\mathcal{O}(h^{4(\lambda-1)/5})$ are proved in the mesh energy norm in terms of data, for two initial functions from the Sobolev and Nikolskii spaces with the smoothness order $\lambda$ and $\lambda-1$ and the free term wit

Sela Fried

A word $w_1w_2\cdots w_n$ is said to be up-down if $w_1 < w_2 >w_3 \cdots$. Carlitz and Scoville found the generating function for the number of up-down words over an alphabet of size $k$. Using properties of the Chebyshev polynomials we derive a closed-form formula for these numbers.

Xuehui Dong, Kai Wan, Shuangyang Li, Robert Caiming Qiu

The Fluid Antenna System (FAS) overcomes the spatial degree-of-freedom limitations of conventional static antenna arrays in wireless communications.This capability critically depends on acquiring full Channel State Information across all accessible ports. Existing studies focus exclusively on narrowband FAS, performing channel estimation solely in the spatia

Ronan Kerr, Juan P. Farias, Lisa Prato, Travis A. Rector

Young associations provide a record that traces the star formation process, and the youngest populations connect progenitor gas dynamics to the resulting stellar populations. We therefore conduct the first comprehensive overview of the Circinus Complex, an under-studied and massive ($\sim$1500 M$_{\odot}$) region consisting of approximately 3100 recently for

Akram Baghban, Mehdi Ghiyasvand

This paper develops a new family of locally recoverable codes for distributed storage systems, Sequential Locally Recoverable Codes (SLRCs) constructed to handle multiple erasures in a sequential recovery approach. We propose a new connection between parallel and sequential recovery, which leads to a general construction of q-ary linear codes with informatio

Selcuk Yazar

Illumination estimation remains a pivotal challenge in image processing, particularly for robotics, where robust environmental perception is essential under varying lighting conditions. Traditional approaches, such as RGB histograms and GIST descriptors, often fail in complex scenarios due to their sensitivity to illumination changes. This study introduces a

Ian Dowding, Christopher A. Schuh

When materials are deformed at extreme strain rates, greater than $10^6 \text{ s}^{-1}$, a counterintuitive mechanical response is seen where the strength and hardness of pure metals increases with increasing temperature. The anti-thermal hardening is due to defects in the material becoming pinned by phonons in the crystal lattice. However, here, using optic

Ata Tuna

We introduce DAMCC (Deep Autoregressive Model for Dynamic Combinatorial Complexes), the first deep learning model designed to generate dynamic combinatorial complexes (CCs). Unlike traditional graph-based models, CCs capture higher-order interactions, making them ideal for representing social networks, biological systems, and evolving infrastructures. While

Amer Al-Hiyasat, Sunghan Ro, Julien Tailleur

We study the dynamics of a zero-temperature particle interacting linearly with a bath of hot Brownian particles. Starting with the most general model of a linearly-coupled bath, we eliminate the bath degrees of freedom exactly to map the tracer dynamics onto a generalized Langevin equation, allowing for an arbitrary external potential on the tracer. We apply

Bryce Cousins, Kristen Schumacher, Adrian Ka-Wai Chung, Colm Talbot

We report the first measurement of the Hubble constant $H_0$ using the stochastic gravitational-wave background arising from binary black hole mergers. This astrophysical background is sensitive to the expansion history of the Universe and thus can be used for cosmological parameter inference independently of not only electromagnetic methods, but also gravit

Yekyung Kim, Jenna Russell, Marzena Karpinska, Mohit Iyyer

We present ONERULER, a multilingual benchmark designed to evaluate long-context language models across 26 languages. ONERULER adapts the English-only RULER benchmark (Hsieh et al., 2024) by including seven synthetic tasks that test both retrieval and aggregation, including new variations of the "needle-in-a-haystack" task that allow for the possibility of a

Roberta M. Humphreys, Terry J. Jones, Kris Davidson, A. M. S. Richards

The red hypergiant VY CMa is remarkable for its very visible record of high mass loss events observed over the range of wavelengths from the optical and infrared to the submillimeter region with ALMA. The SW Clump or SW knots are unique in the ejecta of VY CMa. Except for the central star, they are the brightest sources of dusty infrared emission in its comp

Sade A. Davenport, Scott H. Harrison

Comparing specific types of organisms as they are found across environmental conditions has helped inform how genes and gene products of these organisms relate to phenotypes and adaptation. In this study, we examine metatranscriptomic data as found for oceanic fungi across different oceanic sampling sites. A specific set of three genes was chosen for evaluat

Avishai Gilkis, Eva Laplace, Iair Arcavi, Tomer Shenar

The majority of core-collapse supernova (CCSN) progenitors are massive stars in multiple systems, and their evolution and final fate are affected by interactions with their companions. These interactions can explain the presence of circumstellar material in many CCSNe, and the inferred low mass in stripped-envelope supernova progenitors. Through binary inter

Amir Dehghani, Ghazal Geshnizjani, Jerome Quintin

Cuscuton Gravity is characterized as a scalar field that can be added to general relativity without introducing any new dynamical degrees of freedom on a cosmological background. Yet, it modifies gravity such that spacetime singularities can be avoided. This has led to the Cuscuton bounce, a nonsingular cosmology that has been shown to be linearly stable, wh

Xiangdong Ji, Chen Yang

Momentum conservation in the nucleon is examined in terms of continuous flow of the momentum current density (or in short, momentum flow), which receives contributions from both kinetic motion and interacting forces involving quarks and gluons. While quarks conduct momentum flow through their kinetic motion and the gluon scalar (anomaly) contributes via pure

F. Valentino, K. E. Heintz, G. Brammer, K. Ito

Outflows are a key element in the baryon cycle of galaxies, and their properties provide a fundamental test for our models of how star formation quenches in galaxies. Here we report the detection of outflowing gas in two recently quenched, massive ($M_\star\sim10^{10.2}M_\odot$) galaxies at z=4.106 (NS_274) and z=7.276 (RUBIES-UDS-QG-z7) observed with JWST/N

Devanshu Shekhar, Pragya Shukla

We analyze the effect of varying system conditions on the single-particle entanglement entropy for an arbitrary eigenstate of a complex system that can be described by a multiparametric Gaussian ensemble. Our theoretical analysis leads to the identification of a single functional of the system parameters that governs the entropy dynamics. This reveals a sens

Hridhaan Banerjee, Carmen Isabel Day, Auguste H. Gezalyan, Olga Golovatskaia

Metric spaces defined within convex polygons, such as the Thompson, Funk, reverse Funk, and Hilbert metrics, are subjects of recent exploration and study in computational geometry. This paper contributes an educational piece of software for understanding these unique geometries while also providing a tool to support their research. We provide dynamic softwar

Francesco Fournier-Facio, Bin Sun

We construct finitely generated simple torsion-free groups with strong homological control. Our main result is that every subset of $\mathbb{N} \cup \{\infty\}$, with some obvious exceptions, can be realized as the set of dimensions of subgroups of a finitely generated simple torsion-free group. This is new even for basic cases such as $\{ 0, 1, 3 \}$ and $\

Davis Brown, Prithvi Balehannina, Helen Jin, Shreya Havaldar

Language model evaluations often fail to characterize consequential failure modes, forcing experts to inspect outputs and build new benchmarks. We introduce task elicitation, a method that automatically builds new evaluations to profile model behavior. Task elicitation finds hundreds of natural-language tasks -- an order of magnitude more than prior work --

Sonja Kraiczy, Georgios Papasotiropoulos, Grzegorz Pierczyński, Piotr Skowron

In the classic committee election setting each voter approves a subset of candidates and the goal is to select $k$ winners based on these preferences. A central focus of recent research in the area has been to achieve proportional representation. In this work, we explore notions of proportionality in a more expressive setting that allows voters to vote again

Parth Singh, Hunter Belanger

The Monte Carlo method is typically considered the gold standard for simulating reactor physics problems, as it does not require discretization of the phase space. This is not necessarily true though when simulating multigroup problems, as it has traditionally been a challenge to model anisotropic scattering in such simulations. Multigroup data used in react

Dimitris Chatzigiannakis, Annalisa Pillepich, Aurora Simionescu, Nhut Truong

The new era of high-resolution X-ray spectroscopy will significantly improve our understanding of the intra-cluster medium (ICM) by providing precise constraints on its underlying physical properties. However, spectral fitting requires reasonable assumptions on the thermal and chemical distributions of the gas. We use the output of TNG-Cluster, the newest ad

Timmy Ejdetjärn, Oscar Agertz, Florent Renaud, Göran Östlin

Haro 11 is a metal-poor, starburst galaxy believed to be the result of an ongoing merger, which is shaping the properties of the galaxy. In this study, we carry out a large suite of numerical simulations of a merger between two disc galaxies, to study possible origins of Haro 11 and understand under which conditions various features of the galaxy are formed.

Davide Caffagni, Sara Sarto, Marcella Cornia, Lorenzo Baraldi

Cross-modal retrieval is gaining increasing efficacy and interest from the research community, thanks to large-scale training, novel architectural and learning designs, and its application in LLMs and multimodal LLMs. In this paper, we move a step forward and design an approach that allows for multimodal queries, composed of both an image and a text, and can

Klint Faber, Auguste H. Gezalyan, Adam Martinson, Aniruddh Mutnuru

There are many structures, both classical and modern, involving point-sets and polygons whose deeper understanding can be facilitated through interactive visualizations. The Ipe extensible drawing editor, developed by Otfried Cheong, is a widely used software system for generating geometric figures. One of its features is the capability to extend its functio

H. Riechert, S. Annabi, A. Peugeot, H. Duprez

Gate-tunable transmon qubits are based on quantum conductors used as weak links within hybrid Josephson junctions. These gatemons have been implemented in just a handful of systems, all relying on extended conductors, namely epitaxial semiconductors or exfoliated graphene. Here we present the coherent control of a gatemon based on a single molecule, a one-di

Rahul Sahay, Curt von Keyserlingk, Ruben Verresen, Carolyn Zhang

It is well known that an exponentially localized Hamiltonian must be gapless if its ground state has algebraic correlations. We show that even certain exponentially decaying correlations can imply gaplessness. This is exemplified by the deformed toric code $\propto \exp(\beta \sum_{\ell} Z_{\ell}) |\mathsf{TC}\rangle$, where $|\mathsf{TC}\rangle$ is a fixed-

Brian Hu Zhang, Tao Lin, Yiling Chen, Tuomas Sandholm

We study the problem of learning the utility functions of no-regret learning agents in a repeated normal-form game. Differing from most prior literature, we introduce a principal with the power to observe the agents playing the game, send agents signals, and give agents payments as a function of their actions. We show that the principal can, using a number o

Jared Fier, Henry Han, Bowen Li, Kai Lin

In the framework of Einstein-scalar-Gauss-Bonnet (EsGB) gravity, we systematically study gravitational waves (GWs), first produced by remote compact astrophysical sources and then propagating through the flat homogeneous and isotropic Universe at cosmic distances before arriving at detectors. Assuming that the speed $c_T$ of the spin-2 graviton is the same a

Saiyad Ashanujjaman, Andreas Crivellin, Siddharth P. Maharathy, Anil Thapa

Vector-boson fusion production of new Higgs bosons decaying into pairs of electroweak gauge bosons ($W^\pm W^\pm$, $WZ$ and $ZZ$) is a smoking-gun signature of the Georgi-Machacek (GM) Model. Notably, ATLAS has observed a $3.3\sigma$ excess in $W^\pm W^\pm$ at $\approx 450\,$GeV and a $2.8\sigma$ excess in the $WZ$ channel at $\approx 375\,$GeV, while CMS re

Noah Kubli, Lucio Mayer, Hongping Deng, Douglas N. C. Lin

We present a detailed analysis of the nature of migration of protoplanetary clumps formed via disc instability in self-consistent 3D hydrodynamical (HD) and magneto-hydrodynamical (MHD) simulations of self-gravitating discs. Motivated by the complex structure of protoplanetary clumps we do not introduce sink particles. We find that the orbital evolution of t

Yoram Lithwick

Saturn raises a time-dependent tide on its small moon Enceladus, due to the eccentricity of the orbit. As shown in a companion paper (Goldreich et al.), the resulting tidal heating drives Enceladus into a limit cycle, in which its eccentricity and shell thickness vary in tandem, on a timescale of ~ 10 Myr. The limit cycle explains a variety of observed pheno

A. Tritsis, S. Basu, C. Federrath

Interstellar magnetic fields are thought to play a fundamental role in the evolution of star-forming regions. Polarized thermal dust emission serves as a key probe for understanding the structure of the POS component of the magnetic field. However, inclination effects can significantly influence the apparent morphology of the magnetic field and lead to erron

Lionel Haemmerlé

The formation of the most massive quasars observed at high redshifts requires extreme accretion rates ($>1$ M$_\odot$ yr$^{-1}$). Inflows of $10-1000$ M$_\odot$ yr$^{-1}$ are found in hydrodynamical simulations of galaxy mergers, leading to the formation of supermassive discs (SMDs) with high metallicities ($>$ Z$_\odot$). Supermassive stars (SMSs) born in t

Katrin Lehle, Dylan Nelson, Annalisa Pillepich

In this study, we examine the frequency and physical drivers of transformations from cool-core (CC) to non-cool-core (NCC) clusters, and vice versa, in a sample of 352 massive galaxy clusters (M_vir = 10^14-15.3 M_sun) from the TNG-Cluster magnetohydrodynamical cosmological simulation of galaxies. By identifying transformations based on the evolution of cent

Antonio Matteri, Andrea Pallottini, Andrea Ferrara

JWST is detecting an excess of high-redshift ($z\gtrsim 10$), bright galaxies challenging most theoretical predictions. To address this issue, we investigate the impact of Primordial Black Holes (PBHs) on the halo mass function and UV luminosity function (LF) of super-early galaxies. We explore two key effects: (i) the enhancement of massive halos abundance

Peter Goldreich, Yoram Lithwick, Jing Luan

Enceladus exhibits some remarkable phenomena, including water geysers spraying through surface cracks, a global ice shell that is librating atop an ocean, a large luminosity, and rapid outward orbital migration. Here we model the coupled evolution of Enceladus's orbit and interior structure. We find that Enceladus is driven into a periodic state: a limit cyc

Francesco Scala, Christa Zoufal, Dario Gerace, Francesco Tacchino

Knowing whether a Quantum Machine Learning model would perform well on a given dataset before training it can help to save critical resources. However, gathering a priori information about model performance (e.g., training speed, critical hyperparameters, or inference capabilities on unseen data) is a highly non-trivial task, in general. Recently, the Quantu

Marine Prunier, Annalisa Pillepich, Julie Hlavacek-Larrondo, Dylan Nelson

The TNG-Cluster magnetohydrodynamic cosmological simulations, produce a diverse population of X-ray cavities in the intracluster medium (ICM) of simulated galaxy clusters. These arise from episodic, high velocity, kinetic energy injections from the central active supermassive black hole (AGN, SMBH). Here, we present the first comprehensive comparative analys

Bo Wang, Di Wu, Ellis Ye Yuan

We bootstrap the first-order correction in the curvature expansion of the Virasoro-Shapiro amplitude in AdS spacetime, for arbitrary Kaluza-Klein charges of external operators. By constructing a universal ansatz based on single-valued multiple polylogarithms as well as an AdS$\times$S formalism, and matching it with the low-lying result, we derive a unified

A. Tritsis, S. Basu, C. Federrath

Most of our knowledge regarding molecular clouds and the early stages of star formation stems from molecular spectral-line observations. However, the various chemical and radiative-transfer effects, in combination with projection effects, can lead to a distorted view of molecular clouds. Our objective is to simultaneously study all of these effects by creati

Gabriele Franciolini, Yann Gouttenoire, Ryusuke Jinno

Understanding whether primordial black holes form during strong first-order phase transition (FOPT) is a crucial open question in cosmology. We address this using a fully covariant formalism to study cosmological perturbations, highlighting previously overlooked gauge dependencies. We show that non-covariant treatments can overestimate primordial black holes

Saba Rahimy, Elsa M. Teixeira, Ivonne Zavala

Coupled dark sector models have gained significant attention, motivated by recent advances in cosmology and the pressing need to address unresolved puzzles. In this work, we study coupled scalar dark sector models inspired by ultraviolet-complete frameworks such as supergravity and string theory. These models involve scalar couplings arising either from thei

Milan Staffehl, Dylan Nelson, Mohammadreza Ayromlou, Eric Rohr

In addition to the hot intracluster medium, massive galaxy clusters host complex, multi-phase gaseous halos. In this work, we quantify the abundance, spatial distribution, and origin of the cool T < 10^4.5 K gas within clusters. To do so, we combine the TNG-Cluster and TNG300 cosmological magnetohydrodynamical simulations, yielding a sample of 632 simulated

Victor Montenegro, Sara Dornetti, Alessandro Ferraro, Matteo G. A. Paris

Frequency estimation, a cornerstone of basic and applied sciences, has been significantly enhanced by quantum sensing strategies. Despite breakthroughs in quantum-enhanced frequency estimation, key challenges remain: static probes limit flexibility, and the interplay between resource efficiency, sensing precision, and potential enhancements from nonlinear pr

Jernej Rudi Finžgar, Simone Notarnicola, Madelyn Cain, Mikhail D. Lukin

Counterdiabatic (CD) driving has the potential to speed up adiabatic quantum state preparation by suppressing unwanted excitations. However, existing approaches either require intractable classical computations or are based on approximations which do not have performance guarantees. We propose and analyze a non-variational, system-agnostic CD expansion metho

Jessica Speedie, Ruobing Dong, Richard Teague, Dominique Segura-Cox

Late infall events challenge the traditional view that planet formation occurs without external influence. Here we present deep ALMA $^{12}$CO $J=2-1$ and SO $J_{N}=5_6-4_5$ observations toward AB Aurigae, a Class II disk system with strong signs of gravitational instability and ongoing planet formation. By applying Keplerian and anti-Keplerian masks, we sep

Calvin Osinga, Benedikt Diemer, Francisco Villaescusa-Navarro

Large-scale HI surveys and their cross-correlations with galaxy distributions have immense potential as cosmological probes. Interpreting these measurements requires theoretical models that must incorporate redshift-space distortions (RSDs), such as the Kaiser and fingers-of-God (FoG) effect, and differences in the tracer and matter distributions via the tra

E. Kirkinis, L. Bonds, A. Levchenko, A. V. Andreev

We show that the nonreciprocity of hydrodynamic electron transport in noncentrosymmetric conductors with broken time-reversal symmetry (TRS) is significantly enhanced compared to the disorder-dominated regime. This enhancement is caused by the linear dependence of the viscosity of the electron liquid on the flow velocity, which is allowed in the absence of T

Claude Duhr, Andres Vasquez, Giuseppe Ventura, Eleni Vryonidou

We compute the contributions of CP-conserving operators in the dimension-six SMEFT to the two-loop renormalisation constants of quark and gluon fields in the $\overline{\textrm{MS}}$-scheme. Specifically, we consider the top-quark chromomagnetic operator and the triple gluon operator. We work with the background-field method, which allows us to extract the c

K. Ito, F. Valentino, M. Farcy, G. De Lucia

We report the spectroscopic confirmation of a merging pair of massive quiescent galaxies at $z=3.44$. Using JWST observations, we confirm that the two galaxies lie at a projected separation of 4.5 kpc with a velocity offset of $\sim 680\, {\rm km\, s^{-1}}\ (\delta_z \sim 0.01)$. The pair resides in the core of a known rich overdensity of galaxies, dubbed th

Stewart Morawetz, Anatoli Polkovnikov

Local counterdiabatic (CD) driving provides a systematic way of constructing a control protocol to approximately suppress the excitations resulting from changing some parameter(s) of a quantum system at a finite rate. However, designing CD protocols typically requires knowledge of the original Hamiltonian a priori. In this work, we design local CD driving pr

Giulio Barni, Simone Blasi, Miguel Vanvlasselaer

Building upon the recent findings regarding inverse phase transitions in the early universe, we present the first natural realisation of this phenomenon within a supersymmetry-breaking sector. We demonstrate that inverse hydrodynamics, which is characterized by the fluid being aspirated by the bubble wall rather than being pushed or dragged, is actually not

Jerome P. Gauntlett, Dario Martelli, James Sparks

We review various aspects of Sasaki-Einstein and GK geometry, emphasising their similarities, interconnections and significance for the AdS/CFT correspondence. In particular, we highlight the key role that physical considerations have played in formulating geometric extremization principles, which have been instrumental in both understanding the geometry and

Xuejian Shen, Rahul Kannan, Ewald Puchwein, Aaron Smith

Recent JWST observations hint at unexpectedly intense cosmic star-formation in the early Universe, often attributed to enhanced star-formation efficiencies (SFEs). Here, we analyze the SFE in THESAN-ZOOM, a novel zoom-in radiation-hydrodynamic simulation campaign of high-redshift ($z \gtrsim 3$) galaxies employing a state-of-the-art galaxy formation model re

Kun Xu, Y. P. Jing, Shaun Cole, Carlos S. Frenk

The Photometric Objects Around Cosmic Webs (PAC) method integrates cosmological photometric and spectroscopic surveys, offering valuable insights into galaxy formation. PAC measures the excess surface density of photometric objects, $\bar{n}_2w_{\rm{p}}$, with specific physical properties around spectroscopic tracers. In this study, we improve the PAC method

Akito Nakanishi, Yukie Sano, Geng Liu, Francesco Pierri

In recent years, Large Language Models have attracted growing interest for their significant potential, though concerns have rapidly emerged regarding unsafe behaviors stemming from inherent stereotypes and biases. Most research on stereotypes in LLMs has primarily relied on indirect evaluation setups, in which models are prompted to select between pairs of

Luke Finnerty, Yinzi Xin, Jerry W. Xuan, Julie Inglis

We present atmospheric retrievals from Keck/KPIC phase II observations of the ultra-hot Jupiter KELT-20/MASCARA-2~b. Previous free retrievals of molecular abundances for ultra-hot Jupiters have been impacted by significant model biases due to variations in vertical abundance profiles, which we address by including molecular dissociation into our retrieval fr

Mengyuan Xiao, Pascal A. Oesch, Longji Bing, David Elbaz

Little Red Dots (LRDs) are compact, point-like sources characterized by their red color and broad Balmer lines, which have been debated to be either dominated by active galactic nuclei (AGN) or dusty star-forming galaxies (DSFGs). Here we report two LRDs (ID9094 and ID2756) at z$_{\rm spec}$>7, recently discovered in the JWST FRESCO GOODS-North field. Both s

Kaiwen Zheng, Xingrui Song, Kater W. Murch

Single electrons confined to a free neon surface and manipulated through the circuit quantum electrodynamics (circuit QED) architecture is a promising novel quantum computing platform. Understanding the exact physical nature of the electron-on-neon (eNe) charge states is important for realizing this platform's potential for quantum technologies. We investiga

Rohit Patil, Marcos Rigol

We study the diagonal and off-diagonal matrix elements of observables in the eigenstates of the extended spin-$\frac{1}{2}$ Heisenberg chain, which exhibits the non-Abelian SU(2) symmetry. We explore integrable and nonintegrable regimes, and consider observables that preserve the SU(2) symmetry of the Hamiltonian as well as observables that break it. We stud

Aleksei Zhuravlev, Zorah Lähner, Vladislav Golyanik

Estimating correspondences between pairs of deformable shapes remains a challenging problem. Despite substantial progress, existing methods lack broad generalization capabilities and require category-specific training data. To address these limitations, we propose a fundamentally new approach to shape correspondence based on denoising diffusion models. In ou

Miriam Havin, Timna Wharton Kleinman, Moran Koren, Yaniv Dover

The increasing integration of large language models (LLMs) based conversational agents into everyday life raises critical cognitive and social questions about their potential to influence human opinions. Although previous studies have shown that LLM-based agents can generate persuasive content, these typically involve controlled English-language settings. Ad

Dayal Singh Kalra, John Kirchenbauer, Maissam Barkeshli, Tom Goldstein

Adam is the go-to optimizer for training modern machine learning models, but it requires additional memory to maintain the moving averages of the gradients and their squares. While various low-memory optimizers have been proposed that sometimes match the performance of Adam, their lack of reliability has left Adam as the default choice. In this work, we appl

Hang Liu, Sangli Teng, Ben Liu, Wei Zhang

Hybrid dynamical systems, which include continuous flow and discrete mode switching, can model robotics tasks like legged robot locomotion. Model-based methods usually depend on predefined gaits, while model-free approaches lack explicit mode-switching knowledge. Current methods identify discrete modes via segmentation before regressing continuous flow, but

Damiano Anselmi, Fabio Briscese, Gianluca Calcagni, Leonardo Modesto

In the context of field theories with complex poles, we scrutinize four inequivalent ways of defining the scattering amplitudes, each forfeiting one or more tenets of standard quantum field theory while preserving the others: (i) a textbook Wick rotation by analytic continuation of the external momenta from Euclidean to Lorentzian signature (no optical theor

Yuhui Li, Fangyun Wei, Chao Zhang, Hongyang Zhang

The sequential nature of modern LLMs makes them expensive and slow, and speculative sampling has proven to be an effective solution to this problem. Methods like EAGLE perform autoregression at the feature level, reusing top-layer features from the target model to achieve better results than vanilla speculative sampling. A growing trend in the LLM community

Zhengyuan Jiang, Yuepeng Hu, Yuchen Yang, Yinzhi Cao

Text-to-Image models may generate harmful content, such as pornographic images, particularly when unsafe prompts are submitted. To address this issue, safety filters are often added on top of text-to-image models, or the models themselves are aligned to reduce harmful outputs. However, these defenses remain vulnerable when an attacker strategically designs a

Maria Drencheva, Ivo Petrov, Maximilian Baader, Dimitar I. Dimitrov

Federated learning claims to enable collaborative model training among multiple clients with data privacy by transmitting gradient updates instead of the actual client data. However, recent studies have shown the client privacy is still at risk due to the, so called, gradient inversion attacks which can precisely reconstruct clients' text and image data from

Adrià López Escoriza, Nicklas Hansen, Stone Tao, Tongzhou Mu

Long-horizon tasks in robotic manipulation present significant challenges in reinforcement learning (RL) due to the difficulty of designing dense reward functions and effectively exploring the expansive state-action space. However, despite a lack of dense rewards, these tasks often have a multi-stage structure, which can be leveraged to decompose the overall

Yisen Li, Lingfeng Yang, Wenxuan Shen, Pan Zhou

Distilling advanced Large Language Models' instruction-following capabilities into smaller models using a selected subset has become a mainstream approach in model training. While existing synthetic instruction data selection strategies rely mainly on single-dimensional signals (i.e., reward scores, model perplexity), they fail to capture the complexity of i

Tassilo Wald, Saikat Roy, Fabian Isensee, Constantin Ulrich

Transformers have achieved remarkable success across multiple fields, yet their impact on 3D medical image segmentation remains limited with convolutional networks still dominating major benchmarks. In this work, (A) we analyze current Transformer-based segmentation models and identify critical shortcomings, particularly their over-reliance on convolutional

Roberto Rosati, Sai Shradha, Julian Picker, Andrey Turchanin

Lateral heterostructures built of monolayers of transition metal dichalcogenides (TMDs) are characterized by a thin 1D interface exhibiting a large energy offset. Recently, the formation of spatially separated charge-transfer (CT) excitons at the interface has been demonstrated. The technologically important exciton propagation across the interface and the i

André Jonasson

Transformer-based Large Language Models (LLMs) rely on positional encodings to provide sequence position information to their attention mechanism. Rotary Positional Encodings (RoPE), which encode relative position by rotating queries and keys, have become widely used in modern LLMs. We study the features and patterns that emerge in queries and keys when usin

Lei Su

In this work, we investigate the one-dimensional $XX$ lattice model and its cousins through the lens of momentum and winding $U(1)$ symmetries. We distinguish two closely related $\mathbb{Z}_2$ symmetries based on their relation to the $U(1)$ symmetries, and establish a web of $\mathbb{Z}_2$-gauging relations among these models, rooted in two fundamental see

Badr AlKhamissi, Greta Tuckute, Yingtian Tang, Taha Binhuraib

Large language models (LLMs) exhibit remarkable similarity to neural activity in the human language network. However, the key properties of language shaping brain-like representations, and their evolution during training as a function of different tasks remain unclear. We here benchmark 34 training checkpoints spanning 300B tokens across 8 different model si

Jukka Ruohonen, Jani Koskinen, Søren Harnow Klausen, Anne Gerdes

Context: Dark patterns are user interface or other software designs that deceive or manipulate users to do things they would not otherwise do. Even though dark patterns have been under active research for a long time, including particularly in computer science but recently also in other fields such as law, systematic applied ethical assessments have generall

Anders Sildnes, Nikita Shvetsov, Masoud Tafavvoghi, Vi Ngoc-Nha Tran

Even foundational models that are trained on datasets with billions of data samples may develop shortcuts that lead to overfitting and bias. Shortcuts are non-relevant patterns in data, such as the background color or color intensity. So, to ensure the robustness of deep learning applications, there is a need for methods to detect and remove such shortcuts.

Nemanja Draganić, Peter Keevash, Alp Müyesser

In 1996, in his last paper, Erd\H{o}s asked the following question that he formulated together with Faudree: is there a positive $c$ such that any $(n+1)$-regular graph $G$ on $2n$ vertices contains at least $c 2^{2n}$ distinct vertex-subsets $S$ that are cyclic, meaning that there is a cycle in $G$ using precisely the vertices in $S$. We answer this questio

Matija Bucić, Bryce Frederickson, Alp Müyesser, Alexey Pokrovskiy

We study an old question in combinatorial group theory which can be traced back to a conjecture of Graham from 1971. Given a group $\Gamma$, and some subset $S\subseteq \Gamma$, is it possible to permute $S$ as $s_1, s_2, \ldots, s_d$ so that the partial products $\prod_{1 \leq i \leq t} s_i$, $t\in [d]$ are all distinct? Most of the progress towards this pr

David Klindt, Charles O'Neill, Patrik Reizinger, Harald Maurer

Understanding how information is represented in neural networks is a fundamental challenge in both neuroscience and artificial intelligence. Despite their nonlinear architectures, recent evidence suggests that neural networks encode features in superposition, meaning that input concepts are linearly overlaid within the network's representations. We present a

Vasilis Mageirakos, Bowen Wu, Gustavo Alonso

Retrieval Augmented Generation (RAG) uses vector databases to expand the expertise of an LLM model without having to retrain it. The idea can be applied over data lakes, leading to the notion of embedding data lakes, i.e., a pool of vector databases ready to be used by RAGs. The key component in these systems is the indexes enabling Approximated Nearest Neig

Sai Sumedh R. Hindupur, Ekdeep Singh Lubana, Thomas Fel, Demba Ba

Sparse Autoencoders (SAEs) are widely used to interpret neural networks by identifying meaningful concepts from their representations. However, do SAEs truly uncover all concepts a model relies on, or are they inherently biased toward certain kinds of concepts? We introduce a unified framework that recasts SAEs as solutions to a bilevel optimization problem,

Xingyu Zhu, Abhishek Panigrahi, Sanjeev Arora

We formalize a new concept for LLMs, context-enhanced learning. It involves standard gradient-based learning on text except that the context is enhanced with additional data on which no auto-regressive gradients are computed. This setting is a gradient-based analog of usual in-context learning (ICL) and appears in some recent works. Using a multi-step reason

Yi-Lin Sung, Prateek Yadav, Jialu Li, Jaehong Yoon

Layer-wise quantization is a key technique for efficiently compressing large models without expensive retraining. Previous methods typically quantize the weights of each layer by "uniformly" optimizing the layer reconstruction loss across all output tokens. However, in this paper, we demonstrate that better-quantized models can be obtained by prioritizing le

Adesh Gupta, Abhinav Kumar, Mansi Gupta, Paras Chopra

Generating diverse solutions is key to human-like reasoning, yet autoregressive language models focus on single accurate responses, limiting creativity. GFlowNets optimize solution generation as a flow network, promising greater diversity. Our case study shows their limited zero-shot transferability by fine-tuning small and medium-sized large language models

Aurora Matteo, Ginés Garnés-Morales, Alberto Moreno, Ribeiro Andreia

The Canadian Fire Weather Index (FWI) is widely used to assess wildfire danger and relies on meteorological data at local noon. However, climate models often provide only daily aggregated data, which poses a challenge for accurate FWI calculations in climate change studies. Here, we examine how using daily approximations for FWI95d -- the annual count of ext

Alessandro Daniele, Emile van Krieken

A fundamental challenge in neurosymbolic systems is applying continuous gradient-based optimization to discrete logical domains. While fuzzy relaxations provide differentiability, they often lack a formal structural alignment with classical logic. In this work, we show that G\"odel semantics addresses this limitation through a homomorphism that maps its cont

Jukka Ruohonen, Kalle Hjerppe, Eun-Young Kang

A new Cyber Resilience Act (CRA) was recently agreed upon in the European Union (EU). The paper examines and elaborates what new requirements the CRA entails by contrasting it with the older General Data Protection Regulation (GDPR). According to the results, there are overlaps in terms confidentiality, integrity, and availability guarantees, data minimizati

Jan Dymara, Tadeusz Januszkiewicz

We explain the relation between the Witt class and the universal equicommutative class for PSL(2,K). We discuss an analogue of the Milnor-Wood inequality.

Weizhi Zhang, Liangwei Yang, Wooseong Yang, Henry Peng Zou

Collaborative filtering (CF) is widely adopted in industrial recommender systems (RecSys) for modeling user-item interactions across numerous applications, but often struggles with cold-start and data-sparse scenarios. Recent advancements in pre-trained large language models (LLMs) with rich semantic knowledge, offer promising solutions to these challenges.

Zhixuan Bi, Yunfei Bai, Ying Shi, Tao Sun

Photogenerated carrier transport and recombination in metal halide perovskites are critical to device performance. Despite considerable efforts, sample quality issues and measurement techniques have limited the access to their intrinsic physics. Here, by utilizing high-purity CsPbBr3 single crystals and contact-free transient grating spectroscopy, we directl

Vasily Lubashevskiy, Ihor Lubashevsky, Namik Gusein-zade

A novel model for dynamical traps in intermittent human control is proposed. It describes probabilistic, step-wise transitions between two modes of a subject's behavior - active and passive phases in controlling an object's dynamics - using an original stochastic differential equation. This equation governs time variations of a special variable, denoted as $

Nicholas Carlini, Javier Rando, Edoardo Debenedetti, Milad Nasr

We introduce AutoAdvExBench, a benchmark to evaluate if large language models (LLMs) can autonomously exploit defenses to adversarial examples. Unlike existing security benchmarks that often serve as proxies for real-world tasks, bench directly measures LLMs' success on tasks regularly performed by machine learning security experts. This approach offers a si

Nandini Iyer, Massimiliano Luca, Riccardo Di Clemente

Rural and urban areas exhibit distinct mobility patterns, yet a systematic understanding of how these trends differ across regions and contexts remains underexplored. By using origin-destination matrices from Location-Based Services data in Colombia, India, and Mexico, we delineate urban and rural boundaries through network percolation, reducing reliance on