May 2025 arXiv papers — page 41

Tianyu Fu, Yi Ge, Yichen You, Enshu Liu

Large Language Models (LLMs) achieve impressive reasoning capabilities at the cost of substantial inference overhead, posing substantial deployment challenges. Although distilled Small Language Models (SLMs) significantly enhance efficiency, their performance suffers as they fail to follow LLMs' reasoning paths. Luckily, we reveal that only a small fraction

Kyra Dalbo, Yumna Ahmed, HeuiChan Lim

The complexity of modern Just-In-Time (JIT) compiler optimization poses significant challenges for developers seeking to understand and debug intermediate representation (IR) behavior. This work introduces JITScope, an interactive visualization framework that illustrates how IR nodes and instructions evolve across compilation phases. The system features a fu

Xianling Mu, Joseph Ternasky, Fuat Alican, Yigit Ihlamur

Early-stage startup investment is a high-risk endeavor characterized by scarce data and uncertain outcomes. Traditional machine learning approaches often require large, labeled datasets and extensive fine-tuning, yet remain opaque and difficult for domain experts to interpret or improve. In this paper, we propose a transparent and data-efficient investment d

Yajiao Liu, Congliang Chen, Junchi Yang, Ruoyu Sun

Training large language models with data collected from various domains can improve their performance on downstream tasks. However, given a fixed training budget, the sampling proportions of these different domains significantly impact the model's performance. How can we determine the domain weights across different data domains to train the best-performing

Francesco Cozzi, Marco Pangallo, Alan Perotti, André Panisson

Agent-Based Models (ABMs) are powerful tools for studying emergent properties in complex systems. In ABMs, agent behaviors are governed by local interactions and stochastic rules. However, these rules are, in general, non-differentiable, limiting the use of gradient-based methods for optimization, and thus integration with real-world data. We propose a novel

Naizhu Jin, Zhong Li, Tian Zhang, Qingkai Zeng

With the widespread application of large language models in code generation, recent studies demonstrate that employing additional Chain-of-Thought generation models can significantly enhance code generation performance by providing explicit reasoning steps. However, as external components, CoT models are particularly vulnerable to backdoor attacks, which exi

Abhijit Biswas, Laila S. Busaleh, David I. Ketcheson, Carlos Muñoz-Moncayo

We study a first-order hyperbolic approximation of the nonlinear Schr\"odinger (NLS) equation. We show that the system is strictly hyperbolic and possesses a modified Hamiltonian structure, along with at least three conserved quantities that approximate those of NLS. We provide families of explicit standing-wave solutions to the hyperbolic system, which are

Abdullah Al Mamun, Pollob Chandra Ray, Md Rahat Ul Nasib, Akash Das

The rapid advancement of deep learning in medical image analysis has greatly enhanced the accuracy of skin cancer classification. However, current state-of-the-art models, especially those based on transfer learning like ResNet50, come with significant computational overhead, rendering them impractical for deployment in resource-constrained environments. Thi

Esra Adiyeke, Tianqi Liu, Venkata Sai Dheeraj Naganaboina, Han Li

Traditional methods of surgical decision making heavily rely on human experience and prompt actions, which are variable. A data-driven system generating treatment recommendations based on patient states can be a substantial asset in perioperative decision-making, as in cases of intraoperative hypotension, for which suboptimal management is associated with ac

Shreyas Gururaj, Lars Grüne, Wojciech Samek, Sebastian Lapuschkin

Overfitting is a well-known issue extending even to state-of-the-art (SOTA) Machine Learning (ML) models, resulting in reduced generalization, and a significant train-test performance gap. Mitigation measures include a combination of dropout, data augmentation, weight decay, and other regularization techniques. Among the various data augmentation strategies,

Abbavaram Gowtham Reddy, Celia Rubio-Madrigal, Rebekka Burkholz, Krikamol Muandet

Distribution shifts introduce uncertainty that undermines the robustness and generalization capabilities of machine learning models. While conventional wisdom suggests that learning causal-invariant representations enhances robustness to such shifts, recent empirical studies present a counterintuitive finding: (i) empirical risk minimization (ERM) can rival

Rami Cassia, Rich Kerswell

In compressible fluid flow, reconstructing shocks, discontinuities, rarefactions, and their interactions from sparse measurements is an important inverse problem with practical applications. Moreover, physics-informed machine learning has recently become an increasingly popular approach for performing reconstructions tasks. In this work we explore a machine

Hanzhe Liang

Multimodal feature reconstruction is a promising approach for 3D anomaly detection, leveraging the complementary information from dual modalities. We further advance this paradigm by utilizing multi-modal mentor learning, which fuses intermediate features to further distinguish normal from feature differences. To address these challenges, we propose a novel

Yifan Wang, Kenneth P. Birman

Today's cloud-hosted applications and services are complex systems, and a performance or functional instability can have dozens or hundreds of potential root causes. Our hypothesis is that by combining the pattern matching capabilities of modern AI tools with a natural multi-modal RAG LLM interface, problem identification and resolution can be simplified. AR

Lina Zhao, Zihao Bian, Qingyue Chen, Yafang Li

Focused Ultrasound Ablation Surgery (FUAS) has emerged as a promising non-invasive therapeutic modality, valued for its safety and precision. Nevertheless, its clinical implementation entails intricate tasks such as multimodal image interpretation, personalized dose planning, and real-time intraoperative decision-making processes that demand intelligent assi

Yonggwan Shin, Yire Shin, Jeong-Soo Park

Accurately estimating high quantiles beyond the largest observed value is crucial for risk assessment and devising effective adaptation strategies to prevent a greater disaster. The generalized extreme value distribution is widely used for this purpose, with L-moment estimation (LME) and maximum likelihood estimation (MLE) being the primary methods. However,

Scott C. Egbert, James Greenberg, Brendan M. Heffernan, William F. McGrew

Compact, ultra-low phase noise 10 GHz signals are essential for modern radar, coherent communications, and time-frequency metrology, especially with rising demands for additional spectral purity and portability. Optical frequency division (OFD) of ultra-stable optical references produce the lowest noise microwaves, but typically rely on ultra-low-expansion c

Yanda Geng, Shouvik Mukherjee, Swarnav Banik, Monica Gutierrez Galan

We describe an apparatus that efficiently produces $^{23}$Na Bose-Einstein condensates (BECs) in a hybrid trap that combines a quadrupole magnetic field with a far-detuned optical dipole trap. Using a Bayesian optimization framework, we systematically optimize all BEC production parameters in modest sized batches of highly correlated parameters. Furthermore,

Brett Bissey, Kyle Gatesman, Walker Dimon, Mohammad Alam

This paper introduces a comprehensive framework designed to analyze and secure decision-support systems trained with Deep Reinforcement Learning (DRL), prior to deployment, by providing insights into learned behavior patterns and vulnerabilities discovered through simulation. The introduced framework aids in the development of precisely timed and targeted ob

Xiao Liu, Da Yin, Zirui Wu, Yansong Feng

Large Language Models (LLMs) can enhance their reasoning capabilities by using external tools. However, many tasks lack predefined tools. Prior works have explored instructing LLMs to generate tools on their own, but such approaches depend heavily on internal knowledge and struggle when tasks fall outside the model's knowledge scope. To address this limitati

Giuseppe Conti, Raffaella Paoletti

Geodesic domes, convex polyhedrons with almost spherical shape or parts of them, were the subject of great attention in the twenty years between the mid-1950s and the 1970s, especially thanks to Richard Buckminster Fuller. After a building boom, mostly in the United States, their construction interest declined but their geometric characteristics, studied by

Yehui Tang, Xiaosong Li, Fangcheng Liu, Wei Guo

The surgence of Mixture of Experts (MoE) in Large Language Models promises a small price of execution cost for a much larger model parameter count and learning capacity, because only a small fraction of parameters are activated for each input token. However, it is commonly observed that some experts are activated far more often than others, leading to system

Shashank Sharma, Janina Hoffmann, Vinay Namboodiri

Hierarchical reinforcement learning (HRL) decomposes the policy into a manager and a worker, enabling long-horizon planning but introducing a performance gap on tasks requiring agility. We identify a root cause: in subgoal-based HRL, the manager's goal representation is typically learned without constraints on reachability or temporal distance from the curre

Dario Satriani, Enzo Veltri, Donatello Santoro, Paolo Papotti

Factuality in Large Language Models (LLMs) is a persistent challenge. Current benchmarks often assess short factual answers, overlooking the critical ability to generate structured, multi-record tabular outputs from parametric knowledge. We demonstrate that this relational fact retrieval is substantially more difficult than isolated point-wise queries, even

Fuhai Wang, Zhe Li, Rujing Xiong, Tiebin Mi

Accurate 3D localization is essential for realizing advanced sensing functionalities in next-generation Wi-Fi communication systems. This study investigates the potential of multistatic localization in Wi-Fi networks through the deployment of multiple cooperative antenna arrays. The collaborative gain offered by these arrays is twofold: (i) intra-array coher

Christopher R. Kitching, Lucía S. Ramirez, Maxi San Miguel, Tobias Galla

We study how zealotry and nonlinear social impact affect consensus formation in the nonlinear voter model, evolutionary games, and the partisan voter model. In all three models, consensus is an absorbing state in finite populations, while coexistence is a possible outcome of the deterministic dynamics. We show that sufficiently strong zealotry, i.e. the pres

Abhijit Talluri

Federated Learning (FL) has emerged as a critical paradigm for enabling privacy-preserving machine learning, particularly in regulated sectors such as finance and healthcare. However, standard FL strategies often encounter significant operational challenges related to fault tolerance, system resilience against concurrent client and server failures, and the p

Pedro Pereira, José Gonçalves, João Vitorino, Eva Maia

This work addresses JavaScript malware detection to enhance client-side web application security with a behavior-based system. The ability to detect malicious JavaScript execution sequences is a critical problem in modern web security as attack techniques become more sophisticated. This study introduces a new system for detecting JavaScript malware using a D

A. Granelli, D. Meloni, M. Parriciatu, J. T. Penedo

In the presence of a finite modular flavour symmetry, fermion mass hierarchies may be generated by a slight deviation of the modulus from a symmetric point. We point out that this small parameter governing charged-lepton mass hierarchies may also be responsible for the breaking of lepton number in a symmetry-protected low-scale seesaw, sourcing active neutri

Anas Jnini, Flavio Vella

Natural-gradient methods markedly accelerate the training of Physics-Informed Neural Networks (PINNs), yet their Gauss--Newton update must be solved in the parameter space, incurring a prohibitive $O(n^3)$ time complexity, where $n$ is the number of network trainable weights. We show that exactly the same step can instead be formulated in a generally smaller

Yi-Yao Li, Jing Song, Eulogio Oset, Wei-Hong Liang

We study the decay $\Lambda_c^+ \to \Lambda \pi^+ \pi^+ \pi^-$, focusing on the production of the $\Sigma^*(1430)$ resonance observed by the Belle Collaboration. Interpreted as a dynamically generated state from meson-baryon interactions in the chiral unitary approach, the $\Sigma^*(1430)$ signal is shown to be enhanced by a triangle singularity involving in

Paolo Cosentino, Francesco Malizia

We prove that spike sequences related to a nonlinear problem of Grad-Shafranov type are always simple and always converge toward interior points of the domain. This sharpens the blow-up analysis carried out by Bartolucci-Jevnikar-Wu [Calc. Var. 2025] and provides a converse to the existence result for spike sequences obtained by Wei [Proc. Edinb. Math. Soc.

Wouter Jongeneel

Recently, Kvalheim and Sontag provided a generalized global Hartman-Grobman theorem for equilibria under asymptotically stable continuous vector fields. By leveraging topological properties of Lyapunov functions, their theorem works without assuming hyperbolicity. We extend their theorem to a class of possibly discontinuous vector fields, in particular, to v

Gen Li, Changxiao Cai

Diffusion models have emerged as a powerful paradigm for modern generative modeling, demonstrating strong potential for large language models (LLMs). Unlike conventional autoregressive (AR) models that generate tokens sequentially, diffusion models allow for parallel sampling, offering a promising path to accelerate generation and eliminate the left-to-right

Hovhannes Tamoyan, Subhabrata Dutta, Iryna Gurevych

Factual incorrectness in generated content is one of the primary concerns in ubiquitous deployment of large language models (LLMs). Prior findings suggest LLMs can (sometimes) detect factual incorrectness in their generated content (i.e., fact-checking post-generation). In this work, we provide evidence supporting the presence of LLMs' internal compass that

Maria Cristina Carrisi, Mirko Marras, Sara Vergallo

Younger generations are growing up in a world increasingly shaped by intelligent technologies, making early AI literacy crucial for developing the skills to critically understand and navigate them. However, education in this field often emphasizes tool-based learning, prioritizing usage over understanding the underlying concepts. This lack of knowledge leave

Xiusi Chen, Shanyong Wang, Cheng Qian, Hongru Wang

In high-stakes domains such as healthcare and finance, effective decision-making demands not just accurate outcomes but transparent and explainable reasoning. However, current language models often lack the structured deliberation needed for such tasks, instead generating decisions and justifications in a disconnected, post-hoc manner. To address this, we pr

Xiao Liu, Xinyi Dong, Xinyang Gao, Yansong Feng

Recent advancements in large language models (LLMs) demonstrate strong potential for generating novel research ideas, yet such ideas often struggle with feasibility and effectiveness. In this paper, we investigate whether augmenting LLMs with relevant data during the ideation process can improve idea quality. Our framework integrates data at two stages: (1)

Xingyu Zhou, Yulian Wu, Wenqian Weng, Francesco Orabona

In this paper, we theoretically study the offline alignment of language models with human preference feedback, under both preference label corruption and privacy protections. To this end, we propose Square$\chi$PO, a simple one-line change to $\chi$PO where the standard log-loss is replaced by a new square loss over probability. Thanks to the inherent proper

Tuo Liu, El Mehdi Saad, Wojciech Kotłowski, Francesco Orabona

Dual averaging and gradient descent with their stochastic variants stand as the two canonical recipe books for first-order optimization: Every modern variant can be viewed as a descendant of one or the other. In the convex regime, these algorithms have been deeply studied, and we know that they are essentially equivalent in terms of theoretical guarantees. O

Maoli Liu, Zhuohua Li, Xiangxiang Dai, John C. S. Lui

Conversational recommender systems proactively query users with relevant "key terms" and leverage the feedback to elicit users' preferences for personalized recommendations. Conversational contextual bandits, a prevalent approach in this domain, aim to optimize preference learning by balancing exploitation and exploration. However, several limitations hinder

Thomas Hussenot-Desenonges, Marion Pillas, Sarah Antier, Patrice Hello

In the study of optical transients, parameter inference is the process of extracting physical information, i.e. constraints on the source's characteristics, by comparing the observed lightcurves to the predictions of different models and finding the model and parameter combination that make the closest match. In the developing field of the study of kilonovae

Zixuan Xie, Xinyu Liu, Rohan Chandra, Shangtong Zhang

Linear TD($\lambda$) is one of the most fundamental reinforcement learning algorithms for policy evaluation. Previously, convergence rates are typically established under the assumption of linearly independent features, which does not hold in many practical scenarios. This paper instead establishes the first $L^2$ convergence rates for linear TD($\lambda$) o

Christian Döding, Barbara Verfürth

Temporal metamaterials are artificially manufactured materials with time-dependent material properties that exhibit interesting phenomena when waves propagate through them. The propagation of electromagnetic waves in such time-varying dielectric media is governed by Maxwell's equations, which lead to wave equations with temporal highly oscillatory coefficien

Xuanwen Ding, Chengjun Pan, Zejun Li, Jiwen Zhang

Evaluating multimodal large language models (MLLMs) is increasingly expensive, as the growing size and cross-modality complexity of benchmarks demand significant scoring efforts. To tackle with this difficulty, we introduce AutoJudger, an agent-driven framework for efficient and adaptive benchmarking of MLLMs that tackles this escalating cost. AutoJudger emp

Jingyuan Huang, Dan Luo, Zihe Ye, Weixin Chen

Social recommender systems facilitate social connections by identifying potential friends for users. Each user maintains a local social network centered around themselves, resulting in a naturally distributed social structure. Recent research on distributed modeling for social recommender systems has gained increasing attention, as it naturally aligns with t

Xihong Yang, Siwei Wang, Fangdi Wang, Jiaqi Jin

Leveraging the powerful representation learning capabilities, deep multi-view clustering methods have demonstrated reliable performance by effectively integrating multi-source information from diverse views in recent years. Most existing methods rely on the assumption of clean views. However, noise is pervasive in real-world scenarios, leading to a significa

Guido Carnevale, Filippo Fabiani, Filiberto Fele, Kostas Margellos

We present a fully-distributed algorithm for Nash equilibrium seeking in aggregative games over networks. The proposed scheme endows each agent with a gradient-based scheme equipped with a tracking mechanism to locally reconstruct the aggregative variable, which is not available to the agents. We show that our method falls into the framework of singularly pe

Prajwal Singh, Anupam Sharma, Pankaj Pandey, Krishna Miyapuram

Reconstructing dynamic visual stimuli from EEG recordings is challenging due to the noisy, non-stationary nature of EEG signals and the limited availability of EEG-video datasets. We present EEGVid, a framework that learns EEG representations using triplet loss and reconstructs dynamic videos with a temporally conditioned GAN. We study what these representat

Zhengchang Kou, Junhang Zhang, Chen Gong, Jie Ji

Ultrasound vascular imaging is limited by acoustic diffraction, restricting visualization of microvessels essential for understanding organ function and disease. Label-free super-resolution methods exploiting endogenous red blood cells have faced challenges in acquisition time and complexity. Here we introduce ultrasound phase microscopy (UPM), a label-free

Tomás Ferreira Chase, Diana López Nacir, Nicolás Yunes

The gravitational waves emitted by massive black hole binaries can be affected by a variety of environmental effects, which, if detected, could inform astrophysics and cosmology. We here study how gravitational waves emitted by black holes in quasi-circular orbits are affected by the presence of an ultra-light, vector-field, dark-matter environment that is m

Nastaran Saadati, Zhanhong Jiang, Joshua R. Waite, Shreyan Ganguly

Low-Rank Adaptation (LoRA) has emerged as one of the most effective, computationally tractable fine-tuning approaches for training Vision-Language Models (VLMs) and Large Language Models (LLMs). LoRA accomplishes this by freezing the pre-trained model weights and injecting trainable low-rank matrices, allowing for efficient learning of these foundation model

Linshuang Diao, Sensen Song, Yurong Qian, Dayong Ren

State Space models (SSMs) such as PointMamba enable efficient feature extraction for point cloud self-supervised learning with linear complexity, outperforming Transformers in computational efficiency. However, existing PointMamba-based methods depend on complex token ordering and random masking, which disrupt spatial continuity and local semantic correlatio

Byungjun Kim, Minju Kim, Hyeonchu Park, Bugeun Kim

As malicious users increasingly employ phonetic substitution to evade hate speech detection, researchers have investigated such strategies. However, two key challenges remain. First, existing studies have overlooked the Korean language, despite its vulnerability to phonetic perturbations due to its phonographic nature. Second, prior work has primarily focuse

Francesco D'Andrea

This is a review (in Italian) on aperiodic tilings of the plane intended for a general audience. First, we recall some basic results about lattices and periodic tilings. Then, we move on to one-dimensional (domino) tilings and Wang tilings. We present a beautiful proof of the existence of an aperiodic set of Wang prototiles due to J. Kari. Next, we discuss P

Jules Watson, Xi Wang, Raymond Liu, Suzanne Stevenson

Within the common LLM use case of text revision, we study LLMs' revision of gendered role nouns (e.g., outdoorsperson/woman/man) and their justifications of such revisions. We evaluate their alignment with feminist and trans-inclusive language reforms for English. Drawing on insight from sociolinguistics, we further assess if LLMs are sensitive to the same c

Yidi Li, Jun Xiao, Zhengda Lu, Yiqun Wang

This work presents a novel text-to-vector graphics generation approach, Dream3DVG, allowing for arbitrary viewpoint viewing, progressive detail optimization, and view-dependent occlusion awareness. Our approach is a dual-branch optimization framework, consisting of an auxiliary 3D Gaussian Splatting optimization branch and a 3D vector graphics optimization b

Denis Bernard, Ludwig Hruza

We give a refined definition of the class of random matrix ensembles introduced in our paper "Structured random matrices and cyclic cumulants: A free probability approach" (arXiv:2309.14315) by extending the so-called fourth axiom to deal with cumulants of disjoint cycles. We argue that the theorems concerning the stability of such ensembles under non-linear

Fengxiang Wang, Mingshuo Chen, Yueying Li, Di Wang

Ultra-high-resolution (UHR) remote sensing (RS) imagery offers valuable data for Earth observation but pose challenges for existing multimodal foundation models due to two key bottlenecks: (1) limited availability of UHR training data, and (2) token explosion caused by the large image size. To address data scarcity, we introduce SuperRS-VQA (avg. 8,376$\time

Junhao Cheng, Yuying Ge, Teng Wang, Yixiao Ge

Recent advances in CoT reasoning and RL post-training have been reported to enhance video reasoning capabilities of MLLMs. This progress naturally raises a question: can these models perform complex video reasoning in a manner comparable to human experts? However, existing video benchmarks primarily evaluate visual perception and grounding abilities, with qu

Dušan Đorđević, Danica Kosanović, Jovana Nikolić, Zoran Petrić

A faithful $(1+1)$ TQFT has recently been constructed, but the existence of a faithful $(2+1)$ TQFT remains an open question, that subsumes the hard problem of linearity of mapping class groups of surfaces. To circumvent the latter problem we construct a subcategory of the category of 3-cobordisms, containing disjoint unions of tori and simplest cobordisms b

Andrej Schwanke, Lyubomir Ivanov, David Salinas, Fabio Ferreira

Large Language Models (LLMs) have recently emerged as effective surrogate models and candidate generators within global optimization frameworks for expensive blackbox functions. Despite promising results, LLM-based methods often struggle in high-dimensional search spaces or when lacking domain-specific priors, leading to sparse or uninformative suggestions.

Shu Wang, Zijun Yao, Shuhuai Zhang, Jianuo Gai

Advancements in large language models (LLMs) have sparked a growing interest in measuring and understanding their behavior through experimental economics. However, there is still a lack of established guidelines for designing economic experiments for LLMs. Inspired by principles from experimental economics with insights from LLM research in artificial intell

Xinyuan Wang, Lian Peng, Xiangcheng Li, Yilin He

Object detection in remote sensing imagery remains a challenging task due to extreme scale variation, dense object distributions, and cluttered backgrounds. While recent detectors such as YOLOv8 have shown promising results, their backbone architectures lack explicit mechanisms to guide multi-scale feature refinement, limiting performance on high-resolution

Anshika Bansal, Alexander Khodjamirian, Thomas Mannel

We suggest a new method to analyse the rare $D_{(s)}\to P \ell^+\ell^-$ decays ($P=\pi,K$), combining QCD light-cone sum rules (LCSR) with hadronic dispersion relations. As our main study case, we consider the $D^+\to\pi^+\ell^+\ell^-$ mode which attracts much of interest from the point of view of GIM cancellation and potential new sources of the FCNC $c\to

Shinjirou Kouzuma

Recent studies indicate that the physical properties of eclipsing binaries can be extracted from the derivatives of their light curves. A classification scheme for the derivatives of light curves would be helpful for identifying key characteristics of eclipsing binaries. In this study, we propose a new classification method for the light curves of overcontac

Micha Christoph, Raphael Steiner

There are many intriguing questions in extremal graph theory that are well-understood in the undirected setting and yet remain elusive for digraphs. A natural instance of such a problem was recently studied by Hons, Klimo\v{s}ov\'{a}, Kucheriya, Mik\v{s}an\'{i}k, Tkadlec and Tyomkyn: What are the digraphs that have to appear as a subgraph in all digraphs of

Qi Yu, Zhichen Zeng, Yuchen Yan, Zhining Liu

Network alignment (NA) aims to identify node correspondence across different networks and serves as a critical cornerstone behind various downstream multi-network learning tasks. Despite growing research in NA, there lacks a comprehensive library that facilitates the systematic development and benchmarking of NA methods. In this work, we introduce PLANETALIG

Ara Basmajian, Blanca Marmolejo, Robert Suzzi Valli

A reciprocal geodesic on a (2,k, $\infty$) Hecke surface is a geodesic loop based at an even order cone point p traversing its path an even number of times. Associated to each reciprocal geodesic is the conjugacy class of a hyperbolic element in the (2,k,$\infty$) Hecke group whose axis passes through a cone point that projects to p. Such an element is calle

James Oldfield, Shawn Im, Sharon Li, Mihalis A. Nicolaou

Multilayer perceptrons (MLPs) are an integral part of large language models, yet their dense representations render them difficult to understand, edit, and steer. Recent methods learn interpretable approximations via neuron-level sparsity, yet fail to faithfully reconstruct the original mapping--significantly increasing model's next-token cross-entropy loss.

Anissa Alloula, Charles Jones, Ben Glocker, Bartłomiej W. Papież

Despite the constant development of new bias mitigation methods for machine learning, no method consistently succeeds, and a fundamental question remains unanswered: when and why do bias mitigation techniques fail? In this paper, we hypothesise that a key factor may be the often-overlooked but crucial step shared by many bias mitigation methods: the definiti

Qishuai Zhong, Zongmin Li, Siqi Fan, Aixin Sun

Effective engagement by large language models (LLMs) requires adapting responses to users' sociodemographic characteristics, such as age, occupation, and education level. While many real-world applications leverage dialogue history for contextualization, existing evaluations of LLMs' behavioral adaptation often focus on single-turn prompts. In this paper, we

Andrew Cleary, Jacob Page

Deep autoencoder neural networks can generate highly accurate, low-order representations of turbulence. We design a new family of autoencoders which are a combination of a 'dense-block' encoder-decoder structure (Page et al, J. Fluid Mech. 991, 2024), an 'implicit rank minimization' series of linear layers acting on the embeddings (Zeng et al, Mach. Learn. S

Dhanesh Ramachandram, Ananya Raval

Competing risks are crucial considerations in survival modelling, particularly in healthcare domains where patients may experience multiple distinct event types. We propose CRISP-NAM (Competing Risks Interpretable Survival Prediction with Neural Additive Models), an interpretable neural additive model for competing risks survival analysis which extends the n

Rocco D'Agostino, Francesco Bajardi

In this paper, we investigate the cosmological dynamics of teleparallel dark energy in the presence of nonzero spatial geometry. Extending previous analyses of nonminimal scalar-tensor theories in the torsion-based framework, we consider different scalar field potentials and examine the resulting background evolution and linear perturbations. Adopting a dyna

Jessica E. Libby-Roberts, Aaron Bello-Arufe, Zachory K. Berta-Thompson, Caleb I. Cañas

Kepler-51 is a 500 Myr G dwarf hosting three "super-puffs" and one low-mass non-transiting planet. Kepler-51d, the coolest (T_eq ~ 350 K) transiting planet in this system, is also one of the lowest density super-puffs known to date (rho_p = 0.038 +/- 0.009 g/cm^3). With a planetary mass of Mp = 5.6 +/- 1.2 Earth masses and a radius of Rp = 9.32 +/- 0.18 Eart

Wenyuan Li, Shunlin Liang, Keyan Chen, Yongzhe Chen

Accurate crop mapping fundamentally relies on modeling multi-scale spatiotemporal patterns, where spatial scales range from individual field textures to landscape-level context, and temporal scales capture both short-term phenological transitions and full growing-season dynamics. Transformer-based remote sensing foundation models (RSFMs) offer promising pote

Whenty Ariyanti, Kuan-Yu Chen, Sabato Marco Siniscalchi, Hsin-Min Wang

Perceptual voice quality assessment plays a vital role in diagnosing and monitoring voice disorders. Traditional methods, such as the Consensus Auditory-Perceptual Evaluation of Voice (CAPE-V) and the Grade, Roughness, Breathiness, Asthenia, and Strain (GRBAS) scales, rely on expert raters and are prone to inter-rater variability, emphasizing the need for ob

Muhammad Imran, Wayne G. Brisbane, Li-Ming Su, Jason P. Joseph

Background and objective: Micro-ultrasound (micro-US) is a novel imaging modality with diagnostic accuracy comparable to MRI for detecting clinically significant prostate cancer (csPCa). We investigated whether artificial intelligence (AI) interpretation of micro-US can outperform clinical screening methods using PSA and digital rectal examination (DRE). Met

Bidyarthi Paul, Jalisha Jashim Era, Mirazur Rahman Zim, Tahmid Sattar Aothoi

Solving Bengali Math Word Problems (MWPs) remains a major challenge in natural language processing (NLP) due to the language's low-resource status and the multi-step reasoning required. Existing models struggle with complex Bengali MWPs, largely because no human-annotated Bengali dataset has previously addressed this task. This gap has limited progress in Be

Javier López Miras, Fuensanta Vilches

We introduce $\texttt{mosca}$, a $\texttt{Mathematica}$ package designed to facilitate on-shell calculations in effective field theories (EFTs). This initial release focuses on the reduction of Green's bases to physical bases, as well as transformations between arbitrary operator bases. The core of the package is based on a diagrammatic on-shell matching pro

Chayan Mondal, Sudhanshu Barway

We present the discovery of a peculiar central stellar structure in the collisional ring galaxy AM0644-741 using HST imaging and MUSE integral field unit (IFU) data. We identified two S\'ersic components with a S\'ersic index of 1.72 (inner part) and 1.11 (outer part) in the HST F814W band optical image using \textsc{Galfit}. We utilized the MUSE data cube t

Vsevolod Vladimirovich Deriushkin

In this paper I want to suggest a new solution to the problem of musical tuning. On one hand, I see it as a generalization of Just Intonation (JI) to inharmonic timbers, on another, as a unification of spectral interference and harmonicity contributions to consonance within a single framework. The main achievement of the work is the ability to mathematically

Xupeng Zhu, Yu Qi, Yizhe Zhu, Robin Walters

Transformer architectures can effectively learn language-conditioned, multi-task 3D open-loop manipulation policies from demonstrations by jointly processing natural language instructions and 3D observations. However, although both the robot policy and language instructions inherently encode rich 3D geometric structures, standard transformers lack built-in g

Bernd Heidergott, Frank den Hollander, Ines Lindner, Azadeh Parvaneh

This paper develops a mathematical framework to study signal networks, in which nodes can be active or inactive, and their activation or deactivation is driven by external signals and the states of the nodes to which they are connected via links. The focus is on determining the optimal number of key nodes (= highly connected and structurally important nodes)

Rex Chen, Karen Wu, John McCartney, Norman Sadeh

How can a traffic simulation be designed to faithfully reflect real-world traffic conditions? One crucial step is modeling the volume of traffic demand. But past demand modeling approaches have relied on unrealistic or suboptimal heuristics, and they have failed to adequately account for the effects of noisy and multimodal data on simulation outcomes. In thi

Shunrui Li, Yang Liu

We report a spectral asymmetry effect in the quantum harmonic oscillator, where its partition function is identified as the Chern character. This establishes a direct link between statistical mechanics, and topological invariants (Atiyah-Singer index theorem), revealing the internal energy as a non-SUSY manifestation of the index theorem. We show that the pa

Ziheng Cheng, Yixiao Huang, Hui Xu, Somayeh Sojoudi

Text-to-Image (T2I) models have achieved remarkable success in generating visual content from text inputs. Although multiple safety alignment strategies have been proposed to prevent harmful outputs, they often lead to overly cautious behavior -- rejecting even benign prompts -- a phenomenon known as $\textit{over-refusal}$ that reduces the practical utility

Annika Moucha

Let $f$ be a holomorphic function mapping the open unit disk into itself. We establish a boundary version of Schwarz' lemma in the spirit of a result by Burns and Krantz and provide sufficient conditions on the local behaviour of $f$ near some boundary point that forces $f$ to be a Blaschke product with predescribed critical points. For the proof, a local Ju

Taichi Tsukamoto

We show that the $\theta+$ pentaquark does not exist in the Skyrme model. For the solitons of the theory with topological terms, the standard collective coordinate quantization does not construct the proper low energy effective theory. In the presence of topological terms, zero modes are classified into three groups: dynamical zero modes with constant veloci

Lata Pangtey, Mohammad Zia Ur Rehman, Prasad Chaudhari, Shubhi Bansal

The rapid evolution of social media has generated an overwhelming volume of user-generated content, conveying implicit opinions and contributing to the spread of misinformation. The method aims to enhance the detection of stance where misinformation can polarize user opinions. Stance detection has emerged as a crucial approach to effectively analyze underlyi

Aidan Peppin, Julia Kreutzer, Alice Schoenauer Sebag, Kelly Marchisio

Despite advances in large language model capabilities in recent years, a large gap remains in their capabilities and safety performance for many languages beyond a relatively small handful of globally dominant languages. This paper provides researchers, policymakers and governance experts with an overview of key challenges to bridging the "language gap" in A

Xinyu Feng, Mohammed EL-Hajjar, Chao Xu, Lajos Hanzo

The concept of Compressed Sensing-aided Space-Frequency Index Modulation (CS-SFIM) is conceived for the Large-Scale Multi-User Multiple-Input Multiple-Output Uplink (LS-MU-MIMO-UL) of Next-Generation (NG) networks. Explicitly, in CS-SFIM, the information bits are mapped to both spatial- and frequency-domain indices, where we treat the activation patterns of

Valentin Knappich, Annemarie Friedrich, Anna Hätty, Simon Razniewski

Patent claims define the scope of protection for an invention. If there are ambiguities in a claim, it is rejected by the patent office. In the US, this is referred to as indefiniteness (35 U.S.C {\S} 112(b)) and is among the most frequent reasons for patent application rejection. The development of automatic methods for patent definiteness examination has t

Sergio Leiva-Montecinos, Libor Vojáček, Jing Li, Mairbek Chshiev

The Edelstein effect is a promising mechanism for generating spin and orbital polarization from charge currents in systems without inversion symmetry. In ferroelectric materials, such as Germanium Telluride (GeTe), the combination of bulk Rashba splitting and voltage-controlled ferroelectric polarization provides a pathway for electrical control of the sign

Henryk Mustroph, Michel Kunkler, Stefanie Rinderle-Ma

Suffix prediction of business processes forecasts the remaining sequence of events until process completion. Current approaches focus on predicting the most likely suffix, representing a single scenario. However, when the future course of a process is subject to uncertainty and high variability, the expressiveness of such a single scenario can be limited, si

Yifan Jiang, Fang Rui Lim

We propose a transfer principle to study the adapted 2-Wasserstein distance between stochastic processes. First, we obtain an explicit formula for the distance between real-valued mean-square continuous Gaussian processes by introducing the causal factorization as an infinite-dimensional analogue of the Cholesky decomposition for operators on Hilbert spaces.

Loucas Pillaud-Vivien, Adrien Schertzer

We consider the problem of jointly learning a one-dimensional projection and a univariate function in high-dimensional Gaussian models. Specifically, we study predictors of the form $f(x)=\varphi^\star(\langle w^\star, x \rangle)$, where both the direction $w^\star \in \mathcal{S}_{d-1}$, the sphere of $\mathbb{R}^d$, and the function $\varphi^\star: \mathbb

Takuhiro Kaneko

Recent advancements in neural 3D representations, such as neural radiance fields (NeRF) and 3D Gaussian splatting (3DGS), have enabled the accurate estimation of 3D structures from multiview images. However, this capability is limited to estimating the visible external structure, and identifying the invisible internal structure hidden behind the surface is d

Kele Shao, Keda Tao, Can Qin, Haoxuan You

Video large language models (video LLMs) excel at video comprehension but face significant computational inefficiency due to redundant video tokens. Existing token pruning methods offer solutions. However, approaches operating within the LLM (inner-LLM pruning), such as FastV, incur intrinsic computational overhead in shallow layers. In contrast, methods per