May 2025 arXiv papers — page 29

Mihir Prabhudesai, Lili Chen, Alex Ippoliti, Katerina Fragkiadaki

Reinforcement learning (RL) has enabled machine learning models to achieve significant advances in many fields. Most recently, RL has empowered frontier language models to solve challenging math, science, and coding problems. However, central to any RL algorithm is the reward function, and reward engineering is a notoriously difficult problem in any domain.

Duncan A Clark, Conor J. Kresin, Charlotte M. Jones-Todd

We propose a novel modeling framework for time-evolving networks allowing for long-term dependence in network features that update in continuous time. Dynamic network growth is functionally parameterized via the conditional intensity of a marked point process. This characterization enables flexible, joint modeling of both update timing and the network update

Qi Cai, Jingwen Chen, Yang Chen, Yehao Li

Recent advancements in image generative foundation models have prioritized quality improvements but often at the cost of increased computational complexity and inference latency. To address this critical trade-off, we introduce HiDream-I1, a new open-source image generative foundation model with 17B parameters that achieves state-of-the-art image generation

Brendan P. Marsh, David Atri Schuller, Yunpeng Ji, Henry S. Hunt

We realize a driven-dissipative Ising spin glass using cavity QED in a novel ``4/7" multimode geometry. Gases of ultracold atoms trapped within the cavity by optical tweezers serve as effective spins. They are coupled via randomly signed, all-to-all Ising cavity-mediated interactions. Networks of up to n = 25 spins are holographically imaged via cavity emiss

Wenbo Hu, Yining Hong, Yanjun Wang, Leison Gao

Humans excel at performing complex tasks by leveraging long-term memory across temporal and spatial experiences. In contrast, current Large Language Models (LLMs) struggle to effectively plan and act in dynamic, multi-room 3D environments. We posit that part of this limitation is due to the lack of proper 3D spatial-temporal memory modeling in LLMs. To addre

Yizhou Zhou

In this work, we present a numerical method for the initial-boundary value problem (IBVP) of first-order hyperbolic systems with source terms. The scheme directly solves the relaxation system using a relatively coarse mesh and captures the equilibrium behavior quite well, even in the presence of boundary layers. This method extends the concept of asymptotic-

Michael Kirchhof, Gjergji Kasneci, Enkelejda Kasneci

Large-language models (LLMs) and chatbot agents are known to provide wrong outputs at times, and it was recently found that this can never be fully prevented. Hence, uncertainty quantification plays a crucial role, aiming to quantify the level of ambiguity in either one overall number or two numbers for aleatoric and epistemic uncertainty. This position pape

Ce Zhang, Kaixin Ma, Tianqing Fang, Wenhao Yu

Recent Large Vision-Language Models (LVLMs) have advanced multi-modal understanding by incorporating finer-grained visual perception and encoding. However, such methods incur significant computational costs due to longer visual token sequences, posing challenges for real-time deployment. To mitigate this, prior studies have explored pruning unimportant visua

Ang Lv, Ruobing Xie, Xingwu Sun, Zhanhui Kang

Recent studies on post-training large language models (LLMs) for reasoning through reinforcement learning (RL) typically focus on tasks that can be accurately verified and rewarded, such as solving math problems. In contrast, our research investigates the impact of reward noise, a more practical consideration for real-world scenarios involving the post-train

Matteo Gallet, Georg Grasegger, Matthias Himmelmann, Jan Legerský

We present PyRigi, a novel Python package designed to study the rigidity properties of graphs and frameworks. Among many other capabilities, PyRigi can determine whether a graph admits only finitely many ways, up to isometries, of being drawn in the plane once the edge lengths are fixed, whether it has a unique embedding, or whether it satisfied such propert

Yi Ding, Ruqi Zhang

Reasoning Vision-Language Models (VLMs) have shown promising performance on complex multimodal tasks. However, they still face significant challenges: they are highly sensitive to reasoning errors, require large volumes of annotated data or accurate verifiers, and struggle to generalize beyond specific domains. To address these limitations, we explore self-c

Feng Yao, Zilong Wang, Liyuan Liu, Junxia Cui

Code generation with large language models (LLMs), often termed vibe coding, is increasingly adopted in production but fails to ensure code quality, particularly in security (e.g., SQL injection vulnerabilities) and maintainability (e.g., missing type annotations). Existing methods, such as supervised fine-tuning and rule-based post-processing, rely on labor

Maria-Florina Balcan, Avrim Blum, Zhiyuan Li, Dravyansh Sharma

Chain-of-Thought reasoning has emerged as a powerful approach for solving complex mathematical and logical problems. However, it can often veer off track through incorrect or unsubstantiated inferences. Formal mathematical reasoning, which can be checked with a formal verifier, is one approach to addressing this issue. However, currently LLMs are simply not

Guoxuan Chen, Lianghao Xia, Chao Huang

Modern recommender systems powered by Graph Neural Networks (GNNs) excel at modeling complex user-item interactions, yet increasingly face scenarios requiring selective forgetting of training data. Beyond user requests to remove specific interactions due to privacy concerns or preference changes, regulatory frameworks mandate recommender systems' ability to

Jialong Wu, Baixuan Li, Runnan Fang, Wenbiao Yin

Addressing intricate real-world problems necessitates in-depth information seeking and multi-step reasoning. Recent progress in agentic systems, exemplified by Deep Research, underscores the potential for autonomous multi-step research. In this work, we present a cohesive paradigm for building end-to-end agentic information seeking agents from a data-centric

Zhe Kong, Feng Gao, Yong Zhang, Zhuoliang Kang

Audio-driven human animation methods, such as talking head and talking body generation, have made remarkable progress in generating synchronized facial movements and appealing visual quality videos. However, existing methods primarily focus on single human animation and struggle with multi-stream audio inputs, facing incorrect binding problems between audio

Pardis Semnani, Vincent Guan, Elina Robeva, Darrick Lee

We develop a consistent method for estimating the parameters of a rich class of path-dependent SDEs, called signature SDEs, which can model general path-dependent phenomena. Path signatures are iterated integrals of a given path with the property that any sufficiently nice function of the path can be approximated by a linear functional of its signatures. Thi

Hanjia Lyu, Jiebo Luo, Jian Kang, Allison Koenecke

While the capabilities of Large Language Models (LLMs) have been studied in both Simplified and Traditional Chinese, it is yet unclear whether LLMs exhibit differential performance when prompted in these two variants of written Chinese. This understanding is critical, as disparities in the quality of LLM responses can perpetuate representational harms by ign

Mohamed Aly Bouke

Most classical and post-quantum cryptographic assumptions, including integer factorization, discrete logarithms, and Learning with Errors (LWE), rely on algebraic structures such as rings or vector spaces. While mathematically powerful, these structures can be exploited by quantum algorithms or advanced algebraic attacks, raising a pressing need for structur

Dekai Zhu, Yixuan Hu, Youquan Liu, Dongyue Lu

Leveraging recent diffusion models, LiDAR-based large-scale 3D scene generation has achieved great success. While recent voxel-based approaches can generate both geometric structures and semantic labels, existing range-view methods are limited to producing unlabeled LiDAR scenes. Relying on pretrained segmentation models to predict the semantic maps often re

Younggyo Seo, Carmelo Sferrazza, Haoran Geng, Michal Nauman

Reinforcement learning (RL) has driven significant progress in robotics, but its complexity and long training times remain major bottlenecks. In this report, we introduce FastTD3, a simple, fast, and capable RL algorithm that significantly speeds up training for humanoid robots in popular suites such as HumanoidBench, IsaacLab, and MuJoCo Playground. Our rec

Chengzhi Shi, Stratis Ioannidis

Survival analysis is widely deployed in a diverse set of fields, including healthcare, business, ecology, etc. The Cox Proportional Hazard (CoxPH) model is a semi-parametric model often encountered in the literature. Despite its popularity, wide deployment, and numerous variants, scaling CoxPH to large datasets and deep architectures poses a challenge, espec

Hadrian Heine

Homology is characterized by the Eilenberg-Steenrod axioms. We define homology of higher categories via a categorical analogue of the Eilenberg-Steenrod axioms. We prove a categorical Dold-Kan correspondence, providing a combinatorial presentation of categorical homology in which the Street nerve plays the role of the singular complex. This implies a categor

Nidhi Kalra, Robin Wang, Ismael Arciniegas Rueda

This working paper examines how geopolitical strategies and energy resource management intersect with Artificial Intelligence (AI) development, delineating the AI-energy nexus as critical to sustaining U.S. AI leadership. By analyzing the centralized approaches of authoritarian regimes like China and Gulf nations, alongside market-driven approaches in the U.

Denis Donadel, Gabriele Crestanello, Giulio Morandini, Daniele Antonioli

Industrial Control Systems (ICS) manage critical infrastructures like power grids and water treatment plants. Cyberattacks on ICSs can disrupt operations, causing severe economic, environmental, and safety issues. For example, undetected pollution in a water plant can put the lives of thousands at stake. ICS researchers have increasingly turned to honeypots

Joschka Braun, Carsten Eickhoff, David Krueger, Seyed Ali Bahrainian

Steering vectors are a lightweight method to control language model behavior by adding a learned bias to the activations at inference time. Although steering demonstrates promising performance, recent work shows that it can be unreliable or even counterproductive in some cases. This paper studies the influence of prompt types and the geometry of activation d

Jixin Zhao, Zhouxia Wang, Peiqing Yang, Shangchen Zhou

Object removal requires eliminating not only the target object but also its associated visual effects such as shadows and reflections. However, diffusion-based inpainting and removal methods often introduce artifacts, hallucinate contents, alter background, and struggle to remove object effects accurately. To address these challenges, we propose ObjectClear,

Fangcong Yin, Zeyu Leo Liu, Liu Leqi, Xi Ye

A common approach for teaching large language models (LLMs) to reason is to train on chain-of-thought (CoT) traces of in-distribution reasoning problems, but such annotated data is costly to obtain for every problem of interest. We want reasoning models to generalize beyond their training distribution, and ideally to generalize compositionally: combine atomi

Rui Li, Zixuan Hu, Wenxi Qu, Jinouwen Zhang

Scientific embodied agents play a crucial role in modern laboratories by automating complex experimental workflows. Compared to typical household environments, laboratory settings impose significantly higher demands on perception of physical-chemical transformations and long-horizon planning, making them an ideal testbed for advancing embodied intelligence.

Yida Xue, Zhen Bi, Jinnan Yang, Jungang Lou

Recent advances in Multimodal Large Language Models (MLLMs) have significantly enhanced their capabilities; however, their spatial perception abilities remain a notable limitation. To address this challenge, multimodal data synthesis offers a promising solution. Yet, ensuring that synthesized data adhere to spatial common sense is a non-trivial task. Our app

Chao Ying, Jun Jin, Yi Guo, Xiudi Li

Collecting gold-standard phenotype data via manual extraction is typically labor-intensive and slow, whereas automated computational phenotypes (ACPs) offer a systematic and much faster alternative. However, simply replacing the gold-standard with ACPs, without acknowledging their differences, could lead to biased results and misleading conclusions. Motivate

Atanu Barai, Stephan Eidenbenz, Nandakishore Santhi

To fully leverage the potential of artificial intelligence (AI) systems in a trustworthy manner, it is desirable to couple multiple AI and non-AI systems together seamlessly for constraining and ensuring correctness of the output. This paper introduces a novel parallel discrete event simulation (PDES) based methodology to combine multiple AI and non-AI agent

Yilmaz Ege Gonul, Ceyhun Efe Kayan, Ilknur Mustafazade, Nagarajan Kandasamy

Oscillator-based Ising machines (OIMs) and oscillator-based Potts machines (OPMs) have emerged as promising hardware accelerators for solving NP-hard combinatorial optimization problems by leveraging the phase dynamics of coupled oscillators. In this work, a GPU-accelerated simulated OIM/OPM digital computation framework capable of solving combinatorial opti

Wenceslao Arroyo-Machado, Nicolas Robinson-Garcia, Daniel Torres-Salinas

This study examines the shift in the scientific community from X (formerly Twitter) to Bluesky, its impact on scientific communication, and consequently on social metrics (altmetrics). We analysed 14,497 publications from multidisciplinary and Library and Information Science (LIS) journals between January 2024 and March 2025. The results reveal a notable inc

Ziling Cheng, Meng Cao, Marc-Antoine Rondeau, Jackie Chi Kit Cheung

The widespread success of large language models (LLMs) on NLP benchmarks has been accompanied by concerns that LLMs function primarily as stochastic parrots that reproduce texts similar to what they saw during pre-training, often erroneously. But what is the nature of their errors, and do these errors exhibit any regularities? In this work, we examine irrele

Edward H. Chen, Senrui Chen, Laurin E. Fischer, Andrew Eddins

To successfully perform quantum computations, it is often necessary to first accurately characterize the noise in the underlying hardware. However, it is well known that fundamental limitations prevent the unique identification of the noise. This raises the question of whether these limitations impact the ability to predict noisy dynamics and mitigate errors

Delin Zhang, Ananya Renuka Balakrishna

We present a continuum model for symmetry-breaking phase transformations in intercalation compounds, based on Ericksen's multi-well energy formulation. The model predicts the nucleation and growth of crystallographic microstructures in Li$_{2}$Mn$_{2}$O$_{4}$ -- a representative intercalation compound -- with twin boundary orientations and volume fractions t

Yijun Shen, Delong Chen, Fan Liu, Xingyu Wang

While densely annotated image captions significantly facilitate the learning of robust vision-language alignment, methodologies for systematically optimizing human annotation efforts remain underexplored. We introduce Chain-of-Talkers (CoTalk), an AI-in-the-loop methodology designed to maximize the number of annotated samples and improve their comprehensiven

Yu Zhang, Yuqi Xie, Huihan Liu, Rutav Shah

Imitation learning advances robot capabilities by enabling the acquisition of diverse behaviors from human demonstrations. However, large-scale datasets used for policy training often introduce substantial variability in quality, which can negatively impact performance. As a result, automatically curating datasets by filtering low-quality samples to improve

Qirui Li

We prove both the biquadratic Guo--Jacquet Fundamental Lemma (FL) and the biquadratic linear Arithmetic Fundamental Lemma (AFL) for GL(4) with the unit test function. Our approach relies on a detailed study of pairs of quadratic embeddings, which ultimately enables a reduction from the biquadratic case of GL(4) to the coquadratic case of GL(2). We further id

Ryan J. French, Maria D. Kazachenko, Teodora Mihailescu, Katharine K. Reeves

Despite their somewhat-frequent appearance in EUV imaging of off-limb flares, the origins of Supra-Arcade Downflows (SADs) remain a mystery. Appearing as dark, tendril-like downflows above growing flare loop arcades, SADs themselves are yet to be tied into the standard model of solar flares. The uncertainty of their origin is, in part, due to a lack of spect

Seokjin Moon, David T. Limmer

We study diffusion-controlled processes in nonequilibrium steady states, where standard rate theory assumptions break down. Using transition path theory, we generalize the relations between reactive probability fluxes and measures of the rate of the reaction. Stochastic thermodynamics analysis reveals how work constrains the enhancement of rates relative to

Jiawei Ge, Amanda Wang, Shange Tang, Chi Jin

Modern foundation models exhibit remarkable out-of-distribution (OOD) generalization, solving tasks far beyond the support of their training data. However, the theoretical principles underpinning this phenomenon remain elusive. This paper investigates this problem by examining the compositional generalization abilities of diffusion models in image generation

Noam Soker

I propose a scenario that allows white dwarfs (WDs) to launch relatively powerful jets when they enter a common envelope evolution (CEE) or experience a grazing envelope evolution (GEE) with a red giant branch star (RGB) or an asymptotic giant branch (AGB) star. In this, still a speculative scenario, the accretion for a time is mainly onto an accretion disk

Jan Hubička, Matěj Konečný, Štěpán Vodseďálek, Andy Zucker

Big Ramsey degrees of Fra\"iss\'e limits of finitely constrained free amalgamation classes in finite binary languages have been recently fully characterised by Balko, Chodounsk\'y, Dobrinen, Hubi\v{c}ka, Kone\v{c}n\'y, Vena, and Zucker. A special case of this characterisation is the universal homogeneous $K_4$-free graph. We give a self-contained and relativ

C. G. Liu, P. Bodorik, D. Jutla

Research on blockchains addresses multiple issues, with one being writing smart contracts. In our previous research we described methodology and a tool to generate, in automated fashion, smart contracts from BPMN models. The generated smart contracts provide support for multi-step transactions that facilitate repair/upgrade of smart contracts. In this paper

Ziyue Kang, Weichuan Zhang

A major challenge in rare animal image classification is the scarcity of data, as many species usually have only a small number of labeled samples. To address this challenge, we designed a hybrid deep-learning framework comprising a novel adaptive DCT preprocessing module, ViT-B16 and ResNet50 backbones, and a Bayesian linear classification head. To our know

Chengyue Wu, Hao Zhang, Shuchen Xue, Zhijian Liu

Diffusion-based large language models (Diffusion LLMs) have shown promise for non-autoregressive text generation with parallel decoding capabilities. However, the practical inference speed of open-sourced Diffusion LLMs often lags behind autoregressive models due to the lack of Key-Value (KV) Cache and quality degradation when decoding multiple tokens simult

Ganqu Cui, Yuchen Zhang, Jiacheng Chen, Lifan Yuan

This paper aims to overcome a major obstacle in scaling RL for reasoning with LLMs, namely the collapse of policy entropy. Such phenomenon is consistently observed across vast RL runs without entropy intervention, where the policy entropy dropped sharply at the early training stage, this diminished exploratory ability is always accompanied with the saturatio

Yezhi Shen, Qiuchen Zhai, Fengqing Zhu

Neural rendering methods have gained significant attention for their ability to reconstruct 3D scenes from 2D images. The core idea is to take multiple views as input and optimize the reconstructed scene by minimizing the uncertainty in geometry and appearance across the views. However, the reconstruction quality is limited by the number of input views. This

M. Carlos, A. M. Amarsi, P. E. Nissen, G. Canocchi

Highly-differential spectroscopic studies have revealed that the Sun is deficient in refractory elements relative to solar twins. To investigate the role of giant planets on this signature, we present a high precision abundance analysis of HARPS spectra for 50 F- and G-type stars spanning -0.4<[Fe/H]<+0.5. There are 29 stars in the sample which host planets

Jing-Zhi Zhou, Zhi-Chao Li, Di Wu

In contrast to the large-scale primordial power spectrum $\mathcal{P}_{\zeta}(k)$ and primordial non-Gaussianity $f_{\mathrm{NL}}$, which are strictly constrained, the small-scale $\mathcal{P}_{\zeta}(k)$ and $f_{\mathrm{NL}}$ remain less restricted. Considering local-type primordial non-Gaussianity, we study the PBH and SIGW caused by large-amplitude small-

Yuchi Wang, Yishuo Cai, Shuhuai Ren, Sihan Yang

Image recaptioning is widely used to generate training datasets with enhanced quality for various multimodal tasks. Existing recaptioning methods typically rely on powerful multimodal large language models (MLLMs) to enhance textual descriptions, but often suffer from inaccuracies due to hallucinations and incompleteness caused by missing fine-grained detail

C. G. Liu, P. Bodorik, D. Jutla

This paper addresses the challenge of creating smart contracts for applications represented using Business Process Management and Notation (BPMN) models. In our prior work we presented a methodology that automates the generation of smart contracts from BPMN models. This approach abstracts the BPMN flow control, making it independent of the underlying blockch

Piotr Kamiński, Artur Tyliszczak, Witold Elsner, Paweł Niegodajew

Recently, Kami\'nski et al. [1] demonstrated that a two-dimensional streamwise waviness with carefully selected amplitude and period can be effectively used in postponement of a flow separation at high Reynolds number which is out of reach for other commonly known passive flow control strategies. This paper demonstrates that this approach can be substantiall

Tobias Schwarz, Tobias Kamm, Alexis Engelke

Fast machine code generation is especially important for fast start-up just-in-time compilation, where the compilation time is part of the end-to-end latency. However, widely used compiler frameworks like LLVM do not prioritize fast compilation and require an extra IR translation step increasing latency even further; and rolling a custom code generator is a

Alanna Hazlett, Naomi Ohashi, Timothy Rodriguez, Sodiq Adewole

In this work, we investigate the performance across multiple classification models to classify chest X-ray images into four categories of COVID-19, pneumonia, tuberculosis (TB), and normal cases. We leveraged transfer learning techniques with state-of-the-art pre-trained Convolutional Neural Networks (CNNs) models. We fine-tuned these pre-trained architectur

Haoning Xu, Zhaoqing Li, Youjun Chen, Huimeng Wang

This paper presents a novel approach for speech foundation models compression that tightly integrates model pruning and parameter update into a single stage. Highly compact layer-level tied self-pinching gates each containing only a single learnable threshold are jointly trained with uncompressed models and used in fine-grained neuron level pruning. Experime

Tatsuro Hikawa

Ben Sa\"{\i}d-Kobayashi-Orsted introduced a family of $ \mathfrak{sl}_2 $-triples of differential-difference operators $ \mathbb{H}_{k,a} $, $ \mathbb{E}^+_{k,a} $ and $ \mathbb{E}^-_{k,a} $ on $ \mathbb{R}^N \setminus \{0\} $ indexed by a Dunkl parameter $ k $ and a deformation parameter $ a \neq 0 $. In the present paper, we study the behavior as the param

D. Dominic Briseño-Colunga, Bibek Bhandari, Debmalya Das, Long B. Nguyen

Modern superconducting and semiconducting quantum hardware use external charge and microwave flux drives to both tune and operate devices. However, each external drive is susceptible to low-frequency (e.g., $1/f$) noise that can drastically reduce the decoherence lifetime of the device unless the drive is placed at specific operating points that minimize the

Banafsheh Saber Latibari, Najmeh Nazari, Avesta Sasan, Houman Homayoun

The rise of hardware-level security threats, such as side-channel attacks, hardware Trojans, and firmware vulnerabilities, demands advanced detection mechanisms that are more intelligent and adaptive. Traditional methods often fall short in addressing the complexity and evasiveness of modern attacks, driving increased interest in machine learning-based solut

Aravinda Jatavallabha, Prabhanjan Bharadwaj, Ashish Chander

Graph Neural Networks (GNNs) are powerful tools for recommendation systems, but they often struggle under data sparsity and noise. To address these issues, we implemented LightGCL, a graph contrastive learning model that uses Singular Value Decomposition (SVD) for robust graph augmentation, preserving semantic integrity without relying on stochastic or heuri

Ruixuan Zhang, He Wang, Zhengyu Zhao, Zhiqing Guo

Rapid advances in Artificial Intelligence Generated Images (AIGI) have facilitated malicious use, such as forgery and misinformation. Therefore, numerous methods have been proposed to detect fake images. Although such detectors have been proven to be universally vulnerable to adversarial attacks, defenses in this field are scarce. In this paper, we first ide

Dana Bartošová, David Chodounský, Barbara Csima, Jan Hubička

We show that the big Ramsey degree of the Boolean algebra with 3 atoms within the countable atomless Boolean algebra is infinite.

Mahtab Alizadeh Vandchali, Fangshuo, Liao, Anastasios Kyrillidis

Sequential learning -- where complex tasks are broken down into simpler, hierarchical components -- has emerged as a paradigm in AI. This paper views sequential learning through the lens of low-rank linear regression, focusing specifically on how errors propagate when learning rank-1 subspaces sequentially. We present an analysis framework that decomposes th

Jacob L. Block, Aryan Mokhtari, Sanjay Shakkottai

Machine unlearning algorithms aim to remove the influence of specific training samples, ideally recovering the model that would have resulted from training on the remaining data alone. We study unlearning in the overparameterized setting, where many models interpolate the data, and defining the solution as any loss minimizer over the retained set$\unicode{x2

Kejian Chen, Zhengrong Li, Kohei Inayoshi, Luis C. Ho

Little red dots (LRDs), a population of active galactic nuclei (AGNs) recently identified by JWST, are characterized by their compact morphology and red optical continuum emission, which is often interpreted as evidence for significant dust extinction of $A_V \gtrsim 3$ mag. However, the dust-reddened AGN scenario is increasingly challenged by their faint ne

Jack T. Hughes, Garegin Mazmanyan, Mohammad Ghufran, Hossein Rastgoftar

We present a VR-based teleoperation system for multirotor flight that renders a third-person view (TPV) of the vehicle together with a live 3D reconstruction of its surroundings. The system runs on an embedded GPU (Jetson Orin NX) with ROS2-WebXR integration and streams geometry and video to a headset for closed-loop control in previously unmapped spaces. We

Luca Maria Del Bono, Federico Ricci-Tersenghi, Francesco Zamponi

Recent years have seen a rise in the application of machine learning techniques to aid the simulation of hard-to-sample systems that cannot be studied using traditional methods. Despite the introduction of many different architectures and procedures, a wide theoretical understanding is still lacking, with the risk of suboptimal implementations. As a first st

Ngoc La, Ruaridh Mon-Williams, Julie A. Shah

In recent years, reinforcement learning (RL) methods have been widely tested using tools like OpenAI Gym, though many tasks in these environments could also benefit from hierarchical planning. However, there is a lack of a tool that enables seamless integration of hierarchical planning with RL. Hierarchical Domain Definition Language (HDDL), used in classica

Jiaqi Huang, Zunnan Xu, Jun Zhou, Ting Liu

Leveraging multimodal large models for image segmentation has become a prominent research direction. However, existing approaches typically rely heavily on manually annotated datasets that include explicit reasoning processes, which are costly and time-consuming to produce. Recent advances suggest that reinforcement learning (RL) can endow large models with

A. V. Belitsky, V. A. Smirnov

We study the Sudakov form factor on the Coulomb branch of N=4 sYM, which endows only external states with masses, and implies that the former is off-shell in the traditional sense. Our consideration is performed at three-loop order in the near mass-shell limit. We use a combination of tools to perform required calculations centered around the Method of Regio

Longlin Wang, Yanke Song, Kuanhao Jiang, Pragya Sur

Approximate Message Passing (AMP) algorithms enable precise characterization of certain classes of random objects in the high-dimensional limit, and have found widespread applications in fields such as signal processing, statistics, and communications. In this work, we introduce Multi-Environment Generalized Long AMP, a novel AMP framework that applies to tr

Nabil L. Youssef, Ebtsam H. Taha, A. A. Kotb, S. G. Elgendi

This study presents many special anisotropic conformal changes of a conic pseudo-Finsler surface $(M,F)$, such as $C$-anisotropic and horizontal $C$-anisotropic conformal transformations, which reduce to $C$-conformal when the conformal factor is solely position-dependent. Furthermore, we present vertical $C$-anisotropic conformal changes and demonstrate tha

Yiheng Li, Francisco Carrillo-Perez, Mohammed Alawad, Olivier Gevaert

Lung cancer is the leading cause of cancer mortality worldwide, and non-invasive methods for detecting key mutations and staging are essential for improving patient outcomes. Here, we compare the performance of two machine learning models - FMCIB+XGBoost, a supervised model with domain-specific pretraining, and Dinov2+ABMIL, a self-supervised model with atte

Erxin Yu, Jing Li, Ming Liao, Qi Zhu

Although large language models demonstrate strong performance across various domains, they still struggle with numerous bad cases in mathematical reasoning. Previous approaches to learning from errors synthesize training data by solely extrapolating from isolated bad cases, thereby failing to generalize the extensive patterns inherent within these cases. Thi

Jakub Podolak, Rajeev Verma

We study the source of uncertainty in DeepSeek R1-32B by analyzing its self-reported verbal confidence on question answering (QA) tasks. In the default answer-then-confidence setting, the model is regularly over-confident, whereas semantic entropy - obtained by sampling many responses - remains reliable. We hypothesize that this is because of semantic entrop

Paulo Luz, Filipe C. Mena

We study the initial value problem in Einstein-Cartan theory which includes torsion and, therefore, a non-symmetric connection on the spacetime manifold. Generalizing the path of a classical theorem by Choquet-Bruhat and York for the Einstein equations, we use a $n+1$ splitting of the manifold and compute the evolution and constraint equations for the Einste

Alexey Bychkov, Alexey Litvinov

In this paper, we explore a new class of integrable sigma models, which we refer to as the "dual regime" of Yang-Baxter (YB) deformed $\mathrm{O}(2N)$ sigma models. This dual regime manifests itself in the conformal perturbation approach. Namely, it is well known that conventional YB-deformed $\mathrm{O}(N)$ sigma models are described in the UV by a collecti

Brian Hopkins, James A. Sellers

Kaur, Rana, and Eyyunni recently defined the mex sequence of a partition and established, by analytic methods, connections to two disparate types of partition-related objects. We make a bijection between partitions with certain mex sequences and a uniform family of overpartitions which allows us to provide combinatorial proofs of their results, as they reque

Yoav Gur-Arieh, Clara Suslik, Yihuai Hong, Fazl Barez

Large language models (LLMs) often acquire knowledge during pretraining that is undesirable in downstream deployments, e.g., sensitive information or copyrighted content. Existing approaches for removing such knowledge rely on fine-tuning, training low-rank adapters or fact-level editing, but these are either too coarse, too shallow, or ineffective. In this

N. Piña-León, V. Mantič, S. Jiménez-Alfaro

This study introduces a recursive method for computing asymptotic solutions of the Laplace equation in corner domains with the homogeneous Dirichlet boundary condition on one side and the Robin boundary condition with a power-law coefficient variation with exponent $\alpha\in \mathbb{R}$ on the other side (D-R corner problem). An asymptotic solution of this

Navve Wasserman, Oliver Heinimann, Yuval Golbari, Tal Zimbalist

Rerankers play a critical role in multimodal Retrieval-Augmented Generation (RAG) by refining ranking of an initial set of retrieved documents. Rerankers are typically trained using hard negative mining, whose goal is to select pages for each query which rank high, but are actually irrelevant. However, this selection process is typically passive and restrict

Tobias Lindenbauer, Egor Bogomolov, Yaroslav Zharov

Benchmarks for Software Engineering (SE) AI agents, most notably SWE-bench, have catalyzed progress in programming capabilities of AI agents. However, they overlook critical developer workflows such as Version Control System (VCS) operations. To address this issue, we present GitGoodBench, a novel benchmark for evaluating AI agent performance on VCS tasks. G

Xue Zhang, Yunlong Liang, Fandong Meng, Songming Zhang

Continually expanding new languages for existing large language models (LLMs) is a promising yet challenging approach to building powerful multilingual LLMs. The biggest challenge is to make the model continuously learn new languages while preserving the proficient ability of old languages. To achieve this, recent work utilizes the Mixture-of-Experts (MoE) a

Kartik Kuckreja, Parul Gupta, Injy Hamed, Thamar Solorio

Deepfake generation methods are evolving fast, making fake media harder to detect and raising serious societal concerns. Most deepfake detection and dataset creation research focuses on monolingual content, often overlooking the challenges of multilingual and code-switched speech, where multiple languages are mixed within the same discourse. Code-switching,

Louis Shuo Wang, Jiguang Yu, Zonghao Liu

The main obstacle to effective cancer treatment is the development of drug resistance, which can be divided into two categories: spontaneous and acquired drug resistance. Non-small cell lung cancer (NSCLC) is the main cause of cancer-related deaths worldwide. A subset of lung cancer, adenocarcinomas, is characterised by mutations in the epidermal growth fact

Reid Ferguson, Olaf Hartwig, Guido Mueller

Many years of development have gone into producing instruments that meet the required noise performance of the LISA interferometric detection system. Concurrently, software simulations have been used to extensively develop the data analysis libraries to be used in the LISA pipeline, not least among which are the Time Delay Interferometry (TDI) algorithms. To

Jan Göpfert, Jann M. Weinand, Patrick Kuckertz, Noah Pflugradt

Humanity is progressing towards automated product development, a trend that promises faster creation of better products and thus the acceleration of technological progress. However, increasing reliance on non-human agents for this process introduces many risks. This perspective aims to initiate a discussion on these risks and appropriate mitigation strategie

Elahe Khalili Samani, Marco Radeschi

We prove that a closed, simply connected, positively curved, cohomogeneity-three manifold whose quotient space has no boundary is rationally elliptic, thus providing a generalization of similar results regarding rational ellipticity of homogeneous, cohomogeneity-one, and almost non-negatively curved cohomogeneity-two manifolds.

Qing-He Ni, Christian Hill, Sergei N. Yurchenko, Marco Pezzella

We present the ExoPhoto database (https://exomol.com/exophoto/), an extension of the ExoMol database, specifically developed to address the growing need for high-accuracy, temperature-dependent photodissociation cross section data towards short-UV wavelengths. ExoPhoto combines theoretical models from three major computational databases (ExoMol, UGAMOP and P

Gerard McCaul, Juan Sebastian Totero Gongora, Wendy Otieno, Sergey Savelev

We investigate a minimal architecture for quantum reservoir computing based on Hamiltonian encoding, in which input data is injected via modulation of system parameters rather than state preparation. This approach circumvents many of the experimental overheads typically associated with quantum machine learning, enabling computation without feedback, memory,

Yijie Zhu, Evan Saraivanov, Joshua A. Kable, Artemis Sofia Giannakopoulou

Machine learning can accelerate cosmological inferences that involve many sequential evaluations of computationally expensive data vectors. Previous works in this series have examined how machine learning architectures impact emulator accuracy and training time for optical shear and galaxy clustering 2-point function. In this final manuscript, we explore neu

Guy Moss, Leah Sophie Muhle, Reinhard Drews, Jakob H. Macke

Simulation-based inference (SBI) is an established approach for performing Bayesian inference on scientific simulators. SBI so far works best on low-dimensional parametric models. However, it is difficult to infer function-valued parameters, which frequently occur in disciplines that model spatiotemporal processes such as the climate and earth sciences. Here

Waldemar Chang, Alhassan Yasin

We present Fusion Steering, an activation steering methodology that improves factual accuracy in large language models (LLMs) for question-answering (QA) tasks. This approach introduces flexible steering configurations, including full-layer steering and segmented steering. Unlike traditional methods constrained to single-layer or fixed-layer operations, Fusi

Hoang Pham, Thuy-Duong Nguyen, Khac-Hoai Nam Bui

This paper presents a novel approach for unified retrieval-augmented generation (RAG) systems using the recent emerging large language model (LLM) agent concept. Specifically, Agent LLM, which utilizes LLM as fundamental controllers, has become a promising approach to enable the interpretability of RAG tasks, especially for complex reasoning question-answeri

Iain Moffatt, Maya Thompson

Brylawski's tensor product formula expresses the Tutte polynomial of the tensor product of two graphs in terms of Tutte polynomials arising from the tensor factors. Analogous tensor product formulas are known for the ribbon graph polynomial and transition polynomials of graphs embedded in surfaces, as well as for the Bollob\'as-Riordan polynomial in some spe

Dmitrii Sorokin, Maksim Nakhodnov, Andrey Kuznetsov, Aibek Alanov

Recent advances in diffusion models have led to impressive image generation capabilities, but aligning these models with human preferences remains challenging. Reward-based fine-tuning using models trained on human feedback improves alignment but often harms diversity, producing less varied outputs. In this work, we address this trade-off with two contributi

Aravind R. Krishnan, Thomas Z. Li, Lucas W. Remedios, Michael E. Kim

Reconstruction kernels in computed tomography (CT) affect spatial resolution and noise characteristics, introducing systematic variability in quantitative imaging measurements such as emphysema quantification. Choosing an appropriate kernel is therefore essential for consistent quantitative analysis. We propose a multipath cycleGAN model for CT kernel harmon

Arindam Chaudhuri

Human beings rely heavily on estimation of poses in order to access their body movements. Human pose estimation methods take advantage of computer vision advances in order to track human body movements in real life applications. This comes from videos which are recorded through available devices. These para-digms provide potential to make human movement meas