May 2025 arXiv papers — page 69

Yixuan Guo, Mingliang Xiong, Wen Fang, Qingwei Jiang

The rapid development of IoT technology has led to a shortage of spectrum resources and energy, giving rise to simultaneous wireless information and power transfer (SWIPT) technology. However, traditional multiple input multiple output (MIMO)-based SWIPT faces challenges in target detection. We have designed a passive multi-user resonant beam system (MU-RBS)

Jian Liang, Wenke Huang, Xianda Guo, Guancheng Wan

Low-Rank Adaptation (LoRA) is widely adopted for downstream fine-tuning of foundation models due to its efficiency and zero additional inference cost. Many real-world applications require foundation models to specialize in several specific tasks simultaneously, motivating the need for efficient multi-task downstream adaptation. To address this need, existing

Markus Pössel

The strong evidence for low-frequency gravitational waves from pulsar timing arrays (PTAs), published in 2023, has widened the scope for teaching about gravitational wave astronomy. This article provides a simple, unified overview of the detection of gravitational waves using light waves that encompasses the recent PTA detections, the by-now classic interfer

Md. Tanzib Hosain, Rajan Das Gupta, Md. Kishor Morol

In this work, we provide DZEN, a dataset of parallel Dzongkha and English test questions for Bhutanese middle and high school students. The over 5K questions in our collection span a variety of scientific topics and include factual, application, and reasoning-based questions. We use our parallel dataset to test a number of Large Language Models (LLMs) and fi

Hai-Long Qin, Jincheng Dai, Sixian Wang, Xiaoqi Qin

Semantic communication, leveraging advanced deep learning techniques, emerges as a new paradigm that meets the requirements of next-generation wireless networks. However, current semantic communication systems, which employ neural coding for feature extraction from raw data, have not adequately addressed the fundamental question: Is general feature extractio

Tim G. Zhou, Evan Shelhamer, Geoff Pleiss

The go-to strategy to apply deep networks in settings where uncertainty informs decisions--ensembling multiple training runs with random initializations--is ill-suited for the extremely large-scale models and practical fine-tuning workflows of today. We introduce a new cost-effective strategy for improving the uncertainty quantification and downstream decisi

Luke G Bennetts

Ocean surface waves can propagate long distances through regions containing floating ice covers. The impacts ocean waves have on the ice covers are of interest in the climate change era, as the polar regions experience pressure from rising temperatures. This chapter provides a review of observations and theoretical models for ocean wave propagation through t

NH Wanigasingha, ES Sithpahan, MKA Ariyaratne, PRS De Silva

The recognition and classification of coins are essential in numerous financial and automated systems. This study introduces a comprehensive Sri Lankan coin image dataset and evaluates its impact on machine learning model accuracy for coin classification. We experiment with traditional machine learning classifiers K-Nearest Neighbors (KNN), Support Vector Ma

Makoto Nakamura, Takuma Yoshizumi

Nonexistence of global weak solutions of Klein-Gordon equations with gauge variant semilinear terms are considered in Friedmann-Lema\^itre-Robertson-Walker spacetimes. Effects of spatial expansion or contraction on the solutions are studied through the scale-function and the curved mass.

Yixuan Wang, Shiyu Ji, Yijun Liu, Yuzhuang Xu

Large language models (LLMs) rely on key-value cache (KV cache) to accelerate decoding by reducing redundant computations. However, the KV cache memory usage grows substantially with longer text sequences, posing challenges for efficient deployment. Existing KV cache eviction methods prune tokens using prefilling-stage attention scores, causing inconsistency

Li Wang, Guangqi Yang, Lei Yang, Ziying Song

Safety is a long-standing and the final pursuit in the development of autonomous driving systems, with a significant portion of safety challenge arising from perception. How to effectively evaluate the safety as well as the reliability of perception algorithms is becoming an emerging issue. Despite its critical importance, existing perception methods exhibit

Zhihao Jia, Mingyi Jia, Junwen Duan, Jianxin Wang

Large Language Models (LLMs) demonstrate strong generalization and reasoning abilities, making them well-suited for complex decision-making tasks such as medical consultation (MC). However, existing LLM-based methods often fail to capture the dual nature of MC, which entails two distinct sub-tasks: symptom inquiry, a sequential decision-making process, and d

Yixuan Wang, Yijun Liu, Shiyu ji, Yuzhuang Xu

Large language models (LLMs) suffer from high inference latency due to the auto-regressive decoding process. Speculative decoding accelerates inference by generating multiple draft tokens using a lightweight model and verifying them in parallel. However, existing verification methods rely heavily on distributional consistency while overlooking semantic corre

Han Xiao, Xiaoyan Hu, Wenjie Wang, Kai-Kit Wong

The concept of the frequency diverse reconfigurable intelligent surface (FD-RIS) technology has been introduced, which can enable simultaneous implementation of distance-angle beamforming in far-field communication scenarios. In order to improve the managing ability on undesired harmonic signals and the diversity of frequency offsets, this paper presents a n

Yukun Zhang, Qi Dong

We present Multi-Scale Manifold Alignment(MSMA), an information-geometric framework that decomposes LLM representations into local, intermediate, and global manifolds and learns cross-scale mappings that preserve geometry and information. Across GPT-2, BERT, RoBERTa, and T5, we observe consistent hierarchical patterns and find that MSMA improves alignment me

Szivia Lestyán, William Letrone, Ludovica Robustelli, Gergely Biczók

Anonymization is a foundational principle of data privacy regulation, yet its practical application remains riddled with ambiguity and inconsistency. This paper introduces the concept of anonymity-washing -- the misrepresentation of the anonymity level of ``sanitized'' personal data -- as a critical privacy concern. While both legal and technical critiques o

Ludwig Staiger

Automatic Baire property is a variant of the usual Baire property which is fulfilled for subsets of the Cantor space accepted by finite automata. We consider the family $\mathcal{A}$ of subsets of the Cantor space having the Automatic Baire property. In particular we show that not all finite subsets have the Automatic Baire property, and that already a sligh

Shivam Kumar Jha S, Jaya NN Iyer

This paper proposes a tropical geometry-based edge detection framework that reformulates convolution and gradient computations using min-plus and max-plus algebra. The tropical formulation emphasizes dominant intensity variations, contributing to sharper and more continuous edge representations. Three variants are explored: an adaptive threshold-based method

Arslan Sikandar, M. Jamil Aslam

The radiative $B$ to tensor $\left(K_2^*(1430),\; f_2(1270),\; a_2(1230)\right)$ meson decays are studied at next-to-leading order (NLO) in soft-collinear effective theory (SCET). The SCET allows the systematic treatment of factorizable and non-factorizable contributions along with the resummation of large perturbative logarithms. We performed a two step mat

Lucas Saldyt, Subbarao Kambhampati

This paper aims to understand how neural networks learn algorithmic reasoning by addressing two questions: How faithful are learned algorithms when they are effective, and why do neural networks fail to learn effective algorithms otherwise? To answer these questions, we use neural compilation, a technique that directly encodes a source algorithm into neural

Kourosh Shahnazari, Seyed Moein Ayyoubzadeh, Mohammadali Keshtparvar, Pegah Ghaffari

In recent machine learning systems, confidence scores are being utilized more and more to manage selective prediction, whereby a model can abstain from making a prediction when it is unconfident. Yet, conventional metrics like accuracy, expected calibration error (ECE), and area under the risk-coverage curve (AURC) do not capture the actual reliability of pr

Jiazhen Liu, Ruikun Li, Huandong Wang, Zihan Yu

This position paper argues that next-generation non-equilibrium-inspired generative models will provide the essential foundation for better modeling real-world complex dynamical systems. While many classical generative algorithms draw inspiration from equilibrium physics, they are fundamentally limited in representing systems with transient, irreversible, or

Hao Wu, Daniel F. Agterberg

Parity-time-reversal-symmetric odd-parity antiferromagnetic (AFM1) materials are of interest for their symmetry-enabled quantum transport and optical effects. These materials host odd-parity terms in their band dispersion, leading to asymmetric energy bands and enabling responses such as the magnetopiezoelectric effect, nonreciprocal conductivity, and photoc

Ettore Carretti, Franco Vazza

The Universe's magnetogenesis can be investigated with radio observations of cosmic filaments, where the information on the initial magnetic field seeds is expected to be preserved in time. In this work, we update the comparison between recent observational results in filaments with the predictions from recent cosmological simulations to check whether one of

Chon-Fai Kam

We investigate nonlinear optical analogues of quantum phase transitions within a squeezing-enhanced generalized Lipkin-Meshkov-Glick (LMG) model, focusing on excited-state quantum phase transitions in optical fibers with tetragonal symmetry. Our analysis reveals a novel squeezing effect that induces classical bifurcations in polarization dynamics, even witho

Le Tri Dat, Chi Cuong Nguyen, Nguyen Duy Vy, Amir F. Payam

High-harmonic (HH) frequencies in microcantilever impose several applications in precision detection thanks to the higher sensitivity of the higher modes in comparison to the fundamental modes. In this study, we showed that by tuning the cantilever length via changing the clamped position, the dimensional ratio of the overhang to the main cantilever part is

Li-Hao Xia, Yi-Peng Gao, Zhao-Yang Dong, Jian-Xin Li

The search for Kitaev quantum spin liquids (Kitaev-QSLs) in real materials has mainly focused on $4d$- and $5d$-electron honeycomb systems. A recent experimental study on the $4f^5$ honeycomb iodide $\mathrm{SmI}_3$ reported the absence of long-range magnetic order down to $0.1\ \text{K}$, suggesting a possible Kitaev-QSL phase. Motivated by the interplay be

Alex A. T. Rathke

For the standard lattice model of information structures, we derive a reduced poset representation which provides the same informational content as the complete lattice structure which derives it. Rational agents can recover the complete lattice of events by means of the reduced poset alone. We find that both structures provide isomorphic models under mild c

Faithful Chiagoziem Onwuegbuche, Adelodun Olaoluwa, Anca Delia Jurcut, Liliana Pasquale

Ransomware remains a critical threat to cybersecurity, yet publicly available datasets for training machine learning-based ransomware detection models are scarce and often have limited sample size, diversity, and reproducibility. In this paper, we introduce MLRan, a behavioural ransomware dataset, comprising over 4,800 samples across 64 ransomware families a

Weizhi Zhong, Huan Yang, Zheng Liu, Huiguo He

Personalized text-to-image generation aims to synthesize images of user-provided concepts in diverse contexts. Despite recent progress in multi-concept personalization, most are limited to object concepts and struggle to customize abstract concepts (e.g., pose, lighting). Some methods have begun exploring multi-concept personalization supporting abstract con

B. Hamil, Ahmad Al-Badawi, B. C. Lütfüoğlu

We perform a thorough analysis into a Schwarzschild black hole embedded in a Dehnen-type dark matter halo with a quintessential field. We develop the composite spacetime metric and examine its geometric properties, including horizon structure and curvature invariants. Our findings reveal that increasing both the DM core density $\rho_{s}$ and quintessence pa

Ashwin Sankar, Yoach Lacombe, Sherry Thomas, Praveen Srinivasa Varadhan

We introduce RASMALAI, a large-scale speech dataset with rich text descriptions, designed to advance controllable and expressive text-to-speech (TTS) synthesis for 23 Indian languages and English. It comprises 13,000 hours of speech and 24 million text-description annotations with fine-grained attributes like speaker identity, accent, emotion, style, and bac

Yuetong Fang, Deming Zhou, Ziqing Wang, Hongwei Ren

Spiking Neural Networks promise brain-inspired and energy-efficient computation by transmitting information through binary (0/1) spikes. Yet, their performance still lags behind that of artificial neural networks, often assumed to result from information loss caused by sparse and binary activations. In this work, we challenge this long-standing assumption an

Neil Chaudhary, Zaynah Dhunny

Accurate identification of breast cancer types plays a critical role in guiding treatment decisions and improving patient outcomes. This paper presents an artificial intelligence enabled tool designed to aid in the identification of breast cancer types using histopathological biopsy images. Traditionally additional tests have to be done on women who are dete

Jonathan Leung, Yongjie Wang, Zhiqi Shen

Large Language Models (LLMs) demonstrate impressive general capabilities but often struggle with step-by-step procedural reasoning, a critical challenge in complex interactive environments. While retrieval-augmented methods like GraphRAG attempt to bridge this gap, their fragmented entity-relation graphs hinder the construction of coherent, multi-step plans.

Hao Sun, Yunyi Shen, Mihaela van der Schaar

In the era of large language models (LLMs), high-quality, domain-rich, and continuously evolving datasets capturing expert-level knowledge, core human values, and reasoning are increasingly valuable. This position paper argues that OpenReview -- the continually evolving repository of research papers, peer reviews, author rebuttals, meta-reviews, and decision

Michael Seitz, Nick Garabedian, Ilia Bagov, Christian Greiner

The FAIR (Findable, Accessible, Interoperable, and Reusable) data principles have gained significant attention as a means to enhance data sharing, collaboration, and reuse across various domains. Here, we explore the potential benefits of implementing FAIR data practices within engineering projects, with a monetary focus in the German context, but by conside

G. G. L. Nashed, Waleed El Hanafy

We show that non-minimal coupling between matter and geometry can indeed help in constructing stable, traversable, wormholes (WHs) without requiring exotic matter under certain conditions. In models like $f({\cal Q},{\cal T})={\cal Q}+\beta {\cal T}$ gravity, where ${\cal Q}$ is the non-metricity scalar, and ${\cal T}$ is the trace of the energy-momentum ten

Zhu-yao Jin, Jun Jing

Conventional manipulations over quantum systems for such as coherent population trapping and unidirectional transfer focus on Hamiltonian engineering while regarding the system's manifold geometry and constraint equation as secondary causes. Here we treat them on equal footing in controlling a finite-dimensional quantum system under a time-dependent non-Herm

Xiaohuan Pei, Tao Huang, YanXiang Ma, Chang Xu

Causal attention has become a foundational mechanism in autoregressive vision-language models (VLMs), unifying textual and visual inputs under a single generative framework. However, existing causal mask-based strategies are inherited from large language models (LLMs) where they are tailored for text-only decoding, and their adaptation to vision tokens is in

Hengzhe Zhang, Qi Chen, Bing Xue, Wolfgang Banzhaf

Large language models (LLMs) have revolutionized algorithm development, yet their application in symbolic regression, where algorithms automatically discover symbolic expressions from data, remains limited. In this paper, we propose a meta-learning framework that enables LLMs to automatically design selection operators for evolutionary symbolic regression al

Jongwoo Ko, Sungnyun Kim, Sungwoo Cho, Se-Young Yun

Human-generated reward signals are critical for aligning generative models with human preferences, guiding both training and inference-time evaluations. While large language models (LLMs) employed as proxy evaluators, i.e., LLM-as-a-Judge, significantly reduce the costs associated with manual annotations, they typically require extensive modality-specific tr

Bryan Sangwoo Kim, Jeongsol Kim, Jong Chul Ye

Modern single-image super-resolution (SISR) models deliver photo-realistic results at the scale factors on which they are trained, but collapse when asked to magnify far beyond that regime. We address this scalability bottleneck with Chain-of-Zoom (CoZ), a model-agnostic framework that factorizes SISR into an autoregressive chain of intermediate scale-states

Kaixiang Chen, Naihong Hu, Hengyi Wang

This paper is devoted to studying the centre of the multi-parameter quantum group $U_{q,G}(\mathfrak{g})$ introduced by Okado and Yamane, where $\mathfrak{g}$ is a complex simple Lie algebra, and all parameters lie in general position. We mainly establish the Harish-Chandra theorem, proving that the Harish-Chandra homomorphism is an isomorphism; in particula

PARC E40 Collaboration, T. Sakao, K. Miwa, J. K. Ahn

This paper presents high-precision experimental data of the polarization of the $\Lambda$ hyperon in the $\pi^{-}p \to K^{0} \Lambda$ reaction, measured in the angular range $0.6<\cos \theta ^{CM}_{K0}<1.0$ with a fine bin width of $d\cos \theta ^{CM}_{K0}=0.05$. The data were obtained from the J-PARC E40 experiment at the K1.8 beamline in the J-PARC Hadron

Haojie Wang, Jiuyun Jiang, L. Jeff Hong, Guangxin Jiang

The development of large language models (LLMs) has provided new tools for research in supply chain management (SCM). In this paper, we introduce a retrieval-augmented generation (RAG) framework that dynamically integrates external knowledge into the inference process, and develop a domain-specialized SCM LLM, which demonstrates expert-level competence by pa

Chen Han, Wenzhen Zheng, Xijin Tang

The proliferation of misinformation in digital platforms reveals the limitations of traditional detection methods, which mostly rely on static classification and fail to capture the intricate process of real-world fact-checking. Despite advancements in Large Language Models (LLMs) that enhance automated reasoning, their application to misinformation detectio

The Viet Bui, Tien Mai, Hong Thanh Nguyen

We study offline imitation learning (IL) in cooperative multi-agent settings, where demonstrations have unlabeled mixed quality - containing both expert and suboptimal trajectories. Our proposed solution is structured in two stages: trajectory labeling and multi-agent imitation learning, designed jointly to enable effective learning from heterogeneous, unlab

GuangHao Meng, Sunan He, Jinpeng Wang, Tao Dai

Vision-language retrieval (VLR) has attracted significant attention in both academia and industry, which involves using text (or images) as queries to retrieve corresponding images (or text). However, existing methods often neglect the rich visual semantics knowledge of entities, thus leading to incorrect retrieval results. To address this problem, we propos

Chensheng Wu, Jiao Sun, Qinghe Song, Chunhua Zeng

For warm or hot and dense plasma, ionization potential depression plays a crucial role in determining the ionization balance and understanding the resulting microscopic plasma properties. However, the applicability of the widely used IPD models is currently limited under WDP conditions, where the influence of neighboring ions on IPD becomes nonnegligible. Ne

Junichi Haruna, Keisuke Fujii

We propose and analyze a hierarchical quantum error correction (QEC) scheme that concatenates hypergraph product (HGP) codes with rotated surface codes and that is compatible with quantum computers with only nearest-neighbor interactions. The outer code employs (3,4)-random HGP codes, known for their constant encoding rate and favorable distance scaling, whi

Joery A. de Vries, Jinke He, Mathijs M. de Weerdt, Matthijs T. J. Spaan

Meta-reinforcement learning trains a single reinforcement learning agent on a distribution of tasks to quickly generalize to new tasks outside of the training set at test time. From a Bayesian perspective, one can interpret this as performing amortized variational inference on the posterior distribution over training tasks. Among the various meta-reinforceme

Zehao Wang, Han Zhang, Jingchuan Wang, Weidong Chen

In this work, we propose an optimization-based trajectory planner for tractor-trailer vehicles on curvy roads. The lack of analytical expression for the trailer's errors to the center line pose a great challenge to the trajectory planning for tractor-trailer vehicles. To address this issue, we first use geometric representations to characterize the lateral a

Alejandro Kuratomi, Zed Lee, Guilherme Dinis Chaliane Junior, Tony Lindgren

Several interpretability methods for convolutional network-based classifiers exist. Most of these methods focus on extracting saliency maps for a given sample, providing a local explanation that highlights the main regions for the classification. However, some of these methods lack detailed explanations in the input space due to upscaling issues or may requi

Victor Barroso-Nascimento, Ekaterina Piotrovskaya, Elaine Pimentel

We define base-extension semantics (Bes) using atomic systems based on sequent calculus rather than natural deduction. While traditional Bes aligns naturally with intuitionistic logic due to its constructive foundations, we show that sequent calculi with multiple conclusions yield a Bes framework more suited to classical semantics. The harmony in classical s

Zesheng Shi, Yucheng Zhou, Jing Li

Despite significant progress in safety alignment, large language models (LLMs) remain susceptible to jailbreak attacks. Existing defense mechanisms have not fully deleted harmful knowledge in LLMs, which allows such attacks to bypass safeguards and produce harmful outputs. To address this challenge, we propose a novel safety alignment strategy, Constrained K

Pavan C Shekar, Pawan Soni, Vivek Kanhangad

Deepfakes pose a significant threat to digital media security, with current detection methods struggling to generalize across different manipulation techniques and datasets. While recent approaches combine CNN-based architectures with Vision Transformers or leverage multi-modal learning, they remain limited by the inherent constraints of RGB data. We introdu

Federico Zocco, Andrea Corti, Monica Malvezzi

The demand of finite raw materials will keep increasing as they fuel modern society. Simultaneously, solutions for stopping carbon emissions in the short term are not available, thus making the net zero target extremely challenging to achieve at scale. The circular economy (CE) paradigm is gaining attention as a solution to address climate change and the unc

Chengxi Min, Wei Wang, Yahui Liu, Weixin Ye

Mixture-of-Experts (MoE) models have emerged as a promising direction for scaling vision architectures efficiently. Among them, Soft MoE improves training stability by assigning each token to all experts via continuous dispatch weights. However, current designs overlook the semantic structure which is implicitly encoded in these weights, resulting in subopti

Yewei Liu, Xiyuan Wang, Muhan Zhang

We propose an entirely new meta-learning framework for network pruning. It is a general framework that can be theoretically applied to almost all types of networks with all kinds of pruning and has great generality and transferability. Experiments have shown that it can achieve outstanding results on many popular and representative pruning tasks (including b

Yedi Zhang, Sun Yi Emma, Annabelle Lee Jia En, Jin Song Dong

Large language models (LLMs) have emerged as a dominant AI paradigm due to their exceptional text understanding and generation capabilities. However, their tendency to generate inconsistent or erroneous outputs challenges their reliability, especially in high-stakes domains requiring accuracy and trustworthiness. Existing research primarily focuses on detect

Chaofan Gan, Yuanpeng Tu, Xi Chen, Tieyuan Chen

Pre-trained stable diffusion models (SD) have shown great advances in visual correspondence. In this paper, we investigate the capabilities of Diffusion Transformers (DiTs) for accurate dense correspondence. Distinct from SD, DiTs exhibit a critical phenomenon in which very few feature activations exhibit significantly larger values than others, known as \te

Hongru Song, Yu-an Liu, Ruqing Zhang, Jiafeng Guo

We explore adversarial attacks against retrieval-augmented generation (RAG) systems to identify their vulnerabilities. We focus on generating human-imperceptible adversarial examples and introduce a novel imperceptible retrieve-to-generate attack against RAG. This task aims to find imperceptible perturbations that retrieve a target document, originally exclu

Dongyang Jin, Chao Fan, Jingzhe Ma, Jingkai Zhou

To capture individual gait patterns, excluding identity-irrelevant cues in walking videos, such as clothing texture and color, remains a persistent challenge for vision-based gait recognition. Traditional silhouette- and pose-based methods, though theoretically effective at removing such distractions, often fall short of high accuracy due to their sparse and

Wentao Hu, Wengyu Zhang, Yiyang Jiang, Chen Jason Zhang

Retrieval-Augmented Generation (RAG) enhances factual accuracy by integrating external knowledge, yet it introduces a critical issue: erroneous or biased retrieval can mislead generation, compounding hallucinations, a phenomenon we term Hallucination on Hallucination. To address this, we propose Debate-Augmented RAG (DRAG), a training-free framework that int

Tingpeng Zhang, Xuzhang Peng, Mingyuan Zhou, Guobiao Hu

Structural health monitoring (SHM) involves sensor deployment, data acquisition, and data interpretation, commonly implemented via a tedious wired system. The information processing in current practice majorly depends on electronic computers, albeit with universal applications, delivering challenges such as high energy consumption and low throughput due to t

Celia González-Sánchez, Ignacio Sardinero, Jorge Cuadra, Alfredo Spuri

The combination of superconductivity and magnetic textures represents a promising approach to explore unconventional superconducting phenomena, including new correlated and topological phases. Van der Waals (vdW) materials have emerged in this context as a versatile platform to explore the interplay between these two competing orders. Here, we report on indi

Dongsuk Oh, Miryeong Kwon, Jiseon Kim, Eunjee Na

Integrating compute express link (CXL) with SSDs allows scalable access to large memory but has slower speeds than DRAMs. We present ExPAND, an expander-driven CXL prefetcher that offloads last-level cache (LLC) prefetching from host CPU to CXL-SSDs. ExPAND uses a heterogeneous prediction algorithm for prefetching and ensures data consistency with CXL.mem's

Socratis Petrides, Tucker Hartland, Tzanio Kolev, Chak Shing Lee

Large-scale contact mechanics simulations are crucial in many engineering fields such as structural design and manufacturing. In the frictionless case, contact can be modeled by minimizing an energy functional; however, these problems are often nonlinear, nonconvex, and increasingly difficult to solve as mesh resolution increases. In this work, we employ a N

Yongjie Wang, Yibo Wang, Xin Zhou, Zhiqi Shen

Probing techniques have shown promise in revealing how LLMs encode human-interpretable concepts, particularly when applied to curated datasets. However, the factors governing a dataset's suitability for effective probe training are not well-understood. This study hypothesizes that probe performance on such datasets reflects characteristics of both the LLM's

Charles Hong, Sahil Bhatia, Alvin Cheung, Yakun Sophia Shao

Hardware accelerators, especially those designed for tensor processing, have become ubiquitous in today's computing landscape. However, even with significant efforts in building compilers, programming these tensor accelerators remains challenging, leaving much of their potential underutilized. Recently, large language models (LLMs), trained on large amounts

Mengqi Liao, Xiangyu Xi, Ruinian Chen, Jia Leng

Reasoning large language models (LLMs) excel in complex tasks, which has drawn significant attention to reinforcement learning (RL) for LLMs. However, existing approaches allocate an equal number of rollouts to all questions during the RL process, which is inefficient. This inefficiency stems from the fact that training on simple questions yields limited gai

Yifan Zhu, Chao Zhang, Xin Shi, Xueqiao Zhang

Large Language Models (LLMs)-based Multi-Agent Systems (MAS) exhibit remarkable problem-solving and task planning capabilities across diverse domains due to their specialized agentic roles and collaborative interactions. However, this also amplifies the severity of security risks under MAS attacks. To address this, we introduce MASTER, a novel security resea

Xiao Fu, Jing Xu, Qifan Yang, Xuhe Gong

The rapid development of computational materials science powered by machine learning (ML) is gradually leading to solutions to several previously intractable scientific problems. One of the most prominent is machine learning interatomic potentials (MLIPs), which expedites the study of dynamical methods for large-scale systems. However, a promising field, hig

Tina Khezresmaeilzadeh, Parsa Razmara, Seyedarmin Azizi, Mohammad Erfan Sadeghi

Stock price prediction remains a complex and high-stakes task in financial analysis, traditionally addressed using statistical models or, more recently, language models. In this work, we introduce VISTA (Vision-Language Inference for Stock Time-series Analysis), a novel, training-free framework that leverages Vision-Language Models (VLMs) for multi-modal sto

Sven Raum, Hannes Thiel, Eduard Vilalta

We prove that any reduced twisted group C*-algebra of a selfless group with the rapid decay property is selfless. As an application, we show that twisted group C*-algebras of acylindrically hyperbolic groups (possibly with nontrivial finite radical) and rapid decay are pure, and hence have strict comparison.

Zhikang Chen, Abudukelimu Wuerkaixi, Sen Cui, Haoxuan Li

Deep networks are prone to catastrophic forgetting during sequential task learning, i.e., losing the knowledge about old tasks upon learning new tasks. To this end, continual learning(CL) has emerged, whose existing methods focus mostly on regulating or protecting the parameters associated with the previous tasks. However, parameter protection is often impra

Giovanni Covi, Antti Kujanpää, Jesse Railo

The classical Calder\'on problem with partial data is known to be log-log stable in some special cases, but even the uniqueness problem is open in general. We study the partial data stability of an analogous inverse fractional conductivity problem on bounded smooth domains. Using the fractional Liouville reduction, we obtain a log-log stability estimate when

Afrah Farea, Saiful Khan, Reza Daryani, Emre Cenk Ersan

Physics-informed neural networks (PINNs) have emerged as a promising approach for solving complex fluid dynamics problems, yet their application to fluid-structure interaction (FSI) problems with moving boundaries remains largely unexplored. This work addresses the critical challenge of modeling FSI systems with moving interfaces, where traditional unified P

Alexander A. Borisenko

In this paper, we prove a local rigidity of convex hypersurfaces in the spaces of constant curvature of dimension $n\ge4$. Namely, we show that two convex isometric hypersurfaces are congruent locally around their corresponding under the isometry points of strict convexity. This result extends the result of E.P. Senkin, who showed such rigidity under the add

Yisu Wang, Ruilong Wu, Xinjiao Li, Dirk Kutscher

Large-scale deep neural networks (DNN) exhibit excellent performance for various tasks. As DNNs and datasets grow, distributed training becomes extremely time-consuming and demands larger clusters. A main bottleneck is the resulting gradient aggregation overhead. While gradient compression and sparse collective communication techniques are commonly employed

Xunlian Dai, Li Zhou, Benyou Wang, Haizhou Li

The human-centered word association test (WAT) serves as a cognitive proxy, revealing sociocultural variations through culturally shared semantic expectations and implicit linguistic patterns shaped by lived experiences. We extend this test into an LLM-adaptive, free-relation task to assess the alignment of large language models (LLMs) with cross-cultural co

Yunfu Song, Zhijian Ou

Our examination of deep generative models (DGMs) developed for semi-supervised learning (SSL), mainly GANs and VAEs, reveals two problems. First, mode missing and mode covering phenomenons are observed in genertion with GANs and VAEs. Second, there exists an awkward conflict between good classification and good generation in SSL by employing directed generat

Shiu-hong Kao, Yu-Wing Tai, Chi-Keung Tang

Reasoning Video Object Segmentation is a challenging task, aiming at generating a mask sequence from an input video given a complex and implicit text query. While existing works finetune Multimodal Large Language Models (MLLM) for the task, they still fail in video inputs given complex temporally-sensitive queries, indicating their lack of temporal and spati

Neil Lu, Sizheng Ma, Ornella J. Piccinni, Ling Sun

Measuring quasinormal modes (QNMs) during the ringdown phase of binary black hole coalescences provides key insights into merger dynamics and enables tests of the no-hair theorem. The QNM rational filter has recently been introduced as a technique to identify specific QNMs in ringdown signals without sampling over mode amplitudes and phases. In this work, we

Yu Lu, Guo-Hui Hu

In binary fluid systems under temperature differences, thermophoretic separation and thermal convective mixing are two key mechanisms that affect the processes of transport. The thermophoretic effect, also known as the Soret effect, describes the migration behavior of molecules in the fluid with temperature gradient. Thermophilic and thermophobic molecules t

Wenbo He, Zhijian Ou

Our examination of existing deep generative models (DGMs), including VAEs and GANs, reveals two problems. First, their capability in handling discrete observations and latent codes is unsatisfactory, though there are interesting efforts. Second, both VAEs and GANs optimize some criteria that are indirectly related to the data likelihood. To address these pro

He Zhu, Zhiwen Ruan, Junyou Su, Xingwei He

High-quality instruction data is crucial for developing large language models (LLMs), yet existing approaches struggle to effectively control instruction complexity. We present TAG-INSTRUCT, a novel framework that enhances instruction complexity through structured semantic compression and controlled difficulty augmentation. Unlike previous prompt-based metho

Jun Zhuang, Haibo Jin, Ye Zhang, Zhengjian Kang

Intent detection, a core component of natural language understanding, has considerably evolved as a crucial mechanism in safeguarding large language models (LLMs). While prior work has applied intent detection to enhance LLMs' moderation guardrails, showing a significant success against content-level jailbreaks, the robustness of these intent-aware guardrail

Yiyang Feng, Yichen Wang, Shaobo Cui, Boi Faltings

Large Language Models (LLMs) have demonstrated impressive capabilities in reasoning, positioning them as promising tools for supporting human problem-solving. However, what happens when their performance is affected by misinformation, i.e., incorrect inputs introduced by users due to oversights or gaps in knowledge? Such misinformation is prevalent in real-w

Jaewon Kwon, Yongju Lee, Jiwan Kim, Enhyeok Jang

Modern CPUs suffer from the frontend bottleneck because the instruction footprint of server workloads exceeds the private cache capacity. Prior works have examined the CPU components or private cache to improve the instruction hit rate. The large footprint leads to significant cache misses not only in the core and faster-level cache but also in the last-leve

John Oyekan, Christopher Turner, Michael Bax, Erich Graf

The rapid advancement of Large Language Models (LLMs) has resulted in interest in their potential applications within manufacturing systems, particularly in the context of Industry 5.0. However, determining when to implement LLMs versus other Natural Language Processing (NLP) techniques, ontologies or knowledge graphs, remains an open question. This paper of

Xuan Xue, Yaotian Yang, Zihui Tian, T. C. Chang

The study of vernacular architecture involves recording, ordering, and analysing buildings to probe their physical, social, and cultural explanations. Traditionally, this process is conducted manually and intuitively by researchers. Because human perception is selective and often partial, the resulting interpretations of architecture are invariably broad and

Md Ahsanul Haque, Ismail Hossain, Md Mahmuduzzaman Kamol, Md Jahangir Alam

Machine learning (ML)-based malware detection systems often fail to account for the dynamic nature of real-world training and test data distributions. In practice, these distributions evolve due to frequent changes in the Android ecosystem, adversarial development of new malware families, and the continuous emergence of both benign and malicious applications

Jincheng An, Ajit C. Balram, Udit Khanna, Ganpathy Murthy

We evaluate the transport gaps in the most prominent fractional quantum Hall states in the $\mathbf{n}{=}0$ and $\mathbf{n}{=}1$ Landau Levels of graphene, accounting for the Coulomb interaction, lattice-scale anisotropies, and one-body terms. We find that the fractional phases in the $\mathbf{n}{=}0$ Landau level are bond-ordered, while those in the $\mathb

Baraa Hikal, Mohamed Basem, Islam Oshallah, Ali Hamdi

We present MSA-MathEval, our submission to the BEA 2025 Shared Task on evaluating AI tutor responses across four instructional dimensions: Mistake Identification, Mistake Location, Providing Guidance, and Actionability. Our approach uses a unified training pipeline to fine-tune a single instruction-tuned language model across all tracks, without any task-spe

Sanwoo Lee, Kun Liang, Yunfang Wu

Recent advances in cross-prompt automated essay scoring (AES) typically train models jointly on all source prompts, often requiring additional access to unlabeled target prompt essays simultaneously. However, using all sources is suboptimal in our pilot study, and re-accessing source datasets during adaptation raises privacy concerns. We propose a source-fre

Min Cheng, Fatemeh Doudi, Dileep Kalathil, Mohammad Ghavamzadeh

Reinforcement learning (RL) algorithms have been used recently to align diffusion models with downstream objectives such as aesthetic quality and text-image consistency by fine-tuning them to maximize a single reward function under a fixed KL regularization. However, this approach is inherently restrictive in practice, where alignment must balance multiple,

Dristi Datta, Manoranjan Paul, Manzur Murshed, Shyh Wei Teng

Soil organic carbon (SOC) is a critical indicator of soil health, but its accurate estimation from satellite imagery is hindered in vegetated regions due to spectral contamination from plant cover, which obscures soil reflectance and reduces model reliability. This study proposes the Reflectance Transformation Generative Adversarial Network (ReflectGAN), a n

An Vo, Mohammad Reza Taesiri, Daeyoung Kim, Anh Totti Nguyen

Large language models (LLMs) often exhibit strong biases, e.g, against women or in favor of the number 7. We investigate whether LLMs would be able to output less biased answers when allowed to observe their prior answers to the same question in a multi-turn conversation. To understand which types of questions invite more biased answers, we test LLMs on our