December 2025 arXiv papers — page 75

Aleksandr Arakcheev, Heinz H. Bauschke

Many algorithms in convex optimization and variational analysis can be analyzed using Fej\'er monotone sequences. In 2024, Behling, Bello-Cruz, Iusem, Alves Ribeiro, and Santos introduced a new, more general, notion: Fej\'er* monotonicity. They obtained basic results and discussed applications in optimization. In this work, we complement Behling et al.'s wor

Brandon Marks, Yash Dave, Zixun Wang, Hannah Chung

A scale mixture of normals is a distribution formed by mixing a collection of normal distributions with fixed mean but different variances. A generalized gamma scale mixture draws the variances from a generalized gamma distribution. Generalized gamma scale mixtures of normals have been proposed as an attractive class of parametric priors for Bayesian inferen

Tobias Rüttenauer, Kasimir Dederichs, David Kretschmer

Immigrant residential segregation can profoundly shape access to opportunities, immigrant integration, and inter-group relations. Yet we lack systematic evidence on how segregation varies across Europe, and what structural factors are associated with these patterns. This study addresses the gap by focusing on two questions: (i) how does immigrant-native segr

Yuan-Hung Kuan, Jr-Shin Li

In this paper, we develop a novel framework, Exact Bilinearization Iterative Form (EBIF), for transforming a nonlinear control-affine system into an exact finite-dimensional bilinear representation. In contrast to most existing approaches which generally lead to an infinite-dimensional representation, the proposed EBIF approach yields an iterative procedure

Sara Merino-Aceituno, Carmela Moschella

In this work, we investigate an individual-based model (IBM) for self-propelled agents interacting locally on a plane. Agents are characterized by their position, the angle determining their direction of motion, and their angular velocity. The dynamics combine features of the well-known Vicsek and Kuramoto models, which describe collective dynamics and synch

Chang-Hwan Lee, Chanseung Lee

Non-stationary environments pose a fundamental challenge for deep reinforcement learning, as changes in dynamics or rewards invalidate learned value functions and cause catastrophic forgetting. We propose \emph{Gradient-Boosted Deep Q-Networks (GB-DQN)}, an adaptive ensemble method that addresses model drift through incremental residual learning. Instead of

Joseph Guzman, Jeremiah Murphy, Emma Beasor, Julianne Dalcanton

We infer the ages of three young stellar clusters, NGC 2004, NGC 7419, and NGC 2100, using Stellar Ages, a statistical algorithm designed to infer stellar population properties from color magnitude diagrams. Recent studies have revealed emerging inconsistencies in the inferred ages of very young stellar clusters with ages less than or equal to 50 Myr. Here,

Jay D. Tasson

Additional sensitivities to Lorentz violation can be obtained from existing experiments by considering additional boost-suppressed effects. The additional Lorentz-violating signals arise as variations in experimental observables at the commonly-used sidereal frequency as well as more novel frequencies. In this work we provide some examples that serve to illu

Rhea P. Fernandes, Andrew J. Pizzimenti, Christos N. Gagatsos, Joseph M. Lukens

Non-Gaussian quantum states are critical resources in photonic quantum information processing, rendering their generation and characterization of increasing importance in quantum optics. In this work, we theoretically and numerically analyze the relative efficiency of homodyne versus heterodyne measurements for reconstructing non-Gaussian states, a major out

Patricia A Almeida, George B Martins, Sergio Ulloa

Metallic kagome systems have attracted considerable interest in recent years, as they provide a rich platform for studying phenomena associated with their distinctive band structure. The coexistence of bands with Dirac points similar to those in graphene, along with a completely flat band, makes this an ideal structure for investigating how lattice symmetrie

Istiak Ahmed, Ripan Kumar Kundu, Khaza Anuarul Hoque

Deep learning (DL)-based automated cybersickness detection methods, along with adaptive mitigation techniques, can enhance user comfort and interaction. However, recent studies show that these DL-based systems are susceptible to adversarial attacks; small perturbations to sensor inputs can degrade model performance, trigger incorrect mitigation, and disrupt

Victoria-Elisabeth Gruber, Razvan Marinescu, Diego Fajardo, Amin H. Nassar

As large language models (LLMs) become primary sources of health information for millions, their accuracy in women's health remains critically unexamined. We introduce the Women's Health Benchmark (WHB), the first benchmark evaluating LLM performance specifically in women's health. Our benchmark comprises 96 rigorously validated model stumps covering five me

Erica Coppolillo, Simone Mungari

Encyclopedic knowledge platforms are key gateways through which users explore information online. The recent release of Grokipedia, a fully AI-generated encyclopedia, introduces a new alternative to traditional, well-established platforms like Wikipedia. In this context, search engine mechanisms play an important role in guiding users exploratory paths, yet

Óscar Carrión-González, Sylvestre Lacour, Mathias Nowak

In this white paper, we propose an upgrade to the Very Large Telescope Interferometer (VLTI) consisting of the addition of a new 8m Unit Telescope (UT5). The primary goal of this upgrade is to optimise the VLTI for exoplanet detection by creating four additional baselines of approximately 200m oriented toward the north-west. The inclusion of this telescope w

Amina Kobenova, Piper Stickler, Thaís Alvarenga, Sri Kurniawan

Although serious games have been increasingly used for mental health applications, few explicitly address coping with grief as a core mechanic and narrative experience for patients. Existing grief-related digital games often focus on clinical training for medical professionals rather than immersive storytelling and agency in emotional processing for the pati

Patrick Diehl, Noujoud Nader, Deepti Gupta

Parallel programming remains one of the most challenging aspects of High-Performance Computing (HPC), requiring deep knowledge of synchronization, communication, and memory models. While modern C++ standards and frameworks like OpenMP and MPI have simplified parallelism, mastering these paradigms is still complex. Recently, Large Language Models (LLMs) have

João Ferri, Elisa G. M. Ferreira, Ryo Terasawa

A known problem in cosmic shear two-point statistics is the apparent inconsistency between analyses performed in harmonic space (power spectrum) and real space (angular correlation). This arises mainly from two factors: first, scale cuts in one space correspond to soft cuts in the other, as the relationship between the two spaces is mediated by Bessel functi

Adam Kamel, Tanish Rastogi, Michael Ma, Kailash Ranganathan

Transformer-based large language models (LLMs) have demonstrated strong reasoning abilities across diverse fields, from solving programming challenges to competing in strategy-intensive games such as chess. Prior work has shown that LLMs can develop emergent world models in games of perfect information, where internal representations correspond to latent sta

Martin Schwartz, Fajwel Fogel, Nikola Besic, Damien Robert

Current large-scale satellite-based forest disturbance monitoring systems operate at 10-30~m resolution, too coarse to detect changes at the scale of individual trees and resulting in a systematic underestimation of forest disturbances. Here, we introduce FORMSpoT (Forest Mapping with SPOT Time series), a decade-long (2014-2024), country-scale mapping of for

Saurya Das, Mitja Fridman, Sourav Sur

It is generally assumed that any discrepancy between an object's inertial and gravitational masses, leading to a violation of the equivalence principle, arises from the nature of its internal constituents and their interactions. We show here that the difference can instead be a function of the distance of the object from a gravitating body, and suggest ways

Sérgio Carrôlo, Carolina Figueiredo

We study the leading singularities for pure gluon amplitudes obtained by on-shell gluing of three-particle amplitudes for an arbitrary graph in any number of dimensions. By encoding the polarization vector contractions in a graphical way, on-shell gluing "discovers" curves on surfaces, and we find that the leading singularity is determined by a simple combin

Chengyun Hua, Yadu K. Sarathchandran, Eva Zarkadoula, Wojciech Dmowski

Gallium is a prototypical liquid metal and has gained renewed attention due to its unique properties. Characterizing and elucidating its atomic dynamics remains elusive despite numerous studies, primarily due to the challenges of quantifying atomic-scale dynamics in liquids. Recent developments in inelastic neutron scattering enable us to measure the Van Hov

Euijun Jung, Jingyu Lee, Minji Kim, Youngki Lee

Working with abstract information often relies on static, symbolic representations that constrain exploration. We introduce Explorable Ideas, a framework that externalizes abstract concepts into explorable environments where physical navigation coordinates conceptual exploration. To investigate its practical value, we designed Idea Islands, a VR probe for id

J. M. Santana, L. Heller, R. M. Buehrer

Large Low Earth Orbit (LEO) constellations (e.g., Starlink and Iridium) significantly increase the likelihood of transient, high-power interference events at ground receivers. This report presents SatTrack, a GUI-driven simulation framework that (i) tracks satellite motion relative to a fixed antenna boresight, (ii) predicts reflector gain patterns of a para

Domenico de Gioia, Claudio Pomo, Ludovico Boratto, Tommaso Di Noia

Similarity-based collaborative filtering (CF) models have long demonstrated strong offline performance and conceptual simplicity. However, their scalability is limited by the quadratic cost of maintaining dense item-item similarity matrices. Partitioning-based paradigms have recently emerged as an effective strategy for balancing effectiveness and efficiency

Jonathan Sorce

We construct and characterize canonical purifications for general algebraic states, extending prior constructions by Woronowicz and by Dutta/Faulkner to general quantum theories. Given a state on a $*$-algebra, the canonical purification is a state on a "doubled" algebra that admits an interpretation in terms of CRT reflection. This interpretation holds for

Chengyun Hua, Yadu K. Sarathchandran, Eva Zarkadoula, Wojciech Dmowski

Simplicity in chemical composition does not always translate into simplicity in the structures and dynamics of liquids and solids. Some elementary liquid metals, such as gallium, show unusual behaviors in thermodynamic and transport properties as a result of their complex atomic structure and dynamics. In this work, we study the real-space atomic correlation

Chiao-An Yang, Ryo Hachiuma, Sifei Liu, Subhashree Radhakrishnan

Despite advances in Multimodal LLMs (MLLMs), their ability to reason over 3D structures and temporal dynamics remains limited, constrained by weak 4D perception and temporal understanding. Existing 3D and 4D Video Question Answering (VQA) benchmarks also emphasize static scenes and lack region-level prompting. We tackle these issues by introducing: (a) 4D-RG

Antoine Petitjean, Tilman Plehn, Jonas Spinner, Ullrich Köthe

Modern machine learning is transforming jet tagging at the LHC, but the leading transformer architectures are large, not particularly fast, and training-intensive. We present a slim version of the L-GATr tagger, reduce the number of parameters of jet-tagging transformers, and quantize them. We compare different quantization methods for standard and Lorentz-e

Satyabrata Jana, Sahinur Reja

We uncover four exotic coupled spin-charge ground states in the strong coupling limit of the Kondo lattice model at various electronic fillings on a frustrated decorated honeycomb lattice, where each regular honeycomb sublattice point is occupied by three-site triangular units. We employ a hybrid Markov Chain Monte Carlo (hMCMC) simulation method which combi

João Barata, José Guilherme Milhano, Andrey V. Sadofyev, João M. Silva

Ultrarelativistic heavy-ion collisions provide a unique window into far-from-equilibrium states of QCD matter. The initial stages of these events are characterized by highly anisotropic, nonthermal dynamics that precede hydrodynamization, yet they remain largely inaccessible through conventional soft observables. In this work, we show that the substructure o

Junbo Li, Peng Zhou, Rui Meng, Meet P. Vadera

Reinforcement learning (RL) has re-emerged as a natural approach for training interactive LLM agents in real-world environments. However, directly applying the widely used Group Relative Policy Optimization (GRPO) algorithm to multi-turn tasks exposes notable limitations, particularly in scenarios requiring long-horizon reasoning. To address these challenges

Talia Gillis, Riley Stacy, Sam Brumer, Emily Black

This paper compares two legal frameworks -- disparate impact (DI) and unfair, deceptive, or abusive acts or practices (UDAP) -- as tools for evaluating algorithmic discrimination, focusing on the example of fair lending. While DI has traditionally served as the foundation of fair lending law, recent regulatory efforts have invoked UDAP, a doctrine rooted in

Matthew Golden

Explicit Runge-Kutta schemes become impractical when a stiff linear operator is present in the dynamics. This failure mode is quite common in numerical simulations of fluids and plasmas. Lawson proposed Generalized Runge-Kutta Processes for stiff problems in 1967, in which the stiff linear operator is treated fully implicitly via matrix exponentiation. Any R

Neville Francis, Peter Reinhard Hansen, Chen Tong

We take a new perspective on identification in structural dynamic models: rather than imposing restrictions alone, we optimize an objective. While definitive structural identification ultimately requires exogenous economic insight, a weighted correlation-maximizing objective yields an Order- and Scale-Invariant Scheme (OASIS) that selects the orthogonal rota

Yannis Bennacer, Olivier Mousis, Vincent Hue

The Galilean moons exhibit a decrease in bulk density with distance from Jupiter, which may reflect differences in evolutionary paths and water loss. Early in its history, Jupiter was more luminous and may have driven substantial atmospheric escape on Io and Europa. We investigate whether Io could have lost its water inventory while Europa retained its volat

Andrei Shumilin, Sourav Dey, Denisa Coltuneac, Laurentiu Stoleriu

The progress of magnonics ultimately depends on material platforms that offer precise control of spin waves propagation. Here, we put forward a chemical strategy to create locally tunable magnonic crystals by integrating switchable spin-crossover (SCO) molecules with 2D van der Waals magnets. Specifically, we investigate from first principles a hybrid molecu

Rohit V. Nanavati, Tim J. Glover, Matthew J. Coombes, Cunjia Liu

This paper presents a Multi-Robot Multi-Source Term Estimation (MRMSTE) framework that enables teams of mobile robots to collaboratively sample gas concentrations and infer the parameters of an unknown number of airborne releases. The framework is built on a hybrid Bayesian inference scheme that represents the joint multi-source probability density and incor

D. Bagio, G. A. García, O. Márquez

We introduce a novel approach to compute liftings of bosonizations of Nichols algebras of diagonal braided vector spaces of Cartan type which replaces heavy computations with structural maps related to quantum groups. This provides an answer to a question posed by Andruskiewitsch and Schneider, who classified finite-dimensional complex pointed Hopf algebras

Patrick D. Bolton, Jernej F. Kamenik, Martín Novoa-Brunet

Recent Belle II data on $B^+ \to K^+ E_{\rm miss}$ show an excess consistent with a two-body decay involving a light invisible particle with mass around $2.1\,\mathrm{GeV}$. We present a UV-complete explanation based on a Higgsed $U(1)'$ gauge symmetry with a light vector boson $Z'$ and a vector-like top partner, which naturally enhances $b \to s$ transition

Anna R. Gallazzi, Stefano Zibetti, Mark Sargent, Nicolas Bouche'

The cosmic Middle Ages, spanning the last 8-10 Gyr of the Universe, is a critical period in which massive early-formed systems coexist with global star formation quenching in less massive galaxies, yet galaxies experience further dynamical, morphological and chemical evolution. Understanding the relative role of internal drivers and of interaction with the e

Felipe Espinoza-Arancibia, Bogumił Pilecki, Matylda Łukaszewicz

Aims. This study aims to determine empirical intrinsic edges of the classical Cepheids instability strip (IS) in the Small Magellanic Cloud (SMC) galaxy, considering various effects that alter its shape, and compare them with theoretical models and other galaxies. Methods. We used the data of classical fundamental-mode (F) and first-overtone mode (1O) SMC Ce

Raschid Abedin, Wenjun Niu

In this paper, we construct the dual $Y^*_\hbar(\mathfrak d)$ and double $DY_\hbar (\mathfrak d)$ of the Yangian $Y_\hbar (\mathfrak d)$ associated with a cotangent Lie algebra $\mathfrak d=T^*\mathfrak g$. We define a coherent factorization algebra version of the dual Yangian $Y_\hbar^*(\mathfrak d)^{\mathrm{co-op}}$ with opposite coproduct. Furthermore, we

Stefano Wirth, Luca Morf, Ravit Helled

We present an algorithm to efficiently sample the full space of planetary interior density profiles. Our approach uses as few assumptions as possible to pursue an agnostic algorithm. The algorithm avoids the common Markov Chain Monte Carlo method and instead uses an optimisation-based gradient-descent approach designed for computational efficiency. In this w

Paul M. Neves, Takashi Kurumaji, Joshua P. Wakefield, Chi Ian Jess Ip

EuAg$_4$Sb$_2$ is a rhombohedral europium triangle lattice material that exhibits a rich phase diagram of spin moir\'{e} superlattices (SMS) and single-$q$ magnetic phases. In this paper, we characterize the incommensurate phases accessible with field applied in the plane with small angle neutron scattering (SANS). A variety of phases with unusual SANS patte

Thomas Kupfer, Simone Scaringi, Ingrid Pelisoli, Anna F. Pala

Type Ia supernovae (SNe Ia) are fundamental to cosmology and galactic chemical evolution, yet the nature of their progenitor systems remains unresolved. Multiple evolutionary pathways, including single-degenerate, double-degenerate, and helium-donor systems, are thought to contribute to the SN Ia population, but direct observational constraints are limited.

Ali Ahmad Khostovan, Ryan L. Sanders, Alice E. Shapley, Michael W. Topping

We present new constraints on the Mass -- Metallicity (MZR) and Fundamental Metallicity Relations (FMR) using a sample of 34 galaxies at $1.38\leq~z\leq~3.5$ (median $z=2.28$). These galaxies have direct $T_e$ measurements from [O\sc{iii}]4363\AA~and/or [O\sc{ii}]7320,7331\AA~auroral emission lines detected with \textit{JWST}/NIRSpec as part of the AURORA su

Thomas Kupfer, Simone Scaringi, Paul Groot, Boris Gänsicke

Ultracompact Galactic binaries with orbital periods below an hour are among the strongest persistent gravitational-wave (GW) sources in the mHz band and will constitute the dominant population detected by the Laser Interferometer Space Antenna (LISA). Tens of thousands are predicted to be individually resolved, with a substantial fraction bright enough for e

Raymond T. Co, Siu Cheung Lam, Sai Chaitanya Tadepalli, Tomo Takahashi

Thermal warm dark matter (WDM) particles with $m_{\rm WDM} \leq 1~\mathrm{keV}$ are ruled out at more than $4\sigma$ by multiple observational probes, owing to the strong suppression of small-scale structure induced by early-time free-streaming. Recently, it was highlighted that a small admixture of $\sim1\%$ ($f_{\rm CDM} \sim\!0.01$) cold dark matter (CDM)

A. A. Burkov

We discuss the quantum geometric contribution to the diffusion constant and the DC conductivity in metals and semimetals with linear Dirac dispersion. We demonstrate that, for systems with perfectly linear dispersion, there exists a clear and rigorous separation of the quantum geometric from the ordinary band velocity contributions to the diffusion constant,

Kiana Salehi, Avery Broderick

The bright ring-like structures observed in the images of M87* and SgrA* captured by the Event Horizon Telescope strongly support the validity of general relativity. Lensed images of the emission region, often referred to as photon rings in this context, are a direct consequence of the unstable dynamics of null geodesics near the spherical photon orbit in th

Oscar J. C. Dias, David Sola Gil, Jorge E. Santos

We perform a comprehensive study of the linear stability of rotating BTZ black holes under massive scalar field perturbations with double-trace boundary conditions. While BTZ black holes are stable under standard Dirichlet and Neumann boundary conditions, we demonstrate that they can develop instabilities when subjected to double-trace boundary conditions. O

B. Trefoloni, E. Nardini, S. Carniani, E. Lusso

Quasars (QSOs) emit an enormous amount of light as a result of the accretion of gas onto supermassive black holes (SMBHs). Thanks to their luminosity, the most distant known QSOs allow us to trace the growth of SMBHs deep into the epoch of reionisation. In this work, we employed $JWST$/NIRSpec observations of eight luminous (log$(L_{3000\,A^{\circ}}/(erg \,

David W. Hertzog, Martin Hoferichter

We review the status of the anomalous magnetic moment of the muon as a precision probe of physics beyond the Standard Model (SM) after the release of the final results from the Fermi National Accelerator Laboratory (FNAL) Muon $g-2$ experiment and the second White Paper of the Muon $g-2$ Theory Initiative. While the SM prediction requires further improvement

Hanlin Wang, Hao Ouyang, Qiuyu Wang, Yue Yu

We present WorldCanvas, a framework for promptable world events that enables rich, user-directed simulation by combining text, trajectories, and reference images. Unlike text-only approaches and existing trajectory-controlled image-to-video methods, our multimodal approach combines trajectories -- encoding motion, timing, and visibility -- with natural langu

Chun-Wei Tuan Mu, Cheng-De Fan, Jia-Bin Huang, Yu-Lun Liu

Depth-of-field control is essential in photography, but achieving perfect focus often requires multiple attempts or specialized equipment. Single-image refocusing is still difficult. It involves recovering sharp content and creating realistic bokeh. Current methods have significant drawbacks. They require all-in-focus inputs, rely on synthetic data from simu

Sihan Xu, Ziqiao Ma, Wenhao Chai, Xuweiyi Chen

Inspired by the success of generative pretraining in natural language, we ask whether the same principles can yield strong self-supervised visual learners. Instead of training models to output features for downstream use, we train them to generate embeddings to perform predictive tasks directly. This work explores such a shift from learning representations t

Qihao Liu, Chengzhi Mao, Yaojie Liu, Alan Yuille

Conventional evaluation methods for multimodal LLMs (MLLMs) lack interpretability and are often insufficient to fully disclose significant capability gaps across models. To address this, we introduce AuditDM, an automated framework that actively discovers and rectifies MLLM failure modes by auditing their divergence. AuditDM fine-tunes an MLLM as an auditor

Jinjie Mai, Chaoyang Wang, Guocheng Gordon Qian, Willi Menapace

While image editing has advanced rapidly, video editing remains less explored, facing challenges in consistency, control, and generalization. We study the design space of data, architecture, and control, and introduce \emph{EasyV2V}, a simple and effective framework for instruction-based video editing. On the data side, we compose existing experts with fast

Chaoyang Wang, Kaituo Feng, Dongyang Chen, Zhongyu Wang

Recent advances have shown that multimodal large language models (MLLMs) benefit from multimodal interleaved chain-of-thought (CoT) with vision tool interactions. However, existing open-source models often exhibit blind tool-use reasoning patterns, invoking vision tools even when they are unnecessary, which significantly increases inference overhead and degr

Qihao Liu, Luoxin Ye, Wufei Ma, Yu-Cheng Chou

Large language models (LLMs) with explicit reasoning capabilities excel at mathematical reasoning yet still commit process errors, such as incorrect calculations, brittle logic, and superficially plausible but invalid steps. In this paper, we introduce Generative Adversarial Reasoner, an on-policy joint training framework designed to enhance reasoning by co-

Guibao Shen, Yihua Du, Wenhang Ge, Jing He

The rapid growth of stereoscopic displays, including VR headsets and 3D cinemas, has led to increasing demand for high-quality stereo video content. However, producing 3D videos remains costly and complex, while automatic Monocular-to-Stereo conversion is hindered by the limitations of the multi-stage ``Depth-Warp-Inpaint'' (DWI) pipeline. This paradigm suff

Nikhil Prakash, Donghao Ren, Dominik Moritz, Yannick Assogba

Prior studies investigating the internal workings of LLMs have uncovered sparse subnetworks, often referred to as circuits, that are responsible for performing specific tasks. Additionally, it has been shown that model performance improvement through fine-tuning often results from the strengthening of existing circuits in the model. Taken together, these fin

Xin Lin, Meixi Song, Dizhe Zhang, Wenxuan Lu

In this work, we present a panoramic metric depth foundation model that generalizes across diverse scene distances. We explore a data-in-the-loop paradigm from the view of both data construction and framework design. We collect a large-scale dataset by combining public datasets, high-quality synthetic data from our UE5 simulator and text-to-image models, and

Peter Chen, Xiaopeng Li, Ziniu Li, Wotao Yin

This paper examines the exploration-exploitation trade-off in reinforcement learning with verifiable rewards (RLVR), a framework for improving the reasoning of Large Language Models (LLMs). Recent studies suggest that RLVR can elicit strong mathematical reasoning in LLMs through two seemingly paradoxical mechanisms: spurious rewards, which suppress exploitat

Andrew Wagenmaker, Perry Dong, Raymond Tsao, Chelsea Finn

Standard practice across domains from robotics to language is to first pretrain a policy on a large-scale demonstration dataset, and then finetune this policy, typically with reinforcement learning (RL), in order to improve performance on deployment domains. This finetuning step has proved critical in achieving human or super-human performance, yet while muc

Qihang Rao, Borui Zhang, Wenzhao Zheng, Jie Zhou

Recent advances in multimodal models highlight the pivotal role of image tokenization in high-resolution image generation. By compressing images into compact latent representations, tokenizers enable generative models to operate in lower-dimensional spaces, thereby improving computational efficiency and reducing complexity. Discrete tokenizers naturally alig

Yuanchen Ju, Yongyuan Liang, Yen-Jen Wang, Nandiraju Gireesh

Mobile manipulators in households must both navigate and manipulate. This requires a compact, semantically rich scene representation that captures where objects are, how they function, and which parts are actionable. Scene graphs are a natural choice, yet prior work often separates spatial and functional relations, treats scenes as static snapshots without o

Yuqun Wu, Chih-hao Lin, Henry Che, Aditi Tiwari

We investigate the problem of identifying objects that have been added, removed, or moved between a pair of captures (images or videos) of the same scene at different times. Accurately identifying verifiable changes is extremely challenging -- some objects may appear to be missing because they are occluded or out of frame, while others may appear different d

Mingfei Chen, Yifan Wang, Zhengqin Li, Homanga Bharadhwaj

Prior works on 3D hand trajectory prediction are constrained by datasets that decouple motion from semantic supervision and by models that weakly link reasoning and action. To address these, we first present the EgoMAN dataset, a large-scale egocentric dataset for interaction stage-aware 3D hand trajectory prediction with 219K 6DoF trajectories and 3M struct

Xiaoyan Cong, Haotian Yang, Angtian Wang, Yizhi Wang

Instruction-based video editing aims to modify an input video according to a natural-language instruction while preserving content fidelity and temporal coherence. However, existing diffusion-based approaches are often trained on paired data of simple editing operations, which fundamentally limits their ability to generalize to diverse and complex, real-worl

Hao Li, Daiwei Lu, Xing Yao, Nicholas Kavoussi

In this paper, we present Endo-SemiS, a semi-supervised segmentation framework for providing reliable segmentation of endoscopic video frames with limited annotation. EndoSemiS uses 4 strategies to improve performance by effectively utilizing all available data, particularly unlabeled data: (1) Cross-supervision between two individual networks that supervise

Kaixin Ding, Yang Zhou, Xi Chen, Miao Yang

Recent advances in Text-to-Image (T2I) generative models, such as Imagen, Stable Diffusion, and FLUX, have led to remarkable improvements in visual quality. However, their performance is fundamentally limited by the quality of training data. Web-crawled and synthetic image datasets often contain low-quality or redundant samples, which lead to degraded visual

Pierre Fernandez, Tom Sander, Hady Elsahar, Hongyan Chang

Generation-time text watermarking embeds statistical signals into text for traceability of AI-generated content. We explore *post-hoc watermarking* where an LLM rewrites existing text while applying generation-time watermarking, to protect copyrighted documents, or detect their use in training or RAG via watermark radioactivity. Unlike generation-time approa

G. D'Ambrosio, A. M. Iyer, F. Mahmoudi, S. Neshatpour

Rare kaon decays provide sensitive tests of new physics. In this work, we focus on scalar and pseudoscalar operators, analysing the $K\to \pi \ell^+\ell^-$ and $K\to \ell^+\ell^-$ decays. We highlight the complementary role of different modes: $K^+\to\pi^+\ell^+\ell^-$, in particular the forward-backward asymmetry in the muon channel as a clean probe of scal

Eric Todd, Jannik Brinkmann, Rohit Gandikota, David Bau

We investigate the mechanisms that arise when transformers are trained to solve arithmetic on sequences where tokens are variables whose meaning is determined only through their interactions in-context. While prior work has studied transformers in settings where the answer relies on fixed parametric or geometric information encoded in token embeddings, we de

Rahul Bhargava, Malene Hornstrup Jespersen, Emily Boardman Ndulue, Vivica Dsouza

AI technologies have rapidly moved into business and research applications that involve large text corpora, including computational journalism research and newsroom settings. These models, trained on extant data from various sources, can be conceptualized as historical artifacts that encode decades-old attitudes and stereotypes. This paper investigates one s

Shuyuan Tu, Yueming Pan, Yinming Huang, Xintong Han

Current diffusion-based acceleration methods for long-portrait animation struggle to ensure identity (ID) consistency. This paper presents FlashPortrait, an end-to-end video diffusion transformer capable of synthesizing ID-preserving, infinite-length videos while achieving up to 6x acceleration in inference speed. In particular, FlashPortrait begins by compu

Yushi Hu, Reyhane Askari-Hemmat, Melissa Hall, Emily Dinan

Reward models (RMs) are essential for training large language models (LLMs), but remain underexplored for omni models that handle interleaved image and text sequences. We introduce Multimodal RewardBench 2 (MMRB2), the first comprehensive benchmark for reward models on multimodal understanding and (interleaved) generation. MMRB2 spans four tasks: text-to-ima

Philipp A. Hoehn, Josh Kirklin

In gauge theories, globally charged observables necessarily depend non-locally on the kinematical fields, with this dependence extending to the asymptotic boundary of spacetime. Despite this, we show that a subset of such observables can be consistently regarded as local to the bulk, in a manner that respects microcausality and leaves locality properties of

Manuel Bentele, Onur Altinordu, Jan Körner, Andreas Podelski

The correct use of a Hardware Abstraction Layer (HAL) interface in embedded applications is crucial to prevent malfunctions, crashes, or even hardware damage. Software model checking has been successfully applied to check interface specifications in application programs, but its employment in industrial practice is hindered by its unpredictability (whether i

Jinghuan Shang, Harsh Patel, Ran Gong, Karl Schmeckpeper

Synthetic 3D scenes are essential for developing Physical AI and generative models. Existing procedural generation methods often have low output throughput, creating a significant bottleneck in scaling up dataset creation. In this work, we introduce Sceniris, a highly efficient procedural scene generation framework for rapidly generating large-scale, collisi

Ratip Emin Berker, Emanuel Tewolde, Vincent Conitzer, Mingyu Guo

Core stability is a natural and well-studied notion for group fairness in multi-winner voting, where the task is to select a committee from a pool of candidates. We study the setting where voters either approve or disapprove of each candidate; here, it remains a major open problem whether a core-stable committee always exists. In this work, we develop an app

Nicolas Curien, William Fleurat, Adrianus Twigt

Can we obtain a Brownian CRT of mass $1/2$ from a CRT of mass $1$ by cutting certain branches? In this paper, we will answer that question in the much more general setting of self-similar Markov trees. Self-similar Markov trees (ssMt) are random decorated trees that encode the genealogy of a system of particles carrying positive labels, and where particles u

Daniel Kaplan, Alexander C. Tyner, Eva Y. Andrei, J. H. Pixley

The world of 2D materials is rapidly expanding with new discoveries of stackable and twistable layered systems composed of lattices of different symmetries, orbital character, and structural motifs. Often, however, it is not clear a priori whether a pair of monolayers twisted at a small angle will exhibit correlated or interaction-driven phenomena. The compu

Haichao Zhang, Yao Lu, Lichen Wang, Yunzhe Li

Video Large Language Models (VLLMs) unlock world-knowledge-aware video understanding through pretraining on internet-scale data and have already shown promise on tasks such as movie analysis and video question answering. However, deploying VLLMs for downstream tasks such as video recommendation remains challenging, since real systems require multi-video inpu

Jason Aebischer, Luigi C. Bresciani, Nudzeim Selimovic

The complete set of one-loop anomalous dimensions for general Effective Field Theories (EFTs) is derived using on-shell methods. Combined with previous findings for the bosonic sector, the obtained results conclude the computation of the complete set of leading order Renormalization Group Equations (RGEs) in arbitrary gauge EFTs containing scalar and fermion

Greger Torgrimsson

In this paper, we show how to solve the Dirac equation, $(i\gamma^\mu[\partial_\mu+ieA_\mu(t,{\bf x})]-m)\psi=0$, on a GPU. This is orders of magnitude faster than solving it on CPU and allows us to consider background fields, $A_\mu(t,{\bf x})$, that depend on $2+1$ or even $3+1$ coordinates. Our approach is conveniently implemented using the computational

Julián López, Virginia Mazzone, M. Leticia Rubio Puzzo, Juan Cruz Moreno

The evacuation of pedestrians from enclosed spaces represents a key problem in safety engineering and infrastructure design. Analyzing the collective dynamics that emerge during evacuation processes requires simulation tools capable of capturing individual interactions and spatial constraints realistically. In this work, we present \textit{SiCoBioNa}, an ope

Oliver Hart, David T. Stephen, Evan Wickenden, Rahul Nandkishore

Contextuality is arguably the fundamental property that makes quantum mechanics different from classical physics. It is responsible for quantum computational speedups in both magic-state-injection-based and measurement-based models of computation, and can be directly probed in a many-body setting by multiplayer nonlocal quantum games. Here, we discuss a fami

Norika Wada, Kohei Yamashita, Ryo Kawahara, Ko Nishino

Knowledge of the physical material properties governing the dynamics of a real-world object becomes necessary to accurately anticipate its response to unseen interactions. Existing methods for estimating such physical material parameters from visual data assume homogeneous single-material objects, pre-learned dynamics, or simplistic topologies. Real-world ob

Tzu-Han Lin, Wei-Lin Chen, Chen-An Li, Hung-yi Lee

Equipping large language models (LLMs) with search engines via reinforcement learning (RL) promises effective search agents. However, adaptively balancing internal parametric knowledge with external search remains a challenge, as overreliance on search introduces unnecessary cost and risks exposure to noisy or malicious content, while relying solely on param

Arhan Jain, Mingtong Zhang, Kanav Arora, William Chen

A significant challenge for robot learning research is our ability to accurately measure and compare the performance of robot policies. Benchmarking in robotics is historically challenging due to the stochasticity, reproducibility, and time-consuming nature of real-world rollouts. This challenge is exacerbated for recent generalist policies, which has to be

Valay Bundele, Mehran Hosseinzadeh, Hendrik P. A. Lensch

Accurate surgical instrument segmentation in endoscopy is crucial for computer-assisted interventions, yet remains challenging due to frequent occlusions, rapid motion, and long-term instrument re-entry. While SAM3 provides a powerful spatio-temporal framework for video object segmentation, its performance in surgical scenes is limited by indiscriminate memo

Arnab Adhikary, S. E. Skelton, Alberto Nocera, Mona Berciu

Simulating electron-phonon interactions on quantum computers remains challenging, with most algorithmic effort focused on Hamiltonian simulation and circuit optimization. In this work, we study the single-electron Holstein model and propose an initial-state ansatz that substantially enhances ground state overlap in the strong coupling regime, thereby reducin

Francesco Anna Mele, Filippo Girardi, Senrui Chen, Marco Fanizza

The random purification channel, which, given $n$ copies of an unknown mixed state $\rho$, prepares $n$ copies of an associated random purification, has proved to be an extremely valuable tool in quantum information theory. In this work, we construct a Gaussian version of this channel that, given $n$ copies of a bosonic passive Gaussian state, prepares $n$ c

Edward W. Kolb, Andrew J. Long, Evan McDonough, Jingyuan Wang

We study the cosmological gravitational particle production (CGPP) of spin-3/2 particles during and after cosmic inflation, and map the parameter space that can realize the observed dark matter density in stable spin-3/2 particles. Originally formulated by Rarita and Schwinger, the relativistic theory of a massive spin-3/2 field later found a home in supergr

Alexander List, A. Daniel Boese, Johannes Hoja

Molecular crystals possess a highly complex crystallographic landscape which in many cases results in the experimental observation of multiple crystal structures for the same compound. Accurate results can often be obtained for such systems by employing periodic density functional theory using hybrid functionals; however, this is not always computationally f

Astrid Brull, Sara Aguti, Véronique Bolduc, Ying Hu

The application of Machine Learning (ML) to the diagnosis of rare diseases, such as collagen VI-related dystrophies (COL6-RD), is fundamentally limited by the scarcity and fragmentation of available data. Attempts to expand sampling across hospitals, institutions, or countries with differing regulations face severe privacy, regulatory, and logistical obstacl

Chao Gao, Liren Shan, Vaidehi Srinivas, Aravindan Vijayaraghavan

We study the problem of finding confidence ellipsoids for an arbitrary distribution in high dimensions. Given samples from a distribution $D$ and a confidence parameter $\alpha$, the goal is to find the smallest volume ellipsoid $E$ which has probability mass $\mathbb{P}_{D}[E] \ge 1-\alpha$. Ellipsoids are a highly expressive class of confidence sets as the