id
string | submitter
string | authors
string | title
string | comments
string | journal-ref
string | doi
string | report-no
string | categories
string | license
string | abstract
string | versions
list | update_date
timestamp[s] | authors_parsed
sequence |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
2409.12067 | Tetiana Parshakova | Tetiana Parshakova and Trevor Hastie and Stephen Boyd | Fitting Multilevel Factor Models | null | null | null | null | stat.ML cs.LG cs.MS stat.CO | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We examine a special case of the multilevel factor model, with covariance given by multilevel low rank (MLR) matrix~\cite{parshakova2023factor}. We develop a novel, fast implementation of the expectation-maximization algorithm, tailored for multilevel factor models, to maximize the likelihood of the observed data. This method accommodates any hierarchical structure and maintains linear time and storage complexities per iteration. This is achieved through a new efficient technique for computing the inverse of the positive definite MLR matrix. We show that the inverse of positive definite MLR matrix is also an MLR matrix with the same sparsity in factors, and we use the recursive Sherman-Morrison-Woodbury matrix identity to obtain the factors of the inverse. Additionally, we present an algorithm that computes the Cholesky factorization of an expanded matrix with linear time and space complexities, yielding the covariance matrix as its Schur complement. This paper is accompanied by an open-source package that implements the proposed methods. | [
{
"version": "v1",
"created": "Wed, 18 Sep 2024 15:39:12 GMT"
},
{
"version": "v2",
"created": "Sun, 29 Sep 2024 20:30:42 GMT"
},
{
"version": "v3",
"created": "Sun, 08 Jun 2025 01:51:17 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Parshakova",
"Tetiana",
""
],
[
"Hastie",
"Trevor",
""
],
[
"Boyd",
"Stephen",
""
]
] |
2409.12519 | Chunying Zhou | Chunying Zhou, Xiaoyuan Xie, Gong Chen, Peng He, Bing Li | Multi-View Adaptive Contrastive Learning for Information Retrieval Based Fault Localization | null | null | null | null | cs.SE cs.IR | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Most studies focused on information retrieval-based techniques for fault localization, which built representations for bug reports and source code files and matched their semantic vectors through similarity measurement. However, such approaches often ignore some useful information that might help improve localization performance, such as 1) the interaction relationship between bug reports and source code files; 2) the similarity relationship between bug reports; and 3) the co-citation relationship between source code files. In this paper, we propose a novel approach named Multi-View Adaptive Contrastive Learning for Information Retrieval Fault Localization (MACL-IRFL) to learn the above-mentioned relationships for software fault localization. Specifically, we first generate data augmentations from report-code interaction view, report-report similarity view and code-code co-citation view separately, and adopt graph neural network to aggregate the information of bug reports or source code files from the three views in the embedding process. Moreover, we perform contrastive learning across these views. Our design of contrastive learning task will force the bug report representations to encode information shared by report-report and report-code views,and the source code file representations shared by code-code and report-code views, thereby alleviating the noise from auxiliary information. Finally, to evaluate the performance of our approach, we conduct extensive experiments on five open-source Java projects. The results show that our model can improve over the best baseline up to 28.93%, 25.57% and 20.35% on Accuracy@1, MAP and MRR, respectively. | [
{
"version": "v1",
"created": "Thu, 19 Sep 2024 07:20:10 GMT"
},
{
"version": "v2",
"created": "Mon, 09 Jun 2025 05:51:50 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Zhou",
"Chunying",
""
],
[
"Xie",
"Xiaoyuan",
""
],
[
"Chen",
"Gong",
""
],
[
"He",
"Peng",
""
],
[
"Li",
"Bing",
""
]
] |
2409.12949 | Dingqi Zhang | Dingqi Zhang, Antonio Loquercio, Jerry Tang, Ting-Hao Wang, Jitendra Malik and Mark W. Mueller | A Learning-based Quadcopter Controller with Extreme Adaptation | Accepted for the Transaction on Robotics (T-RO), April 2025 | null | 10.1109/TRO.2025.3577037 | null | cs.RO | http://creativecommons.org/licenses/by/4.0/ | This paper introduces a learning-based low-level controller for quadcopters, which adaptively controls quadcopters with significant variations in mass, size, and actuator capabilities. Our approach leverages a combination of imitation learning and reinforcement learning, creating a fast-adapting and general control framework for quadcopters that eliminates the need for precise model estimation or manual tuning. The controller estimates a latent representation of the vehicle's system parameters from sensor-action history, enabling it to adapt swiftly to diverse dynamics. Extensive evaluations in simulation demonstrate the controller's ability to generalize to unseen quadcopter parameters, with an adaptation range up to 16 times broader than the training set. In real-world tests, the controller is successfully deployed on quadcopters with mass differences of 3.7 times and propeller constants varying by more than 100 times, while also showing rapid adaptation to disturbances such as off-center payloads and motor failures. These results highlight the potential of our controller in extreme adaptation to simplify the design process and enhance the reliability of autonomous drone operations in unpredictable environments. The video and code are at: https://github.com/muellerlab/xadapt_ctrl | [
{
"version": "v1",
"created": "Thu, 19 Sep 2024 17:56:30 GMT"
},
{
"version": "v2",
"created": "Sun, 08 Jun 2025 19:33:04 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Zhang",
"Dingqi",
""
],
[
"Loquercio",
"Antonio",
""
],
[
"Tang",
"Jerry",
""
],
[
"Wang",
"Ting-Hao",
""
],
[
"Malik",
"Jitendra",
""
],
[
"Mueller",
"Mark W.",
""
]
] |
2409.13353 | Mohamed Ali Hamza | Ahmad Z. Fino, Mohamed Ali Hamza | Global existence for wave equations with scale-invariant time-dependent damping and time derivative nonlinearity | We have identified an idea that may significantly improve our results, and we are currently working on it | null | null | null | math.AP | http://creativecommons.org/publicdomain/zero/1.0/ | This paper addresses the Cauchy problem for wave equations with scale-invariant time-dependent damping and nonlinear time-derivative terms, modeled as $$\partial_{t}^2u- \Delta u +\frac{\mu}{1+t}\partial_tu= f(\partial_tu), \quad x\in \mathbb{R}^n, t>0,$$ where $f(\partial_tu)=|\partial_tu|^p $ or $|\partial_tu|^{p-1}\partial_tu$ with $p>1$ and $\mu>0$. We prove global existence of small data solutions in low dimensions $1\leq n\leq 3$ by using energy estimates in appropriate Sobolev spaces. Our primary contribution is an existence result for $p>1+\frac2{\mu}$, in the one-dimensional case, when $\mu \le 2$, which in conjunction with prior blow-up results from \cite{Our2}, establish that the critical exponent for small data solutions in one dimension is $p_G(1,\mu)=1+\frac2{\mu}$, when $\mu \le 2$. To the best of our knowledge, this is the first identification of the critical exponent range for the time-dependent damped wave equations with scale-invariant and time-derivative nonlinearity. | [
{
"version": "v1",
"created": "Fri, 20 Sep 2024 09:34:30 GMT"
},
{
"version": "v2",
"created": "Sun, 08 Jun 2025 10:38:32 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Fino",
"Ahmad Z.",
""
],
[
"Hamza",
"Mohamed Ali",
""
]
] |
2409.13414 | Christian Olivera | Wladimir Neves and Christian Olivera | An Eulerian-Lagrangian formulation for compressible Euler Equations | null | null | null | null | math.AP | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | In this paper, we present a novel Eulerian-Lagrangian formulation for the compressible isentropic Euler equations with a general pressure law. Using the developed Lagrangian formulation, we show a short-time solution which is unique in the framework of Bessel space $H_{p}^{\beta}(\mathbb{T}^{d})$ for $\beta>\frac{d}{p}+1$. The proof is precise and relatively simple, relying on the application of the Implicit Function Theorem in Banach spaces combined with a carefully designed regularization procedure. To the best of our knowledge, this is the first time the Lagrangian formulation has been employed to show local wellposedness for compressible Euler equations. Moreover, it can be adaptableto $C^{k,\alpha}$ spaces, and notably Besov type spaces. | [
{
"version": "v1",
"created": "Fri, 20 Sep 2024 11:25:20 GMT"
},
{
"version": "v2",
"created": "Wed, 23 Apr 2025 11:56:38 GMT"
},
{
"version": "v3",
"created": "Sat, 07 Jun 2025 12:39:18 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Neves",
"Wladimir",
""
],
[
"Olivera",
"Christian",
""
]
] |
2409.13654 | Parham Oveissi | Parham Oveissi, Turibius Rozario, Ankit Goel | A Novel Neural Filter to Improve Accuracy of Neural Network Models of Dynamic Systems | null | null | null | null | cs.LG math.DS | http://creativecommons.org/licenses/by/4.0/ | The application of neural networks in modeling dynamic systems has become prominent due to their ability to estimate complex nonlinear functions. Despite their effectiveness, neural networks face challenges in long-term predictions, where the prediction error diverges over time, thus degrading their accuracy. This paper presents a neural filter to enhance the accuracy of long-term state predictions of neural network-based models of dynamic systems. Motivated by the extended Kalman filter, the neural filter combines the neural network state predictions with the measurements from the physical system to improve the estimated state's accuracy. The neural filter's improvements in prediction accuracy are demonstrated through applications to four nonlinear dynamical systems. Numerical experiments show that the neural filter significantly improves prediction accuracy and bounds the state estimate covariance, outperforming the neural network predictions. Furthermore, it is also shown that the accuracy of a poorly trained neural network model can be improved to the same level as that of an adequately trained neural network model, potentially decreasing the training cost and required data to train a neural network. | [
{
"version": "v1",
"created": "Fri, 20 Sep 2024 17:03:04 GMT"
},
{
"version": "v2",
"created": "Sat, 07 Jun 2025 18:42:29 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Oveissi",
"Parham",
""
],
[
"Rozario",
"Turibius",
""
],
[
"Goel",
"Ankit",
""
]
] |
2409.14453 | Xingping Zhou | Hengxuan Jiang, Xiumei Wang, and Xingping Zhou | Self-Attention Assistant Classification of Non-Hermitian Phases in Two-Dimensional Lattice | null | null | null | null | cond-mat.other | http://creativecommons.org/licenses/by/4.0/ | Classification of the non-Hermitian phases in high-dimensional lattice becomes challenging due to interplay of the band topology and non-Hermiticity. The significant increase in data dimensions and the number of categories has rendered traditional supervised learning and unsupervised manifold learning failed. Here, we propose the self-attention assistant machine learning for clustering non-Hermitian phases in two-dimensional lattice. By incorporating the self-attention mechanism, the model can effectively capture long-range dependencies and important patterns, resulting in a more compact and information-rich latent space. It can achieve Altland-Zirnbauer classification with Bloch vector dataset and distinguish the phases of eigenstates' localized behavior with the competition between non-Hermitian skin effect and topological localization. Our results provide a general method for characterizing non-Hermitian phases in two-dimensional lattice via machine learning. | [
{
"version": "v1",
"created": "Sun, 22 Sep 2024 14:06:57 GMT"
},
{
"version": "v2",
"created": "Fri, 27 Sep 2024 06:55:12 GMT"
},
{
"version": "v3",
"created": "Sun, 06 Oct 2024 08:19:00 GMT"
},
{
"version": "v4",
"created": "Mon, 09 Jun 2025 08:12:52 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Jiang",
"Hengxuan",
""
],
[
"Wang",
"Xiumei",
""
],
[
"Zhou",
"Xingping",
""
]
] |
2409.15104 | Zeyu Zhang | Zeyu Zhang, Haiying Shen | PecSched: Preemptive and Efficient Cluster Scheduling for LLM Inference | null | null | null | null | cs.DC cs.LG | http://creativecommons.org/licenses/by/4.0/ | The scaling of transformer-based Large Language Models (LLMs) has significantly expanded their context lengths, enabling applications where inputs exceed 100K tokens. Our analysis of a recent Azure LLM inference trace reveals a highly skewed long-tail distribution of input lengths, with approximately 80% of inputs shorter than 2K tokens. Long inputs constitute only a small fraction. Existing cluster-level LLM scheduling strategies, including First-In-First-Out (FIFO), reservation-based, and priority-based approaches, primarily target short-input requests with lengths below 2K and fail to address this heterogeneity, leading to inefficiencies such as head-of-line blocking, resource underutilization, and starvation of long-input requests. We propose PecSched, a Preemptive and Efficient Cluster SCHEDuling system for LLM inference. PecSched introduces the following key techniques: 1) preemptive scheduling that prioritizes short-input requests for their performance; 2) coordinated prefill-decode colocation and disaggregation, which reduces both the duration and frequency of preemptions; 3) fast Sequence Parallelism (SP) that minimizes the prefill time of long-input requests to further reduce the likelihood and frequency of preemptions. Evaluations based on Azure LLM inference trace show that, compared to state-of-the-art cluster-level LLM inference schedulers, PecSched reduces the 99th percentile queueing delay of short-input requests by up to 92% and improves their throughput by up to 595%, without significantly affecting the Job Completion Time (JCT) of long-input requests. We open-sourced our code. | [
{
"version": "v1",
"created": "Mon, 23 Sep 2024 15:16:29 GMT"
},
{
"version": "v2",
"created": "Sun, 08 Jun 2025 20:13:28 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Zhang",
"Zeyu",
""
],
[
"Shen",
"Haiying",
""
]
] |
2409.15134 | Ernesto Lax | Ernesto Lax | SCMAlgebras: a Macaulay2 package to check sequential Cohen-Macaulayness | 11 pages, Updated paper to match new release of the package | null | null | null | math.AC math.CO | http://creativecommons.org/licenses/by-sa/4.0/ | We introduce the Macaulay2 package SCMAlgebras. It provides functions for computing the modules of deficiency and the filter ideals, in order to check whether a module or an ideal is sequentially Cohen-Macaulay. After the basic algebraic notions and results are recalled, the main features of the package are described through examples | [
{
"version": "v1",
"created": "Mon, 23 Sep 2024 15:38:40 GMT"
},
{
"version": "v2",
"created": "Mon, 30 Sep 2024 09:42:49 GMT"
},
{
"version": "v3",
"created": "Sun, 06 Oct 2024 16:49:34 GMT"
},
{
"version": "v4",
"created": "Mon, 09 Jun 2025 10:25:44 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Lax",
"Ernesto",
""
]
] |
2409.16613 | Eric Yanchenko | Eric Yanchenko | Oral exams in introductory statistics class with non-native English speakers | null | null | null | null | stat.OT | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Oral exams are a powerful tool to assess student's learning. This is particularly important in introductory statistics classes where students struggle to grasp various topics like the interpretation of probability, $p$-values and more. The challenge of acquiring conceptual understanding is only heightened when students are learning in a second language. In this paper, I share my experience administering oral exams to an introductory statistics class of non-native English speakers at a Japanese university. I explain the context of the university and course, before detailing the exam. Of particular interest is the relationship between exam performance and English proficiency. The results showed little relationship between the two, meaning the exam seemed to truly test student's statistical knowledge rather than their English ability. I close with encouragements and recommendations for practitioners hoping to implement similar oral exams, focusing on the unique difficulties faced by students not learning in their mother tongue. | [
{
"version": "v1",
"created": "Wed, 25 Sep 2024 04:31:07 GMT"
},
{
"version": "v2",
"created": "Mon, 18 Nov 2024 23:20:56 GMT"
},
{
"version": "v3",
"created": "Mon, 09 Jun 2025 01:56:55 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Yanchenko",
"Eric",
""
]
] |
2409.16814 | Jong-In Kim | Jong-in Kim, Donghyun Lee | On the large amplitude solution of the Boltzmann equation with large external potential and boundary effects | 99 pages | J. Stat. Phys., Vol 192, No 77, (2025) | 10.1007/s10955-025-03459-0 | null | math.AP | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | The Boltzmann equation is a fundamental equation in kinetic theory that describes the motion of rarefied gases. In this study, we examine the Boltzmann equation within a $C^1$ bounded domain, subject to a large external potential $\Phi(x)$ and diffuse reflection boundary conditions. Initially, we prove the asymptotic stability of small perturbations near the local Maxwellian $\mu_E (x,v)$. Subsequently, we demonstrate the asymptotic stability of large amplitude solutions with initial data that is arbitrarily large in (weighted) $L^\infty$, but sufficiently small in the sense of relative entropy. Specifically, we extend the results for large amplitude solutions of the Boltzmann equation (with or without external potential) [10, 11, 12, 23] to scenarios involving significant external potentials [19, 28] under diffuse reflection boundary conditions. | [
{
"version": "v1",
"created": "Wed, 25 Sep 2024 11:09:39 GMT"
},
{
"version": "v2",
"created": "Sat, 07 Jun 2025 08:12:07 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Kim",
"Jong-in",
""
],
[
"Lee",
"Donghyun",
""
]
] |
2409.17458 | Yifan Jiang | Yifan Jiang, Kriti Aggarwal, Tanmay Laud, Kashif Munir, Jay Pujara, Subhabrata Mukherjee | RED QUEEN: Safeguarding Large Language Models against Concealed Multi-Turn Jailbreaking | Accepted in ACL 2025 Findings | null | null | null | cs.CR cs.CL cs.LG | http://creativecommons.org/licenses/by/4.0/ | The rapid progress of Large Language Models (LLMs) has opened up new opportunities across various domains and applications; yet it also presents challenges related to potential misuse. To mitigate such risks, red teaming has been employed as a proactive security measure to probe language models for harmful outputs via jailbreak attacks. However, current jailbreak attack approaches are single-turn with explicit malicious queries that do not fully capture the complexity of real-world interactions. In reality, users can engage in multi-turn interactions with LLM-based chat assistants, allowing them to conceal their true intentions in a more covert manner. To bridge this gap, we, first, propose a new jailbreak approach, RED QUEEN ATTACK. This method constructs a multi-turn scenario, concealing the malicious intent under the guise of preventing harm. We craft 40 scenarios that vary in turns and select 14 harmful categories to generate 56k multi-turn attack data points. We conduct comprehensive experiments on the RED QUEEN ATTACK with four representative LLM families of different sizes. Our experiments reveal that all LLMs are vulnerable to RED QUEEN ATTACK, reaching 87.62% attack success rate on GPT-4o and 75.4% on Llama3-70B. Further analysis reveals that larger models are more susceptible to the RED QUEEN ATTACK, with multi-turn structures and concealment strategies contributing to its success. To prioritize safety, we introduce a straightforward mitigation strategy called RED QUEEN GUARD, which aligns LLMs to effectively counter adversarial attacks. This approach reduces the attack success rate to below 1% while maintaining the model's performance across standard benchmarks. Full implementation and dataset are publicly accessible at https://github.com/kriti-hippo/red_queen. | [
{
"version": "v1",
"created": "Thu, 26 Sep 2024 01:24:17 GMT"
},
{
"version": "v2",
"created": "Sat, 07 Jun 2025 08:11:07 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Jiang",
"Yifan",
""
],
[
"Aggarwal",
"Kriti",
""
],
[
"Laud",
"Tanmay",
""
],
[
"Munir",
"Kashif",
""
],
[
"Pujara",
"Jay",
""
],
[
"Mukherjee",
"Subhabrata",
""
]
] |
2409.17641 | Amir Masoud Ghalamzan Esfahani | Venkatesh Sripada, Samuel Carter, Frank Guerin, Amir Ghalamzan | Scene Exploration by Vision-Language Models | null | null | null | null | cs.RO | http://creativecommons.org/licenses/by-sa/4.0/ | Active perception enables robots to dynamically gather information by adjusting their viewpoints, a crucial capability for interacting with complex, partially observable environments. In this paper, we present AP-VLM, a novel framework that combines active perception with a Vision-Language Model (VLM) to guide robotic exploration and answer semantic queries. Using a 3D virtual grid overlaid on the scene and orientation adjustments, AP-VLM allows a robotic manipulator to intelligently select optimal viewpoints and orientations to resolve challenging tasks, such as identifying objects in occluded or inclined positions. We evaluate our system on two robotic platforms: a 7-DOF Franka Panda and a 6-DOF UR5, across various scenes with differing object configurations. Our results demonstrate that AP-VLM significantly outperforms passive perception methods and baseline models, including Toward Grounded Common Sense Reasoning (TGCSR), particularly in scenarios where fixed camera views are inadequate. The adaptability of AP-VLM in real-world settings shows promise for enhancing robotic systems' understanding of complex environments, bridging the gap between high-level semantic reasoning and low-level control. | [
{
"version": "v1",
"created": "Thu, 26 Sep 2024 08:44:49 GMT"
},
{
"version": "v2",
"created": "Mon, 09 Jun 2025 11:32:42 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Sripada",
"Venkatesh",
""
],
[
"Carter",
"Samuel",
""
],
[
"Guerin",
"Frank",
""
],
[
"Ghalamzan",
"Amir",
""
]
] |
2409.17645 | Alberto Guandalini | Alberto Guandalini, Giovanni Caldarelli, Francesco Macheda, Francesco Mauri | High- and low-energy many-body effects of graphene in a unified approach | null | null | null | null | physics.atm-clus | http://creativecommons.org/licenses/by/4.0/ | We show that the many-body features of graphene band structure and electronic response can be accurately evaluated by applying many-body perturbation theory to a tight-binding (TB) model. In particular, we compare TB results for the optical conductivity with previous ab-initio calculations, showing a nearly perfect agreement both in the low energy region near the Dirac cone ($\sim 100$ meV), and at the higher energies of the {\pi} plasmon ($\sim 5$ eV). A reasonable agreement is reached also for the density-density response at the Brillouin zone corner. With the help of the reduced computational cost of the TB model, we study the effect of self-consistency on the screened interaction (W) and on the quasi-particle corrections, a task that is not yet achievable in ab-initio frameworks. We find that self-consistency is important to reproduce the experimental results on the divergence of the Fermi velocity, while it marginally affects the optical conductivity. Finally, we study the robustness of our results against doping or the introduction of a uniform dielectric environment. | [
{
"version": "v1",
"created": "Thu, 26 Sep 2024 08:50:16 GMT"
},
{
"version": "v2",
"created": "Fri, 31 Jan 2025 10:51:31 GMT"
},
{
"version": "v3",
"created": "Mon, 09 Jun 2025 09:12:13 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Guandalini",
"Alberto",
""
],
[
"Caldarelli",
"Giovanni",
""
],
[
"Macheda",
"Francesco",
""
],
[
"Mauri",
"Francesco",
""
]
] |
2409.18122 | Yuezhan Tao | Yuezhan Tao, Dexter Ong, Varun Murali, Igor Spasojevic, Pratik Chaudhari and Vijay Kumar | RT-GuIDE: Real-Time Gaussian splatting for Information-Driven Exploration | null | null | null | null | cs.RO | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We propose a framework for active mapping and exploration that leverages Gaussian splatting for constructing dense maps. Further, we develop a GPU-accelerated motion planning algorithm that can exploit the Gaussian map for real-time navigation. The Gaussian map constructed onboard the robot is optimized for both photometric and geometric quality while enabling real-time situational awareness for autonomy. We show through simulation experiments that our method yields comparable Peak Signal-to-Noise Ratio (PSNR) and similar reconstruction error to state-of-the-art approaches, while being orders of magnitude faster to compute. In real-world experiments, our algorithm achieves better map quality (at least 0.8dB higher PSNR and more than 16% higher geometric reconstruction accuracy) than maps constructed by a state-of-the-art method, enabling semantic segmentation using off-the-shelf open-set models. Experiment videos and more details can be found on our project page: https://tyuezhan.github.io/RT_GuIDE/ | [
{
"version": "v1",
"created": "Thu, 26 Sep 2024 17:58:05 GMT"
},
{
"version": "v2",
"created": "Sun, 08 Jun 2025 02:21:16 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Tao",
"Yuezhan",
""
],
[
"Ong",
"Dexter",
""
],
[
"Murali",
"Varun",
""
],
[
"Spasojevic",
"Igor",
""
],
[
"Chaudhari",
"Pratik",
""
],
[
"Kumar",
"Vijay",
""
]
] |
2409.18367 | Shaozong Wang | Shaozong Wang | A Framework for Gluing Harmonic Maps | null | null | null | null | math.DG | http://creativecommons.org/licenses/by/4.0/ | In this paper, we study the gluing construction of the extended harmonic maps between Riemannian manifolds. Harmonic maps are critical points of the energy functional. We construct the gluing map of the extended harmonic maps from Riemann surfaces to some Riemannian manifold $N$ under certain conditions. | [
{
"version": "v1",
"created": "Fri, 27 Sep 2024 00:58:48 GMT"
},
{
"version": "v2",
"created": "Mon, 09 Jun 2025 02:32:32 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Wang",
"Shaozong",
""
]
] |
2409.18421 | Oskar Henriksson | Oskar Henriksson, Kristian Ranestad, Lisa Seccia, Teresa Yu | Moment varieties of the inverse Gaussian and gamma distributions are nondefective | 24 pages. Minor corrections and expository improvements. To appear in J. Symb. Comput | null | 10.1016/j.jsc.2025.102460 | null | math.AG math.ST stat.TH | http://creativecommons.org/licenses/by/4.0/ | We show that the parameters of a $k$-mixture of inverse Gaussian or gamma distributions are algebraically identifiable from the first $3k-1$ moments, and rationally identifiable from the first $3k+2$ moments. Our proofs are based on Terracini's classification of defective surfaces, careful analysis of the intersection theory of moment varieties, and a recent result on sufficient conditions for rational identifiability of secant varieties by Massarenti--Mella. | [
{
"version": "v1",
"created": "Fri, 27 Sep 2024 03:23:05 GMT"
},
{
"version": "v2",
"created": "Thu, 24 Oct 2024 06:24:21 GMT"
},
{
"version": "v3",
"created": "Sat, 07 Jun 2025 20:45:40 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Henriksson",
"Oskar",
""
],
[
"Ranestad",
"Kristian",
""
],
[
"Seccia",
"Lisa",
""
],
[
"Yu",
"Teresa",
""
]
] |
2409.18433 | Chenghao Deng | Mucong Ding, Chenghao Deng, Jocelyn Choo, Zichu Wu, Aakriti Agrawal, Avi Schwarzschild, Tianyi Zhou, Tom Goldstein, John Langford, Anima Anandkumar, Furong Huang | Easy2Hard-Bench: Standardized Difficulty Labels for Profiling LLM Performance and Generalization | NeurIPS 2024 Datasets and Benchmarks Track | null | null | null | cs.LG cs.AI cs.CL | http://creativecommons.org/licenses/by-sa/4.0/ | While generalization over tasks from easy to hard is crucial to profile language models (LLMs), the datasets with fine-grained difficulty annotations for each problem across a broad range of complexity are still blank. Aiming to address this limitation, we present Easy2Hard-Bench, a consistently formatted collection of 6 benchmark datasets spanning various domains, such as mathematics and programming problems, chess puzzles, and reasoning questions. Each problem within these datasets is annotated with numerical difficulty scores. To systematically estimate problem difficulties, we collect abundant performance data on attempts to each problem by humans in the real world or LLMs on the prominent leaderboard. Leveraging the rich performance data, we apply well-established difficulty ranking systems, such as Item Response Theory (IRT) and Glicko-2 models, to uniformly assign numerical difficulty scores to problems. Moreover, datasets in Easy2Hard-Bench distinguish themselves from previous collections by a higher proportion of challenging problems. Through extensive experiments with six state-of-the-art LLMs, we provide a comprehensive analysis of their performance and generalization capabilities across varying levels of difficulty, with the aim of inspiring future research in LLM generalization. The datasets are available at https://huggingface.co/datasets/furonghuang-lab/Easy2Hard-Bench. | [
{
"version": "v1",
"created": "Fri, 27 Sep 2024 03:49:56 GMT"
},
{
"version": "v2",
"created": "Mon, 09 Jun 2025 17:15:44 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Ding",
"Mucong",
""
],
[
"Deng",
"Chenghao",
""
],
[
"Choo",
"Jocelyn",
""
],
[
"Wu",
"Zichu",
""
],
[
"Agrawal",
"Aakriti",
""
],
[
"Schwarzschild",
"Avi",
""
],
[
"Zhou",
"Tianyi",
""
],
[
"Goldstein",
"Tom",
""
],
[
"Langford",
"John",
""
],
[
"Anandkumar",
"Anima",
""
],
[
"Huang",
"Furong",
""
]
] |
2409.19080 | Nathan Drouillard | Nathan G. Drouillard, TJ Hammond | Stimulated Raman Spectroscopy with Tunable Visible Broadband Probe Pulse Generated by Kerr-Instability Amplification | null | null | null | null | physics.optics physics.app-ph physics.chem-ph | http://creativecommons.org/licenses/by/4.0/ | Femtosecond, broadband stimulated Raman spectroscopy is a popular approach to measuring molecular dynamics with excellent signal-to-noise and spectral resolution. We present a new method for broadband stimulated Raman spectroscopy that employs Kerr-instability amplification to amplify the supercontinuum spectrum from sapphire and create a highly tunable Raman probe spectrum spanning from 530 to 1000 nm (-6000 to 2800 cm^(-1)). Our method, called Kerr-instability amplification for broadband stimulated Raman spectroscopy (KAB-SRS) provides an alternative to optical parametric amplifiers by producing a broader and more tunable spectrum at a fraction of the cost. We demonstrate the effectiveness of our method by measuring the stimulated Raman loss spectrum of 1-decanol. | [
{
"version": "v1",
"created": "Fri, 27 Sep 2024 18:27:22 GMT"
},
{
"version": "v2",
"created": "Thu, 06 Feb 2025 20:18:31 GMT"
},
{
"version": "v3",
"created": "Mon, 09 Jun 2025 13:06:01 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Drouillard",
"Nathan G.",
""
],
[
"Hammond",
"TJ",
""
]
] |
2409.19431 | Gholamali Aminian | Gholamali Aminian, Amir R. Asadi, Tian Li, Ahmad Beirami, Gesine Reinert, Samuel N. Cohen | Generalization and Robustness of the Tilted Empirical Risk | Accepted in ICML 2025 | null | null | null | stat.ML cs.IT cs.LG math.IT | http://creativecommons.org/licenses/by/4.0/ | The generalization error (risk) of a supervised statistical learning algorithm quantifies its prediction ability on previously unseen data. Inspired by exponential tilting, \citet{li2020tilted} proposed the {\it tilted empirical risk} (TER) as a non-linear risk metric for machine learning applications such as classification and regression problems. In this work, we examine the generalization error of the tilted empirical risk in the robustness regime under \textit{negative tilt}. Our first contribution is to provide uniform and information-theoretic bounds on the {\it tilted generalization error}, defined as the difference between the population risk and the tilted empirical risk, under negative tilt for unbounded loss function under bounded $(1+\epsilon)$-th moment of loss function for some $\epsilon\in(0,1]$ with a convergence rate of $O(n^{-\epsilon/(1+\epsilon)})$ where $n$ is the number of training samples, revealing a novel application for TER under no distribution shift. Secondly, we study the robustness of the tilted empirical risk with respect to noisy outliers at training time and provide theoretical guarantees under distribution shift for the tilted empirical risk. We empirically corroborate our findings in simple experimental setups where we evaluate our bounds to select the value of tilt in a data-driven manner. | [
{
"version": "v1",
"created": "Sat, 28 Sep 2024 18:31:51 GMT"
},
{
"version": "v2",
"created": "Thu, 17 Oct 2024 12:23:07 GMT"
},
{
"version": "v3",
"created": "Sat, 07 Jun 2025 17:10:54 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Aminian",
"Gholamali",
""
],
[
"Asadi",
"Amir R.",
""
],
[
"Li",
"Tian",
""
],
[
"Beirami",
"Ahmad",
""
],
[
"Reinert",
"Gesine",
""
],
[
"Cohen",
"Samuel N.",
""
]
] |
2410.00117 | Alan Papalia | Alan Papalia, Yulun Tian, David M. Rosen, Jonathan P. How, John J. Leonard | An Overview of the Burer-Monteiro Method for Certifiable Robot Perception | Accepted to 2024 Robotics: Science and Systems (RSS) Safe Autonomy Workshop | null | null | null | cs.RO cs.CV cs.LG | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | This paper presents an overview of the Burer-Monteiro method (BM), a technique that has been applied to solve robot perception problems to certifiable optimality in real-time. BM is often used to solve semidefinite programming relaxations, which can be used to perform global optimization for non-convex perception problems. Specifically, BM leverages the low-rank structure of typical semidefinite programs to dramatically reduce the computational cost of performing optimization. This paper discusses BM in certifiable perception, with three main objectives: (i) to consolidate information from the literature into a unified presentation, (ii) to elucidate the role of the linear independence constraint qualification (LICQ), a concept not yet well-covered in certifiable perception literature, and (iii) to share practical considerations that are discussed among practitioners but not thoroughly covered in the literature. Our general aim is to offer a practical primer for applying BM towards certifiable perception. | [
{
"version": "v1",
"created": "Mon, 30 Sep 2024 18:00:24 GMT"
},
{
"version": "v2",
"created": "Mon, 09 Jun 2025 13:32:24 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Papalia",
"Alan",
""
],
[
"Tian",
"Yulun",
""
],
[
"Rosen",
"David M.",
""
],
[
"How",
"Jonathan P.",
""
],
[
"Leonard",
"John J.",
""
]
] |
2410.00527 | Duc Dung Nguyen | The Hieu Pham, Phuong Thanh Tran Nguyen, Xuan Tho Nguyen, Tan Dat Nguyen, Duc Dung Nguyen | Wanna hear your voice? A sample is all we need! | work in progress | null | null | null | eess.AS | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Research on audio clue-based target speaker extraction (TSE) has focused on modeling mixtures and reference speech, achieving strong results in English due to abundant datasets. However, cross-lingual properties remain underexplored, as low-resource languages face challenges from limited annotated data and linguistic resources. To bridge this gap, we propose WHYV (Wanna Hear Your Voice), a cross-lingual TSE framework enabling zero-shot adaptation without fine-tuning. WHYV employs a frequency-modulated gating mechanism that dynamically adjusts the acoustic features of the target speaker, minimizing reliance on language-specific cues. Evaluations demonstrate state-of-the-art zero-shot performance: 13.8 dB (Libri2Mix mix-both), 18.1 dB (mix-clean), and 14.8 dB on Vietnamese data. | [
{
"version": "v1",
"created": "Tue, 01 Oct 2024 09:11:03 GMT"
},
{
"version": "v2",
"created": "Thu, 03 Oct 2024 03:52:35 GMT"
},
{
"version": "v3",
"created": "Mon, 10 Feb 2025 03:07:44 GMT"
},
{
"version": "v4",
"created": "Mon, 09 Jun 2025 08:06:13 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Pham",
"The Hieu",
""
],
[
"Nguyen",
"Phuong Thanh Tran",
""
],
[
"Nguyen",
"Xuan Tho",
""
],
[
"Nguyen",
"Tan Dat",
""
],
[
"Nguyen",
"Duc Dung",
""
]
] |
2410.00775 | Paloma Piot | Paloma Piot, Javier Parapar | Decoding Hate: Exploring Language Models' Reactions to Hate Speech | null | Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies | 10.18653/v1/2025.naacl-long.45 | null | cs.CL | http://creativecommons.org/licenses/by-nc-sa/4.0/ | Hate speech is a harmful form of online expression, often manifesting as derogatory posts. It is a significant risk in digital environments. With the rise of Large Language Models (LLMs), there is concern about their potential to replicate hate speech patterns, given their training on vast amounts of unmoderated internet data. Understanding how LLMs respond to hate speech is crucial for their responsible deployment. However, the behaviour of LLMs towards hate speech has been limited compared. This paper investigates the reactions of seven state-of-the-art LLMs (LLaMA 2, Vicuna, LLaMA 3, Mistral, GPT-3.5, GPT-4, and Gemini Pro) to hate speech. Through qualitative analysis, we aim to reveal the spectrum of responses these models produce, highlighting their capacity to handle hate speech inputs. We also discuss strategies to mitigate hate speech generation by LLMs, particularly through fine-tuning and guideline guardrailing. Finally, we explore the models' responses to hate speech framed in politically correct language. | [
{
"version": "v1",
"created": "Tue, 01 Oct 2024 15:16:20 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Piot",
"Paloma",
""
],
[
"Parapar",
"Javier",
""
]
] |
2410.00990 | Jian Yang | Jian Yang, Xukun Wang, Wentao Wang, Guoming Li, Qihang Fang, Ruihong Yuan, Tianyang Wang, Xiaomei Zhang, Yeying Jin, Zhaoxin Fan | Lipschitz-Driven Noise Robustness in VQ-AE for High-Frequency Texture Repair in ID-Specific Talking Heads | null | null | null | null | cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Audio-driven IDentity-specific Talking Head Generation (ID-specific THG) has shown increasing promise for applications in filmmaking and virtual reality. Existing approaches are generally constructed as end-to-end paradigms, and have achieved significant progress. However, they often struggle to capture high-frequency textures due to limited model capacity. To address these limitations, we adopt a simple yet efficient post-processing framework -- unlike previous studies that focus solely on end-to-end training -- guided by our theoretical insights. Specifically, leveraging the \textit{Lipschitz Continuity Theory} of neural networks, we prove a crucial noise tolerance property for the Vector Quantized AutoEncoder (VQ-AE), and establish the existence of a Noise Robustness Upper Bound (NRoUB). This insight reveals that we can efficiently obtain an identity-specific denoiser by training an identity-specific neural discrete representation, without requiring an extra network. Based on this theoretical foundation, we propose a plug-and-play Space-Optimized VQ-AE (SOVQAE) with enhanced NRoUB to achieve temporally-consistent denoising. For practical deployment, we further introduce a cascade pipeline combining a pretrained Wav2Lip model with SOVQAE to perform ID-specific THG. Our experiments demonstrate that this pipeline achieves \textit{state-of-the-art} performance in video quality and robustness for out-of-distribution lip synchronization, surpassing existing identity-specific THG methods. In addition, the pipeline requires only a couple of consumer GPU hours and runs in real time, which is both efficient and practical for industry applications. | [
{
"version": "v1",
"created": "Tue, 01 Oct 2024 18:32:02 GMT"
},
{
"version": "v2",
"created": "Fri, 21 Mar 2025 06:17:16 GMT"
},
{
"version": "v3",
"created": "Sun, 08 Jun 2025 14:25:14 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Yang",
"Jian",
""
],
[
"Wang",
"Xukun",
""
],
[
"Wang",
"Wentao",
""
],
[
"Li",
"Guoming",
""
],
[
"Fang",
"Qihang",
""
],
[
"Yuan",
"Ruihong",
""
],
[
"Wang",
"Tianyang",
""
],
[
"Zhang",
"Xiaomei",
""
],
[
"Jin",
"Yeying",
""
],
[
"Fan",
"Zhaoxin",
""
]
] |
2410.01223 | Chengpu Wang | Chengpu Wang | Statistical Taylor Expansion | 66 pages, 54 figures | null | null | null | stat.CO cs.LG | http://creativecommons.org/licenses/by/4.0/ | Statistical Taylor expansion replaces the input precise variables in a conventional Taylor expansion with random variables each with known distribution, to calculate the result mean and deviation. It is based on the uncorrelated uncertainty assumption: Each input variable is measured independently with fine enough statistical precision, so that their uncertainties are independent of each other. It requires each sampling count to be sufficiently large for accurate results. Statistical Taylor expansion reviews that the intermediate analytic expressions can no longer be regarded as independent of each other, and the result of analytic expression should be path independent. This conclusion differs fundamentally from the conventional common approach in applied mathematics to find the best execution path for a result. This paper also presents an implementation of statistical Taylor expansion called variance arithmetic, and the tests on variance arithmetic. | [
{
"version": "v1",
"created": "Wed, 02 Oct 2024 04:02:21 GMT"
},
{
"version": "v2",
"created": "Mon, 07 Oct 2024 23:49:02 GMT"
},
{
"version": "v3",
"created": "Tue, 15 Oct 2024 13:48:14 GMT"
},
{
"version": "v4",
"created": "Thu, 26 Dec 2024 17:27:23 GMT"
},
{
"version": "v5",
"created": "Thu, 15 May 2025 13:37:18 GMT"
},
{
"version": "v6",
"created": "Wed, 21 May 2025 20:08:24 GMT"
},
{
"version": "v7",
"created": "Mon, 02 Jun 2025 14:24:37 GMT"
},
{
"version": "v8",
"created": "Mon, 09 Jun 2025 12:52:06 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Wang",
"Chengpu",
""
]
] |
2410.01947 | Leonie Heinze | L. Heinze, T. Kotte, R. Rausch, A. Demuer, S. Luther, R. Feyerherm, E. L. Q. N. Ammerlaan, U. Zeitler, D. I. Gorbunov, M. Uhlarz, K. C. Rule, A. U. B. Wolter, H. K\"uhne, J. Wosnitza, C. Karrasch, and S. S\"ullow | Atacamite Cu$_2$Cl(OH)$_3$ in High Magnetic Fields: Quantum Criticality and Dimensional Reduction of a Sawtooth-Chain Compound | null | Phys. Rev. Lett. 134, 216701 (2025) | 10.1103/PhysRevLett.134.216701 | null | cond-mat.str-el | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We report an extensive high-field study of atacamite Cu$_2$Cl(OH)$_3$, a material realization of quantum sawtooth chains with weak interchain couplings, in continuous and pulsed magnetic fields up to 58 T. In particular, we have mapped the entropy landscape for fields as high as 35 T and have identified a field-induced quantum critical point at 21.9(1) T for $\mathbf{H} \parallel c$ axis. The quantum critical point separates field regions with and without magnetic order, evidenced by our thermodynamic study and $^1$H nuclear magnetic resonance spectroscopy, but lies far below full saturation of the magnetization. Corroborated by numerical results using density-matrix renormalization group (DMRG) calculations, we find this behavior associated with a dimensional reduction of the spin system: the sawtooth chain effectively decouples into an antiferromagnetic spin-$1/2$ chain (backbone of the sawtooth chain) in the presence of an exchange field produced by the remaining field-polarized spins. | [
{
"version": "v1",
"created": "Wed, 02 Oct 2024 18:48:33 GMT"
},
{
"version": "v2",
"created": "Wed, 23 Apr 2025 18:37:05 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Heinze",
"L.",
""
],
[
"Kotte",
"T.",
""
],
[
"Rausch",
"R.",
""
],
[
"Demuer",
"A.",
""
],
[
"Luther",
"S.",
""
],
[
"Feyerherm",
"R.",
""
],
[
"Ammerlaan",
"E. L. Q. N.",
""
],
[
"Zeitler",
"U.",
""
],
[
"Gorbunov",
"D. I.",
""
],
[
"Uhlarz",
"M.",
""
],
[
"Rule",
"K. C.",
""
],
[
"Wolter",
"A. U. B.",
""
],
[
"Kühne",
"H.",
""
],
[
"Wosnitza",
"J.",
""
],
[
"Karrasch",
"C.",
""
],
[
"Süllow",
"S.",
""
]
] |
2410.02367 | Jintao Zhang | Jintao Zhang, Jia Wei, Haofeng Huang, Pengle Zhang, Jun Zhu, Jianfei Chen | SageAttention: Accurate 8-Bit Attention for Plug-and-play Inference Acceleration | @inproceedings{zhang2025sageattention, title={SageAttention: Accurate 8-Bit Attention for Plug-and-play Inference Acceleration}, author={Zhang, Jintao and Wei, Jia and Zhang, Pengle and Zhu, Jun and Chen, Jianfei}, booktitle={International Conference on Learning Representations (ICLR)}, year={2025} } | The Thirteenth International Conference on Learning Representations (ICLR 2025) | null | null | cs.LG | http://creativecommons.org/licenses/by/4.0/ | The transformer architecture predominates across various models. As the heart of the transformer, attention has a computational complexity of $O(N^2)$, compared to $O(N)$ for linear transformations. When handling large sequence lengths, attention becomes the primary time-consuming component. Although quantization has proven to be an effective method for accelerating model inference, existing quantization methods primarily focus on optimizing the linear layer. In response, we first analyze the feasibility of quantization in attention detailedly. Following that, we propose SageAttention, a highly efficient and accurate quantization method for attention. The OPS (operations per second) of our approach outperforms FlashAttention2 and xformers by about 2.1 times and 2.7 times, respectively. SageAttention also achieves superior accuracy performance over FlashAttention3. Comprehensive experiments confirm that our approach incurs almost no end-to-end metrics loss across diverse models, including those for large language processing, image generation, and video generation. The codes are available at https://github.com/thu-ml/SageAttention. | [
{
"version": "v1",
"created": "Thu, 03 Oct 2024 10:25:23 GMT"
},
{
"version": "v2",
"created": "Thu, 14 Nov 2024 08:39:54 GMT"
},
{
"version": "v3",
"created": "Tue, 24 Dec 2024 02:29:17 GMT"
},
{
"version": "v4",
"created": "Wed, 25 Dec 2024 02:23:02 GMT"
},
{
"version": "v5",
"created": "Wed, 26 Feb 2025 14:28:20 GMT"
},
{
"version": "v6",
"created": "Wed, 30 Apr 2025 10:31:29 GMT"
},
{
"version": "v7",
"created": "Sat, 31 May 2025 16:19:09 GMT"
},
{
"version": "v8",
"created": "Sun, 08 Jun 2025 09:01:56 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Zhang",
"Jintao",
""
],
[
"Wei",
"Jia",
""
],
[
"Huang",
"Haofeng",
""
],
[
"Zhang",
"Pengle",
""
],
[
"Zhu",
"Jun",
""
],
[
"Chen",
"Jianfei",
""
]
] |
2410.02372 | Haowei Hua | Haowei Hua, Jingwen Yang, Wanyu Lin, Pan Zhou, Ming Yang, Yew-Soon Ong | Revisiting the Canonicalization for Fast and Accurate Crystal Tensor Property Prediction | null | null | null | null | cs.CE | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Predicting the tensor properties of crystalline materials is a fundamental task in materials science. Unlike single-value property prediction, which is inherently invariant, tensor property prediction requires maintaining O(3) group tensor equivariance. Such equivariance constraint often requires specialized architecture designs to achieve effective predictions, inevitably introducing tremendous computational costs. Canonicalization, a classical technique for geometry, has recently been explored for efficient learning with symmetry. In this work, we revisit the problem of crystal tensor property prediction through the lens of canonicalization. Specifically, we demonstrate how polar decomposition, a simple yet efficient algebraic method, can serve as a form of canonicalization and be leveraged to ensure equivariant tensor property prediction. Building upon this insight, we propose a general O(3)-equivariant framework for fast and accurate crystal tensor property prediction, referred to as GoeCTP. By utilizing canonicalization, GoeCTP achieves high efficiency without requiring the explicit incorporation of equivariance constraints into the network architecture. Experimental results indicate that GoeCTP achieves the best prediction accuracy and runs up to 13 times faster compared to existing state-of-the-art methods in benchmarking datasets, underscoring its effectiveness and efficiency. | [
{
"version": "v1",
"created": "Thu, 03 Oct 2024 10:48:03 GMT"
},
{
"version": "v2",
"created": "Fri, 04 Oct 2024 10:00:38 GMT"
},
{
"version": "v3",
"created": "Sun, 08 Jun 2025 03:02:49 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Hua",
"Haowei",
""
],
[
"Yang",
"Jingwen",
""
],
[
"Lin",
"Wanyu",
""
],
[
"Zhou",
"Pan",
""
],
[
"Yang",
"Ming",
""
],
[
"Ong",
"Yew-Soon",
""
]
] |
2410.02411 | Chen-Kai Qiao | Ping Su and Chen-Kai Qiao | Gravitational Lensing of Euler-Heisenberg Black Hole Surrounded by Perfect Fluid Dark Matter | 29 pages, 8 figures, 1 table, 2 appendices. V2: major revision, some unphysical parameter choices are removed. V3: a new appendix is added, some errors are corrected | null | null | null | gr-qc | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | In this work, we study the gravitational lensing of Euler-Heisenberg black hole surrounded by perfect fluid dark matter. This kind of black hole solution enables us to investigate the nontrivial interplay between the dark matter effects and nonlinear electrodynamics effects (or quantum electrodynamics effects) on charged black hole systems. The important observables in gravitational lensings are calculated and discussed in our work, including the gravitational deflection angle of light and time delay of light. Additionally, we also explore the massive orbit's bound orbits (and their precession angles) and black hole shadow radius for Euler-Heisenberg black hole in the presence of dark matter. The results indicate that the Euler-Heisenberg black hole with a larger perfect fluid dark matter parameter could greatly reduce the gravitational deflection angle of light, time delay of light, and precession angle of massive object's bound orbit, while the nonlinear electrodynamics effects do not have large influences on these observables.
Keywords: Euler Heisenberg Black Hole; Gravitational Lensing; Perfect Fluid Dark Matter; Nonlinear Electrodynamics | [
{
"version": "v1",
"created": "Thu, 03 Oct 2024 11:47:33 GMT"
},
{
"version": "v2",
"created": "Sun, 09 Feb 2025 19:01:24 GMT"
},
{
"version": "v3",
"created": "Sat, 07 Jun 2025 15:11:41 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Su",
"Ping",
""
],
[
"Qiao",
"Chen-Kai",
""
]
] |
2410.02676 | Pranjal Ralegankar | Pranjal Ralegankar, Enrico Garaldi, and Matteo Viel | Matter power spectrum induced by primordial magnetic fields: from the linear to the non-linear regime | 23 pages + appendices; matches version accepted in JCAP | null | null | null | astro-ph.CO | http://creativecommons.org/licenses/by/4.0/ | Linear theory predicts that primordial magnetic fields (PMFs) enhance the matter power spectrum on small scales. However, the linear approximation breaks down on sufficiently small scales where PMF-induced baryon perturbations back-react onto the magnetic fields. Previous studies assumed that the baryon power spectrum would be sharply suppressed in this non-linear regime, based on arguments related to the magnetic Jeans scale. For the first time, we perform dedicated magnetohydrodynamic (MHD) simulations to investigate the transition from the linear to the non-linear regime. Our simulations confirm the expected linear behavior on large scales. In the non-linear regime, however, we find that the dimensionless baryon power spectrum saturates to an $\mathcal{O}(1)$ value, which contrasts with previous analytical expectations. Additionally, our results show that several past studies overestimated the total matter power spectrum by orders of magnitude near the transition to non-linearity. Thus, the results presented in this work are useful to obtain more accurate constraints on PMFs from structure formation processes and/or different tracers of cosmic structures. | [
{
"version": "v1",
"created": "Thu, 03 Oct 2024 17:03:20 GMT"
},
{
"version": "v2",
"created": "Sun, 08 Jun 2025 11:54:10 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Ralegankar",
"Pranjal",
""
],
[
"Garaldi",
"Enrico",
""
],
[
"Viel",
"Matteo",
""
]
] |
2410.02734 | Zachary Porat | Zachary Porat | Computations directly on the cuspidal cohomology of congruence subgroups of $\mathrm{SL}(3, \mathbb{Z})$ | 11 pages, 2 tables, comments welcome; added additional data (v2) | null | null | null | math.NT | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | In their 1984 work, Ash, Grayson, and Green compute the action of Hecke operators on a certain subspace of the cohomology of low-level congruence subgroups of $\mathsf{SL}(3, \mathbb{Z})$. This subspace contains the cuspidal cohomology, which is of primary interest. We extend their work, introducing a method that allows for computing the action of Hecke operators directly on the cuspidal cohomology. Using this method, we obtain data for prime level less than 2400, finding six additional levels at which nonzero cuspidal classes appear and calculating local factors for five of these levels. | [
{
"version": "v1",
"created": "Thu, 03 Oct 2024 17:51:36 GMT"
},
{
"version": "v2",
"created": "Sat, 07 Jun 2025 17:36:00 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Porat",
"Zachary",
""
]
] |
2410.03483 | Zixi Chen | Zixi Chen, Qinghua Guan, Josie Hughes, Arianna Menciassi, Cesare Stefanini | A Versatile Neural Network Configuration Space Planning and Control Strategy for Modular Soft Robot Arms | 14 pages, 16 figures, 5 tables; accepted by IEEE T-Ro | null | null | null | cs.RO | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Modular soft robot arms (MSRAs) are composed of multiple modules connected in a sequence, and they can bend at different angles in various directions. This capability allows MSRAs to perform more intricate tasks than single-module robots. However, the modular structure also induces challenges in accurate planning and control. Nonlinearity and hysteresis complicate the physical model, while the modular structure and increased DOFs further lead to cumulative errors along the sequence. To address these challenges, we propose a versatile configuration space planning and control strategy for MSRAs, named S2C2A (State to Configuration to Action). Our approach formulates an optimization problem, S2C (State to Configuration planning), which integrates various loss functions and a forward model based on biLSTM to generate configuration trajectories based on target states. A configuration controller C2A (Configuration to Action control) based on biLSTM is implemented to follow the planned configuration trajectories, leveraging only inaccurate internal sensing feedback. We validate our strategy using a cable-driven MSRA, demonstrating its ability to perform diverse offline tasks such as position and orientation control and obstacle avoidance. Furthermore, our strategy endows MSRA with online interaction capability with targets and obstacles. Future work focuses on addressing MSRA challenges, such as more accurate physical models. | [
{
"version": "v1",
"created": "Fri, 04 Oct 2024 14:54:53 GMT"
},
{
"version": "v2",
"created": "Sun, 08 Jun 2025 19:06:45 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Chen",
"Zixi",
""
],
[
"Guan",
"Qinghua",
""
],
[
"Hughes",
"Josie",
""
],
[
"Menciassi",
"Arianna",
""
],
[
"Stefanini",
"Cesare",
""
]
] |
2410.03815 | Mohammad Mirtaba | Mohammad Mirtaba, Parham Oveissi, Juan Augusto Paredes Salaza, Ankit Goel | Single-Shot Learning of Multirotor Controller Gains: A Data-Driven Approach with Experimental Validation | null | null | null | null | eess.SY cs.SY | http://creativecommons.org/licenses/by-nc-sa/4.0/ | This paper demonstrates the single-shot learning capabilities of retrospective cost optimization based data-driven control applied to learning multirotor controller gains for trajectory tracking. In particular, the proposed control approach is first used within a simple multirotor simulation environment to learn appropriate multirotor controller gains to follow a trajectory. Then, the gains resulting from a single simulation run are used in a more complex multirotor simulation environment based on Simulink for performance verification. Finally, the resulting gains are implemented in a physical quadrotor and the results for waypoint and trajectory tracking are reported in this paper. The proposed control approach is the continuous-time version of the widely used discrete-time retrospective control adaptive control algorithm, which is simpler to implement within continuous-time simulation environments and whose performance does not depend on appropriate sampling time choice. | [
{
"version": "v1",
"created": "Fri, 04 Oct 2024 15:46:59 GMT"
},
{
"version": "v2",
"created": "Fri, 06 Jun 2025 19:31:22 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Mirtaba",
"Mohammad",
""
],
[
"Oveissi",
"Parham",
""
],
[
"Salaza",
"Juan Augusto Paredes",
""
],
[
"Goel",
"Ankit",
""
]
] |
2410.04196 | Tuan Truong | Tuan Truong, Quyen Tran, Quan Pham-Ngoc, Nhat Ho, Dinh Phung, Trung Le | Improving Generalization with Flat Hilbert Bayesian Inference | Accepted (ICML 2025) | null | null | null | cs.LG stat.ML | http://creativecommons.org/licenses/by/4.0/ | We introduce Flat Hilbert Bayesian Inference (FHBI), an algorithm designed to enhance generalization in Bayesian inference. Our approach involves an iterative two-step procedure with an adversarial functional perturbation step and a functional descent step within a reproducing kernel Hilbert space. This methodology is supported by a theoretical analysis that extends previous findings on generalization ability from finite-dimensional Euclidean spaces to infinite-dimensional functional spaces. To evaluate the effectiveness of FHBI, we conduct comprehensive comparisons against nine baseline methods on the \texttt{VTAB-1K} benchmark, which encompasses 19 diverse datasets across various domains with diverse semantics. Empirical results demonstrate that FHBI consistently outperforms the baselines by notable margins, highlighting its practical efficacy. | [
{
"version": "v1",
"created": "Sat, 05 Oct 2024 15:17:37 GMT"
},
{
"version": "v2",
"created": "Sun, 08 Jun 2025 16:20:56 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Truong",
"Tuan",
""
],
[
"Tran",
"Quyen",
""
],
[
"Pham-Ngoc",
"Quan",
""
],
[
"Ho",
"Nhat",
""
],
[
"Phung",
"Dinh",
""
],
[
"Le",
"Trung",
""
]
] |
2410.04571 | Aakriti Agrawal | Aakriti Agrawal, Mucong Ding, Zora Che, Chenghao Deng, Anirudh Satheesh, John Langford, Furong Huang | EnsemW2S: Can an Ensemble of LLMs be Leveraged to Obtain a Stronger LLM? | superalignment, weak-to-strong generalization on unseen OOD task; formerly appeared as arXiv:2505.21959v1 which was uploaded as a new submission in error | null | null | null | cs.LG | http://creativecommons.org/licenses/by/4.0/ | How can we harness the collective capabilities of multiple Large Language Models (LLMs) to create an even more powerful model? This question forms the foundation of our research, where we propose an innovative approach to weak-to-strong (w2s) generalization-a critical problem in AI alignment. Our work introduces an easy-to-hard (e2h) framework for studying the feasibility of w2s generalization, where weak models trained on simpler tasks collaboratively supervise stronger models on more complex tasks. This setup mirrors real-world challenges, where direct human supervision is limited. To achieve this, we develop a novel AdaBoost-inspired ensemble method, demonstrating that an ensemble of weak supervisors can enhance the performance of stronger LLMs across classification and generative tasks on difficult QA datasets. In several cases, our ensemble approach matches the performance of models trained on ground-truth data, establishing a new benchmark for w2s generalization. We observe an improvement of up to 14% over existing baselines and average improvements of 5% and 4% for binary classification and generative tasks, respectively. This research points to a promising direction for enhancing AI through collective supervision, especially in scenarios where labeled data is sparse or insufficient. | [
{
"version": "v1",
"created": "Sun, 06 Oct 2024 18:06:42 GMT"
},
{
"version": "v2",
"created": "Thu, 05 Jun 2025 22:40:11 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Agrawal",
"Aakriti",
""
],
[
"Ding",
"Mucong",
""
],
[
"Che",
"Zora",
""
],
[
"Deng",
"Chenghao",
""
],
[
"Satheesh",
"Anirudh",
""
],
[
"Langford",
"John",
""
],
[
"Huang",
"Furong",
""
]
] |
2410.04790 | Xinyu Wang | Xinyu Wang, Yanzheng Xiang, Lin Gui, Yulan He | PECAN: LLM-Guided Dynamic Progress Control with Attention-Guided Hierarchical Weighted Graph for Long-Document QA | null | null | null | null | cs.CL | http://creativecommons.org/licenses/by/4.0/ | Long-document QA presents challenges with large-scale text and long-distance dependencies. Recent advances in Large Language Models (LLMs) enable entire documents to be processed in a single pass. However, their computational cost is significantly high. Retrieval-Augmented Generation (RAG) methods split text into smaller chunks, but they often yield inferior results and may lose global context. Recent approaches that integrate LLMs into RAG via iterative summarization either underutilize LLM capabilities or still incur high computational costs. In this paper, we combine the high accuracy of LLMs with the efficiency of RAG and propose LLM-Guided Dynamic Progress Control with Attention-Based Hierarchical Weighted Graph (PECAN). Our method introduces two key improvements: (1) LLM-Guided Dynamic Progress Control: We leverage LLMs to dynamically control the retrieval process, adjusting the amount of retrieved information based on different queries to achieve a better balance of effectiveness and efficiency. (2) Attention-Guided Retrieval: We propose a novel retrieval method that constructs a hierarchical graph where edges are derived by LLM attention weights. Experimental results demonstrate that PECAN achieves LLM-level performance while maintaining computational complexity comparable to that of RAG methods on two single-document and two multi-document QA datasets. | [
{
"version": "v1",
"created": "Mon, 07 Oct 2024 07:02:09 GMT"
},
{
"version": "v2",
"created": "Sun, 08 Jun 2025 14:53:35 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Wang",
"Xinyu",
""
],
[
"Xiang",
"Yanzheng",
""
],
[
"Gui",
"Lin",
""
],
[
"He",
"Yulan",
""
]
] |
2410.04811 | Zhiyu Zhu | Zhiyu Zhu, Jinhui Hou, Hui Liu, Huanqiang Zeng, and Junhui Hou | Learning Efficient and Effective Trajectories for Differential Equation-based Image Restoration | null | null | null | null | cs.CV | http://creativecommons.org/licenses/by-nc-sa/4.0/ | The differential equation-based image restoration approach aims to establish learnable trajectories connecting high-quality images to a tractable distribution, e.g., low-quality images or a Gaussian distribution. In this paper, we reformulate the trajectory optimization of this kind of method, focusing on enhancing both reconstruction quality and efficiency. Initially, we navigate effective restoration paths through a reinforcement learning process, gradually steering potential trajectories toward the most precise options. Additionally, to mitigate the considerable computational burden associated with iterative sampling, we propose cost-aware trajectory distillation to streamline complex paths into several manageable steps with adaptable sizes. Moreover, we fine-tune a foundational diffusion model (FLUX) with 12B parameters by using our algorithms, producing a unified framework for handling 7 kinds of image restoration tasks. Extensive experiments showcase the $\textit{significant}$ superiority of the proposed method, achieving a maximum PSNR improvement of 2.1 dB over state-of-the-art methods, while also greatly enhancing visual perceptual quality. Project page: https://zhu-zhiyu.github.io/FLUX-IR/. | [
{
"version": "v1",
"created": "Mon, 07 Oct 2024 07:46:08 GMT"
},
{
"version": "v2",
"created": "Mon, 09 Jun 2025 12:14:25 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Zhu",
"Zhiyu",
""
],
[
"Hou",
"Jinhui",
""
],
[
"Liu",
"Hui",
""
],
[
"Zeng",
"Huanqiang",
""
],
[
"Hou",
"Junhui",
""
]
] |
2410.04959 | Emanuele Sansone | Emanuele Sansone, Tim Lebailly, Tinne Tuytelaars | Collapse-Proof Non-Contrastive Self-Supervised Learning | ICML 2025 | null | null | null | cs.LG stat.ML | http://creativecommons.org/licenses/by/4.0/ | We present a principled and simplified design of the projector and loss function for non-contrastive self-supervised learning based on hyperdimensional computing. We theoretically demonstrate that this design introduces an inductive bias that encourages representations to be simultaneously decorrelated and clustered, without explicitly enforcing these properties. This bias provably enhances generalization and suffices to avoid known training failure modes, such as representation, dimensional, cluster, and intracluster collapses. We validate our theoretical findings on image datasets, including SVHN, CIFAR-10, CIFAR-100, and ImageNet-100. Our approach effectively combines the strengths of feature decorrelation and cluster-based self-supervised learning methods, overcoming training failure modes while achieving strong generalization in clustering and linear classification tasks. | [
{
"version": "v1",
"created": "Mon, 07 Oct 2024 11:58:56 GMT"
},
{
"version": "v2",
"created": "Tue, 08 Oct 2024 07:00:01 GMT"
},
{
"version": "v3",
"created": "Sat, 07 Jun 2025 21:37:07 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Sansone",
"Emanuele",
""
],
[
"Lebailly",
"Tim",
""
],
[
"Tuytelaars",
"Tinne",
""
]
] |
2410.05132 | Ian Fleschler | Ian Fleschler | On the uniqueness of tangent cones to area minimizing currents at boundaries with arbitrary multiplicity | Corrected typos and improved exposition | null | null | null | math.AP math.DG | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We consider an area minimizing current $T$ in a $C^2$ submanifold $\Sigma$ of $\mathbb{R}^{m+n}$, with arbitrary integer boundary multiplicity $\partial T = Q [\![ \Gamma ]\!]$ where $\Gamma$ is a $C^2$ submanifold of $\Sigma$. We show that at every density $Q/2$ boundary point the tangent cone to $T$ is unique and there is a power rate of convergence to the unique tangent cone.
In particular, if $\Gamma$ is a closed manifold which lies at the boundary of a uniformly convex set $\Omega$ and $\Sigma=\mathbb{R}^{m+n}$ then $T$ has a unique tangent cone at every boundary point.
As a structural consequence of the uniqueness of the tangent cone, we obtain a decomposition theorem which is the starting point of the boundary regularity theory we develop in another paper in collaboration with Reinaldo Resende. The regularity theory we obtain generalizes Allard's boundary regularity theorem to a higher multiplicity setting. | [
{
"version": "v1",
"created": "Mon, 07 Oct 2024 15:34:56 GMT"
},
{
"version": "v2",
"created": "Mon, 09 Jun 2025 12:01:07 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Fleschler",
"Ian",
""
]
] |
2410.05548 | Manan Saxena | Manan Saxena, Tinghua Chen, Justin D. Silverman | Scalable Inference for Bayesian Multinomial Logistic-Normal Dynamic Linear Models | null | null | null | null | stat.AP stat.ME stat.ML | http://creativecommons.org/licenses/by/4.0/ | Many scientific fields collect longitudinal count compositional data. Each observation is a multivariate count vector, where the total counts are arbitrary, and the information lies in the relative frequency of the counts. Multiple authors have proposed Bayesian Multinomial Logistic-Normal Dynamic Linear Models (MLN-DLMs) as a flexible approach to modeling these data. However, adoption of these methods has been limited by computational challenges. This article develops an efficient and accurate approach to posterior state estimation, called $\textit{Fenrir}$. Our approach relies on a novel algorithm for MAP estimation and an accurate approximation to a key posterior marginal of the model. As there are no equivalent methods against which we can compare, we also develop an optimized Stan implementation of MLN-DLMs. Our experiments suggest that Fenrir can be three orders of magnitude more efficient than Stan and can even be incorporated into larger sampling schemes for joint inference of model hyperparameters. Our methods are made available to the community as a user-friendly software library written in C++ with an R interface. | [
{
"version": "v1",
"created": "Mon, 07 Oct 2024 23:20:14 GMT"
},
{
"version": "v2",
"created": "Sun, 08 Jun 2025 13:52:22 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Saxena",
"Manan",
""
],
[
"Chen",
"Tinghua",
""
],
[
"Silverman",
"Justin D.",
""
]
] |
2410.05823 | Claudio Bonati | Claudio Bonati and Andrea Pelissetto and Ettore Vicari | Three-dimensional Abelian and non-Abelian gauge Higgs theories | 117 pages, 50 pdf figures. References and comments added. Accepted for publication by Physics Reports | Physics Reports 1133, 1-92 (2025) | 10.1016/j.physrep.2025.05.004 | null | cond-mat.stat-mech hep-lat hep-th | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Gauge symmetries and Higgs mechanisms are key features of theories describing high-energy particle physics and collective phenomena in statistical and condensed-matter physics. In this review we address the collective behavior of systems of multicomponent scalar fields interacting with gauge fields, which can be already present in the underlying microscopic system or emerge only at criticality. The interplay between local gauge and global symmetries determines the phase diagram, the nature of the Higgs phases, and the nature of phase transitions between the high-temperature disordered and the low-temperature Higgs phases. However, additional crucial features determine the universal properties of the critical behavior at continuous transitions. Specifically, their nature also depends on the role played by the gauge modes at criticality. Effective (Abelian or non-Abelian) gauge Higgs field theories emerge when gauge modes develop critical correlations. On the other hand, a more standard critical behavior, which admits an effective description in terms of Landau-Ginzburg-Wilson $\Phi^4$ theories, occurs when gauge-field modes are short ranged at the transition. In the latter case, gauge fields only prevent non-gauge invariant correlation functions from becoming critical. This review covers the recent progress made in the study of Higgs systems with Abelian and non-Abelian gauge fields. We discuss the equilibrium thermodynamic properties of systems with a classical partition function, focusing mainly on three-dimensional systems, and only briefly discussing two-dimensional models. However, by using the quantum-to-classical mapping, the results on the critical behavior for classical systems in $D=d+1$ dimensions can be extended to quantum transitions in $d$ dimensions. | [
{
"version": "v1",
"created": "Tue, 08 Oct 2024 08:52:10 GMT"
},
{
"version": "v2",
"created": "Thu, 07 Nov 2024 13:04:59 GMT"
},
{
"version": "v3",
"created": "Tue, 27 May 2025 09:12:32 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Bonati",
"Claudio",
""
],
[
"Pelissetto",
"Andrea",
""
],
[
"Vicari",
"Ettore",
""
]
] |
2410.05880 | Guy Kornowski | Guy Kornowski, Daogao Liu, Kunal Talwar | Improved Sample Complexity for Private Nonsmooth Nonconvex Optimization | Accepted to ICML 2025; some fixes following reviews | null | null | null | cs.LG cs.CR math.OC stat.ML | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We study differentially private (DP) optimization algorithms for stochastic and empirical objectives which are neither smooth nor convex, and propose methods that return a Goldstein-stationary point with sample complexity bounds that improve on existing works. We start by providing a single-pass $(\epsilon,\delta)$-DP algorithm that returns an $(\alpha,\beta)$-stationary point as long as the dataset is of size $\widetilde{\Omega}(\sqrt{d}/\alpha\beta^{3}+d/\epsilon\alpha\beta^{2})$, which is $\Omega(\sqrt{d})$ times smaller than the algorithm of Zhang et al. [2024] for this task, where $d$ is the dimension. We then provide a multi-pass polynomial time algorithm which further improves the sample complexity to $\widetilde{\Omega}\left(d/\beta^2+d^{3/4}/\epsilon\alpha^{1/2}\beta^{3/2}\right)$, by designing a sample efficient ERM algorithm, and proving that Goldstein-stationary points generalize from the empirical loss to the population loss. | [
{
"version": "v1",
"created": "Tue, 08 Oct 2024 10:15:49 GMT"
},
{
"version": "v2",
"created": "Sat, 07 Jun 2025 18:12:38 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Kornowski",
"Guy",
""
],
[
"Liu",
"Daogao",
""
],
[
"Talwar",
"Kunal",
""
]
] |
2410.06511 | Wanchao Liang | Wanchao Liang, Tianyu Liu, Less Wright, Will Constable, Andrew Gu, Chien-Chin Huang, Iris Zhang, Wei Feng, Howard Huang, Junjie Wang, Sanket Purandare, Gokul Nadathur and Stratos Idreos | TorchTitan: One-stop PyTorch native solution for production ready LLM pre-training | null | null | null | null | cs.CL cs.AI cs.DC cs.LG | http://creativecommons.org/licenses/by/4.0/ | The development of large language models (LLMs) has been instrumental in advancing state-of-the-art natural language processing applications. Training LLMs with billions of parameters and trillions of tokens require sophisticated distributed systems that enable composing and comparing several state-of-the-art techniques in order to efficiently scale across thousands of accelerators. However, existing solutions are complex, scattered across multiple libraries/repositories, lack interoperability, and are cumbersome to maintain. Thus, curating and empirically comparing training recipes require non-trivial engineering effort.
This paper introduces TorchTitan, an open-source, PyTorch-native distributed training system that unifies state-of-the-art techniques, streamlining integration and reducing overhead. TorchTitan enables 3D parallelism in a modular manner with elastic scaling, providing comprehensive logging, checkpointing, and debugging tools for production-ready training. It also incorporates hardware-software co-designed solutions, leveraging features like Float8 training and SymmetricMemory. As a flexible test bed, TorchTitan facilitates custom recipe curation and comparison, allowing us to develop optimized training recipes for Llama 3.1 and provide guidance on selecting techniques for maximum efficiency based on our experiences.
We thoroughly assess TorchTitan on the Llama 3.1 family of LLMs, spanning 8 billion to 405 billion parameters, and showcase its exceptional performance, modular composability, and elastic scalability. By stacking training optimizations, we demonstrate accelerations of 65.08% with 1D parallelism at the 128-GPU scale (Llama 3.1 8B), an additional 12.59% with 2D parallelism at the 256-GPU scale (Llama 3.1 70B), and an additional 30% with 3D parallelism at the 512-GPU scale (Llama 3.1 405B) on NVIDIA H100 GPUs over optimized baselines. | [
{
"version": "v1",
"created": "Wed, 09 Oct 2024 03:26:11 GMT"
},
{
"version": "v2",
"created": "Mon, 04 Nov 2024 13:52:23 GMT"
},
{
"version": "v3",
"created": "Sat, 07 Jun 2025 19:16:22 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Liang",
"Wanchao",
""
],
[
"Liu",
"Tianyu",
""
],
[
"Wright",
"Less",
""
],
[
"Constable",
"Will",
""
],
[
"Gu",
"Andrew",
""
],
[
"Huang",
"Chien-Chin",
""
],
[
"Zhang",
"Iris",
""
],
[
"Feng",
"Wei",
""
],
[
"Huang",
"Howard",
""
],
[
"Wang",
"Junjie",
""
],
[
"Purandare",
"Sanket",
""
],
[
"Nadathur",
"Gokul",
""
],
[
"Idreos",
"Stratos",
""
]
] |
2410.06648 | Xing Lei | Xing Lei, Xuetao Zhang, Zifeng Zhuang, Donglin Wang | Q-WSL: Optimizing Goal-Conditioned RL with Weighted Supervised Learning via Dynamic Programming | null | null | null | null | cs.LG | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | A novel class of advanced algorithms, termed Goal-Conditioned Weighted Supervised Learning (GCWSL), has recently emerged to tackle the challenges posed by sparse rewards in goal-conditioned reinforcement learning (RL). GCWSL consistently delivers strong performance across a diverse set of goal-reaching tasks due to its simplicity, effectiveness, and stability. However, GCWSL methods lack a crucial capability known as trajectory stitching, which is essential for learning optimal policies when faced with unseen skills during testing. This limitation becomes particularly pronounced when the replay buffer is predominantly filled with sub-optimal trajectories. In contrast, traditional TD-based RL methods, such as Q-learning, which utilize Dynamic Programming, do not face this issue but often experience instability due to the inherent difficulties in value function approximation. In this paper, we propose Q-learning Weighted Supervised Learning (Q-WSL), a novel framework designed to overcome the limitations of GCWSL by incorporating the strengths of Dynamic Programming found in Q-learning. Q-WSL leverages Dynamic Programming results to output the optimal action of (state, goal) pairs across different trajectories within the replay buffer. This approach synergizes the strengths of both Q-learning and GCWSL, effectively mitigating their respective weaknesses and enhancing overall performance. Empirical evaluations on challenging goal-reaching tasks demonstrate that Q-WSL surpasses other goal-conditioned approaches in terms of both performance and sample efficiency. Additionally, Q-WSL exhibits notable robustness in environments characterized by binary reward structures and environmental stochasticity. | [
{
"version": "v1",
"created": "Wed, 09 Oct 2024 08:00:12 GMT"
},
{
"version": "v2",
"created": "Thu, 10 Oct 2024 09:27:29 GMT"
},
{
"version": "v3",
"created": "Fri, 11 Oct 2024 07:35:43 GMT"
},
{
"version": "v4",
"created": "Tue, 22 Oct 2024 03:35:54 GMT"
},
{
"version": "v5",
"created": "Sat, 07 Jun 2025 02:05:57 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Lei",
"Xing",
""
],
[
"Zhang",
"Xuetao",
""
],
[
"Zhuang",
"Zifeng",
""
],
[
"Wang",
"Donglin",
""
]
] |
2410.06864 | No First Name Rakesh | Lauri Oksanen and Rakesh and Mikko Salo | Rigidity in fixed angle inverse scattering for Riemannian metrics | Major changes are the following: More details in the proofs of Propositions 2.2, 2.3, 3.1; a correction in the proof of Proposition 2.3; part of the proof of Proposition 3.1 moved to the appendix | null | null | null | math.AP | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | The fixed angle inverse scattering problem for a velocity consists in determining a sound speed, or a Riemannian metric up to diffeomorphism, from measurements obtained by probing the medium with a single plane wave. This is a formally determined inverse problem that is open in general. In this article we consider the rigidity question of distinguishing a sound speed or a Riemannian metric from the Euclidean metric. We prove that a general smooth metric that is Euclidean outside a ball can be distinguished from the Euclidean metric. The methods involve distorted plane waves and a combination of geometric, topological and unique continuation arguments. | [
{
"version": "v1",
"created": "Wed, 09 Oct 2024 13:23:20 GMT"
},
{
"version": "v2",
"created": "Sun, 08 Jun 2025 19:37:54 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Oksanen",
"Lauri",
""
],
[
"Rakesh",
"",
""
],
[
"Salo",
"Mikko",
""
]
] |
2410.07029 | Paul Manuel Schindler | Paul M. Schindler and Marin Bukov | Geometric Floquet theory | 9 + 10 pages, 5 + 5 figures | null | null | null | quant-ph cond-mat.quant-gas | http://creativecommons.org/licenses/by/4.0/ | We derive Floquet theory from quantum geometry. We identify quasienergy folding as a consequence of a broken gauge group of the adiabatic gauge potential $U(1){\mapsto}\mathbb{Z}$. Fixing instead the gauge freedom using the parallel-transport gauge uniquely decomposes Floquet dynamics into a purely geometric and a purely dynamical evolution. The dynamical average-energy operator provides an unambiguous sorting of the quasienergy spectrum, identifying a Floquet ground state and suggesting a way to define the filling of Floquet-Bloch bands. We exemplify the features of geometric Floquet theory using an exactly solvable XY model and a non-integrable kicked Ising chain. We elucidate the geometric origin of inherently nonequilibrium effects, like the $\pi$-quasienergy splitting in discrete time crystals or $\pi$-edge modes in anomalous Floquet topological insulators. The spectrum of the average-energy operator is a susceptible indicator for both heating and spatiotemporal symmetry-breaking transitions. Last, we demonstrate that the periodic lab frame Hamiltonian generates transitionless counterdiabatic driving for Floquet eigenstates. This work directly bridges seemingly unrelated areas of nonequilibrium physics. | [
{
"version": "v1",
"created": "Wed, 09 Oct 2024 16:12:15 GMT"
},
{
"version": "v2",
"created": "Thu, 16 Jan 2025 11:04:36 GMT"
},
{
"version": "v3",
"created": "Mon, 09 Jun 2025 08:46:54 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Schindler",
"Paul M.",
""
],
[
"Bukov",
"Marin",
""
]
] |
2410.07694 | Abhishek Naskar | Anish Ghoshal, Abhishek Naskar and Nobuchika Okada | Unified Origin of Curvature Perturbation and Baryon Asymmetry of the Universe | 18 pages, 5 figures and 1 table, Appendix added | null | null | null | astro-ph.CO hep-ph | http://creativecommons.org/licenses/by/4.0/ | We propose a unified framework that describes both the curvaton mechanism for generating primordial density fluctuations and the Affleck-Dine (AD) mechanism for baryogenesis. By introducing a complex scalar field (AD field) carrying a baryon/lepton number and its potential consisting of quadratic and quartic terms with a small baryon/lepton-number-violating mass term, we investigate the evolution of the scalar field during the radiation-dominated era following inflation. We set the initial conditions such that the quartic term dominates the scalar potential, and the angular component of the AD field is non-zero. We focus on a scenario where the AD field sufficiently dominates the energy density of the universe before its decay. We show that the radial component of the AD field can be identified with the curvaton to solely produce the Planck normalized scalar power spectrum while the evolution of the angular component is crucial for generating the observed baryon asymmetry of the universe. Additionally, we find that the amplitude of scalar bispectrum $f_{NL}$ is negative, which is consistent with the current Planck data and testable in future observations such as CMB-S4, LiteBIRD, LSS, and 21-cm experiments. In our estimation of the scalar power spectrum and bispectrum, we develop a novel analytical scheme for computing scalar fluctuations based on the $\delta N$ formalism, which allows us to deal with the evolution of curvaton with polynomial potential more accurately in comparison to the existing analytical methods. | [
{
"version": "v1",
"created": "Thu, 10 Oct 2024 08:08:17 GMT"
},
{
"version": "v2",
"created": "Mon, 09 Jun 2025 05:56:06 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Ghoshal",
"Anish",
""
],
[
"Naskar",
"Abhishek",
""
],
[
"Okada",
"Nobuchika",
""
]
] |
2410.08102 | Tianyi Bai | Tianyi Bai, Ling Yang, Zhen Hao Wong, Fupeng Sun, Jiahui Peng, Xinlin Zhuang, Chi Zhang, Lijun Wu, Jiantao Qiu, Wentao Zhang, Binhang Yuan, Conghui He | Efficient Pretraining Data Selection for Language Models via Multi-Actor Collaboration | null | null | null | null | cs.CL | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Efficient data selection is crucial to accelerate the pretraining of language model (LMs). While various methods have been proposed to enhance data efficiency, limited research has addressed the inherent conflicts between these approaches to achieve optimal data selection for LM pretraining. To tackle this problem, we propose a multi-actor collaborative data selection mechanism: each data selection method independently prioritizes data based on its criterion and updates its prioritization rules using the current state of the model, functioning as an independent actor for data selection; and a console is designed to adjust the impacts of different actors at various stages and dynamically integrate information from all actors throughout the LM pretraining process. We conduct extensive empirical studies to evaluate our multi-actor framework. The experimental results demonstrate that our approach significantly improves data efficiency, accelerates convergence in LM pretraining, and achieves an average relative performance gain up to $10.5\%$ across multiple language model benchmarks compared to the state-of-the-art methods. | [
{
"version": "v1",
"created": "Thu, 10 Oct 2024 16:45:28 GMT"
},
{
"version": "v2",
"created": "Mon, 14 Oct 2024 14:22:30 GMT"
},
{
"version": "v3",
"created": "Sun, 08 Jun 2025 17:32:07 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Bai",
"Tianyi",
""
],
[
"Yang",
"Ling",
""
],
[
"Wong",
"Zhen Hao",
""
],
[
"Sun",
"Fupeng",
""
],
[
"Peng",
"Jiahui",
""
],
[
"Zhuang",
"Xinlin",
""
],
[
"Zhang",
"Chi",
""
],
[
"Wu",
"Lijun",
""
],
[
"Qiu",
"Jiantao",
""
],
[
"Zhang",
"Wentao",
""
],
[
"Yuan",
"Binhang",
""
],
[
"He",
"Conghui",
""
]
] |
2410.08272 | Nudzeim Selimovic | Barbara Anna Erdelyi, Ramona Gr\"ober, Nudzeim Selimovic | How large can the Light Quark Yukawa couplings be? | 38 pages + appendices, v2: matches published version | null | 10.1007/JHEP05(2025)189 | null | hep-ph hep-ex | http://creativecommons.org/licenses/by/4.0/ | We investigate models that can induce significant modifications to the couplings of first- and second-generation quarks with Higgs bosons. Specifically, we identify all simplified models featuring two vector-like quark states which can lead to substantial enhancements in these couplings. In addition, these models generate operators in Standard Model Effective Field Theory, both at tree-level and one-loop, that are constrained by electroweak precision and Higgs data. We show how to evade constraints from flavour physics and consider direct searches for vector-like quarks. Ultimately, we demonstrate that viable ultraviolet models can be found with first-generation quark Yukawa couplings enhanced by several hundred times their Standard Model value, while the Higgs couplings to charm (strange) quarks can be increased by factors of a few (few tens). Given the importance of electroweak precision data in constraining these models, we also discuss projections for future measurements at the Tera-$Z$ FCC-ee machine. | [
{
"version": "v1",
"created": "Thu, 10 Oct 2024 18:00:01 GMT"
},
{
"version": "v2",
"created": "Mon, 09 Jun 2025 11:06:37 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Erdelyi",
"Barbara Anna",
""
],
[
"Gröber",
"Ramona",
""
],
[
"Selimovic",
"Nudzeim",
""
]
] |
2410.08476 | Fan Yang | Fan Yang, Zhan Wang, Ning Kang, Zhenlong Ma, Jianxiong Li, Guojun Yuan and Guangming Tan (Institute of Computing Technology, Chinese Academy of Sciences) | JingZhao: A Framework for Rapid NIC Prototyping in the Domain-Specific-Network Era | 20 pages. 14 figures | null | null | null | cs.NI | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | The network is becoming domain-specific, which requires on-demand design of the network protocols, as well as the microarchitecture of the NIC. However, to develop such a NIC is not that easy. Since the scissor gap between network speed and the growth of CPU frequency is expanding, most of the protocols need to be offloaded to hardware. The process of designing, verifying and optimizing a domain-specific NIC usually takes great effort, which hinders the rapid iteration of new protocols and algorithms. In this paper, we propose JingZhao, an open-sourced framework for NIC prototyping, which could be leveraged to rapidly implement and verify a domain-specific NIC. Using this framework, we implement a fully-functional RDMA NIC (RNIC). To the best of our knowledge, this represents the first open-source RDMA NIC solution with complete compatibility to the standard OFED communication library. The RNIC was also taped out using TSMC's 28nm process to validate our design. Our evaluation results show that new network functions can be easily integrated into the framework, and achieve nearly line-rate packet processing. | [
{
"version": "v1",
"created": "Fri, 11 Oct 2024 03:09:08 GMT"
},
{
"version": "v2",
"created": "Tue, 15 Oct 2024 02:08:21 GMT"
},
{
"version": "v3",
"created": "Sun, 08 Jun 2025 03:03:49 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Yang",
"Fan",
"",
"Institute of Computing Technology, Chinese Academy of Sciences"
],
[
"Wang",
"Zhan",
"",
"Institute of Computing Technology, Chinese Academy of Sciences"
],
[
"Kang",
"Ning",
"",
"Institute of Computing Technology, Chinese Academy of Sciences"
],
[
"Ma",
"Zhenlong",
"",
"Institute of Computing Technology, Chinese Academy of Sciences"
],
[
"Li",
"Jianxiong",
"",
"Institute of Computing Technology, Chinese Academy of Sciences"
],
[
"Yuan",
"Guojun",
"",
"Institute of Computing Technology, Chinese Academy of Sciences"
],
[
"Tan",
"Guangming",
"",
"Institute of Computing Technology, Chinese Academy of Sciences"
]
] |
2410.08525 | Tomohiko Sano | Takara Abe, Isamu Hashiguchi, Yukitake Nakahara, Shunsuke Kobayashi, Ryuichi Tarumi, Hidetoshi Takahashi, Genya Ishigami and Tomohiko G. Sano | Snap and Jump: How Elastic Shells Pop Out | Main text: 18 pages, 4 figures. Supplementary Information: 14 pages, 3 figures | null | 10.1002/adrr.202500041 | null | cond-mat.soft cond-mat.mtrl-sci cond-mat.stat-mech | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Grip, walk, crawl, and jump. Soft robots are integrated functional structures composed of compliant mechanisms, whose activity spans various industrial applications such as surgery, healthcare, surveillance, and even planetary exploration. One of their promising mobility mechanism is snap-buckling; the instability mode of flexible structures passing from one equilibrium state to another can instantaneously generate large power for its motion. Predicting their performance with even simple geometry requires disentangling material, geometric nonlinearity, and contact, thereby still being a challenging problem to date. Here, we study the jumping dynamics of hemispherical elastic shells driven by snap-buckling, as a model system of soft jumping mechanisms, combining experiments, simulations, and analytical theory. We find that the contact transition dynamics trigger the jumping phenomenon upon snap-buckling by constructing the analytical predictions with shell elasticity in excellent agreement with both experiments and simulations. Despite the simple geometry of the shell, its dynamical performance primarily relies on a complex interplay between elasticity, geometry, and contact friction. By elucidating the dynamics of the building blocks of soft robots that undergo large deformations, we can build their predictive experimental and numerical framework. Our research paves the way for designing soft robots suitable for the required loading conditions or structural requirements without empirical methods. | [
{
"version": "v1",
"created": "Fri, 11 Oct 2024 04:55:55 GMT"
},
{
"version": "v2",
"created": "Tue, 10 Dec 2024 05:31:54 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Abe",
"Takara",
""
],
[
"Hashiguchi",
"Isamu",
""
],
[
"Nakahara",
"Yukitake",
""
],
[
"Kobayashi",
"Shunsuke",
""
],
[
"Tarumi",
"Ryuichi",
""
],
[
"Takahashi",
"Hidetoshi",
""
],
[
"Ishigami",
"Genya",
""
],
[
"Sano",
"Tomohiko G.",
""
]
] |
2410.08646 | Andrew Wang | Andrew Wang and Mike Davies | Fully Unsupervised Dynamic MRI Reconstruction via Diffeo-Temporal Equivariance | Conference paper at ISBI 2025 | null | 10.1109/ISBI60581.2025.10980940 | null | eess.IV cs.CV | http://creativecommons.org/licenses/by-nc-sa/4.0/ | Reconstructing dynamic MRI image sequences from undersampled accelerated measurements is crucial for faster and higher spatiotemporal resolution real-time imaging of cardiac motion, free breathing motion and many other applications. Classical paradigms, such as gated cine MRI, assume periodicity, disallowing imaging of true motion. Supervised deep learning methods are fundamentally flawed as, in dynamic imaging, ground truth fully-sampled videos are impossible to truly obtain. We propose an unsupervised framework to learn to reconstruct dynamic MRI sequences from undersampled measurements alone by leveraging natural geometric spatiotemporal equivariances of MRI. Dynamic Diffeomorphic Equivariant Imaging (DDEI) significantly outperforms state-of-the-art unsupervised methods such as SSDU on highly accelerated dynamic cardiac imaging. Our method is agnostic to the underlying neural network architecture and can be used to adapt the latest models and post-processing approaches. Our code and video demos are at https://github.com/Andrewwango/ddei. | [
{
"version": "v1",
"created": "Fri, 11 Oct 2024 09:16:30 GMT"
},
{
"version": "v2",
"created": "Thu, 13 Feb 2025 12:10:11 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Wang",
"Andrew",
""
],
[
"Davies",
"Mike",
""
]
] |
2410.08850 | Yuchen Zhu | Lorenzo Magnino, Yuchen Zhu, Mathieu Lauri\`ere | Learning to Stop: Deep Learning for Mean Field Optimal Stopping | Accepted to ICML 2025 | null | null | null | math.OC cs.LG | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Optimal stopping is a fundamental problem in optimization with applications in risk management, finance, robotics, and machine learning. We extend the standard framework to a multi-agent setting, named multi-agent optimal stopping (MAOS), where agents cooperate to make optimal stopping decisions in a finite-space, discrete-time environment. Since solving MAOS becomes computationally prohibitive as the number of agents is very large, we study the mean-field optimal stopping (MFOS) problem, obtained as the number of agents tends to infinity. We establish that MFOS provides a good approximation to MAOS and prove a dynamic programming principle (DPP) based on mean-field control theory. We then propose two deep learning approaches: one that learns optimal stopping decisions by simulating full trajectories and another that leverages the DPP to compute the value function and to learn the optimal stopping rule using backward induction. Both methods train neural networks to approximate optimal stopping policies. We demonstrate the effectiveness and the scalability of our work through numerical experiments on 6 different problems in spatial dimension up to 300. To the best of our knowledge, this is the first work to formalize and computationally solve MFOS in discrete time and finite space, opening new directions for scalable MAOS methods. | [
{
"version": "v1",
"created": "Fri, 11 Oct 2024 14:27:17 GMT"
},
{
"version": "v2",
"created": "Mon, 09 Jun 2025 16:11:54 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Magnino",
"Lorenzo",
""
],
[
"Zhu",
"Yuchen",
""
],
[
"Laurière",
"Mathieu",
""
]
] |
2410.08898 | Yang Chen | Yang Chen, Long Yang, Yitao Liang, Zhouchen Lin | Low-Dimension-to-High-Dimension Generalization And Its Implications for Length Generalization | null | null | null | null | cs.LG | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Low-Dimension-to-High-Dimension (LDHD) generalization is a special case of Out-of-Distribution (OOD) generalization, where the training data are restricted to a low-dimensional subspace of the high-dimensional testing space. Assuming that each instance is generated from a latent variable and the dimension of the latent variable reflects the problem scale, the inherent scaling challenge in length generalization can be captured by the LDHD generalization in the latent space. We theoretically demonstrate that LDHD generalization is generally unattainable without exploiting prior knowledge to provide appropriate inductive bias. Specifically, we explore LDHD generalization in Boolean functions. We verify that different architectures trained with (S)GD converge to \emph{min-degree interpolators w.r.t. different independent sets}. LDHD generalization is achievable if and only if the target function coincides with this inductive bias. Applying the insights from LDHD generalization to length generalization, we explain the effectiveness of CoT as changing the structure latent space to enable better LDHD generalization. We also propose a principle for position embedding design to handle both the inherent LDHD generalization and the nuisances such as the data format. Following the principle, we propose a novel position embedding called RPE-Square that remedies the RPE for dealing with the data format nuisance. | [
{
"version": "v1",
"created": "Fri, 11 Oct 2024 15:18:43 GMT"
},
{
"version": "v2",
"created": "Mon, 09 Jun 2025 07:19:42 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Chen",
"Yang",
""
],
[
"Yang",
"Long",
""
],
[
"Liang",
"Yitao",
""
],
[
"Lin",
"Zhouchen",
""
]
] |
2410.08911 | Marcus Kessel | Marcus Kessel | Test-driven Software Experimentation with LASSO: an LLM Prompt Benchmarking Example | under review | null | null | null | cs.SE cs.AI | http://creativecommons.org/licenses/by/4.0/ | Empirical software engineering faces a critical gap: the lack of standardized tools for rapid development and execution of Test-Driven Software Experiments (TDSEs) -- that is, experiments that involve the execution of software subjects and the observation and analysis of their "de facto" run-time behavior. In this paper we present a general-purpose analysis platform called LASSO that provides a minimal set of domain-specific languages and data structures to conduct TDSEs. By empowering users with an executable scripting language to design and execute TDSEs, LASSO enables efficient evaluation of run-time semantics and execution characteristics in addition to statically determined properties. We present an example TDSE that demonstrates the practical benefits of LASSO's scripting capabilities for assessing the reliability of LLMs for code generation by means of a self-contained, reusable and extensible study script. The LASSO platform and live pipeline examples are publicly available at: https://softwareobservatorium.github.io/. | [
{
"version": "v1",
"created": "Fri, 11 Oct 2024 15:32:48 GMT"
},
{
"version": "v2",
"created": "Sun, 08 Jun 2025 11:32:28 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Kessel",
"Marcus",
""
]
] |
2410.09016 | Wonjun Kang | Kevin Galim, Wonjun Kang, Yuchen Zeng, Hyung Il Koo, Kangwook Lee | Parameter-Efficient Fine-Tuning of State Space Models | Accepted at ICML 2025. Code is available at https://github.com/furiosa-ai/ssm-peft | null | null | null | cs.LG cs.CL | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Deep State Space Models (SSMs), such as Mamba (Gu & Dao, 2024), have become powerful tools for language modeling, offering high performance and linear scalability with sequence length. However, the application of parameter-efficient fine-tuning (PEFT) methods to SSM-based models remains largely underexplored. We start by investigating two fundamental questions on existing PEFT methods: (i) How do they perform on SSM-based models? (ii) Which parameters should they target for optimal results? Our analysis shows that LoRA and its variants consistently outperform all other PEFT methods. While LoRA is effective for linear projection matrices, it fails on SSM modules-yet still outperforms other methods applicable to SSMs, indicating their limitations. This underscores the need for a specialized SSM tuning approach. To address this, we propose Sparse Dimension Tuning (SDT), a PEFT method tailored for SSM modules. Combining SDT for SSMs with LoRA for linear projection matrices, we achieve state-of-the-art performance across extensive experiments. | [
{
"version": "v1",
"created": "Fri, 11 Oct 2024 17:30:28 GMT"
},
{
"version": "v2",
"created": "Fri, 14 Mar 2025 01:26:57 GMT"
},
{
"version": "v3",
"created": "Mon, 09 Jun 2025 06:07:50 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Galim",
"Kevin",
""
],
[
"Kang",
"Wonjun",
""
],
[
"Zeng",
"Yuchen",
""
],
[
"Koo",
"Hyung Il",
""
],
[
"Lee",
"Kangwook",
""
]
] |
2410.11005 | Fangru Lin | Fangru Lin, Shaoguang Mao, Emanuele La Malfa, Valentin Hofmann, Adrian de Wynter, Xun Wang, Si-Qing Chen, Michael Wooldridge, Janet B. Pierrehumbert, Furu Wei | Assessing Dialect Fairness and Robustness of Large Language Models in Reasoning Tasks | ACL 2025 main | null | null | null | cs.CL cs.LG | http://creativecommons.org/licenses/by/4.0/ | Language is not monolithic. While benchmarks, including those designed for multiple languages, are often used as proxies to evaluate the performance of Large Language Models (LLMs), they tend to overlook the nuances of within-language variation and thus fail to model the experience of speakers of non-standard dialects. Focusing on African American Vernacular English (AAVE), we present the first study aimed at objectively assessing the fairness and robustness of LLMs in handling dialects across canonical reasoning tasks, including algorithm, math, logic, and integrated reasoning. We introduce ReDial (Reasoning with Dialect Queries), a benchmark containing 1.2K+ parallel query pairs in Standardized English and AAVE. We hire AAVE speakers, including experts with computer science backgrounds, to rewrite seven popular benchmarks, such as HumanEval and GSM8K. With ReDial, we evaluate widely used LLMs, including GPT, Claude, Llama, Mistral, and the Phi model families. Our findings reveal that almost all of these widely used models show significant brittleness and unfairness to queries in AAVE. Our work establishes a systematic and objective framework for analyzing LLM bias in dialectal queries. Moreover, it highlights how mainstream LLMs provide unfair service to dialect speakers in reasoning tasks, laying a critical foundation for future research. | [
{
"version": "v1",
"created": "Mon, 14 Oct 2024 18:44:23 GMT"
},
{
"version": "v2",
"created": "Tue, 14 Jan 2025 09:52:50 GMT"
},
{
"version": "v3",
"created": "Mon, 09 Jun 2025 11:20:58 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Lin",
"Fangru",
""
],
[
"Mao",
"Shaoguang",
""
],
[
"La Malfa",
"Emanuele",
""
],
[
"Hofmann",
"Valentin",
""
],
[
"de Wynter",
"Adrian",
""
],
[
"Wang",
"Xun",
""
],
[
"Chen",
"Si-Qing",
""
],
[
"Wooldridge",
"Michael",
""
],
[
"Pierrehumbert",
"Janet B.",
""
],
[
"Wei",
"Furu",
""
]
] |
2410.12006 | Christopher Irwin | Annalisa Chiocchetti, Marco Dossena, Christopher Irwin and Luigi Portinale | Beyond Labels: A Self-Supervised Framework with Masked Autoencoders and Random Cropping for Breast Cancer Subtype Classification | null | null | 10.1609/aaaiss.v5i1.35546 | null | cs.CV cs.AI cs.LG | http://creativecommons.org/licenses/by/4.0/ | This work contributes to breast cancer sub-type classification using histopathological images. We utilize masked autoencoders (MAEs) to learn a self-supervised embedding tailored for computer vision tasks in this domain. This embedding captures informative representations of histopathological data, facilitating feature learning without extensive labeled datasets. During pre-training, we investigate employing a random crop technique to generate a large dataset from WSIs automatically. Additionally, we assess the performance of linear probes for multi-class classification tasks of cancer sub-types using the representations learnt by the MAE. Our approach aims to achieve strong performance on downstream tasks by leveraging the complementary strengths of ViTs and autoencoders. We evaluate our model's performance on the BRACS dataset and compare it with existing benchmarks. | [
{
"version": "v1",
"created": "Tue, 15 Oct 2024 19:13:05 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Chiocchetti",
"Annalisa",
""
],
[
"Dossena",
"Marco",
""
],
[
"Irwin",
"Christopher",
""
],
[
"Portinale",
"Luigi",
""
]
] |
2410.12204 | Xin-Yuan Gao | Xin-Yuan Gao and Yangqian Yan | Fate of thermalization of ultracold fermions with two-body dissipation | 7 pages, 2 figures, companion paper | Phys. Rev. Lett. 134, 153402 (2025) | 10.1103/PhysRevLett.134.153402 | null | cond-mat.quant-gas | http://creativecommons.org/licenses/by/4.0/ | Two-body dissipation due to chemical reactions occurs in both ultracold fermionic and bosonic molecular gases. Despite recent advances in achieving quantum degeneracy, the loss dynamics are typically described phenomenologically using rate equations, often assuming thermalization during chemical reactions. From the first principles, we analyze particle loss, temperature evolution, and momentum distributions in single-component Fermi gases using the inelastic quantum Boltzmann equation. Our results prove that the conventional two-body loss model is valid for trapped systems, though it fails to describe the dynamics in homogeneous systems accurately. Interestingly, we find that systems prepared near or above quantum degeneracy can thermalize spontaneously, even in the absence of elastic collisions, while systems initialized deep in degeneracy display non-equilibrium behavior. Our calculations are in good agreement with recent experimental data from trapped systems and could be further tested in atomic systems with induced two-body loss in box potentials. | [
{
"version": "v1",
"created": "Wed, 16 Oct 2024 03:55:59 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Gao",
"Xin-Yuan",
""
],
[
"Yan",
"Yangqian",
""
]
] |
2410.13147 | Khiem Le | Khiem Le, Ting Hua, Nitesh V. Chawla | AgentDrug: Utilizing Large Language Models in an Agentic Workflow for Zero-Shot Molecular Optimization | null | null | null | null | cs.LG cs.AI cs.CV | http://creativecommons.org/licenses/by/4.0/ | Molecular optimization -- modifying a given molecule to improve desired properties -- is a fundamental task in drug discovery. While LLMs hold the potential to solve this task using natural language to drive the optimization, straightforward prompting achieves limited accuracy. In this work, we propose AgentDrug, an agentic workflow that leverages LLMs in a structured refinement process to achieve significantly higher accuracy. AgentDrug defines a nested refinement loop: the inner loop uses feedback from cheminformatics toolkits to validate molecular structures, while the outer loop guides the LLM with generic feedback and a gradient-based objective to steer the molecule toward property improvement. We evaluate AgentDrug on benchmarks with both single- and multi-property optimization under loose and strict thresholds. Results demonstrate significant performance gains over previous methods. With Qwen-2.5-3B, AgentDrug improves accuracy by 20.7\% (loose) and 16.8\% (strict) on six single-property tasks, and by 7.0\% and 5.3\% on eight multi-property tasks. With larger model Qwen-2.5-7B, AgentDrug further improves accuracy on 6 single-property objectives by 28.9\% (loose) and 29.0\% (strict), and on 8 multi-property objectives by 14.9\% (loose) and 13.2\% (strict). | [
{
"version": "v1",
"created": "Thu, 17 Oct 2024 02:04:57 GMT"
},
{
"version": "v2",
"created": "Mon, 21 Oct 2024 00:38:07 GMT"
},
{
"version": "v3",
"created": "Sat, 26 Oct 2024 01:31:08 GMT"
},
{
"version": "v4",
"created": "Wed, 30 Oct 2024 14:54:25 GMT"
},
{
"version": "v5",
"created": "Wed, 06 Nov 2024 05:18:04 GMT"
},
{
"version": "v6",
"created": "Mon, 18 Nov 2024 15:41:01 GMT"
},
{
"version": "v7",
"created": "Sun, 08 Dec 2024 04:33:07 GMT"
},
{
"version": "v8",
"created": "Thu, 23 Jan 2025 20:19:40 GMT"
},
{
"version": "v9",
"created": "Sun, 08 Jun 2025 10:39:08 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Le",
"Khiem",
""
],
[
"Hua",
"Ting",
""
],
[
"Chawla",
"Nitesh V.",
""
]
] |
2410.13253 | Jintao Zhang | Jintao Zhang, Mingyue Cheng, Xiaoyu Tao, Zhiding Liu, Daoyu Wang | Conditional Denoising Meets Polynomial Modeling: A Flexible Decoupled Framework for Time Series Forecasting | null | null | null | null | cs.LG | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Time series forecasting models are becoming increasingly prevalent due to their critical role in decision-making across various domains. However, most existing approaches represent the coupled temporal patterns, often neglecting the distinction between their specific components. In particular, fluctuating patterns and smooth trends within time series exhibit distinct characteristics. In this work, to model complicated temporal patterns, we propose a Conditional Denoising Polynomial Modeling (CDPM) framework, where probabilistic diffusion models and deterministic linear models are trained end-to-end. Instead of modeling the coupled time series, CDPM decomposes it into trend and seasonal components for modeling them separately. To capture the fluctuating seasonal component, we employ a probabilistic diffusion model based on statistical properties from the historical window. For the smooth trend component, a module is proposed to enhance linear models by incorporating historical dependencies, thereby preserving underlying trends and mitigating noise distortion. Extensive experiments conducted on six benchmarks demonstrate the effectiveness of our framework, highlighting the potential of combining probabilistic and deterministic models. Our code is available at https://github.com/zjt-gpu/CDPM. | [
{
"version": "v1",
"created": "Thu, 17 Oct 2024 06:20:43 GMT"
},
{
"version": "v2",
"created": "Tue, 22 Oct 2024 04:58:21 GMT"
},
{
"version": "v3",
"created": "Thu, 24 Oct 2024 03:27:06 GMT"
},
{
"version": "v4",
"created": "Thu, 31 Oct 2024 11:30:52 GMT"
},
{
"version": "v5",
"created": "Sun, 23 Feb 2025 07:08:49 GMT"
},
{
"version": "v6",
"created": "Wed, 28 May 2025 08:26:49 GMT"
},
{
"version": "v7",
"created": "Sun, 08 Jun 2025 12:43:41 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Zhang",
"Jintao",
""
],
[
"Cheng",
"Mingyue",
""
],
[
"Tao",
"Xiaoyu",
""
],
[
"Liu",
"Zhiding",
""
],
[
"Wang",
"Daoyu",
""
]
] |
2410.13263 | Weishan Cai | Weishan Cai, Wenjun Ma and Yuncheng Jiang | A Simplifying and Learnable Graph Convolutional Attention Network for Unsupervised Knowledge Graphs Alignment | There is more work being done: the author hopes to resubmit the paper to provide better research after more research or further analysis | null | null | null | cs.AI cs.LG | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | The success of current Entity Alignment (EA) task depends largely on the supervision information provided by labeled data. Considering the cost of labeled data, most supervised methods are difficult to apply in practical scenarios. Therefore, more and more works based on contrastive learning, active learning or other deep learning techniques have been developed, to solve the performance bottleneck caused by the lack of labeled data. However, the existing unsupervised EA methods still have some limitations, either their modeling complexity is high or they cannot balance the effectiveness and practicality of alignment. To overcome these issues, we propose a Simplifying and Learnable graph convolutional attention network for Unsupervised Knowledge Graphs alignment method (SLU). Specifically, we first introduce LCAT, a new and simple framework as the backbone network to model the graph structure of two KGs. Then we design a reconstruction method of relation structure based on potential matching relations for efficiently filtering invalid neighborhood information of aligned entities, to improve the usability and scalability of SLU. Impressively, a similarity function based on consistency is proposed to better measure the similarity of candidate entity pairs. Finally, we conduct extensive experiments on three datasets of different sizes (15K and 100K) and different types (cross-lingual and monolingual) to verify the superiority of SLU. Experimental results show that SLU significantly improves alignment accuracy, outperforming 25 supervised or unsupervised methods, and improving 6.4% in Hits@1 over the best baseline in the best case. | [
{
"version": "v1",
"created": "Thu, 17 Oct 2024 06:37:46 GMT"
},
{
"version": "v2",
"created": "Mon, 09 Jun 2025 08:46:25 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Cai",
"Weishan",
""
],
[
"Ma",
"Wenjun",
""
],
[
"Jiang",
"Yuncheng",
""
]
] |
2410.13695 | Mervyn Tong | Mervyn Tong | Zarankiewicz bounds from distal regularity lemma | 16 pages. Relationship with semialgebraic results added, definition of strong distal regularity tuple updated to align with combinatorial literature, and Question 5.6 in the previous version replaced by an answer | null | null | null | math.CO math.LO | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Since K\H{o}v\'ari, S\'os, and Tur\'an proved upper bounds for the Zarankiewicz problem in 1954, much work has been undertaken to improve these bounds, and some have done so by restricting to particular classes of graphs. In 2017, Fox, Pach, Sheffer, Suk, and Zahl proved better bounds for semialgebraic binary relations, and this work was extended by Do in the following year to arbitrary semialgebraic relations. In this paper, we show that Zarankiewicz bounds in the shape of Do's are enjoyed by all relations satisfying the distal regularity lemma, an improved version of the Szemer\'edi regularity lemma satisfied by relations definable in distal structures (a vast generalisation of o-minimal structures). | [
{
"version": "v1",
"created": "Thu, 17 Oct 2024 16:03:35 GMT"
},
{
"version": "v2",
"created": "Mon, 09 Jun 2025 15:28:24 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Tong",
"Mervyn",
""
]
] |
2410.13794 | Da Long | Da Long, Zhitong Xu, Guang Yang, Akil Narayan, Shandian Zhe | Arbitrarily-Conditioned Multi-Functional Diffusion for Multi-Physics Emulation | null | null | null | null | cs.LG | http://creativecommons.org/licenses/by/4.0/ | Modern physics simulation often involves multiple functions of interests, and traditional numerical approaches are known to be complex and computationally costly. While machine learning-based surrogate models can offer significant cost reductions, most focus on a single task, such as forward prediction, and typically lack uncertainty quantification -- an essential component in many applications. To overcome these limitations, we propose Arbitrarily-Conditioned Multi-Functional Diffusion (ACM-FD), a versatile probabilistic surrogate model for multi-physics emulation. ACM-FD can perform a wide range of tasks within a single framework, including forward prediction, various inverse problems, and simulating data for entire systems or subsets of quantities conditioned on others. Specifically, we extend the standard Denoising Diffusion Probabilistic Model (DDPM) for multi-functional generation by modeling noise as Gaussian processes (GP). We propose a random-mask based, zero-regularized denoising loss to achieve flexible and robust conditional generation. We induce a Kronecker product structure in the GP covariance matrix, substantially reducing the computational cost and enabling efficient training and sampling. We demonstrate the effectiveness of ACM-FD across several fundamental multi-physics systems. | [
{
"version": "v1",
"created": "Thu, 17 Oct 2024 17:34:06 GMT"
},
{
"version": "v2",
"created": "Sat, 07 Jun 2025 02:53:05 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Long",
"Da",
""
],
[
"Xu",
"Zhitong",
""
],
[
"Yang",
"Guang",
""
],
[
"Narayan",
"Akil",
""
],
[
"Zhe",
"Shandian",
""
]
] |
2410.13879 | Philippe Chlenski | Philippe Chlenski, Quentin Chu, Raiyan R. Khan, Kaizhu Du, Antonio Khalil Moretti, Itsik Pe'er | Mixed-curvature decision trees and random forests | 30 pages, 12 figures, 13 tables. Camera-ready version for ICML 2025 | null | null | null | cs.LG | http://creativecommons.org/licenses/by-sa/4.0/ | Decision trees (DTs) and their random forest (RF) extensions are workhorses of classification and regression in Euclidean spaces. However, algorithms for learning in non-Euclidean spaces are still limited. We extend DT and RF algorithms to product manifolds: Cartesian products of several hyperbolic, hyperspherical, or Euclidean components. Such manifolds handle heterogeneous curvature while still factorizing neatly into simpler components, making them compelling embedding spaces for complex datasets. Our novel angular reformulation respects manifold geometry while preserving the algorithmic properties that make decision trees effective. In the special cases of single-component manifolds, our method simplifies to its Euclidean or hyperbolic counterparts, or introduces hyperspherical DT algorithms, depending on the curvature. In benchmarks on a diverse suite of 57 classification, regression, and link prediction tasks, our product RFs ranked first on 29 tasks and came in the top 2 for 41. This highlights the value of product RFs as straightforward yet powerful new tools for data analysis in product manifolds. Code for our method is available at https://github.com/pchlenski/manify. | [
{
"version": "v1",
"created": "Thu, 03 Oct 2024 00:48:07 GMT"
},
{
"version": "v2",
"created": "Thu, 13 Feb 2025 16:43:47 GMT"
},
{
"version": "v3",
"created": "Fri, 06 Jun 2025 21:26:35 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Chlenski",
"Philippe",
""
],
[
"Chu",
"Quentin",
""
],
[
"Khan",
"Raiyan R.",
""
],
[
"Du",
"Kaizhu",
""
],
[
"Moretti",
"Antonio Khalil",
""
],
[
"Pe'er",
"Itsik",
""
]
] |
2410.14073 | Mahdi Teimouri Yanesari | Mahdi Teimouri | Digesting Gibbs Sampling Using R | null | null | null | null | stat.CO | http://creativecommons.org/licenses/by/4.0/ | In general, the statistical simulation approaches are referred to as the Monte Carlo methods as a whole. The broad class of the Monte Carlo methods involves the Markov chain Monte Carlo (MCMC) techniques that attract the attention of researchers from a wide variety of study fields. The main focus of this report is to provide a framework for all users who are interested in implementing the MCMC approaches in their investigations, especially the Gibbs sampling. I have tried, if possible, to eliminate the proofs, but reader is expected to know some topics in elementary calculus (including mathematical function, limit, derivative, partial derivative, simple integral) and statistics (including random variables, expected value and variance, moment generating function, multivariate distribution, distribution of a functions of random variable, and the central limit theorem). | [
{
"version": "v1",
"created": "Thu, 17 Oct 2024 22:46:19 GMT"
},
{
"version": "v2",
"created": "Mon, 23 Dec 2024 15:36:14 GMT"
},
{
"version": "v3",
"created": "Sun, 08 Jun 2025 08:36:25 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Teimouri",
"Mahdi",
""
]
] |
2410.14182 | Yujun Zhou | Yujun Zhou, Jingdong Yang, Yue Huang, Kehan Guo, Zoe Emory, Bikram Ghosh, Amita Bedar, Sujay Shekar, Zhenwen Liang, Pin-Yu Chen, Tian Gao, Werner Geyer, Nuno Moniz, Nitesh V Chawla and Xiangliang Zhang | LabSafety Bench: Benchmarking LLMs on Safety Issues in Scientific Labs | null | null | null | null | cs.CL cs.LG | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Artificial Intelligence (AI) is revolutionizing scientific research, yet its growing integration into laboratory environments presents critical safety challenges. While large language models (LLMs) increasingly assist in tasks ranging from procedural guidance to autonomous experiment orchestration, an "illusion of understanding" may lead researchers to overestimate their reliability. Such overreliance is particularly dangerous in high-stakes laboratory settings, where failures in hazard identification or risk assessment can result in severe accidents. To address these concerns, we propose the Laboratory Safety Benchmark (LabSafety Bench), a comprehensive framework that evaluates large language models and vision language models (VLMs) on their ability to identify potential hazards, assess risks, and predict the consequences of unsafe actions in lab environments. LabSafety Bench comprises 765 multiple-choice questions aligned with US Occupational Safety and Health Administration (OSHA) protocols, along with 404 realistic laboratory scenarios featuring dual evaluation tasks: the Hazards Identification Test and the Consequence Identification Test, with 3128 open-ended questions in total. Evaluations across eight proprietary models, seven open-weight LLMs, and four VLMs reveal that, despite advanced performance on structured assessments, no model achieves the safety threshold required for reliable operation -- none scoring above 70% on the Hazards Identification Test. Moreover, while proprietary models tend to excel in multiple-choice evaluations, their performance in open-ended, real-world scenario responses is comparable to that of open-source models. These findings underscore the urgent need for specialized evaluation frameworks to ensure the safe and responsible deployment of AI in laboratory settings. | [
{
"version": "v1",
"created": "Fri, 18 Oct 2024 05:21:05 GMT"
},
{
"version": "v2",
"created": "Wed, 26 Feb 2025 09:17:27 GMT"
},
{
"version": "v3",
"created": "Fri, 06 Jun 2025 21:15:53 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Zhou",
"Yujun",
""
],
[
"Yang",
"Jingdong",
""
],
[
"Huang",
"Yue",
""
],
[
"Guo",
"Kehan",
""
],
[
"Emory",
"Zoe",
""
],
[
"Ghosh",
"Bikram",
""
],
[
"Bedar",
"Amita",
""
],
[
"Shekar",
"Sujay",
""
],
[
"Liang",
"Zhenwen",
""
],
[
"Chen",
"Pin-Yu",
""
],
[
"Gao",
"Tian",
""
],
[
"Geyer",
"Werner",
""
],
[
"Moniz",
"Nuno",
""
],
[
"Chawla",
"Nitesh V",
""
],
[
"Zhang",
"Xiangliang",
""
]
] |
2410.14676 | Muhao Chen | Qin Liu, Fei Wang, Chaowei Xiao, Muhao Chen | SudoLM: Learning Access Control of Parametric Knowledge with Authorization Alignment | ACL 2025 | null | null | null | cs.CL cs.AI | http://creativecommons.org/licenses/by/4.0/ | Existing preference alignment is a one-size-fits-all alignment mechanism, where the part of the large language model (LLM) parametric knowledge with non-preferred features is uniformly blocked to all the users. However, this part of knowledge can be useful to advanced users whose expertise qualifies them to handle these information. The one-size-fits-all alignment mechanism undermines LLM's utility for these qualified users. To address this problem, we propose SudoLM, a framework that lets LLMs learn access control over specific parametric knowledge for users with different credentials via authorization alignment. SudoLM allows authorized users to unlock their access to all the parametric knowledge with an assigned SUDO key while blocking access to non-qualified users. Experiments on two application scenarios demonstrate that SudoLM effectively controls the user's access to the parametric knowledge and maintains its general utility. | [
{
"version": "v1",
"created": "Fri, 18 Oct 2024 17:59:51 GMT"
},
{
"version": "v2",
"created": "Thu, 27 Feb 2025 01:05:43 GMT"
},
{
"version": "v3",
"created": "Sat, 07 Jun 2025 05:42:06 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Liu",
"Qin",
""
],
[
"Wang",
"Fei",
""
],
[
"Xiao",
"Chaowei",
""
],
[
"Chen",
"Muhao",
""
]
] |
2410.14886 | Chaoxi Niu | Chaoxi Niu, Hezhe Qiao, Changlu Chen, Ling Chen, Guansong Pang | Zero-shot Generalist Graph Anomaly Detection with Unified Neighborhood Prompts | Accepted by IJCAI25 | null | null | null | cs.LG | http://creativecommons.org/licenses/by-sa/4.0/ | Graph anomaly detection (GAD), which aims to identify nodes in a graph that significantly deviate from normal patterns, plays a crucial role in broad application domains. However, existing GAD methods are one-model-for-one-dataset approaches, i.e., training a separate model for each graph dataset. This largely limits their applicability in real-world scenarios. To overcome this limitation, we propose a novel zero-shot generalist GAD approach UNPrompt that trains a one-for-all detection model, requiring the training of one GAD model on a single graph dataset and then effectively generalizing to detect anomalies in other graph datasets without any retraining or fine-tuning. The key insight in UNPrompt is that i) the predictability of latent node attributes can serve as a generalized anomaly measure and ii) generalized normal and abnormal graph patterns can be learned via latent node attribute prediction in a properly normalized node attribute space. UNPrompt achieves a generalist mode for GAD through two main modules: one module aligns the dimensionality and semantics of node attributes across different graphs via coordinate-wise normalization, while another module learns generalized neighborhood prompts that support the use of latent node attribute predictability as an anomaly score across different datasets. Extensive experiments on real-world GAD datasets show that UNPrompt significantly outperforms diverse competing methods under the generalist GAD setting, and it also has strong superiority under the one-model-for-one-dataset setting. Code is available at https://github.com/mala-lab/UNPrompt. | [
{
"version": "v1",
"created": "Fri, 18 Oct 2024 22:23:59 GMT"
},
{
"version": "v2",
"created": "Sat, 07 Jun 2025 03:22:18 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Niu",
"Chaoxi",
""
],
[
"Qiao",
"Hezhe",
""
],
[
"Chen",
"Changlu",
""
],
[
"Chen",
"Ling",
""
],
[
"Pang",
"Guansong",
""
]
] |
2410.15021 | Hidetaka Kamigaito | Hidetaka Kamigaito, Hiroyuki Deguchi, Yusuke Sakai, Katsuhiko Hayashi, Taro Watanabe | Diversity Explains Inference Scaling Laws: Through a Case Study of Minimum Bayes Risk Decoding | Accepted to ACL 2025 Main | null | null | null | cs.CL | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Inference methods play an important role in eliciting the performance of large language models (LLMs). Currently, LLMs use inference methods utilizing generated multiple samples, which can be derived from Minimum Bayes Risk (MBR) Decoding. Previous studies have conducted empirical analyses to clarify the improvements in generation performance achieved by MBR decoding and have reported various observations. However, the theoretical underpinnings of these findings remain uncertain. To address this, we offer a new theoretical interpretation of MBR decoding from the perspective of bias-diversity decomposition. In this interpretation, the error in the quality estimation of hypotheses by MBR decoding is decomposed into two main factors: bias, which considers the closeness between the utility function and human evaluation, and diversity, which represents the variability in the quality estimation of the utility function. The theoretical analysis reveals the difficulty of simultaneously improving bias and diversity, confirming the validity of enhancing MBR decoding performance by increasing diversity. Furthermore, we reveal that diversity can explain one aspect of inference scaling laws that describe performance improvement by increasing sample size. Moreover, experiments across multiple NLP tasks yielded results consistent with these theoretical characteristics. Our code is available at https://github.com/naist-nlp/mbr-bias-diversity. | [
{
"version": "v1",
"created": "Sat, 19 Oct 2024 07:32:10 GMT"
},
{
"version": "v2",
"created": "Fri, 06 Jun 2025 20:14:03 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Kamigaito",
"Hidetaka",
""
],
[
"Deguchi",
"Hiroyuki",
""
],
[
"Sakai",
"Yusuke",
""
],
[
"Hayashi",
"Katsuhiko",
""
],
[
"Watanabe",
"Taro",
""
]
] |
2410.15068 | Hrishav Bakul Barua | Hrishav Bakul Barua, Kalin Stefanov, Lemuel Lai En Che, Abhinav Dhall, KokSheik Wong, Ganesh Krishnasamy | LLM-HDR: Bridging LLM-based Perception and Self-Supervision for Unpaired LDR-to-HDR Image Reconstruction | null | null | null | null | cs.CV cs.AI cs.GR cs.LG cs.RO | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | The translation of Low Dynamic Range (LDR) to High Dynamic Range (HDR) images is an important computer vision task. There is a significant amount of research utilizing both conventional non-learning methods and modern data-driven approaches, focusing on using both single-exposed and multi-exposed LDR for HDR image reconstruction. However, most current state-of-the-art methods require high-quality paired {LDR,HDR} datasets for model training. In addition, there is limited literature on using unpaired datasets for this task, that is, the model learns a mapping between domains, i.e., {LDR,HDR}. This paper proposes LLM-HDR, a method that integrates the perception of Large Language Models (LLM) into a modified semantic- and cycle-consistent adversarial architecture that utilizes unpaired {LDR,HDR} datasets for training. The method introduces novel artifact- and exposure-aware generators to address visual artifact removal and an encoder and loss to address semantic consistency, another under-explored topic. LLM-HDR is the first to use an LLM for the {LDR,HDR} translation task in a self-supervised setup. The method achieves state-of-the-art performance across several benchmark datasets and reconstructs high-quality HDR images. The official website of this work is available at: https://github.com/HrishavBakulBarua/LLM-HDR | [
{
"version": "v1",
"created": "Sat, 19 Oct 2024 11:11:58 GMT"
},
{
"version": "v2",
"created": "Tue, 11 Mar 2025 06:46:42 GMT"
},
{
"version": "v3",
"created": "Sun, 08 Jun 2025 05:49:47 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Barua",
"Hrishav Bakul",
""
],
[
"Stefanov",
"Kalin",
""
],
[
"Che",
"Lemuel Lai En",
""
],
[
"Dhall",
"Abhinav",
""
],
[
"Wong",
"KokSheik",
""
],
[
"Krishnasamy",
"Ganesh",
""
]
] |
2410.15464 | Lance Calvin Gamboa | Lance Calvin Lim Gamboa and Mark Lee | A Novel Interpretability Metric for Explaining Bias in Language Models: Applications on Multilingual Models from Southeast Asia | Accepted for oral presentation at PACLIC 38 (38th Pacific Asia Conference on Language, Information, and Computation) | https://aclanthology.org/2024.paclic-1.29/ | null | null | cs.CL | http://creativecommons.org/licenses/by/4.0/ | Work on bias in pretrained language models (PLMs) focuses on bias evaluation and mitigation and fails to tackle the question of bias attribution and explainability. We propose a novel metric, the $\textit{bias attribution score}$, which draws from information theory to measure token-level contributions to biased behavior in PLMs. We then demonstrate the utility of this metric by applying it on multilingual PLMs, including models from Southeast Asia which have not yet been thoroughly examined in bias evaluation literature. Our results confirm the presence of sexist and homophobic bias in Southeast Asian PLMs. Interpretability and semantic analyses also reveal that PLM bias is strongly induced by words relating to crime, intimate relationships, and helping among other discursive categories, suggesting that these are topics where PLMs strongly reproduce bias from pretraining data and where PLMs should be used with more caution. | [
{
"version": "v1",
"created": "Sun, 20 Oct 2024 18:31:05 GMT"
},
{
"version": "v2",
"created": "Thu, 24 Oct 2024 15:58:57 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Gamboa",
"Lance Calvin Lim",
""
],
[
"Lee",
"Mark",
""
]
] |
2410.16201 | Niclas Dern | Niclas Dern, John P. Cunningham, Geoff Pleiss | Theoretical Limitations of Ensembles in the Age of Overparameterization | Accepted for publication at ICML 2025. 33 pages, 17 figures | null | null | null | stat.ML cs.LG | http://creativecommons.org/licenses/by/4.0/ | Classic ensembles generalize better than any single component model. In contrast, recent empirical studies find that modern ensembles of (overparameterized) neural networks may not provide any inherent generalization advantage over single but larger neural networks. This paper clarifies how modern overparameterized ensembles differ from their classic underparameterized counterparts, using ensembles of random feature (RF) regressors as a basis for developing theory. In contrast to the underparameterized regime, where ensembling typically induces regularization and increases generalization, we prove with minimal assumptions that infinite ensembles of overparameterized RF regressors become pointwise equivalent to (single) infinite-width RF regressors, and finite width ensembles rapidly converge to single models with the same parameter budget. These results, which are exact for ridgeless models and approximate for small ridge penalties, imply that overparameterized ensembles and single large models exhibit nearly identical generalization. We further characterize the predictive variance amongst ensemble members, demonstrating that it quantifies the expected effects of increasing capacity rather than capturing any conventional notion of uncertainty. Our results challenge common assumptions about the advantages of ensembles in overparameterized settings, prompting a reconsideration of how well intuitions from underparameterized ensembles transfer to deep ensembles and the overparameterized regime. | [
{
"version": "v1",
"created": "Mon, 21 Oct 2024 17:03:20 GMT"
},
{
"version": "v2",
"created": "Mon, 09 Jun 2025 09:30:56 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Dern",
"Niclas",
""
],
[
"Cunningham",
"John P.",
""
],
[
"Pleiss",
"Geoff",
""
]
] |
2410.16316 | Md Faizul Bari | Md Faizul Bari, Meghna Roy Chowdhury, and Shreyas Sen | A Computational Harmonic Detection Algorithm to Detect Data Leakage through EM Emanation | This is the extended version of our previously published conference paper (DOI: 10.23919/DATE56975.2023.10137263) which can be found here: https://ieeexplore.ieee.org/abstract/document/10137263 | null | 10.1109/JIOT.2025.3578511 | null | cs.CR eess.SP | http://creativecommons.org/licenses/by-nc-sa/4.0/ | Unintended electromagnetic emissions, called EM emanations, can be exploited to recover sensitive information, posing security risks. Metal shielding, used by defense organizations to prevent data leakage, is costly and impractical for widespread use. This issue is particularly significant for IoT devices due to their sheer volume and varied deployment environments. Therefore, there is a research need for an automated detection method to monitor facilities and address data leakage promptly. To resolve this challenge, in the preliminary version of this work [1], a CNN-based detection method was proposed using HDMI cable emanations that provided ~95% accuracy up to 22.5 m but had limitations due to training data. In this extended version, we augment the initial study by collecting and characterizing emanation data from IoT devices, everyday electronics, and cables. We propose a harmonic-based emanation detection method by developing a computational harmonic detection algorithm. The proposed method addresses the limitations of the CNN-based method and provides ~100% accuracy not only for HDMI emanation (compared to ~95% in the earlier CNN method) but also for all other tested devices and cables. Finally, it has also been tested in different environments to prove its efficacy in practical scenarios. | [
{
"version": "v1",
"created": "Wed, 09 Oct 2024 14:40:15 GMT"
},
{
"version": "v2",
"created": "Mon, 09 Jun 2025 04:38:04 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Bari",
"Md Faizul",
""
],
[
"Chowdhury",
"Meghna Roy",
""
],
[
"Sen",
"Shreyas",
""
]
] |
2410.16558 | Alexei Bylinskii | Daniel Gonzalez-Cuadra, Majd Hamdan, Torsten V. Zache, Boris Braverman, Milan Kornjaca, Alexander Lukin, Sergio H. Cantu, Fangli Liu, Sheng-Tao Wang, Alexander Keesling, Mikhail D. Lukin, Peter Zoller, Alexei Bylinskii | Observation of string breaking on a (2 + 1)D Rydberg quantum simulator | null | null | 10.1038/s41586-025-09051-6 | null | quant-ph cond-mat.str-el hep-lat | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Lattice gauge theories (LGTs) describe a broad range of phenomena in condensed matter and particle physics. A prominent example is confinement, responsible for bounding quarks inside hadrons such as protons or neutrons. When quark-antiquark pairs are separated, the energy stored in the string of gluon fields connecting them grows linearly with their distance, until there is enough energy to create new pairs from the vacuum and break the string. While such phenomena are ubiquitous in LGTs, simulating the resulting dynamics is a challenging task. Here, we report the observation of string breaking in synthetic quantum matter using a programmable quantum simulator based on neutral atom arrays. We show that a (2+1)D LGT with dynamical matter can be efficiently implemented when the atoms are placed on a Kagome geometry, with a local U(1) symmetry emerging from the Rydberg blockade, while long-range Rydberg interactions naturally give rise to a linear confining potential for a pair of charges, allowing us to tune both their masses as well as the string tension. We experimentally map out the corresponding phase diagram by adiabatically preparing the ground state of the atom array in the presence of defects, and observe substructure of the confined phase, distinguishing regions dominated by fluctuating strings or by broken string configurations. Finally, by harnessing local control over the atomic detuning, we quench string states and observe string breaking dynamics exhibiting a many-body resonance phenomenon. Our work paves a way to explore phenomena in high-energy physics using programmable quantum simulators. | [
{
"version": "v1",
"created": "Mon, 21 Oct 2024 22:33:16 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Gonzalez-Cuadra",
"Daniel",
""
],
[
"Hamdan",
"Majd",
""
],
[
"Zache",
"Torsten V.",
""
],
[
"Braverman",
"Boris",
""
],
[
"Kornjaca",
"Milan",
""
],
[
"Lukin",
"Alexander",
""
],
[
"Cantu",
"Sergio H.",
""
],
[
"Liu",
"Fangli",
""
],
[
"Wang",
"Sheng-Tao",
""
],
[
"Keesling",
"Alexander",
""
],
[
"Lukin",
"Mikhail D.",
""
],
[
"Zoller",
"Peter",
""
],
[
"Bylinskii",
"Alexei",
""
]
] |
2410.16860 | Eddy Keming Chen | Eddy Keming Chen, Roderich Tumulka | Typical Quantum States of the Universe are Observationally Indistinguishable | 23 pages LaTeX, no figures; v2 major revision and extension | null | null | null | quant-ph cond-mat.stat-mech physics.hist-ph | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We establish three new impossibility results regarding our knowledge of the quantum state of the universe -- a central object in quantum theory. We show that, if the universal quantum state is a typical unit vector from a high-dimensional subspace H_0 of Hilbert space H (such as the one defined by a low-entropy macro-state as prescribed by the Past Hypothesis), then no observation can determine or just significantly narrow down which vector it is. In other words, the overwhelming majority of possible state vectors are observationally indistinguishable from each other (and from the density matrix of H_0). Moreover, we show that for any observation that isn't too unlikely and most pairs of unit vectors from H_0, the observation will not significantly favor one vector over the other. We further show that the uniform distribution over the unit sphere in H_0, after Bayesian updating in the light of any observation that isn't too unlikely, is still extremely close to uniform. Our arguments rely on a typicality theorem from quantum statistical mechanics. We also discuss how theoretical considerations beyond empirical evidence might inform our understanding of this fact and our knowledge of the universal quantum state. | [
{
"version": "v1",
"created": "Tue, 22 Oct 2024 09:59:57 GMT"
},
{
"version": "v2",
"created": "Sun, 08 Jun 2025 21:37:33 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Chen",
"Eddy Keming",
""
],
[
"Tumulka",
"Roderich",
""
]
] |
2410.17754 | Jaron Skovsted Gundersen | Jaron Skovsted Gundersen, Ren\'e B{\o}dker Christensen, Markus Grassl, Petar Popovski, Rafa{\l} Wisniewski | Puncturing Quantum Stabilizer Codes | Accepted version for IEEE Journal on Selected Areas in Information Theory | IEEE Journal on Selected Areas in Information Theory, vol. 6, pp. 74-84, 2025 | 10.1109/JSAIT.2025.3562287 | null | cs.IT math.IT math.RA quant-ph | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Classical coding theory contains several techniques to obtain new codes from other codes, including puncturing and shortening. For quantum codes, a form of puncturing is known, but its description is based on the code space rather than its generators. In this work, we generalize the puncturing procedure to allow more freedom in the choice of which coded states are kept and which are removed. We describe this puncturing by focusing on the stabilizer matrix containing the generators of the code. In this way, we are able to explicitly describe the stabilizer matrix of the punctured code given the stabilizer matrix of the original stabilizer code. The additional freedom in the procedure also opens up new ways to construct new codes from old, and we present several ways to utilize this for the search of codes with good or even optimal parameters. In particular, we use the construction to obtain codes whose parameters exceed the best previously known. Lastly, we generalize the proof of the Griesmer bound from the classical setting to stabilizer codes since the proof relies heavily on the puncturing technique. | [
{
"version": "v1",
"created": "Wed, 23 Oct 2024 10:31:34 GMT"
},
{
"version": "v2",
"created": "Thu, 21 Nov 2024 13:48:15 GMT"
},
{
"version": "v3",
"created": "Wed, 16 Apr 2025 09:50:33 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Gundersen",
"Jaron Skovsted",
""
],
[
"Christensen",
"René Bødker",
""
],
[
"Grassl",
"Markus",
""
],
[
"Popovski",
"Petar",
""
],
[
"Wisniewski",
"Rafał",
""
]
] |
2410.17897 | Zhanchao Zhou | Zhanchao Zhou, Tianyi Wu, Zhiyun Jiang, Fares Obeid, Zhenzhong Lan | Value Residual Learning | null | null | null | null | cs.CL | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | While Transformer models have achieved remarkable success in various domains, the effectiveness of information propagation through deep networks remains a critical challenge. Standard hidden state residuals often fail to adequately preserve initial token-level information in deeper layers. This paper introduces ResFormer, a novel architecture that enhances information flow by incorporating value residual connections in addition to hidden state residuals. And a variant is SVFormer, where all layers share the first layer's value embedding. Comprehensive empirical evidence demonstrates ResFormer achieves equivalent validation loss with 16.11\% fewer model parameters and 20.3\% less training data compared to Transformer, while maintaining similar memory usage and computational cost. Besides, SVFormer reduces KV cache size by nearly half with only a small performance penalty and can be integrated with other KV-efficient methods, yielding further reductions in KV cache, with performance influenced by sequence length and cumulative learning rate. | [
{
"version": "v1",
"created": "Wed, 23 Oct 2024 14:15:07 GMT"
},
{
"version": "v2",
"created": "Thu, 14 Nov 2024 17:46:04 GMT"
},
{
"version": "v3",
"created": "Tue, 03 Dec 2024 12:36:19 GMT"
},
{
"version": "v4",
"created": "Wed, 19 Feb 2025 17:53:11 GMT"
},
{
"version": "v5",
"created": "Sun, 08 Jun 2025 16:04:59 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Zhou",
"Zhanchao",
""
],
[
"Wu",
"Tianyi",
""
],
[
"Jiang",
"Zhiyun",
""
],
[
"Obeid",
"Fares",
""
],
[
"Lan",
"Zhenzhong",
""
]
] |
2410.17968 | Wei-Chih Huang | Bhaskar Dutta, Bai-Shan Hu, Wei-Chih Huang, and Richard G. Van de Water | Novel approach to investigate ATOMKI anomaly using Coherent CAPTAIN-Mills detectors | 8 pages, 6 figures | null | null | null | hep-ph nucl-th | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | ATOMKI nuclear anomaly has suggested a new BSM (Beyond the Standard Model) boson with mass $\sim17$ MeV emitted from excited nuclei and quickly decays into a pair of $e^+e^-$. In order to search for the new particle, we propose a new approach that utilizes the ongoing Coherent CAPTAIN-Mills (CCM) 10-ton LAr (liquid argon) detectors. The neutrons from the Lujan target can scatter inelastically by the PMT glass in the CCM detector can produce the new boson which solves the ATOMKI anomaly. The new boson can be detected from its decay to a $e^+e^-$ pair. We find that CCM probes a large area of the anomaly-allowable parameter space. We also show the prediction for a 100-ton LAr detector and 5-ton EOS water detector. | [
{
"version": "v1",
"created": "Wed, 23 Oct 2024 15:34:52 GMT"
},
{
"version": "v2",
"created": "Sat, 07 Jun 2025 13:09:25 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Dutta",
"Bhaskar",
""
],
[
"Hu",
"Bai-Shan",
""
],
[
"Huang",
"Wei-Chih",
""
],
[
"Van de Water",
"Richard G.",
""
]
] |
2410.18239 | Hossein Rajabzadeh | Maryam Dialameh, Hossein Rajabzadeh, Moslem Sadeghi-Goughari, Jung Suk Sim, Hyock Ju Kwon | E2E-Swin-Unet++: An Enhanced End-to-End Swin-Unet Architecture With Dual Decoders For PTMC Segmentation | null | null | null | null | eess.IV cs.AI cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Precise segmentation of papillary thyroid microcarcinoma (PTMC) during ultrasound-guided radiofrequency ablation (RFA) is critical for effective treatment but remains challenging due to acoustic artifacts, small lesion size, and anatomical variability. In this study, we propose \textbf{DualSwinUnet++}, a dual-decoder transformer-based architecture designed to enhance PTMC segmentation by incorporating thyroid gland context. DualSwinUnet++ employs independent linear projection heads for each decoder and a residual information flow mechanism that passes intermediate features from the first (thyroid) decoder to the second (PTMC) decoder via concatenation and transformation. These design choices allow the model to condition tumor prediction explicitly on gland morphology without shared gradient interference. Trained on a clinical ultrasound dataset with 691 annotated RFA images and evaluated against state-of-the-art models, DualSwinUnet++ achieves superior Dice and Jaccard scores while maintaining sub-200ms inference latency. The results demonstrate the model's suitability for near real-time surgical assistance and its effectiveness in improving segmentation accuracy in challenging PTMC cases. | [
{
"version": "v1",
"created": "Wed, 23 Oct 2024 19:33:33 GMT"
},
{
"version": "v2",
"created": "Sun, 08 Jun 2025 03:11:31 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Dialameh",
"Maryam",
""
],
[
"Rajabzadeh",
"Hossein",
""
],
[
"Sadeghi-Goughari",
"Moslem",
""
],
[
"Sim",
"Jung Suk",
""
],
[
"Kwon",
"Hyock Ju",
""
]
] |
2410.19202 | Mitchell Linegar | Mitchell Linegar, Betsy Sinclair, Sander van der Linden, R. Michael Alvarez | Towards Generalizable AI-Assisted Misinformation Inoculation: Protecting Confidence Against False Election Narratives | Version 2. Updates from previous: main independent variables are now post-treatment confidence scores, and we include pre-treatment confidence scores as a control, rather than work with difference scores. This improves power, but conclusions should otherwise be identical. Clarifications of experimental design, scalability claims, more precise phrasing of results | null | null | null | econ.GN q-fin.EC | http://creativecommons.org/licenses/by/4.0/ | We present a generalizable AI-assisted framework for rapidly generating effective "prebunking" interventions against misinformation. Like mRNA vaccine platforms, our approach uses a stable template structure that can be quickly adapted to counter emerging false narratives. In a preregistered two-wave experiment with 4,293 U.S. registered voters, we test this framework against politically-charged election misinformation -- one of the most challenging domains for misinformation intervention. Our design directly tests scalability by comparing human-reviewed and purely AI-generated inoculation messages. We find that LLM-generated prebunking significantly reduced belief in election rumors (persisting for at least one week) and increased confidence in election integrity across partisan lines. Purely AI-generated messages proved as effective as human-reviewed versions, with some achieving larger protective effects, demonstrating that effective misinformation inoculation can be achieved at machine speed without proportional human effort, offering a scalable defense against the accelerating threat of false narratives across all domains. | [
{
"version": "v1",
"created": "Thu, 24 Oct 2024 23:24:35 GMT"
},
{
"version": "v2",
"created": "Fri, 06 Jun 2025 20:55:07 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Linegar",
"Mitchell",
""
],
[
"Sinclair",
"Betsy",
""
],
[
"van der Linden",
"Sander",
""
],
[
"Alvarez",
"R. Michael",
""
]
] |
2410.19214 | Xueqi Cheng | Xueqi Cheng, Catherine Yang, Yuying Zhao, Yu Wang, Hamid Karimi, Tyler Derr | BTS: A Comprehensive Benchmark for Tie Strength Prediction | null | null | null | null | cs.SI | http://creativecommons.org/licenses/by/4.0/ | The rapid rise of online social networks underscores the need to understand the heterogeneous strengths of online relationships. Yet, efforts to assess tie strength (TS) are hindered by the lack of ground-truth labels, differing research perspectives, and limited model performance in real-world settings. To address this gap, we introduce BTS, a comprehensive Benchmark for Tie Strength prediction, aiming to establish a standardized foundation for evaluating and advancing TS prediction methodologies. Specifically, our contributions are: TS Pseudo-Label Techniques -- we categorize TS into seven standardized pseudo-labeling techniques based on prior literature; TS Dataset Collection -- we present a representative collection of three social networks and perform data analysis by investigating the class distributions and correlations across the generated pseudo-labels; TS Pseudo-Label Evaluation Framework -- we propose a standardized framework to evaluate the pseudo-label quality from the perspective of tie resilience; Benchmarking -- we evaluate existing tie strength prediction model performance using the BTS dataset collection, exploring the effects of different experiment settings, models, and evaluation criteria on the results. Furthermore, we derive key insights to enhance existing methods and shed light on promising directions for future research in this domain. The BTS dataset collection, along with the curation codes and experimental scripts, is all available at: https://github.com/XueqiC/Awesome-Tie-Strength-Prediction. | [
{
"version": "v1",
"created": "Thu, 24 Oct 2024 23:52:08 GMT"
},
{
"version": "v2",
"created": "Tue, 29 Oct 2024 04:37:15 GMT"
},
{
"version": "v3",
"created": "Tue, 10 Dec 2024 14:12:03 GMT"
},
{
"version": "v4",
"created": "Mon, 02 Jun 2025 04:04:58 GMT"
},
{
"version": "v5",
"created": "Sat, 07 Jun 2025 14:20:59 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Cheng",
"Xueqi",
""
],
[
"Yang",
"Catherine",
""
],
[
"Zhao",
"Yuying",
""
],
[
"Wang",
"Yu",
""
],
[
"Karimi",
"Hamid",
""
],
[
"Derr",
"Tyler",
""
]
] |
2410.19583 | Attila Jo\'o | Attila Jo\'o | Hindrance from a wasteful partial linkage | null | null | null | null | math.CO | http://creativecommons.org/licenses/by/4.0/ | Let $ D=(V,E) $ be a (possibly infinite) digraph and $ A,B\subseteq V $. A hindrance consists of an $ AB $-separator $ S $ together with a set of disjoint $ AS $-paths linking a proper subset of $ A $ onto $ S $. Hindrances and configurations guaranteeing the existence of hindrances play an essential role in the proof of the infinite version of Menger's theorem and are important in the context of certain open problems as well. This motivates the investigation of circumstances under which hindrances appear. In this paper we show that if there is a ``wasteful partial linkage'', i.e. a set $ \mathcal{P} $ of disjoint $ AB $-paths with fewer unused vertices in $ B $ than in $ A $, then there exists a hindrance. | [
{
"version": "v1",
"created": "Fri, 25 Oct 2024 14:23:12 GMT"
},
{
"version": "v2",
"created": "Sat, 07 Jun 2025 07:34:28 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Joó",
"Attila",
""
]
] |
2410.19789 | Silvia Seidlitz | Jan Sellner, Alexander Studier-Fischer, Ahmad Bin Qasim, Silvia Seidlitz, Nicholas Schreck, Minu Tizabi, Manuel Wiesenfarth, Annette Kopp-Schneider, Janne Heinecke, Jule Brandt, Samuel Kn\"odler, Caelan Max Haney, Gabriel Salg, Berkin \"Ozdemir, Maximilian Dietrich, Maurice Stephan Michel, Felix Nickel, Karl-Friedrich Kowalewski, Lena Maier-Hein | Xeno-learning: knowledge transfer across species in deep learning-based spectral image analysis | Jan Sellner and Alexander Studier-Fischer contributed equally to this work | null | null | null | cs.CV cs.AI cs.LG | http://creativecommons.org/licenses/by/4.0/ | Novel optical imaging techniques, such as hyperspectral imaging (HSI) combined with machine learning-based (ML) analysis, have the potential to revolutionize clinical surgical imaging. However, these novel modalities face a shortage of large-scale, representative clinical data for training ML algorithms, while preclinical animal data is abundantly available through standardized experiments and allows for controlled induction of pathological tissue states, which is not ethically possible in patients. To leverage this situation, we propose a novel concept called "xeno-learning", a cross-species knowledge transfer paradigm inspired by xeno-transplantation, where organs from a donor species are transplanted into a recipient species. Using a total of 13,874 HSI images from humans as well as porcine and rat models, we show that although spectral signatures of organs differ substantially across species, relative changes resulting from pathologies or surgical manipulation (e.g., malperfusion; injection of contrast agent) are comparable. Such changes learnt in one species can thus be transferred to a new species via a novel "physiology-based data augmentation" method, enabling the large-scale secondary use of preclinical animal data for humans. The resulting ethical, monetary, and performance benefits promise a high impact of the proposed knowledge transfer paradigm on future developments in the field. | [
{
"version": "v1",
"created": "Tue, 15 Oct 2024 16:25:16 GMT"
},
{
"version": "v2",
"created": "Sat, 07 Jun 2025 11:19:06 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Sellner",
"Jan",
""
],
[
"Studier-Fischer",
"Alexander",
""
],
[
"Qasim",
"Ahmad Bin",
""
],
[
"Seidlitz",
"Silvia",
""
],
[
"Schreck",
"Nicholas",
""
],
[
"Tizabi",
"Minu",
""
],
[
"Wiesenfarth",
"Manuel",
""
],
[
"Kopp-Schneider",
"Annette",
""
],
[
"Heinecke",
"Janne",
""
],
[
"Brandt",
"Jule",
""
],
[
"Knödler",
"Samuel",
""
],
[
"Haney",
"Caelan Max",
""
],
[
"Salg",
"Gabriel",
""
],
[
"Özdemir",
"Berkin",
""
],
[
"Dietrich",
"Maximilian",
""
],
[
"Michel",
"Maurice Stephan",
""
],
[
"Nickel",
"Felix",
""
],
[
"Kowalewski",
"Karl-Friedrich",
""
],
[
"Maier-Hein",
"Lena",
""
]
] |
2410.19934 | Tobias Oliver | Tobias G. Oliver, Eric G. Blackman, John A. Tarduno, and Michael A. Calkins | Turbulence in Earth's core generates large topographic torques on the mantle | 25 pages, 9 figures | null | null | null | physics.geo-ph | http://creativecommons.org/licenses/by/4.0/ | Seismic and geodynamic studies indicate that the boundary between the Earth's liquid outer core and solid mantle is not spherical, but is likely characterized by topography in the form of inverted mountains and valleys that have typical amplitudes of several kilometers. One of the dynamical consequences of these deformations is that turbulent flow in the core can exert pressure torques on the mantle, thereby resulting in a transfer of angular momentum between the outer core and the mantle. Understanding this transfer of angular momentum is important for explaining variations in the Earth's rotation rate, or length of day. Whether kilometer-sized topography can explain observed variations in length of day is a longstanding question in geophysics. Here we use a suite of state-of-the-art numerical simulations of hydrodynamic convection in a rotating spherical shell with boundary topography to show that topographic torques exhibit a linear dependence on topographic amplitude and approach a quadratic dependence on flow speeds. This observation is explained with the asymptotic theory of rapidly rotating convection. These results imply that topographic torques are of sufficient magnitude to explain length of day variations. | [
{
"version": "v1",
"created": "Fri, 25 Oct 2024 19:11:53 GMT"
},
{
"version": "v2",
"created": "Thu, 27 Mar 2025 19:59:25 GMT"
},
{
"version": "v3",
"created": "Mon, 09 Jun 2025 17:06:11 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Oliver",
"Tobias G.",
""
],
[
"Blackman",
"Eric G.",
""
],
[
"Tarduno",
"John A.",
""
],
[
"Calkins",
"Michael A.",
""
]
] |
2410.20159 | Xiaofeng Chen | Xiongtao Xiao, Xiaofeng Chen, Feiyan Jiang, Songming Zhang, Wenming Cao, Cheng Tan, Zhangyang Gao, Zhongshan Li | GATES: Graph Attention Network with Global Expression Fusion for Deciphering Spatial Transcriptome Architectures | null | null | null | null | q-bio.GN | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Single-cell spatial transcriptomics (ST) offers a unique approach to measuring gene expression profiles and spatial cell locations simultaneously. However, most existing ST methods assume that cells in closer spatial proximity exhibit more similar gene expression patterns. Such assumption typically results in graph structures that prioritize local spatial information while overlooking global patterns, limiting the ability to fully capture the broader structural features of biological tissues. To overcome this limitation, we propose GATES (Graph Attention neTwork with global Expression fuSion), a novel model designed to capture structural details in spatial transcriptomics data. GATES first constructs an expression graph that integrates local and global information by leveraging both spatial proximity and gene expression similarity. The model then employs an autoencoder with adaptive attention to assign proper weights for neighboring nodes, enhancing its capability of feature extraction. By fusing features of both the spatial and expression graphs, GATES effectively balances spatial context with gene expression data. Experimental results across multiple datasets demonstrate that GATES significantly outperforms existing methods in identifying spatial domains, highlighting its potential for analyzing complex biological tissues. Our code can be accessed on GitHub at https://github.com/xiaoxiongtao/GATES. | [
{
"version": "v1",
"created": "Sat, 26 Oct 2024 12:18:23 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Xiao",
"Xiongtao",
""
],
[
"Chen",
"Xiaofeng",
""
],
[
"Jiang",
"Feiyan",
""
],
[
"Zhang",
"Songming",
""
],
[
"Cao",
"Wenming",
""
],
[
"Tan",
"Cheng",
""
],
[
"Gao",
"Zhangyang",
""
],
[
"Li",
"Zhongshan",
""
]
] |
2410.20188 | Siyong Tao | Siyong Tao, Zida Xiao, Huaiqing Zuo | Bernstein-Sato functional equations for ideals in positive characteristic | null | null | null | null | math.AC math.AG | http://creativecommons.org/licenses/by/4.0/ | For an ideal of a regular $\cc$-algebra, its Bernstein-Sato polynomial is the monic polynomial of the lowest degree satisfying an Bernstein-Sato functional equation. We generalize the notion of Bernstein-Sato functional equations to the case of ideals in an $F$-finite ring of positive characteristic $p$, and show the relationship between these equations and Bernstein-Sato roots. By applying this theory, we provide an explicit description of Bernstein-Sato roots of a weighted homogeneous polynomial with an isolated singularity at the origin in characteristic $p$. Moreover, we give multiplicative and additive Thom-Sebastiani properties for the set of Bernstein-Sato roots, which prove the characteristic $p$ analogue of Budur and Popa's question. | [
{
"version": "v1",
"created": "Sat, 26 Oct 2024 14:29:22 GMT"
},
{
"version": "v2",
"created": "Sun, 08 Jun 2025 14:13:47 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Tao",
"Siyong",
""
],
[
"Xiao",
"Zida",
""
],
[
"Zuo",
"Huaiqing",
""
]
] |
2410.20250 | Amir Najafi | Amir Najafi, Samin Mahdizadeh Sani, Farzan Farnia | Certifiably Robust Model Evaluation in Federated Learning under Meta-Distributional Shifts | Published at the International Conference on Machine Learning (ICML) 2025. 36 pages, 10 figures | null | null | null | stat.ML cs.LG | http://creativecommons.org/licenses/by/4.0/ | We address the challenge of certifying the performance of a federated learning model on an unseen target network using only measurements from the source network that trained the model. Specifically, consider a source network "A" with $K$ clients, each holding private, non-IID datasets drawn from heterogeneous distributions, modeled as samples from a broader meta-distribution $\mu$. Our goal is to provide certified guarantees for the model's performance on a different, unseen network "B", governed by an unknown meta-distribution $\mu'$, assuming the deviation between $\mu$ and $\mu'$ is bounded either in Wasserstein distance or an $f$-divergence. We derive worst-case uniform guarantees for both the model's average loss and its risk CDF, the latter corresponding to a novel, adversarially robust version of the Dvoretzky-Kiefer-Wolfowitz (DKW) inequality. In addition, we show how the vanilla DKW bound enables principled certification of the model's true performance on unseen clients within the same (source) network. Our bounds are efficiently computable, asymptotically minimax optimal, and preserve clients' privacy. We also establish non-asymptotic generalization bounds that converge to zero as $K$ grows and the minimum per-client sample size exceeds $\mathcal{O}(\log K)$. Empirical evaluations confirm the practical utility of our bounds across real-world tasks. The project code is available at: github.com/samin-mehdizadeh/Robust-Evaluation-DKW | [
{
"version": "v1",
"created": "Sat, 26 Oct 2024 18:45:15 GMT"
},
{
"version": "v2",
"created": "Sat, 07 Jun 2025 01:54:52 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Najafi",
"Amir",
""
],
[
"Sani",
"Samin Mahdizadeh",
""
],
[
"Farnia",
"Farzan",
""
]
] |
2410.20580 | Yan Zhang | Zongyi Xiang, Yan Zhang, Lixin Duan, Hongzhi Yin, Ivor W. Tsang | Coherence-guided Preference Disentanglement for Cross-domain Recommendations | 28 pages | null | null | null | cs.IR | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Discovering user preferences across different domains is pivotal in cross-domain recommendation systems, particularly when platforms lack comprehensive user-item interactive data. The limited presence of shared users often hampers the effective modeling of common preferences. While leveraging shared items' attributes, such as category and popularity, can enhance cross-domain recommendation performance, the scarcity of shared items between domains has limited research in this area. To address this, we propose a Coherence-guided Preference Disentanglement (CoPD) method aimed at improving cross-domain recommendation by i) explicitly extracting shared item attributes to guide the learning of shared user preferences and ii) disentangling these preferences to identify specific user interests transferred between domains. CoPD introduces coherence constraints on item embeddings of shared and specific domains, aiding in extracting shared attributes. Moreover, it utilizes these attributes to guide the disentanglement of user preferences into separate embeddings for interest and conformity through a popularity-weighted loss. Experiments conducted on real-world datasets demonstrate the superior performance of our proposed CoPD over existing competitive baselines, highlighting its effectiveness in enhancing cross-domain recommendation performance. | [
{
"version": "v1",
"created": "Sun, 27 Oct 2024 20:21:14 GMT"
},
{
"version": "v2",
"created": "Mon, 09 Jun 2025 12:41:13 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Xiang",
"Zongyi",
""
],
[
"Zhang",
"Yan",
""
],
[
"Duan",
"Lixin",
""
],
[
"Yin",
"Hongzhi",
""
],
[
"Tsang",
"Ivor W.",
""
]
] |
2410.20734 | Saverio Spagnolie | Jingyi Li, Laurel Ohm and Saverio E. Spagnolie | Arrested development and traveling waves of active suspensions in nematic liquid crystals | 5 figures | null | null | null | cond-mat.soft q-bio.CB | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Active particles in anisotropic, viscoelastic fluids experience competing stresses which guide their trajectories. An aligned suspension of particles can trigger a hydrodynamic bend instability, but the elasticity of the fluid can drive particle orientations back towards alignment. To study these competing effects, we examine a dilute suspension of active particles in an Ericksen-Leslie model nematic liquid crystal. An anchoring strength linking the active and passive media tunes the system between active suspension theory in Newtonian fluids in one limit, and active nematic theory in another. For extensile active stresses, beyond a critical active Ericksen number or particle concentration, the suspension first comes into alignment, then buckles via a classical bend instability. Rather than entering the fully developed roiling state observed in isotropic fluids, the development is arrested into a steady, flowing state by fluid elasticity. Arrested states of higher wavenumber appear at yet larger extensile activity, and a phase transition is identified at finite anchoring strength. If the active particles are motile, the particles can surf along the bent environment of their own creation. More exotic states are also observed, including a oscillatory 'thrashing' mode. Moment equations are derived, compared to kinetic theory simulations, and analyzed in asymptotic limits which admit exact expressions for the traveling wave speed and both particle orientation and director fields in the first arrested state. | [
{
"version": "v1",
"created": "Mon, 28 Oct 2024 04:52:47 GMT"
},
{
"version": "v2",
"created": "Tue, 18 Feb 2025 17:59:37 GMT"
},
{
"version": "v3",
"created": "Sun, 08 Jun 2025 15:59:52 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Li",
"Jingyi",
""
],
[
"Ohm",
"Laurel",
""
],
[
"Spagnolie",
"Saverio E.",
""
]
] |
2410.20950 | Joni Laulainen | Ilkka Helenius, Joni O. Laulainen, Christian T. Preuss | Multi-Jet Production in Deep Inelastic Scattering with Pythia | 28 pages, 12 figures | J. High Energ. Phys. 2025, 153 (2025) | 10.1007/JHEP05(2025)153 | null | hep-ph | http://creativecommons.org/licenses/by-nc-nd/4.0/ | We introduce multi-jet merging for deep inelastic scattering in the Vincia parton shower in the Monte Carlo event generator Pythia 8. Merging combines event samples of different parton multiplicities with logarithmically enhanced parton-shower radiation. We consider up to five outgoing partons using two different merging algorithms. We vary the relevant scale choices and compare to experimental data by H1 collaboration at the HERA collider. Results show that multi-jet merging improves description to data, especially for low virtuality events, and that jet cross-sections converge when the partonic multiplicity exceeds the number of reconstructed jets. | [
{
"version": "v1",
"created": "Mon, 28 Oct 2024 12:10:25 GMT"
},
{
"version": "v2",
"created": "Tue, 14 Jan 2025 16:19:28 GMT"
},
{
"version": "v3",
"created": "Mon, 09 Jun 2025 14:47:26 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Helenius",
"Ilkka",
""
],
[
"Laulainen",
"Joni O.",
""
],
[
"Preuss",
"Christian T.",
""
]
] |
2410.21225 | Eishica Chand | Eishica Chand, Arka Banerjee, Simon Foreman and Francisco Villaescusa-Navarro | Boosting HI-Galaxy Cross-Clustering Signal through Higher-Order Cross-Correlations | 14 pages, 9 figures (with 2 figures included in the appendix), 2 tables. Published in MNRAS | null | 10.1093/mnras/staf433 | null | astro-ph.CO | http://creativecommons.org/licenses/by/4.0/ | After reionization, neutral hydrogen (HI) traces the large-scale structure (LSS) of the Universe, enabling HI intensity mapping (IM) to capture the LSS in 3D and constrain key cosmological parameters. We present a new framework utilizing higher-order cross-correlations to study HI clustering around galaxies, tested using real-space data from the IllustrisTNG300 simulation. This approach computes the joint distributions of $k$-nearest neighbor ($k$NN) optical galaxies and the HI brightness temperature field smoothed at relevant scales (the $k$NN-field framework), providing sensitivity to all higher-order cross-correlations, unlike two-point statistics. To simulate HI data from actual surveys, we add random thermal noise and apply a simple foreground cleaning model, filtering out Fourier modes of the brightness temperature field with $k_\parallel < k_{\rm min,\parallel}$. Under current levels of thermal noise and foreground cleaning, typical of a Canadian Hydrogen Intensity Mapping Experiment (CHIME)-like survey, the HI-galaxy cross-correlation signal in our simulations, using the $k$NN-field framework, is detectable at $>30\sigma$ across $r = [3,12] \, h^{-1}$Mpc. In contrast, the detectability of the standard two-point correlation function (2PCF) over the same scales depends strongly on the foreground filter: a sharp $k_\parallel$ filter can spuriously boost detection to $8\sigma$ due to position-space ringing, whereas a less sharp filter yields no detection. Nonetheless, we conclude that $k$NN-field cross-correlations are robustly detectable across a broad range of foreground filtering and thermal noise conditions, suggesting their potential for enhanced constraining power over 2PCFs. | [
{
"version": "v1",
"created": "Mon, 28 Oct 2024 17:10:56 GMT"
},
{
"version": "v2",
"created": "Mon, 09 Jun 2025 11:52:53 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Chand",
"Eishica",
""
],
[
"Banerjee",
"Arka",
""
],
[
"Foreman",
"Simon",
""
],
[
"Villaescusa-Navarro",
"Francisco",
""
]
] |
2410.21527 | Benjamin Brindle | Benjamin Brindle, Thomas Derrick Hull, Matteo Malgaroli, Nicolas Charon | VISTA-SSM: Varying and Irregular Sampling Time-series Analysis via State Space Models | null | null | null | null | stat.AP | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We introduce VISTA, a clustering approach for multivariate and irregularly sampled time series based on a parametric state space mixture model. VISTA is specifically designed for the unsupervised identification of groups in datasets originating from healthcare and psychology where such sampling issues are commonplace. Our approach adapts linear Gaussian state space models (LGSSMs) to provide a flexible parametric framework for fitting a wide range of time series dynamics. The clustering approach itself is based on the assumption that the population can be represented as a mixture of a fixed number of LGSSMs. VISTA's model formulation allows for an explicit derivation of the log-likelihood function, from which we develop an expectation-maximization scheme for fitting model parameters to the observed data samples. Our algorithmic implementation is designed to handle populations of multivariate time series that can exhibit large changes in sampling rate as well as irregular sampling. We evaluate the versatility and accuracy of our approach on simulated and real-world datasets, including demographic trends, wearable sensor data, epidemiological time series, and ecological momentary assessments. Our results indicate that VISTA outperforms most comparable standard times series clustering methods. We provide an open-source implementation of VISTA in Python. | [
{
"version": "v1",
"created": "Mon, 28 Oct 2024 20:54:05 GMT"
},
{
"version": "v2",
"created": "Sun, 08 Jun 2025 19:28:33 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Brindle",
"Benjamin",
""
],
[
"Hull",
"Thomas Derrick",
""
],
[
"Malgaroli",
"Matteo",
""
],
[
"Charon",
"Nicolas",
""
]
] |
2410.21842 | Yiming Ji | Yiming Ji, Kaijie Yun, Yang Liu, Zhengpu Wang, Boyu Ma, Zongwu Xie, and Hong Liu | Diffusion as Reasoning: Enhancing Object Navigation via Diffusion Model Conditioned on LLM-based Object-Room Knowledge | null | null | null | null | cs.CV cs.AI | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | The Object Navigation (ObjectNav) task aims to guide an agent to locate target objects in unseen environments using partial observations. Prior approaches have employed location prediction paradigms to achieve long-term goal reasoning, yet these methods often struggle to effectively integrate contextual relation reasoning. Alternatively, map completion-based paradigms predict long-term goals by generating semantic maps of unexplored areas. However, existing methods in this category fail to fully leverage known environmental information, resulting in suboptimal map quality that requires further improvement. In this work, we propose a novel approach to enhancing the ObjectNav task, by training a diffusion model to learn the statistical distribution patterns of objects in semantic maps, and using the map of the explored regions during navigation as the condition to generate the map of the unknown regions, thereby realizing the long-term goal reasoning of the target object, i.e., diffusion as reasoning (DAR). Meanwhile, we propose the Room Guidance method, which leverages commonsense knowledge derived from large language models (LLMs) to guide the diffusion model in generating room-aware object distributions. Based on the generated map in the unknown region, the agent sets the predicted location of the target as the goal and moves towards it. Experiments on Gibson and MP3D show the effectiveness of our method. | [
{
"version": "v1",
"created": "Tue, 29 Oct 2024 08:10:06 GMT"
},
{
"version": "v2",
"created": "Fri, 06 Jun 2025 02:18:14 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Ji",
"Yiming",
""
],
[
"Yun",
"Kaijie",
""
],
[
"Liu",
"Yang",
""
],
[
"Wang",
"Zhengpu",
""
],
[
"Ma",
"Boyu",
""
],
[
"Xie",
"Zongwu",
""
],
[
"Liu",
"Hong",
""
]
] |
2410.22351 | Xue-Ping Wang | Feng-qing Zhu and Xue-Ping Wang | The characterization of general pseudo-homogeneous t-norms | 18 | Fuzzy Sets and Systems 517 (2025) 109459 | 10.1016/j.fss.2025.109459 | 13 | math.GM | http://creativecommons.org/licenses/by-nc-nd/4.0/ | This article first introduces the concept of a general pseudo-homogeneous triangular norm. It then gives some properties of general pseudo-homogeneous triangular norms. Finally, it characterizes all general pseudo-homogeneous triangular norms completely. | [
{
"version": "v1",
"created": "Tue, 15 Oct 2024 06:40:57 GMT"
},
{
"version": "v2",
"created": "Thu, 07 Nov 2024 02:59:14 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Zhu",
"Feng-qing",
""
],
[
"Wang",
"Xue-Ping",
""
]
] |
2410.22405 | Jannis Gebhard | Jannis Gebhard, Aleksas Mazeliauskas, Adam Takacs | No-quenching baseline for energy loss signals in oxygen-oxygen collisions | 26 pages, 15 figures. v2: Added nNNPDF3.0 results at LO | JHEP 04 (2025) 034 | 10.1007/JHEP04(2025)034 | null | hep-ph nucl-ex nucl-th | http://creativecommons.org/licenses/by/4.0/ | In this work, we perform computations of inclusive jet, and semi-inclusive jet-hadron cross sections for minimum bias oxygen-oxygen collisions at RHIC and LHC collision energies. We compute the no-quenching baseline for the jet nuclear modification factor $R_\mathrm{AA}$ and jet-, and hadron-triggered semi-inclusive nuclear modification factors $I_\mathrm{AA}$. We do this with state-of-the-art nuclear parton distribution functions (nPDFs), next-to-leading-order matrix elements, parton shower, and hadronization. We observe deviations from unity due to cold-nuclear matter effects, even without quenching. We demonstrate that the parton distribution uncertainties constitute a significant obstacle in detecting energy loss in small collision systems. Hadron-triggered observables are particularly sensitive to uncertainties due to correlations between the trigger and analyzed particles. For jet-triggered $I_\mathrm{AA}$, there exists a kinematic window in which nPDF, scale, shower, and hadronization uncertainties cancel down to percent level, overcoming a major limiting factor for energy loss discovery in small systems. | [
{
"version": "v1",
"created": "Tue, 29 Oct 2024 18:00:01 GMT"
},
{
"version": "v2",
"created": "Thu, 12 Dec 2024 13:50:13 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Gebhard",
"Jannis",
""
],
[
"Mazeliauskas",
"Aleksas",
""
],
[
"Takacs",
"Adam",
""
]
] |
2410.22430 | Robert Shrock | Yue Chen and Robert Shrock | Exact Potts/Tutte Polynomials for Hammock Chain Graphs | 57 pages, latex, 26 figures | Journal of Statistical Physics 192, 78 (2025) | 10.1007/s10955-025-03457-2 | YITP-SB-2024-15 | cond-mat.stat-mech | http://creativecommons.org/licenses/by/4.0/ | We present exact calculations of the $q$-state Potts model partition functions and the equivalent Tutte polynomials for chain graphs comprised of $m$ repeated hammock subgraphs $H_{e_1,...,e_r}$ connected with line graphs of length $e_g$ edges, such that the chains have open or cyclic boundary conditions (BC). Here, $H_{e_1,...,e_r}$ is a hammock (series-parallel) subgraph with $r$ separate paths along ``ropes'' with respective lengths $e_1, ..., e_r$ edges, connecting the two end vertices. We denote the resultant chain graph as $G_{\{e_1,...,e_r\},e_g,m;BC}$. We discuss special cases, including chromatic, flow, and reliability polynomials. In the case of cyclic boundary conditions, the zeros of the Potts partition function in the complex $q$ function accumulate, in the limit $m \to \infty$, onto curves forming a locus ${\cal B}$, and we study this locus. | [
{
"version": "v1",
"created": "Tue, 29 Oct 2024 18:17:27 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Chen",
"Yue",
""
],
[
"Shrock",
"Robert",
""
]
] |
2410.22656 | Tian Li | Tian Li, Tianyi Zhou, Jeffrey A. Bilmes | Tilted Sharpness-Aware Minimization | Accepted by ICML 2025 | null | null | null | cs.LG | http://creativecommons.org/licenses/by/4.0/ | Sharpness-Aware Minimization (SAM) has been demonstrated to improve the generalization performance of overparameterized models by seeking flat minima on the loss landscape through optimizing model parameters that incur the largest loss within a neighborhood. Nevertheless, such min-max formulations are computationally challenging especially when the problem is highly non-convex. Additionally, focusing only on the worst-case local solution while ignoring potentially many other local solutions may be suboptimal when searching for flat minima. In this work, we propose Tilted SAM (TSAM), a smoothed generalization of SAM inspired by exponential tilting that effectively assigns higher priority to local solutions that incur larger losses. TSAM is parameterized by a tilt hyperparameter $t$ and reduces to SAM as $t$ approaches infinity. We show that TSAM is smoother than SAM and thus easier to optimize, and it explicitly favors flatter minima. We develop algorithms motivated by the discretization of Hamiltonian dynamics to solve TSAM. Empirically, TSAM arrives at flatter local minima and results in superior test performance than the baselines of SAM and ERM across a range of image and text tasks. | [
{
"version": "v1",
"created": "Wed, 30 Oct 2024 02:49:48 GMT"
},
{
"version": "v2",
"created": "Sun, 08 Jun 2025 16:30:11 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Li",
"Tian",
""
],
[
"Zhou",
"Tianyi",
""
],
[
"Bilmes",
"Jeffrey A.",
""
]
] |
2410.22889 | Francesco Macheda | Giovanni Caldarelli, Alberto Guandalini, Francesco Macheda, Francesco Mauri | Variational formulation of dynamical electronic response functions in presence of nonlocal exchange interactions | 23 pages, 3 figures. Reuploaded with CC BY licence | Physical Review B 111, 075137 (2025) | 10.1103/PhysRevB.111.075137 | null | cond-mat.mtrl-sci cond-mat.other cond-mat.str-el | http://creativecommons.org/licenses/by/4.0/ | We consider the dynamical electronic response function in theoretical frameworks that include nonlocal exchange interactions, such as the Bethe-Salpeter equation with the frequency independent approximation of the screened interaction, Hartree-Fock, and range-separated Hybrid DFT approaches. Within these pictures, we demonstrate that any time-dependent electronic linear response function allows for a formulation which is variational in the electronic density matrix. To achieve our goal, we consider the usual form of a response function, written in terms of a screened and a bare electronic vertices (`bare-screen'), and perform an exact rewriting in terms of purely screened electronic vertices (`screen-screen'). Within the `screen-screen' formulation, the response function can be written as a stationary point of a functional of the exact density matrix. Further, we show that the imaginary part of any electronic response can be written in the form of a generalized Fermi Golden Rule, by introducing an exact complementary rewriting in terms of vertices related by complex conjugation (`screen*-screen'). The screen-screen formulation can be further extended partitioning the electronic interaction in separate contributions, expressing the response in terms of partially screened electronic vertices (`partial screen-partial screen'), preserving the stationary properties. We numerically validate the effectiveness of our formalism by calculating the optical conductivity of graphene, which exhibits strong excitonic effects. To do so, we solve the Bethe-Salpeter Equation on a tight-binding model, including exchange effects in the response of graphene. Our findings show the advantages of the variationality of the screen-screen formulation over the others both in convergence properties and robustness with density-matrix approximations. | [
{
"version": "v1",
"created": "Wed, 30 Oct 2024 10:33:47 GMT"
},
{
"version": "v2",
"created": "Thu, 14 Nov 2024 15:15:36 GMT"
},
{
"version": "v3",
"created": "Mon, 20 Jan 2025 17:23:12 GMT"
},
{
"version": "v4",
"created": "Wed, 19 Feb 2025 09:41:17 GMT"
},
{
"version": "v5",
"created": "Sat, 07 Jun 2025 14:09:28 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Caldarelli",
"Giovanni",
""
],
[
"Guandalini",
"Alberto",
""
],
[
"Macheda",
"Francesco",
""
],
[
"Mauri",
"Francesco",
""
]
] |
2410.22931 | Thien-Minh Nguyen | Thien-Minh Nguyen, Ziyu Cao, Kailai Li, William Talbot, Tongxing Jin, Shenghai Yuan, Timothy D. Barfoot, Lihua Xie | A Third-Order Gaussian Process Trajectory Representation Framework with Closed-Form Kinematics for Continuous-Time Motion Estimation | The source code has been released. All feedbacks are welcome | null | null | null | cs.RO | http://creativecommons.org/licenses/by-nc-nd/4.0/ | In this paper, we propose a third-order, i.e., white-noise-on-jerk, Gaussian Process (GP) Trajectory Representation (TR) framework for continuous-time (CT) motion estimation (ME) tasks. Our framework features a unified trajectory representation that encapsulates the kinematic models of both $SO(3)\times\mathbb{R}^3$ and $SE(3)$ pose representations. This encapsulation strategy allows users to use the same implementation of measurement-based factors for either choice of pose representation, which facilitates experimentation and comparison to achieve the best model for the ME task. In addition, unique to our framework, we derive the kinematic models with the closed-form temporal derivatives of the local variable of $SO(3)$ and $SE(3)$, which so far has only been approximated based on the Taylor expansion in the literature. Our experiments show that these kinematic models can improve the estimation accuracy in high-speed scenarios. All analytical Jacobians of the interpolated states with respect to the support states of the trajectory representation, as well as the motion prior factors, are also provided for accelerated Gauss-Newton (GN) optimization. Our experiments demonstrate the efficacy and efficiency of the framework in various motion estimation tasks such as localization, calibration, and odometry, facilitating fast prototyping for ME researchers. We release the source code for the benefit of the community. Our project is available at https://github.com/brytsknguyen/gptr. | [
{
"version": "v1",
"created": "Wed, 30 Oct 2024 11:37:47 GMT"
},
{
"version": "v2",
"created": "Thu, 31 Oct 2024 15:04:25 GMT"
},
{
"version": "v3",
"created": "Sat, 02 Nov 2024 06:09:23 GMT"
},
{
"version": "v4",
"created": "Sat, 07 Jun 2025 11:35:11 GMT"
}
] | 2025-06-10T00:00:00 | [
[
"Nguyen",
"Thien-Minh",
""
],
[
"Cao",
"Ziyu",
""
],
[
"Li",
"Kailai",
""
],
[
"Talbot",
"William",
""
],
[
"Jin",
"Tongxing",
""
],
[
"Yuan",
"Shenghai",
""
],
[
"Barfoot",
"Timothy D.",
""
],
[
"Xie",
"Lihua",
""
]
] |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.