May 2025 arXiv papers — page 96

Evangelia E. Zavvou, Alexander Jarosik, Hajnalka Nádasi, Christoforos A. Krontiras

The recent discovery of ferroelectric nematics-genuine 3D ferroelectric fluids-has underscored the importance of electrostatic interactions in shaping the physical behaviour of soft matter systems. In this paper, we investigate the mechanical properties of ferroelectric nematics by directly comparing the splay and twist elastic constants in a liquid crystal

Caio Machado Fernandes, Enric Brillas, Mauro C. Santos, Sergi Garcia-Segura

This work reports the efficient degradation and mineralization of benzophenone-4 (BP-4), a widely used UV filter associated with endocrine-disrupting effects, via the electro-Fenton process. Key operating parameters including pH, current density, Fe2+ dosage, and initial pollutant concentration were optimized. At pH = 3.0, the best performance was obtained.

Yuan Yuan, Muyu He, Muhammad Adil Shahid, Jiani Huang

This paper introduces TurnaboutLLM, a novel framework and dataset for evaluating the deductive reasoning abilities of Large Language Models (LLMs) by leveraging the interactive gameplay of detective games Ace Attorney and Danganronpa. The framework tasks LLMs with identifying contradictions between testimonies and evidences within long narrative contexts, a

Tianqi Du, Zeming Wei, Quan Chen, Chenheng Zhang

The rapid advancement of large language models (LLMs) has demonstrated milestone success in a variety of tasks, yet their potential for generating harmful content has raised significant safety concerns. Existing safety evaluation approaches typically operate directly on textual responses, overlooking the rich information embedded in the model's internal repr

Marcelo A. Pires, Constantino Tsallis, Evaldo M. F. Curado

We introduce the $\alpha$-Gauss-Logistic map, a new nonlinear dynamics constructed by composing the logistic and $\alpha$-Gauss maps. Explicitly, our model is given by $x_{t+1} = f_L(x_t)x_t^{-\alpha} - \lfloor f_L(x_t)x_t^{-\alpha} \rfloor $ where $f_L(x_t) = r x_t (1-x_t)$ is the logistic map and $ \lfloor \ldots \rfloor $ is the integer part function. Our

Shuang Qian, Devran Ugurlu, Elliot Fairweather, Richard E Jones

Cardiac resynchronization therapy (CRT) guidelines are based on clinical trials with limited female representation and inconsistent left bundle branch block (LBBB) definitions. Conventional QRS duration (QRSd) criteria show variable diagnostic accuracy between sexes, partly due to differences in heart size and remodeling. We evaluated the influence of sex, h

Souvik Dey

Finitely generated reflexive modules over commutative Noetherian rings form a key component of Auslander and Bridger's stable module theory and are likewise essential in the study of Cohen--Macaulay representations. Recently, H. Dao characterized Arf local rings as exactly those one-dimensional Cohen--Macaulay local rings over which every finitely generated

Java Darleen Villano

In this paper, we apply the machinery developed in arXiv:2401.06641(2) to study the behavior of computable categoricity relativized to non-c.e. degrees. In particular, we show that we can build a computable structure which is not computably categorical but is computably categorical relative to a $1$-generic degree. Additionally, we show that other classes of

Pronama Biswas, Asmita Saha, Bhoomika Sridhar, Anwesha Patel

Quantum dots (QDs) have emerged as promising nanomaterials with unique optical and physical properties, making them highly attractive for various applications in biomedicine. This review provides a comprehensive overview of the types, modes of synthesis, characterization, applications, and recent advances of QDs in the field of biomedicine, with a primary fo

Shahriyar Jafarzade, Richard F. Lebed

The dynamical diquark model describes multiquark exotic hadrons in terms of diquark components nucleated by heavy quarks, and successfully explains multiple features of hidden-charm and -bottom exotics. Here we apply the model to the marginally heavy case of hidden-strange states to probe whether mesons near 2 GeV with peculiar properties, such as $\phi(2170

Xiaodong Mei, Sheng Wang, Jie Cheng, Yingbing Chen

Motion forecasting represents a critical challenge in autonomous driving systems, requiring accurate prediction of surrounding agents' future trajectories. While existing approaches predict future motion states with the extracted scene context feature from historical agent trajectories and road layouts, they suffer from the information degradation during the

Peng Wang, Biyu Zhou, Xuehai Tang, Jizhong Han

Large Language Models often contain factually incorrect or outdated knowledge, giving rise to model editing methods for precise knowledge updates. However, current mainstream locate-then-edit approaches exhibit a progressive performance decline during sequential editing, due to inadequate mechanisms for long-term knowledge preservation. To tackle this, we mo

Kefan Song, Amir Moeini, Peng Wang, Lei Gong

Reinforcement learning (RL) is a framework for solving sequential decision-making problems. In this work, we demonstrate that, surprisingly, RL emerges during the inference time of large language models (LLMs), a phenomenon we term in-context RL (ICRL). To reveal this capability, we introduce a simple multi-round prompting framework, we call ICRL prompting,

Pingqing Zheng, Jiayin Qin, Fuqi Zhang, Niraj Chitla

Retrieval Augmented Generation (RAG) is an essential agent for Large Language Model (LLM) aided Description Language (HDL) tasks, addressing the challenges of limited training data and prohibitively long prompts. However, its performance in handling ambiguous queries and real-world, repository-level HDL projects containing thousands or even tens of thousands

Alkis Koudounas, Claudio Savelli, Flavio Giobergia, Elena Baralis

Machine unlearning, the process of efficiently removing specific information from machine learning models, is a growing area of interest for responsible AI. However, few studies have explored the effectiveness of unlearning methods on complex tasks, particularly speech-related ones. This paper introduces UnSLU-BENCH, the first benchmark for machine unlearnin

Ahsan Sanaullah, Degui Zhi, Shaojie Zhang

In this paper, we describe a new type of match between a pattern and a text that aren't necessarily maximal in the query, but still contain useful matching information: locally maximal exact matches (LEMs). There are usually a large amount of LEMs, so we only consider those above some length threshold $\mathcal{L}$. These are referred to as long LEMs. The pu

Lennart Rohlfs, Julien Weiss

Wall shear stress quantification is fundamental in fluid dynamics but remains challenging in wind-tunnel experiments. Sensor-based methods offer high accuracy but lack spatial resolution for capturing complex three-dimensional effects. Conversely, oil-film visualization is a simple method to obtain high-resolution surface flow topology by processing a sequen

Maike Behrendt, Stefan Sylvius Wagner, Stefan Harmeling

The [CLS] token in BERT is commonly used as a fixed-length representation for classification tasks, yet prior work has shown that both other tokens and intermediate layers encode valuable contextual information. In this work, we study lightweight extensions to BERT that refine the [CLS] representation by aggregating information across layers and tokens. Spec

Ryang Heo, Yongsik Seo, Junseong Lee, Dongha Lee

The surge of user-generated online content presents a wealth of insights into customer preferences and market trends. However, the highly diverse, complex, and context-rich nature of such contents poses significant challenges to traditional opinion mining approaches. To address this, we introduce Online Opinion Mining Benchmark (OOMB), a novel dataset and ev

Xingyu Zhou, Yulian Wu, Francesco Orabona

In this paper, we theoretically investigate the effects of noisy labels in offline alignment, with a focus on the interplay between privacy and robustness against adversarial corruption. Specifically, under linear modeling assumptions, we present a unified analysis covering both reinforcement learning from human feedback (RLHF) and direct preference optimiza

Milad Kazemi, Mateo Perez, Fabio Somenzi, Sadegh Soudjani

Recent advances in reinforcement learning (RL) have renewed interest in reward design for shaping agent behavior, but manually crafting reward functions is tedious and error-prone. A principled alternative is to specify behavioral requirements in a formal, unambiguous language and automatically compile them into learning objectives. $\omega$-regular language

Claudia Maraston, Marco Limongi, Justus Neumann, Lorenzo Roberti

We update our stellar population models for the time evolution of the number and mass of massive remnants - neutron stars and black holes - with a new initial mass-remnant mass relation for core collapse supernovae. The calculations are based on hydrodynamical simulations and induced explosions of a subset of previously published pre-supernovae models spanni

Marquita Ellis, Iksha Gurung, Muthukumaran Ramasubramanian, Rahul Ramachandran

Is natural-language-driven earth observation data analysis now feasible with the assistance of Large Language Models (LLMs)? For open science in service of public interest, feasibility requires reliably high accuracy, interactive latencies, low (sustainable) costs, open LLMs, and openly maintainable software -- hence, the challenge. What are the techniques a

Ali Gholami, Kamal Aghazade, Akshay Vishwakarma

The Lagrange multiplier method has proven highly effective for mitigating the ill-conditioning of full waveform inversion (FWI), enabling robust and computationally efficient algorithms that converge to accurate velocity models even from poor initial estimates. Classical multiplier-based FWI methods optimize an augmented Lagrangian (AL) functional with a sca

Leonardo N. Coregliano, Maryanthe Malliaris

Recently, the authors introduced the theory of high-arity PAC learning, which is well-suited for learning graphs, hypergraphs and relational structures. In the same initial work, the authors proved a high-arity analogue of the Fundamental Theorem of Statistical Learning that almost completely characterizes all notions of high-arity PAC learning in terms of a

Zhe Xu, Cheng Jin, Yihui Wang, Ziyi Liu

Multimodal pathological image understanding has garnered widespread interest due to its potential to improve diagnostic accuracy and enable personalized treatment through integrated visual and textual data. However, existing methods exhibit limited reasoning capabilities, which hamper their ability to handle complex diagnostic scenarios. Additionally, the en

Zhengji Feng, Hengxiang Chen, Liqun Chen, Heyan Li

Autonomous underwater vehicles (AUVs) are increasingly used in marine research, military applications, and undersea exploration. However, their operational range is significantly affected by battery performance. In this paper, a framework for a wireless energy sharing system among AUVs is proposed, enabling rapid energy replenishment. Path planning plays a c

Xiuchao Sui, Daiying Tian, Qi Sun, Ruirui Chen

Foundation models (FMs) are increasingly used to bridge language and action in embodied agents, yet the operational characteristics of different FM integration strategies remain under-explored -- particularly for complex instruction following and versatile action generation in changing environments. This paper examines three paradigms for building robotic sy

Gengyang Li, Yifeng Gao, Yuming Li, Yunfang Wu

While Chain-of-Thought (CoT) prompting improves reasoning in large language models (LLMs), the excessive length of reasoning tokens increases latency and KV cache memory usage, and may even truncate final answers under context limits. We propose ThinkLess, an inference-efficient framework that terminates reasoning generation early and maintains output qualit

Zishuai Zhang, Hainan zhang, Weihua Li, Qinnan zhang

Private data holds promise for improving LLMs due to its high quality, but its scattered distribution across data silos and the high computational demands of LLMs limit their deployment in federated environments. To address this, the transformer-based federated split models are proposed, which offload most model parameters to the server (or distributed clien

Cosimo Iaia, Bhavin Choksi, Emily Wiebers, Gemma Roig

The nouns of our language refer to either concrete entities (like a table) or abstract concepts (like justice or love), and cognitive psychology has established that concreteness influences how words are processed. Accordingly, understanding how concreteness is represented in our mind and brain is a central question in psychology, neuroscience, and computati

Gourav Kumawat, Arvind K. Dattatrey, R. K. S Yadav

We conducted a spectroscopic study of 39 blue straggler stars in the globular cluster NGC 3201. The spectra of these stars were collected from the literature. We determined the radial velocity, atmospheric parameters (Teff, log g), and the abundance of Mg, as well as the metallicity ([Fe/H]) of the blue straggler population. The mean radial velocity and [Fe/

Ailier Rivero-Acosta, H. García-Tecocoatzi, A. Ramirez-Morales, E. Santopinto

In this work, we investigate the radiative decays of the $\Lambda_b$ and $\Xi_b$ bottom baryons, which belong to the flavor anti-triplet ($\mathbf{\bar{3}}_{\rm F}$), within the constituent quark model formalism. The electromagnetic transitions are calculated from the second-shell states to both the ground and $P$-wave final states. These decays play a cruci

Kang Ding, Chunxuan Jiao, Yunze Hu, Kangjie Zhou

Swarm robotic trajectory planning faces challenges in computational efficiency, scalability, and safety, particularly in complex, obstacle-dense environments. To address these issues, we propose SwarmDiff, a hierarchical and scalable generative framework for swarm robots. We model the swarm's macroscopic state using Probability Density Functions (PDFs) and l

Caio Machado Fernandes, Gabriel A. Cerron-Calle, Enric Brillas, Mauro C. Santos

The electro-Fenton (EF) process using a boron-doped diamond (BDD) anode and a carbon cloth air-diffusion cathode was optimized for efficient fipronil degradation. The system achieved high H2O2 electrogeneration with close to 80% current efficiency operating between 10 and 50 mA cm-2, with hydroxil radicals formed from BDD oxidation and Fenton's reaction that

Rajah P. Nutakki, Sylvain Capponi, Ludovic D. C. Jaubert, Lode Pollet

The low-energy physics of quantum spin ice is known to support an emergent form of quantum electrodynamics (QED), where magnetic monopoles exist and the fine structure constant is material dependent. In this article, we show how this QED is modified via a coupling to dynamical matter on the centered pyrochlore lattice, a structure which has recently been syn

Aitor Balmaseda, Julio I. de Vicente

Quantum network states are multipartite states built from distributing pairwise entanglement among parties and underpin the paradigm of quantum networks for quantum information processing. In this work we introduce the problem of partial distillability in noisy quantum networks. This corresponds to the possibility of distilling by local operations and classi

Ai-Wei Guan, Dong-Jie Wu, Chuan-Fu Yang, Natalia P. Bondarenko

In this paper, we study an inverse spectral operator for the higher-order differential equation $(-1)^my^{(2m)}+ q y = \lambda y$, where $q \in L^2(0,\pi)$. We prove that if $\|q\|_2$ is sufficiently small, the two spectra corresponding to the both Dirichlet boundary conditions and to the Dirichlet-Neumann ones uniquely determine the potential $q$. The resul

Miao Yu, Liang Lin, Guibin Zhang, Xinfeng Li

Large language models (LLMs) require iterative updates to address the outdated information problem, where LLM unlearning offers an approach for selective removal. However, mainstream unlearning methods primarily rely on fine-tuning techniques, which often lack precision in targeted unlearning and struggle to balance unlearning efficacy with general ability u

Sorina Popescu

We present a study of high-$\beta$* optics configurations at Interaction Point 2 (IP2) of the Large Hadron Collider (LHC), developed to enable forward and diffractive physics measurements with the ALICE experiment during Runs 3 and 4. Using MAD-X, we designed a $\beta$* = 30 m optics scheme that satisfies beam stability and aperture requirements, while offer

Pavel Drozdov, Giorgio Gubbiotti, Danilo Latini

We present a detailed study of a parametric Lie algebra encompassing the symmetry algebras of various models, both continuous and discrete. This algebraic structure characterizes the isotropic oscillator (with positive, purely imaginary, and zero frequency) and one of its possible nonlinear deformations. We demonstrate a novel occurrence of this Lie algebra

Hamzeh Asgharnezhad, Afshar Shamsi, Roohallah Alizadehsani, Arash Mohammadi

Knowing the uncertainty associated with the output of a deep neural network is of paramount importance in making trustworthy decisions, particularly in high-stakes fields like medical diagnosis and autonomous systems. Monte Carlo Dropout (MCD) is a widely used method for uncertainty quantification, as it can be easily integrated into various deep architectur

Ke Hu, Ehsan Hosseini-Asl, Chen Chen, Edresson Casanova

Spoken dialogue is an intuitive form of human-computer interaction, yet current speech language models often remain constrained to turn-based exchanges, lacking real-time adaptability such as user barge-in. We propose a novel duplex speech to speech (S2S) architecture featuring continuous user inputs and codec agent outputs with channel fusion that directly

Ying Zhu, Heng Zhou, Rui Su, Peiqin Zhuang

Recently, many approaches, such as Chain-of-Thought (CoT) prompting and Multi-Agent Debate (MAD), have been proposed to further enrich Large Language Models' (LLMs) complex problem-solving capacities in reasoning scenarios. However, these methods may fail to solve complex problems due to the lack of ability to find optimal solutions. Swarm Intelligence has b

Katharina Brechtelsbauer, Friederike Butt, David F. Locher, Santiago Higuera Quintero

In this paper, we derive optimized measurement-free protocols for quantum error correction and the implementation of a universal gate set optimized for an error model that is noise biased . The noise bias is adapted for neutral atom platforms, where two- and multi-qubit gates are realized with Rydberg interactions and are thus expected to be the dominating s

Davide Scassola, Sebastiano Saccani, Luca Bortolussi

Data synthesis is gaining momentum as a privacy-enhancing technology. While single-table tabular data generation has seen considerable progress, current methods for multi-table data often lack the flexibility and expressiveness needed to capture complex relational structures. In particular, they struggle with long-range dependencies and complex foreign-key r

Nicholas Sanders, Yuanchao Li, Korin Richmond, Simon King

Quantization in SSL speech models (e.g., HuBERT) improves compression and performance in tasks like language modeling, resynthesis, and text-to-speech but often discards prosodic and paralinguistic information (e.g., emotion, prominence). While increasing codebook size mitigates some loss, it inefficiently raises bitrates. We propose Segmentation-Variant Cod

Julian Max Salchegger, Rajdeep Adhikari, Bogdan Faina, Alberta Bonanni

High magnetic field and low temperature transport is carried out in order to characterize the charge carriers of $\text{PtSe}_2$. In particular, the Shubnikov-de Haas oscillations arising at applied magnetic field strengths $\gtrsim 4.5\,\text{T}$ are found to occur exclusively in plane and emerge at a layer thickness of $\approx 18\,\text{nm}$, increasing i

Henrik Kiefer, Benjamin J. A. Héry, Lucas Tepper, Benjamin A. Dalton

The generalized Langevin equation (GLE) is a useful framework for analyzing and modeling the dynamics of many-body systems in terms of low-dimensional reaction coordinates, with its specific form determined by the choice of projection formalism. We compare parameters derived from different GLE formulations using molecular dynamics simulations of butane's dih

Yuhang Ge, Yachuan Liu, Zhangyan Ye, Yuren Mao

Data preparation (DP) transforms raw data into a form suitable for downstream applications, typically by composing operations into executable pipelines. Building such pipelines is time-consuming and requires sophisticated programming skills, posing a significant barrier for non-experts. To lower this barrier, we introduce Text-to-Pipeline, a new task that tr

Hiranya Kishore Dey

A classical result in design theory, known as Fisher's inequality, states that if every pair of clubs in a town shares the same number of members, then the number of clubs cannot exceed the number of inhabitants in the town. In this short note, we establish a $q$-analogue of Fisher's inequality. Additionally, we present a $q$-analogue of the oddtown theorem

Kseniia Mosina, Aljoscha Söll, Jiri Sturala, Martin Veselý

Alkali metal intercalation is an important strategy for doping van der Waals materials. Lithium, in particular, was shown to achieve exceptional charge carrier densities, reaching levels at which fundamental electrical, optical, and magnetic material properties begin to be strongly modified. While lithium is known to be highly volatile, its migration dynamic

Jianlong Lu, Hanqiu Peng, Hongrui Zhang, Ying Chen

We introduce Transformer-based Neural Quantum Digital Twins (Tx-NQDTs) to reconstruct the low-energy spectral evolution of many-body quantum systems along quantum-annealing paths, including ground- and first-excited-state energies, spectral gaps, and transition matrix elements, at efficient computational cost. Tx-NQDTs employ a graph-informed Transformer neu

Sina Mohammad-Taheri, Matthew J. Colbrook, Simone Brugiapaglia

Gradient-based learning imposes (deep) neural networks to be differentiable at all steps. This includes model-based architectures constructed by unrolling iterations of an iterative algorithm onto layers of a neural network, known as algorithm unrolling. However, greedy sparse recovery algorithms depend on the non-differentiable argsort operator, which hinde

Jiaming Zhou, Ke Ye, Jiayi Liu, Teli Ma

The generalization capabilities of vision-language-action (VLA) models to unseen tasks are crucial to achieving general-purpose robotic manipulation in open-world settings. However, the cross-task generalization capabilities of existing VLA models remain significantly underexplored. To address this gap, we introduce AGNOSTOS, a novel simulation benchmark des

Matthew DeLorenzo, Kevin Tieu, Prithwish Jana, Piyush Jha

Large language models (LLMs) have achieved impressive proficiency on logic and programming tasks, often rivaling expert-level performance. However, generating functionally correct hardware description language (HDL) code from natural language specifications remains challenging, primarily in data-scarce domains. Therefore, we present Abstractions-of-Thought (

Ruijie Zheng, Jing Wang, Scott Reed, Johan Bjorck

We introduce $\textbf{F}$uture $\textbf{LA}$tent $\textbf{RE}$presentation Alignment ($\textbf{FLARE}$), a novel framework that integrates predictive latent world modeling into robot policy learning. By aligning features from a diffusion transformer with latent embeddings of future observations, $\textbf{FLARE}$ enables a diffusion transformer policy to anti

Šárka Nečasová, Tong Tang, Emil Wiedemann, Lu Zhu

In this paper, we consider the problem of energy conservation for weak solutions of the inviscid Primitive Equations (PE) in a bounded domain. Based on the work [Bardos et al., Onsager's conjecture with physical boundaries and an application to the vanishing viscosity limit, Comm. Math. Phys., 2019, 291-310], we prove the energy conservation for PE with boun

Cheng Yan, Felix Mohr, Tom Viering

Sample-wise learning curves plot performance versus training set size. They are useful for studying scaling laws and speeding up hyperparameter tuning and model selection. Learning curves are often assumed to be well-behaved: monotone (i.e. improving with more data) and convex. By constructing the Learning Curves Database 1.1 (LCDB 1.1), a large-scale databa

Zhexin Zhang, Yuhao Sun, Junxiao Yang, Shiyao Cui

Fine-tuning on open-source Large Language Models (LLMs) with proprietary data is now a standard practice for downstream developers to obtain task-specific LLMs. Surprisingly, we reveal a new and concerning risk along with the practice: the creator of the open-source LLMs can later extract the private downstream fine-tuning data through simple backdoor traini

Jakub Gajarský, Jeremi Gładkowski, Jan Jedelský, Michał Pilipczuk

We prove several negative results about first-order transducibility for classes of sparse graphs: - for every $t \in \mathbb{N}$, the class of graphs of treewidth at most $t+1$ is not transducible from the class of graphs of treewidth at most $t$; - for every $t \in \mathbb{N}$, the class of graphs with Hadwiger number at most $t+2$ is not transducible from

Seri Khoury, Aaron Schild

In this work, we present an $\Omega\left(\min\{\log \Delta, \sqrt{\log n}\}\right)$ lower bound for Maximal Matching (MM) in $\Delta$-ary trees against randomized algorithms. By a folklore reduction, the same lower bound applies to Maximal Independent Set (MIS), albeit not in trees. As a function of $n$, this is the first advancement in our understanding of

Yong See Foo, Adriana Zanca, Jennifer A. Flegg, Ivo Siekmann

Dynamical systems in biology are complex, and one often does not have comprehensive knowledge about the interactions involved. Chemical reaction network (CRN) inference aims to identify, from observing species concentrations over time, the unknown reactions between the species. Existing approaches such as sparse regularisation largely focus on identifying a

Seri Khoury, Aaron Schild

We study the problem of finding a maximal independent set (MIS) in the standard LOCAL model of distributed computing. Classical algorithms by Luby [JACM'86] and Alon, Babai, and Itai [JALG'86] find an MIS in $O(\log n)$ rounds in $n$-node graphs with high probability. Despite decades of research, the existence of any $o(\log n)$-round algorithm for general g

Yousef Bisabr

The Hubble tension, as a persistent discrepancy between early-time and late-time measurements of the Hubble constant, motivates explorations of new physics in the early Universe. In a recent early dark energy (EDE) model, we introduced a scalar field interacting with the radiation sector at early-time before recombination. We showed that such a scalar-photon

Jia-Xin Lin, Jing Song, Miguel Albaladejo, Albert Feijoo

We study the feasibility of having the $\Sigma^*(1430)$ state, predicted within the chiral unitary approach and recently reported by the Belle Collaboration, as corresponding to a state of non-molecular nature. Starting from this assumption, since the state is observed in the $\pi \Lambda$ channel, we allow the coupling to this state and relate the coupling

Tianjiao Cao, Jiahao Lyu, Weichao Zeng, Weimin Mu

Scene text detection has seen the emergence of high-performing methods that excel on academic benchmarks. However, these detectors often fail to replicate such success in real-world scenarios. We uncover two key factors contributing to this discrepancy through extensive experiments. First, a \textit{Fine-tuning Gap}, where models leverage \textit{Dataset-Spe

Harmender Gahlawat, Meirav Zehavi

Decision trees are a fundamental tool in machine learning for representing, classifying, and generalizing data. It is desirable to construct ``small'' decision trees, by minimizing either the \textit{size} ($s$) or the \textit{depth} $(d)$ of the \textit{decision tree} (\textsc{DT}). Recently, the parameterized complexity of \textsc{Decision Tree Learning} h

Youming Tao, Zuyuan Zhang, Dongxiao Yu, Xiuzhen Cheng

We investigate the problem of finding second-order stationary points (SOSP) in differentially private (DP) stochastic non-convex optimization. Existing methods suffer from two key limitations: (i) inaccurate convergence error rate due to overlooking gradient variance in the saddle point escape analysis, and (ii) dependence on auxiliary private model selectio

Ke Hu, Krishna Puvvada, Elena Rastorgueva, Zhehuai Chen

We introduce a data-driven approach for enabling word-level timestamp prediction in the Canary model. Accurate timestamp information is crucial for a variety of downstream tasks such as speech content retrieval and timed subtitles. While traditional hybrid systems and end-to-end (E2E) models may employ external modules for timestamp prediction, our approach

Stefano Zazzera, José Fonseca, Tessa Baker, Chris Clarkson

Future third-generation gravitational wave detectors like the Einstein Telescope (ET) and Cosmic Explorer (CE) are expected to detect millions of binary black hole (BBH) mergers. Alongside these advances, upcoming radio surveys, such as the Square Kilometer Array Observatory (SKAO) will provide new sets of cosmological tracers. These include mapping the larg

Behnam Shobiri, Sajjad Pourali, Daniel Migault, Ioana Boureanu

In many web applications, such as Content Delivery Networks (CDNs), TLS credentials are shared, e.g., between the website's TLS origin server and the CDN's edge servers, which can be distributed around the globe. To enhance the security and trust for TLS 1.3 in such scenarios, we propose LURK-T, a provably secure framework which allows for limited use of rem

Zhen Sun, Ziyi Zhang, Zeren Luo, Zhiyuan Zhong

Fine-grained detection and localization of localized image edits is crucial for assessing content authenticity, especially as modern diffusion models and image editors can produce highly realistic manipulations. However, this problem faces three key challenges: (1) most AIGC detectors produce only a global real-or-fake label without indicating where edits oc

Lan V. Truong

We study best-arm identification in stochastic multi-armed bandits under the fixed-confidence setting, focusing on instances with multiple optimal arms. Unlike prior work that addresses the unknown-number-of-optimal-arms case, we consider the setting where the number of optimal arms is known in advance. We derive a new information-theoretic lower bound on th

Rahul Bhattacharjee, Henning Struchtrup, Anirudh Singh Rana

We investigate interfacial fluid dynamics and heat transfer at nanoscales using an improved diffuse interface approach for liquid-vapor interfaces in non-equilibrium. Conventional Navier-Stokes-Korteweg (NSK) formulations often fail to accurately capture transport phenomena across extremely thin interfaces due to underestimation of interface resistances. In

Zhengjia Zhuo, Viswanath Nagarajan

Optimal decision tree (\odt) is a fundamental problem arising in applications such as active learning, entity identification, and medical diagnosis. An instance of \odt is given by $m$ hypotheses, out of which an unknown ``true'' hypothesis is drawn according to some probability distribution. An algorithm needs to identify the true hypothesis by making queri

Ninoslav Truhar, Krešimir Veselić

This paper investigates two optimization criteria for damping optimization in a multi-body oscillator system with arbitrary degrees of freedom ($n$), resembling string/rod free vibrations. The total average energy over all possible initial data and the total average displacement over all possible initial data. Our first result shows that both criteria are eq

Fausto Colantoni, Mirko D'Ovidio, Gianni Pagnini

In this paper, we study reflecting Brownian motion with Poissonian resetting. After providing a probabilistic description of the phenomenon using jump diffusions and semigroups, we analyze the time-reversed process starting from the stationary measure. We prove that the time-reversed process is a Brownian motion with a negative drift and non-local boundary c

Daniel Waxman, Fernando Llorente, Petar M. Djurić

We revisit the classical problem of Bayesian ensembles and address the challenge of learning optimal combinations of Bayesian models in an online, continual learning setting. To this end, we reinterpret existing approaches such as Bayesian model averaging (BMA) and Bayesian stacking through a novel empirical Bayes lens, shedding new light on the limitations

Pujun Xue, Junyi Ge, Xiaotong Jiang, Siyang Song

Malocclusion is a major challenge in orthodontics, and its complex presentation and diverse clinical manifestations make accurate localization and diagnosis particularly important. Currently, one of the major shortcomings facing the field of dental image analysis is the lack of large-scale, accurately labeled datasets dedicated to malocclusion issues, which

Yousef Al-Jazzazi, Haya Diwan, Jinrui Gou, Cameron Musco

Nearest neighbor search is central in machine learning, information retrieval, and databases. For high-dimensional datasets, graph-based methods such as HNSW, DiskANN, and NSG have become popular thanks to their empirical accuracy and efficiency. These methods construct a directed graph over the dataset and perform beam search on the graph to find nodes clos

T. J. Volkoff

The SU(1,1) interferometer introduced by Yurke, McCall, Klauder is reformulated starting from the Hamiltonian of two identical optical downconversion processes with opposite pump phases. From the four optical modes, two are singled out up to a relative phase by the assumption of exact alignment of the interferometer (i.e., mode indistinguishability). The sta

Zihao Li, Xu Wang, Yuzhe Yang, Ziyu Yao

Large Language Models (LLMs) demonstrate the ability to solve reasoning and mathematical problems using the Chain-of-Thought (CoT) technique. Expanding CoT length, as seen in models such as DeepSeek-R1, significantly enhances this reasoning for complex problems, but requires costly and high-quality long CoT data and fine-tuning. This work, inspired by the de

David Thulke, Jakob Kemmler, Christian Dugast, Hermann Ney

Large language models that use retrieval augmented generation have the potential to unlock valuable knowledge for researchers, policymakers, and the public by making long and technical climate-related documents more accessible. While this approach can help alleviate factual hallucinations by relying on retrieved passages as additional context, its effectiven

Xavier Pic, Raja Appuswamy

The exponential increase in storage demand and low lifespan of data storage devices has resulted in long-term archival and preservation emerging as a critical bottlenecks in data storage. In order to meet this demand, researchers are now investigating novel forms of data storage media. The high density, long lifespan and low energy needs of synthetic DNA mak

Nick Kocher, Christian Wassermann, Leona Hennig, Jonas Seng

Neural Architecture Search (NAS) accelerates progress in deep learning through systematic refinement of model architectures. The downside is increasingly large energy consumption during the search process. Surrogate-based benchmarking mitigates the cost of full training by querying a pre-trained surrogate to obtain an estimate for the quality of the model. S

Colin Defant

We construct and analyze several new families of permutons arising from random processes involving the Demazure product on the symmetric group. First, we consider Demazure products associated to random pipe dreams, generalizing the Grothendieck permutons introduced by Morales, Panova, Petrov, and Yeliussizov by replacing staircase shapes with arbitrary order

Takuto Nabeoka, Yijun Duan, Qiang Ma

Social networking services (SNS) contain vast amounts of image-text posts, necessitating effective analysis of their relationships for improved information retrieval. This study addresses the classification of image-text pairs in SNS, overcoming prior limitations in distinguishing relationships beyond similarity. We propose a graph-based method to classify i

Iuliia Kotseruba, John K. Tsotsos

Generalization of deep-learning-based (DL) computer vision algorithms to various image perturbations is hard to establish and remains an active area of research. The majority of past analyses focused on the images already captured, whereas effects of the image formation pipeline and environment are less studied. In this paper, we address this issue by analyz

Raoni Arroyo

This chapter acknowledges a gap between the ``non-individuals'' interpretation of quantum mechanics and our world of experience, and begins to bridge it. Section 1 states the problem with Abner Shimony's ``Phenomenological principle''; section 2 briefly presents the interpretation with connection to standard quantum mechanics; section 3 presents the measurem

Jacopo Teneggi, Zhenzhen Wang, Paul H. Yi, Tianmin Shu

Concept-based explanation methods aim at making machine learning models more transparent by finding the most important semantic features of an input (e.g., colors, patterns, shapes) for a given prediction task. However, these methods generally ignore the communicative context of explanations, such as the preferences of a listener. For example, medical doctor

Hongqian Wu, Hongzhong Deng, Jichao Li, Chengxing Wu

The phenomenon of group cooperation constitutes a fundamental mechanism underlying various social and biological systems. Complex networks provide a structural framework for group interactions, where individuals can not only obtain information from their neighbors but also choose neighbors as cooperative partners. However, traditional evolutionary game theor

J. Ruotsalainen, A. Jaries, M. Stryjczyk, A. Kankainen

The masses of neutron-rich $^{104-106}$Tc isotopes were measured using the JYFLTRAP double Penning trap and found to deviate from the Atomic Mass Evaluation 2020 by $-79(25)$, $40(12)$ and $94(41)$ keV, respectively. In the case of $^{105,106}$Tc, the updated $Q_\beta$ values are in agreement with a previous JYFLTRAP measurement, disagreeing with the values

H. V. AlquBoj, Hilal AlQuabeh, Velibor Bojkovic, Munachiso Nwadike

Grokking, a delayed generalization in neural networks after perfect training performance, has been observed in Transformers and MLPs, but the components driving it remain underexplored. We show that embeddings are central to grokking: introducing them into MLPs induces delayed generalization in modular arithmetic tasks, whereas MLPs without embeddings can ge

Tiasa Singha Roy, Aditeya Baral, Ayush Rajesh Jhaveri, Yusuf Baig

Large language models (LLMs) demonstrate considerable potential in various natural language tasks but face significant challenges in mathematical reasoning, particularly in executing precise, multi-step logic. However, current evaluation frameworks judge their performance solely based on accuracy, which only accounts for the final answer. This study explores

Pietro Bartoli, Christian Veronesi, Andrea Giudici, David Siorpaes

The rise of IoT has increased the need for on-edge machine learning, with TinyML emerging as a promising solution for resource-constrained devices such as MCU. However, evaluating their performance remains challenging due to diverse architectures and application scenarios. Current solutions have many non-negligible limitations. This work introduces an altern

Shuyin Ouyang, Dong Huang, Jingwen Guo, Zeyu Sun

We introduce DSCodeBench, a new benchmark designed to evaluate large language models (LLMs) on complicated and realistic data science code generation tasks. DSCodeBench consists of 1,000 carefully constructed problems sourced from realistic problems from GitHub across ten widely used Python data science libraries. DSCodeBench offers a more challenging and re

BESIII Collaboration, M. Ablikim, M. N. Achasov, P. Adlarson

By analyzing $(2712.4\pm14.3)\times10^6$ $\psi(3686)$ events collected with the BESIII detector operating at the BEPCII collider, the decays $\chi_{c0,1,2} \to 3K_S^0K^\pm\pi^\mp$ are observed for the first time with statistical significances greater than $10\sigma$. The branching fractions of these decays are determined to be $\mathcal{B}(\chi_{c0}\to 3K_S^

Kan Nakazono, Shion Chen, Hajime Fukuda, Yutaro Iiyama

We report the results of a search for dark photon dark matter using a cavity that employs a transmon qubit as a frequency tuner. The tuning mechanism utilizes the energy level shift arising from the mode mixing between the qubit and the cavity mode. This method is advantageous as it avoids the frictional heating and electromagnetic leakage associated with me

Bernard Derrida

These notes are a written version of lectures given in the 2024 Les Houches Summer School on {\it Large deviations and applications}. They are are based on a series of works published over the last 25 years on steady properties of non-equilibrium systems in contact with several heat baths at different temperatures or several reservoirs of particles at differ