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 |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
2412.15484 | Saehyung Lee | Saehyung Lee, Seunghyun Yoon, Trung Bui, Jing Shi, Sungroh Yoon | Toward Robust Hyper-Detailed Image Captioning: A Multiagent Approach and Dual Evaluation Metrics for Factuality and Coverage | ICML 2025 | null | null | null | cs.CV | http://creativecommons.org/licenses/by/4.0/ | Multimodal large language models (MLLMs) excel at generating highly detailed captions but often produce hallucinations. Our analysis reveals that existing hallucination detection methods struggle with detailed captions. We attribute this to the increasing reliance of MLLMs on their generated text, rather than the input image, as the sequence length grows. To address this issue, we propose a multiagent approach that leverages LLM-MLLM collaboration to correct given captions. Additionally, we introduce an evaluation framework and a benchmark dataset to facilitate the systematic analysis of detailed captions. Our experiments demonstrate that our proposed evaluation method better aligns with human judgments of factuality than existing metrics and that existing approaches to improve the MLLM factuality may fall short in hyper-detailed image captioning tasks. In contrast, our proposed method significantly enhances the factual accuracy of captions, even improving those generated by GPT-4V. Finally, we highlight a limitation of VQA-centric benchmarking by demonstrating that an MLLM's performance on VQA benchmarks may not correlate with its ability to generate detailed image captions. | [
{
"version": "v1",
"created": "Fri, 20 Dec 2024 01:37:22 GMT"
},
{
"version": "v2",
"created": "Tue, 24 Dec 2024 04:42:49 GMT"
},
{
"version": "v3",
"created": "Thu, 29 May 2025 13:54:51 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Lee",
"Saehyung",
""
],
[
"Yoon",
"Seunghyun",
""
],
[
"Bui",
"Trung",
""
],
[
"Shi",
"Jing",
""
],
[
"Yoon",
"Sungroh",
""
]
] |
2412.15819 | Cheng Wang | Cheng Wang, Ziyang Feng, Pin Zhang, Manjiang Cao, Yiming Yuan, Tengfei Chang | Robustness-enhanced Myoelectric Control with GAN-based Open-set Recognition | 11 pages, 14 figures | null | null | null | cs.CV cs.HC eess.SP | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Electromyography (EMG) signals are widely used in human motion recognition and medical rehabilitation, yet their variability and susceptibility to noise significantly limit the reliability of myoelectric control systems. Existing recognition algorithms often fail to handle unfamiliar actions effectively, leading to system instability and errors. This paper proposes a novel framework based on Generative Adversarial Networks (GANs) to enhance the robustness and usability of myoelectric control systems by enabling open-set recognition. The method incorporates a GAN-based discriminator to identify and reject unknown actions, maintaining system stability by preventing misclassifications. Experimental evaluations on publicly available and self-collected datasets demonstrate a recognition accuracy of 97.6\% for known actions and a 23.6\% improvement in Active Error Rate (AER) after rejecting unknown actions. The proposed approach is computationally efficient and suitable for deployment on edge devices, making it practical for real-world applications. | [
{
"version": "v1",
"created": "Fri, 20 Dec 2024 12:01:01 GMT"
},
{
"version": "v2",
"created": "Thu, 29 May 2025 14:39:27 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Wang",
"Cheng",
""
],
[
"Feng",
"Ziyang",
""
],
[
"Zhang",
"Pin",
""
],
[
"Cao",
"Manjiang",
""
],
[
"Yuan",
"Yiming",
""
],
[
"Chang",
"Tengfei",
""
]
] |
2412.16039 | Jiadong Pan | Jiadong Pan, Liang Li, Hongcheng Gao, Zheng-Jun Zha, Qingming Huang, Jiebo Luo | SafeCFG: Controlling Harmful Features with Dynamic Safe Guidance for Safe Generation | null | null | null | null | cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Diffusion models (DMs) have demonstrated exceptional performance in text-to-image tasks, leading to their widespread use. With the introduction of classifier-free guidance (CFG), the quality of images generated by DMs is significantly improved. However, one can use DMs to generate more harmful images by maliciously guiding the image generation process through CFG. Existing safe alignment methods aim to mitigate the risk of generating harmful images but often reduce the quality of clean image generation. To address this issue, we propose SafeCFG to adaptively control harmful features with dynamic safe guidance by modulating the CFG generation process. It dynamically guides the CFG generation process based on the harmfulness of the prompts, inducing significant deviations only in harmful CFG generations, achieving high quality and safety generation. SafeCFG can simultaneously modulate different harmful CFG generation processes, so it could eliminate harmful elements while preserving high-quality generation. Additionally, SafeCFG provides the ability to detect image harmfulness, allowing unsupervised safe alignment on DMs without pre-defined clean or harmful labels. Experimental results show that images generated by SafeCFG achieve both high quality and safety, and safe DMs trained in our unsupervised manner also exhibit good safety performance. | [
{
"version": "v1",
"created": "Fri, 20 Dec 2024 16:40:11 GMT"
},
{
"version": "v2",
"created": "Thu, 29 May 2025 12:01:59 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Pan",
"Jiadong",
""
],
[
"Li",
"Liang",
""
],
[
"Gao",
"Hongcheng",
""
],
[
"Zha",
"Zheng-Jun",
""
],
[
"Huang",
"Qingming",
""
],
[
"Luo",
"Jiebo",
""
]
] |
2412.16197 | Wenhui Cui | Wenhui Cui, Haleh Akrami, Anand A. Joshi, Richard M. Leahy | Generalizable Representation Learning for fMRI-based Neurological Disorder Identification | Accepted by TMLR | null | null | null | eess.IV cs.CE cs.CV cs.LG | http://creativecommons.org/licenses/by-nc-nd/4.0/ | Despite the impressive advances achieved using deep learning for functional brain activity analysis, the heterogeneity of functional patterns and the scarcity of imaging data still pose challenges in tasks such as identifying neurological disorders. For functional Magnetic Resonance Imaging (fMRI), while data may be abundantly available from healthy controls, clinical data is often scarce, especially for rare diseases, limiting the ability of models to identify clinically-relevant features. We overcome this limitation by introducing a novel representation learning strategy integrating meta-learning with self-supervised learning to improve the generalization from normal to clinical features. This approach enables generalization to challenging clinical tasks featuring scarce training data. We achieve this by leveraging self-supervised learning on the control dataset to focus on inherent features that are not limited to a particular supervised task and incorporating meta-learning to improve the generalization across domains. To explore the generalizability of the learned representations to unseen clinical applications, we apply the model to four distinct clinical datasets featuring scarce and heterogeneous data for neurological disorder classification. Results demonstrate the superiority of our representation learning strategy on diverse clinically-relevant tasks. Code is publicly available at https://github.com/wenhui0206/MeTSK/tree/main | [
{
"version": "v1",
"created": "Mon, 16 Dec 2024 22:07:35 GMT"
},
{
"version": "v2",
"created": "Wed, 28 May 2025 18:27:12 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Cui",
"Wenhui",
""
],
[
"Akrami",
"Haleh",
""
],
[
"Joshi",
"Anand A.",
""
],
[
"Leahy",
"Richard M.",
""
]
] |
2412.16359 | Nilanjana Das | Nilanjana Das, Edward Raff, Aman Chadha, Manas Gaur | Human-Readable Adversarial Prompts: An Investigation into LLM Vulnerabilities Using Situational Context | arXiv admin note: text overlap with arXiv:2407.14644 | null | null | null | cs.CL cs.AI | http://creativecommons.org/licenses/by/4.0/ | As the AI systems become deeply embedded in social media platforms, we've uncovered a concerning security vulnerability that goes beyond traditional adversarial attacks. It becomes important to assess the risks of LLMs before the general public use them on social media platforms to avoid any adverse impacts. Unlike obvious nonsensical text strings that safety systems can easily catch, our work reveals that human-readable situation-driven adversarial full-prompts that leverage situational context are effective but much harder to detect. We found that skilled attackers can exploit the vulnerabilities in open-source and proprietary LLMs to make a malicious user query safe for LLMs, resulting in generating a harmful response. This raises an important question about the vulnerabilities of LLMs. To measure the robustness against human-readable attacks, which now present a potent threat, our research makes three major contributions. First, we developed attacks that use movie scripts as situational contextual frameworks, creating natural-looking full-prompts that trick LLMs into generating harmful content. Second, we developed a method to transform gibberish adversarial text into readable, innocuous content that still exploits vulnerabilities when used within the full-prompts. Finally, we enhanced the AdvPrompter framework with p-nucleus sampling to generate diverse human-readable adversarial texts that significantly improve attack effectiveness against models like GPT-3.5-Turbo-0125 and Gemma-7b. Our findings show that these systems can be manipulated to operate beyond their intended ethical boundaries when presented with seemingly normal prompts that contain hidden adversarial elements. By identifying these vulnerabilities, we aim to drive the development of more robust safety mechanisms that can withstand sophisticated attacks in real-world applications. | [
{
"version": "v1",
"created": "Fri, 20 Dec 2024 21:43:52 GMT"
},
{
"version": "v2",
"created": "Tue, 11 Mar 2025 21:41:19 GMT"
},
{
"version": "v3",
"created": "Thu, 29 May 2025 05:54:54 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Das",
"Nilanjana",
""
],
[
"Raff",
"Edward",
""
],
[
"Chadha",
"Aman",
""
],
[
"Gaur",
"Manas",
""
]
] |
2412.16555 | Yanxu Mao | Yanxu Mao, Peipei Liu, Tiehan Cui, Zhaoteng Yan, Congying Liu, Datao You | Divide and Conquer: A Hybrid Strategy Defeats Multimodal Large Language Models | null | null | null | null | cs.CL | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Large language models (LLMs) are widely applied in various fields of society due to their powerful reasoning, understanding, and generation capabilities. However, the security issues associated with these models are becoming increasingly severe. Jailbreaking attacks, as an important method for detecting vulnerabilities in LLMs, have been explored by researchers who attempt to induce these models to generate harmful content through various attack methods. Nevertheless, existing jailbreaking methods face numerous limitations, such as excessive query counts, limited coverage of jailbreak modalities, low attack success rates, and simplistic evaluation methods. To overcome these constraints, this paper proposes a multimodal jailbreaking method: JMLLM. This method integrates multiple strategies to perform comprehensive jailbreak attacks across text, visual, and auditory modalities. Additionally, we contribute a new and comprehensive dataset for multimodal jailbreaking research: TriJail, which includes jailbreak prompts for all three modalities. Experiments on the TriJail dataset and the benchmark dataset AdvBench, conducted on 13 popular LLMs, demonstrate advanced attack success rates and significant reduction in time overhead. | [
{
"version": "v1",
"created": "Sat, 21 Dec 2024 09:43:51 GMT"
},
{
"version": "v2",
"created": "Thu, 22 May 2025 13:27:09 GMT"
},
{
"version": "v3",
"created": "Thu, 29 May 2025 14:05:50 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Mao",
"Yanxu",
""
],
[
"Liu",
"Peipei",
""
],
[
"Cui",
"Tiehan",
""
],
[
"Yan",
"Zhaoteng",
""
],
[
"Liu",
"Congying",
""
],
[
"You",
"Datao",
""
]
] |
2412.17031 | Lovisa Hagstr\"om | Lovisa Hagstr\"om, Sara Vera Marjanovi\'c, Haeun Yu, Arnav Arora, Christina Lioma, Maria Maistro, Pepa Atanasova, Isabelle Augenstein | A Reality Check on Context Utilisation for Retrieval-Augmented Generation | Accepted at ACL 2025 | null | null | null | cs.CL cs.AI | http://creativecommons.org/licenses/by/4.0/ | Retrieval-augmented generation (RAG) helps address the limitations of parametric knowledge embedded within a language model (LM). In real world settings, retrieved information can vary in complexity, yet most investigations of LM utilisation of context has been limited to synthetic text. We introduce DRUID (Dataset of Retrieved Unreliable, Insufficient and Difficult-to-understand contexts) with real-world queries and contexts manually annotated for stance. The dataset is based on the prototypical task of automated claim verification, for which automated retrieval of real-world evidence is crucial. We compare DRUID to synthetic datasets (CounterFact, ConflictQA) and find that artificial datasets often fail to represent the complexity and diversity of realistically retrieved context. We show that synthetic datasets exaggerate context characteristics rare in real retrieved data, which leads to inflated context utilisation results, as measured by our novel ACU score. Moreover, while previous work has mainly focused on singleton context characteristics to explain context utilisation, correlations between singleton context properties and ACU on DRUID are surprisingly small compared to other properties related to context source. Overall, our work underscores the need for real-world aligned context utilisation studies to represent and improve performance in real-world RAG settings. | [
{
"version": "v1",
"created": "Sun, 22 Dec 2024 14:16:38 GMT"
},
{
"version": "v2",
"created": "Thu, 29 May 2025 09:10:25 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Hagström",
"Lovisa",
""
],
[
"Marjanović",
"Sara Vera",
""
],
[
"Yu",
"Haeun",
""
],
[
"Arora",
"Arnav",
""
],
[
"Lioma",
"Christina",
""
],
[
"Maistro",
"Maria",
""
],
[
"Atanasova",
"Pepa",
""
],
[
"Augenstein",
"Isabelle",
""
]
] |
2412.17333 | Jaeheun Jung | Jaeheun Jung, Jaehyuk Lee, Changhae Jung, Hanyoung Kim, Bosung Jung, Donghun Lee | Broadband Ground Motion Synthesis by Diffusion Model with Minimal Condition | Accepted to ICML 2025 | null | null | null | cs.LG cs.AI physics.geo-ph | http://creativecommons.org/licenses/by/4.0/ | Shock waves caused by earthquakes can be devastating. Generating realistic earthquake-caused ground motion waveforms help reducing losses in lives and properties, yet generative models for the task tend to generate subpar waveforms. We present High-fidelity Earthquake Groundmotion Generation System (HEGGS) and demonstrate its superior performance using earthquakes from North American, East Asian, and European regions. HEGGS exploits the intrinsic characteristics of earthquake dataset and learns the waveforms using an end-to-end differentiable generator containing conditional latent diffusion model and hi-fidelity waveform construction model. We show the learning efficiency of HEGGS by training it on a single GPU machine and validate its performance using earthquake databases from North America, East Asia, and Europe, using diverse criteria from waveform generation tasks and seismology. Once trained, HEGGS can generate three dimensional E-N-Z seismic waveforms with accurate P/S phase arrivals, envelope correlation, signal-to-noise ratio, GMPE analysis, frequency content analysis, and section plot analysis. | [
{
"version": "v1",
"created": "Mon, 23 Dec 2024 06:56:28 GMT"
},
{
"version": "v2",
"created": "Thu, 29 May 2025 14:13:30 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Jung",
"Jaeheun",
""
],
[
"Lee",
"Jaehyuk",
""
],
[
"Jung",
"Changhae",
""
],
[
"Kim",
"Hanyoung",
""
],
[
"Jung",
"Bosung",
""
],
[
"Lee",
"Donghun",
""
]
] |
2412.18202 | Yining Zhou | Zhuohuan Hu, Richard Yu, Zizhou Zhang, Haoran Zheng, Qianying Liu, Yining Zhou | Developing Cryptocurrency Trading Strategy Based on Autoencoder-CNN-GANs Algorithms | The paper was accepted by 2024 4th International Conference on Artificial Intelligence, Robotics, and Communication(ICAIRC 2024) | null | null | null | cs.LG q-fin.ST | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | This paper leverages machine learning algorithms to forecast and analyze financial time series. The process begins with a denoising autoencoder to filter out random noise fluctuations from the main contract price data. Then, one-dimensional convolution reduces the dimensionality of the filtered data and extracts key information. The filtered and dimensionality-reduced price data is fed into a GANs network, and its output serve as input of a fully connected network. Through cross-validation, a model is trained to capture features that precede large price fluctuations. The model predicts the likelihood and direction of significant price changes in real-time price sequences, placing trades at moments of high prediction accuracy. Empirical results demonstrate that using autoencoders and convolution to filter and denoise financial data, combined with GANs, achieves a certain level of predictive performance, validating the capabilities of machine learning algorithms to discover underlying patterns in financial sequences. Keywords - CNN;GANs; Cryptocurrency; Prediction. | [
{
"version": "v1",
"created": "Tue, 24 Dec 2024 06:14:34 GMT"
},
{
"version": "v2",
"created": "Fri, 27 Dec 2024 05:28:43 GMT"
},
{
"version": "v3",
"created": "Wed, 22 Jan 2025 18:21:07 GMT"
},
{
"version": "v4",
"created": "Sat, 01 Feb 2025 07:22:53 GMT"
},
{
"version": "v5",
"created": "Thu, 29 May 2025 05:52:32 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Hu",
"Zhuohuan",
""
],
[
"Yu",
"Richard",
""
],
[
"Zhang",
"Zizhou",
""
],
[
"Zheng",
"Haoran",
""
],
[
"Liu",
"Qianying",
""
],
[
"Zhou",
"Yining",
""
]
] |
2412.18293 | Shaofei Cai | Shaofei Cai, Zhancun Mu, Kaichen He, Bowei Zhang, Xinyue Zheng, Anji Liu, Yitao Liang | MineStudio: A Streamlined Package for Minecraft AI Agent Development | null | null | null | null | cs.AI | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Minecraft's complexity and diversity as an open world make it a perfect environment to test if agents can learn, adapt, and tackle a variety of unscripted tasks. However, the development and validation of novel agents in this setting continue to face significant engineering challenges. This paper presents MineStudio, an open-source software package designed to streamline the development of autonomous agents in Minecraft. MineStudio represents the first comprehensive integration of seven critical engineering components: simulator, data, model, offline pre-training, online fine-tuning, inference, and benchmark, thereby allowing users to concentrate their efforts on algorithm innovation. We provide a user-friendly API design accompanied by comprehensive documentation and tutorials. Our project is released at https://github.com/CraftJarvis/MineStudio. | [
{
"version": "v1",
"created": "Tue, 24 Dec 2024 09:01:43 GMT"
},
{
"version": "v2",
"created": "Wed, 25 Dec 2024 04:51:04 GMT"
},
{
"version": "v3",
"created": "Thu, 29 May 2025 05:36:12 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Cai",
"Shaofei",
""
],
[
"Mu",
"Zhancun",
""
],
[
"He",
"Kaichen",
""
],
[
"Zhang",
"Bowei",
""
],
[
"Zheng",
"Xinyue",
""
],
[
"Liu",
"Anji",
""
],
[
"Liang",
"Yitao",
""
]
] |
2412.19020 | Jorge Passamani Zubelli | Jorge P. Zubelli, Kuldeep Singh, Vinicius Albani and Ioannis Kourakis | Travelling wave solutions of an equation of Harry Dym type arising in the Black-Scholes framework | 13 pages, 3 figures (with subfigures) | null | null | null | math.NA cs.NA q-fin.MF | http://creativecommons.org/licenses/by/4.0/ | The Black-Scholes framework is crucial in pricing a vast number of financial instruments that permeate the complex dynamics of world markets. Associated with this framework, we consider a second-order differential operator $L(x, {\partial_x}) := v^2(x,t) (\partial_x^2 -\partial_x)$ that carries a variable volatility term $v(x,t)$ and which is dependent on the underlying log-price $x$ and a time parameter $t$ motivated by the celebrated Dupire local volatility model. In this context, we ask and answer the question of whether one can find a non-linear evolution equation derived from a zero-curvature condition for a time-dependent deformation of the operator $L$. The result is a variant of the Harry Dym equation for which we can then find a family of travelling wave solutions. This brings in extensive machinery from soliton theory and integrable systems. As a by-product, it opens up the way to the use of coherent structures in financial-market volatility studies. | [
{
"version": "v1",
"created": "Thu, 26 Dec 2024 02:09:47 GMT"
},
{
"version": "v2",
"created": "Thu, 29 May 2025 16:12:37 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Zubelli",
"Jorge P.",
""
],
[
"Singh",
"Kuldeep",
""
],
[
"Albani",
"Vinicius",
""
],
[
"Kourakis",
"Ioannis",
""
]
] |
2412.19297 | Davide Venturelli | Gabriele Bandini and Davide Venturelli and Sarah A. M. Loos and Asja Jelic and Andrea Gambassi | The XY model with vision cone: non-reciprocal vs. reciprocal interactions | 19+5 pages, 13+4 figures | J. Stat. Mech. (2025) 053205 | 10.1088/1742-5468/adc243 | null | cond-mat.stat-mech cond-mat.soft | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We study the behavior of the classical XY model on a two-dimensional square lattice, with interactions occurring within a vision cone of each spin. Via Monte Carlo simulations, we explore one non-reciprocal and two reciprocal implementations of these interactions. The corresponding energy involves couplings that depend non-trivially on the system's configuration, leading to both long-range and quasi-long-range ordered phases at low temperatures. Our results demonstrate that non-reciprocity is not essential for achieving long-range order at low temperatures. Using symmetry arguments, we provide a theoretical framework to explain these findings, and additionally we uncover an unexpected order-by-disorder transition. | [
{
"version": "v1",
"created": "Thu, 26 Dec 2024 17:45:10 GMT"
},
{
"version": "v2",
"created": "Thu, 29 May 2025 13:42:32 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Bandini",
"Gabriele",
""
],
[
"Venturelli",
"Davide",
""
],
[
"Loos",
"Sarah A. M.",
""
],
[
"Jelic",
"Asja",
""
],
[
"Gambassi",
"Andrea",
""
]
] |
2412.19529 | Zijian Liu | Zijian Liu, Zhengyuan Zhou | Nonconvex Stochastic Optimization under Heavy-Tailed Noises: Optimal Convergence without Gradient Clipping | ICLR 2025 | null | null | null | math.OC cs.LG stat.ML | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Recently, the study of heavy-tailed noises in first-order nonconvex stochastic optimization has gotten a lot of attention since it was recognized as a more realistic condition as suggested by many empirical observations. Specifically, the stochastic noise (the difference between the stochastic and true gradient) is considered to have only a finite $\mathfrak{p}$-th moment where $\mathfrak{p}\in\left(1,2\right]$ instead of assuming it always satisfies the classical finite variance assumption. To deal with this more challenging setting, people have proposed different algorithms and proved them to converge at an optimal $\mathcal{O}(T^{\frac{1-\mathfrak{p}}{3\mathfrak{p}-2}})$ rate for smooth objectives after $T$ iterations. Notably, all these new-designed algorithms are based on the same technique - gradient clipping. Naturally, one may want to know whether the clipping method is a necessary ingredient and the only way to guarantee convergence under heavy-tailed noises. In this work, by revisiting the existing Batched Normalized Stochastic Gradient Descent with Momentum (Batched NSGDM) algorithm, we provide the first convergence result under heavy-tailed noises but without gradient clipping. Concretely, we prove that Batched NSGDM can achieve the optimal $\mathcal{O}(T^{\frac{1-\mathfrak{p}}{3\mathfrak{p}-2}})$ rate even under the relaxed smooth condition. More interestingly, we also establish the first $\mathcal{O}(T^{\frac{1-\mathfrak{p}}{2\mathfrak{p}}})$ convergence rate in the case where the tail index $\mathfrak{p}$ is unknown in advance, which is arguably the common scenario in practice. | [
{
"version": "v1",
"created": "Fri, 27 Dec 2024 08:46:46 GMT"
},
{
"version": "v2",
"created": "Fri, 21 Feb 2025 20:44:52 GMT"
},
{
"version": "v3",
"created": "Mon, 17 Mar 2025 22:23:21 GMT"
},
{
"version": "v4",
"created": "Thu, 29 May 2025 03:42:49 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Liu",
"Zijian",
""
],
[
"Zhou",
"Zhengyuan",
""
]
] |
2412.20647 | Aaditya Panigrahi | Aaditya Panigrahi, Alexei Tsvelik and Piers Coleman | Microscopic theory of pair density waves in spin-orbit coupled Kondo lattice | null | null | null | null | cond-mat.str-el cond-mat.supr-con | http://creativecommons.org/licenses/by/4.0/ | We demonstrate that the discommensuration between the Fermi surfaces of a conduction sea and an underlying spin liquid provides a natural mechanism for the spontaneous formation of pair density waves. Using a recent formulation of the Kondo lattice model which incorporates a Yao Lee spin liquid proposed by the authors, we demonstrate that doping away from half-filling induces finite-momentum electron-Majorana pair condensation, resulting in amplitude-modulated PDWs. Our approach provides a precise, analytically tractable pathway for understanding the spontaneous formation of PDWs in higher dimensions and offers a natural mechanism for PDW formation in the absence of a Zeeman field. | [
{
"version": "v1",
"created": "Mon, 30 Dec 2024 01:47:10 GMT"
},
{
"version": "v2",
"created": "Thu, 29 May 2025 11:38:33 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Panigrahi",
"Aaditya",
""
],
[
"Tsvelik",
"Alexei",
""
],
[
"Coleman",
"Piers",
""
]
] |
2412.21125 | Eugenio Clerico | Eugenio Clerico | Optimal e-value testing for properly constrained hypotheses | null | null | null | null | math.ST stat.TH | http://creativecommons.org/licenses/by/4.0/ | Hypothesis testing via e-variables can be framed as a sequential betting game, where a player each round picks an e-variable. A good player's strategy results in an effective statistical test that rejects the null hypothesis as soon as sufficient evidence arises. Building on recent advances, we address the question of restricting the pool of e-variables to simplify strategy design without compromising effectiveness. We extend the results of Clerico(2024), by characterising optimal sets of e-variables for a broad class of non-parametric hypothesis tests, defined by finitely many regular constraints. As an application, we discuss this notion of optimality in algorithmic mean estimation, including for heavy-tailed random variables. | [
{
"version": "v1",
"created": "Mon, 30 Dec 2024 17:57:38 GMT"
},
{
"version": "v2",
"created": "Wed, 26 Mar 2025 18:55:42 GMT"
},
{
"version": "v3",
"created": "Thu, 29 May 2025 14:05:04 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Clerico",
"Eugenio",
""
]
] |
2412.21165 | Daniel Britzger | Fazila Ahmadova, Daniel Britzger, Xuan Chen, Johannes G\"a{\ss}ler, Aude Gehrmann-De Ridder, Thomas Gehrmann, Nigel Glover, Claire Gwenlan, Gudrun Heinrich, Alexander Huss, Lucas Kunz, Jo\~ao Pires, Klaus Rabbertz, and Mark Sutton | Precise Determination of the Strong Coupling Constant from Dijet Cross Sections up to the Multi-TeV Range | 5 pages, 3 figures, appendix with 2 pages, supplementary material with 3 pages and 3 figures, to be published in Phys.Rev.Lett | null | null | CERN-TH-2024-224, KA-TP-25-2024, MPP-2024-260, ZU-TH/01/25 | hep-ph hep-ex | http://creativecommons.org/licenses/by/4.0/ | We determine the value of the strong coupling $\alpha_\text{s}$ and study its running over a wide range of scales as probed by the dijet production process at hadron colliders, based on an NNLO QCD analysis of LHC dijet data. From a large subset of these data a value of $\alpha_\text{s} (m_\text{Z}) = 0.1178 \pm 0.0022$ is obtained for the strong coupling at the scale of the Z-boson mass $m_\text{Z}$, using the invariant mass of the dijet system to select the scale where $\alpha_\text{s}$ is probed. The combination of different data sets enhances the reach and precision of the analysis in the mutli-TeV range and allows for the first determination of $\alpha_\text{s}$ up to scales of 7 TeV. Complementing the LHC data with dijet cross sections measured at the HERA electron-proton collider, the kinematic range is extended to test the running of the strong coupling towards smaller scales. Our results exhibit excellent agreement with predictions based on the renormalization group equation of QCD, and represent a comprehensive test of the asymptotic behavior of QCD, spanning more than three orders of magnitude in energy scale. | [
{
"version": "v1",
"created": "Mon, 30 Dec 2024 18:43:59 GMT"
},
{
"version": "v2",
"created": "Thu, 29 May 2025 15:59:38 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Ahmadova",
"Fazila",
""
],
[
"Britzger",
"Daniel",
""
],
[
"Chen",
"Xuan",
""
],
[
"Gäßler",
"Johannes",
""
],
[
"Ridder",
"Aude Gehrmann-De",
""
],
[
"Gehrmann",
"Thomas",
""
],
[
"Glover",
"Nigel",
""
],
[
"Gwenlan",
"Claire",
""
],
[
"Heinrich",
"Gudrun",
""
],
[
"Huss",
"Alexander",
""
],
[
"Kunz",
"Lucas",
""
],
[
"Pires",
"João",
""
],
[
"Rabbertz",
"Klaus",
""
],
[
"Sutton",
"Mark",
""
]
] |
2501.00349 | Shun Fujii | Tatsuki Murakami, Koshiro Wada, Soma Kogure, Ryomei Takabayashi, Liu Yang, Riku Shibata, Hajime Kumazaki, Shinichi Watanabe, Atsushi Ishizawa, Takasumi Tanabe, and Shun Fujii | Architecture for coherent dual-comb spectroscopy and low-noise photonic microwave generation using mechanically actuated soliton microcombs | null | Optics Letters, 50, 4, 1417-1420 (2025) | 10.1364/OL.554432 | null | physics.optics physics.app-ph | http://creativecommons.org/licenses/by/4.0/ | Dissipative Kerr soliton microcombs have inspired various intriguing applications such as spectroscopy, ranging, telecommunication, and high purity microwave generation. Mechanically actuated soliton microcombs provide enhanced controllability and flexibility for Kerr solitons, thus enabling technological progress to be made on such practical applications. Here, we present architectures for coherent dual-comb techniques and ultralow-noise microwave generation by exploiting the mechanical actuation of ultrahigh-Q crystalline microresonators. By unifying a pump laser, we demonstrate highly coherent dual-soliton combs using distinct resonators with slightly different repetition rates. We also report significant phase noise reduction achieved by directly generating Kerr solitons from a sub-Hz linewidth ultrastable laser. This study paves the way for further advancements in a wide variety of applications based on Kerr soliton microcombs. | [
{
"version": "v1",
"created": "Tue, 31 Dec 2024 08:55:36 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Murakami",
"Tatsuki",
""
],
[
"Wada",
"Koshiro",
""
],
[
"Kogure",
"Soma",
""
],
[
"Takabayashi",
"Ryomei",
""
],
[
"Yang",
"Liu",
""
],
[
"Shibata",
"Riku",
""
],
[
"Kumazaki",
"Hajime",
""
],
[
"Watanabe",
"Shinichi",
""
],
[
"Ishizawa",
"Atsushi",
""
],
[
"Tanabe",
"Takasumi",
""
],
[
"Fujii",
"Shun",
""
]
] |
2501.00417 | Hiroyuki Masuyama Dr. | Hiroyuki Masuyama | PureRank: A Parameter-Free Recursive Importance Measure for Network Nodes | This work is subject to a pending patent application | null | null | null | cs.SI | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Classical parameter-free centrality measures based on the recursive definition of importance (RDI), such as eigenvector centrality and Seeley centrality, are limited to strongly connected networks, while widely used methods like Katz centrality and PageRank rely on free parameters, such as the damping factor, to handle general networks. This motivates our central question: can an RDI-based importance (centrality) measure be defined for arbitrary networks without any tunable parameters? We answer this by introducing $PureRank$, a parameter-free recursive importance measure that faithfully reflects intrinsic network structure. PureRank classifies nodes into recurrent, transient, and dangling classes using strongly connected component (SCC) decomposition, computes local importance vectors for these classes, and aggregates them globally via the RDI principle. This modular, RDI-based design enables parallel and incremental computation -- ensuring scalability even for large or dynamic networks -- and, crucially, reduces to Seeley centrality (a classical and $pure$ RDI-based measure) on strongly connected networks. Furthermore, PureRank admits a probabilistic interpretation via a random-surfer model. Our numerical experiments illustrate that, unlike PageRank, PureRank yields unique and consistent importance scores for any input network. We also present theoretical extensions to signed networks. These results indicate that PureRank is a robust and transparent alternative to classical RDI-based centralities including PageRank. | [
{
"version": "v1",
"created": "Tue, 31 Dec 2024 12:37:49 GMT"
},
{
"version": "v2",
"created": "Sat, 04 Jan 2025 06:41:04 GMT"
},
{
"version": "v3",
"created": "Wed, 28 May 2025 22:20:27 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Masuyama",
"Hiroyuki",
""
]
] |
2501.00558 | Moreno Bevilacqua | Xavier Emery, Moreno Bevilacqua, Emilio Porcu | Matern and Generalized Wendland correlation models that parameterize hole effect, smoothness, and support | null | null | null | null | stat.ME | http://creativecommons.org/licenses/by/4.0/ | A huge literature in statistics and machine learning is devoted to parametric families of correlation functions, where the correlation parameters are used to understand the properties of an associated spatial random process in terms of smoothness and global or compact support. However, most of current parametric correlation functions attain only non-negative values. This work provides two new families that parameterize negative dependencies (aka hole effects), along with smoothness, and global or compact support. They generalize the celebrated Mat\'ern and Generalized Wendland models, respectively, which are attained as special cases. A link between the two new families is also established, showing that a specific reparameterization of the latter includes the former as a special limit case. Their performance in terms of estimation accuracy and goodness of best linear unbiased prediction is illustrated through synthetic and real data. | [
{
"version": "v1",
"created": "Tue, 31 Dec 2024 17:42:39 GMT"
},
{
"version": "v2",
"created": "Thu, 29 May 2025 12:21:38 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Emery",
"Xavier",
""
],
[
"Bevilacqua",
"Moreno",
""
],
[
"Porcu",
"Emilio",
""
]
] |
2501.01482 | Muhammad Ahmad Sultan | Muhammad Ahmad Sultan, Chong Chen, Yingmin Liu, Katarzyna Gil, Karolina Zareba, Rizwan Ahmad | An unsupervised method for MRI recovery: Deep image prior with structured sparsity | Magn Reson Mater Phy (2025) | null | 10.1007/s10334-025-01257-z | null | eess.IV cs.CV cs.LG eess.SP | http://creativecommons.org/licenses/by-nc-sa/4.0/ | Objective: To propose and validate an unsupervised MRI reconstruction method that does not require fully sampled k-space data. Materials and Methods: The proposed method, deep image prior with structured sparsity (DISCUS), extends the deep image prior (DIP) by introducing group sparsity to frame-specific code vectors, enabling the discovery of a low-dimensional manifold for capturing temporal variations. \discus was validated using four studies: (I) simulation of a dynamic Shepp-Logan phantom to demonstrate its manifold discovery capabilities, (II) comparison with compressed sensing and DIP-based methods using simulated single-shot late gadolinium enhancement (LGE) image series from six distinct digital cardiac phantoms in terms of normalized mean square error (NMSE) and structural similarity index measure (SSIM), (III) evaluation on retrospectively undersampled single-shot LGE data from eight patients, and (IV) evaluation on prospectively undersampled single-shot LGE data from eight patients, assessed via blind scoring from two expert readers. Results: DISCUS outperformed competing methods, demonstrating superior reconstruction quality in terms of NMSE and SSIM (Studies I--III) and expert reader scoring (Study IV). Discussion: An unsupervised image reconstruction method is presented and validated on simulated and measured data. These developments can benefit applications where acquiring fully sampled data is challenging. | [
{
"version": "v1",
"created": "Thu, 02 Jan 2025 18:13:35 GMT"
},
{
"version": "v2",
"created": "Thu, 15 May 2025 17:15:14 GMT"
},
{
"version": "v3",
"created": "Wed, 28 May 2025 18:06:12 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Sultan",
"Muhammad Ahmad",
""
],
[
"Chen",
"Chong",
""
],
[
"Liu",
"Yingmin",
""
],
[
"Gil",
"Katarzyna",
""
],
[
"Zareba",
"Karolina",
""
],
[
"Ahmad",
"Rizwan",
""
]
] |
2501.04837 | Alberto Carnelli | Carnelli Alberto | Proposal for simplified template cross-sections extension using $\cal{CP}$ observables in $t\overline{t}H$ | Poster at the 17th International Workshop on Top Quark Physics (Top2024), 22-27 September 2024 | null | null | null | hep-ph | http://creativecommons.org/licenses/by/4.0/ | The Large Hadron Collider (LHC) offers a unique opportunity to investigate $\cal{CP}$ violation in the Yukawa coupling between the Higgs boson and the top quark by studying Higgs production in association with top quarks; this is of fundamental importance, seeing that the $\cal{CP}$ properties of the Higgs boson are yet to measure with high precision. To address this, the focus of this work has been an extension of the simplified template cross-section (STXS) framework, devised to be sensitive to $\cal{CP}$ effects. Our study focused on $\cal{CP}$-sensitive observables across multiple Higgs decay channels, comparing their performances. The result indicates that the most efficient extension of the current binning used in the STXS framework, which currently uses the Higgs boson's transverse momentum $p_{T,H}$, requires adding one further split using $\cal{CP}$-sensitive observables. Between these observables, one of the best is the Collins-Soper angle $|\cos\theta^*|$, a variable derived from momenta information of the top quarks. We have investigated the improvement brought by our two-dimensional STXS setup and compared it to the currently employed methodologies, finding an increase in performances at an integrated luminosity of $300$ fb$^{-1}$. Moreover, our results highlight that this advantage seems to be present also at $3000$ fb$^{-1}$. | [
{
"version": "v1",
"created": "Wed, 08 Jan 2025 20:47:24 GMT"
},
{
"version": "v2",
"created": "Thu, 29 May 2025 17:27:08 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Alberto",
"Carnelli",
""
]
] |
2501.05290 | Konstantin L. Metlov | Konstantin L. Metlov | Elliptical stability of hopfions in bulk helimagnets | 6 pages, 3 figures, accepted to Phys. Rev. B | Phys. Rev. B 111, 174450 (2025) | 10.1103/PhysRevB.111.174450 | null | cond-mat.mes-hall nlin.PS | http://creativecommons.org/licenses/by/4.0/ | Magnetic hopfions are three-dimensional topological solitons with non-zero Hopf index ${\cal H}$ in the vector field of material's local magnetization. In this Letter elliptical stability of hopfions with ${\cal H}=1$ in a classical helimagnet is studied on the basis of a variational model. It is shown that, depending on their internal structure (vortex and antivortex tubes ordering), the hopfions can either be stable in a bulk magnet or unstable with respect to elongation along their central axis. It is found that the energy of stable hopfions is always below the energy of the $2\pi$-skyrmion lattice in the same material, suggesting the possibility to use $2\pi$-skyrmions as a precursor for hopfion nucleation. Stability diagram for hopfions on the magnetic anisotropy-field phase diagram is computed numerically. Explicit analytical expressions for some of its critical lines are derived. | [
{
"version": "v1",
"created": "Thu, 09 Jan 2025 14:53:29 GMT"
},
{
"version": "v2",
"created": "Sat, 17 May 2025 10:00:30 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Metlov",
"Konstantin L.",
""
]
] |
2501.05418 | Guoqing Zhang | Guoqing Zhang, Zihan Chen and Long Wang | Integrated Shape and Force Sensing for Continuum Instruments via Polynomial Curvature Modeling and Torque-Cell-Based Actuation | null | null | null | null | cs.RO | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Continuum instruments are integral to robot-assisted minimally invasive surgery (MIS), with cable-driven mechanisms being the most common. Real-time tension feedback is crucial for precise articulation but remains a challenge in compact actuation unit designs. Additionally, accurate shape and external force sensing of continuum instruments are essential for advanced control and manipulation. This paper presents a compact actuation unit that integrates a torque cell directly into the pulley module to provide real-time tension feedback. Building on this unit, we propose a novel shape-force sensing framework incorporating polynomial curvature kinematics and virtual-work-based force sensing model. The framework combines pose sensor measurements at the instrument tip and actuation tension feedback at the developed actuation unit. Experimental results demonstrate the improved performance of the framework in terms of shape reconstruction accuracy and force estimation reliability compared to conventional constant curvature methods. | [
{
"version": "v1",
"created": "Thu, 09 Jan 2025 18:20:57 GMT"
},
{
"version": "v2",
"created": "Wed, 28 May 2025 18:34:23 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Zhang",
"Guoqing",
""
],
[
"Chen",
"Zihan",
""
],
[
"Wang",
"Long",
""
]
] |
2501.05419 | Saulius Vaitiek\.enas | M. T. Deng, Carlos Pay\'a, Pablo San-Jose, Elsa Prada, C. M. Marcus, and S. Vaitiek\.enas | Caroli-de Gennes-Matricon Analogs in Full-Shell Hybrid Nanowires | 5 pages, 4 figures | Phys. Rev. Lett. 134, 206302 (2025) | 10.1103/PhysRevLett.134.206302 | NBI QDEV 2024 | cond-mat.mes-hall | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We report tunneling spectroscopy of Andreev subgap states in hybrid nanowires with a thin superconducting full-shell surrounding a semiconducting core. The combination of the quantized fluxoid of the shell and the Andreev reflection at the superconductor-semiconductor interface gives rise to analogs of Caroli-de Gennes-Matricon (CdGM) states found in Abrikosov vortices in type-II superconductors. Unlike in metallic superconductors, CdGM analogs in full-shell hybrid nanowires manifest as one-dimensional van Hove singularities with energy spacings comparable to the superconducting gap and independent of the Fermi energy, making them readily observable. Evolution of these analogs with axial magnetic field, skewed within the Little-Parks lobe structure, is consistent with theory and yields information about the radial distribution and angular momenta of the corresponding subbands. | [
{
"version": "v1",
"created": "Thu, 09 Jan 2025 18:22:08 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Deng",
"M. T.",
""
],
[
"Payá",
"Carlos",
""
],
[
"San-Jose",
"Pablo",
""
],
[
"Prada",
"Elsa",
""
],
[
"Marcus",
"C. M.",
""
],
[
"Vaitiekėnas",
"S.",
""
]
] |
2501.05440 | Joanna Morgan | Joanna Piper Morgan and Braxton Cuneo and Ilham Variansyah and Kyle E. Niemeyer | Enabling GPU Portability into the Numba-JITed Monte Carlo Particle Transport Code MC/DC | 10 pages, 3 figures. Special Session on Research Activities of the Center for Exascale Monte Carlo Neutron Transport | Proceedings of the International Conference on Mathematics and Computational Methods Applied to Nuclear Science and Engineering (M&C 2025), pp.1934-1943. (2025). Denver, CO, USA | 10.13182/MC25-47142 | null | physics.comp-ph | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | The Center for Exascale Monte Carlo Neutron Transport is developing Monte Carlo / Dynamic Code (MC/DC) as a portable Monte Carlo neutron transport package for rapid numerical methods exploration on CPU- and GPU-based high-performance computers. In this paper, we describe MC/DC's current event-based GPU algorithm as well as the just-in-time (JIT) compilation scheme we use to enable GPU operability on Nvidia and AMD GPUs from MC/DC's Python source. To analyze performance, we conduct runtime tests of the C5G7 k-eigenvalue benchmark problem and a continuous-energy infinite pin cell on Nvidia Tesla V100 GPU, AMD MI250X GPU, and the AMD MI300A APU and make comparison to a dual-socket Intel Xeon Sapphire Rapid CPU node. We found that for the multi-group C5G7 benchmark problem, we respectively see a 15$\times$, 0.7$\times$, 12$\times$ speedup on a V100, MI250X, and MI300A over 112 Intel Xeon CPU cores. For the continuous-energy infinite pin-cell benchmark, we found speedups of 5$\times$, 3$\times$, 4$\times$ on a V100, MI250X, and MI300A, respectively, over the same CPU node. | [
{
"version": "v1",
"created": "Thu, 09 Jan 2025 18:50:27 GMT"
},
{
"version": "v2",
"created": "Thu, 29 May 2025 17:54:38 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Morgan",
"Joanna Piper",
""
],
[
"Cuneo",
"Braxton",
""
],
[
"Variansyah",
"Ilham",
""
],
[
"Niemeyer",
"Kyle E.",
""
]
] |
2501.05683 | Dhayaa Anbajagane | D. Anbajagane, M. Tabbutt, J. Beas-Gonzalez, B. Yanny, S. Everett, M. R. Becker, M. Yamamoto, E. Legnani, J. De Vicente, K. Bechtol, J. Elvin-Poole, G. M. Bernstein, A. Choi, M. Gatti, G. Giannini, R. A. Gruendl, M. Jarvis, S. Lee, J. Mena-Fern\'andez, A. Porredon, M. Rodriguez-Monroy, E. Rozo, E. S. Rykoff, T. Schutt, E. Sheldon, M. A. Troxel, N. Weaverdyck, V. Wetzell, M. Aguena, A. Alarcon, S. Allam, A. Amon, F. Andrade-Oliveira, J. Blazek, D. Brooks, A. Carnero Rosell, J. Carretero, C. Chang, M. Crocce, L. N. da Costa, M. E. S. Pereira, T. M. Davis, S. Desai, H. T. Diehl, S. Dodelson, P. Doel, A. Drlica-Wagner, A. Fert\'e, J. Frieman, J. Garc\'ia-Bellido, E. Gaztanaga, D. Gruen, G. Gutierrez, W. G. Hartley, K. Herner, S. R. Hinton, D. L. Hollowood, K. Honscheid, D. Huterer, D. J. James, E. Krause, K. Kuehn, O. Lahav, J. L. Marshall, R. Miquel, J. Muir, J. Myles, A. Pieres, A. A. Plazas Malag\'on, J. Prat, M. Raveri, S. Samuroff, E. Sanchez, D. Sanchez Cid, I. Sevilla-Noarbe, M. Smith, E. Suchyta, G. Tarle, D. L. Tucker, A. R. Walker, P. Wiseman, Y. Zhang (DES Collaboration) | Dark Energy Survey Year 6 Results: Synthetic-source Injection Across the Full Survey Using Balrog | v2: accepted to OJA | null | 10.33232/001c.138627 | null | astro-ph.CO astro-ph.GA | http://creativecommons.org/licenses/by/4.0/ | Synthetic source injection (SSI), the insertion of sources into pixel-level on-sky images, is a powerful method for characterizing object detection and measurement in wide-field, astronomical imaging surveys. Within the Dark Energy Survey (DES), SSI plays a critical role in characterizing all necessary algorithms used in converting images to catalogs, and in deriving quantities needed for the cosmology analysis, such as object detection rates, galaxy redshift estimation, galaxy magnification, star-galaxy classification, and photometric performance. We present here a source injection catalog of $146$ million injections spanning the entire 5000 deg$^2$ DES footprint, generated using the Balrog SSI pipeline. Through this sample, we demonstrate that the DES Year 6 (Y6) image processing pipeline provides accurate estimates of the object properties, for both galaxies and stars, at the percent-level, and we highlight specific regimes where the accuracy is reduced. We then show the consistency between SSI and data catalogs, for all galaxy samples developed within the weak lensing and galaxy clustering analyses of DES Y6. The consistency between the two catalogs also extends to their correlations with survey observing properties (seeing, airmass, depth, extinction, etc.). Finally, we highlight a number of applications of this catalog to the DES Y6 cosmology analysis. This dataset is the largest SSI catalog produced at this fidelity and will serve as a key testing ground for exploring the utility of SSI catalogs in upcoming surveys such as the Vera C. Rubin Observatory Legacy Survey of Space and Time. | [
{
"version": "v1",
"created": "Fri, 10 Jan 2025 03:13:49 GMT"
},
{
"version": "v2",
"created": "Wed, 28 May 2025 16:22:03 GMT"
},
{
"version": "v3",
"created": "Thu, 29 May 2025 15:03:02 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Anbajagane",
"D.",
"",
"DES Collaboration"
],
[
"Tabbutt",
"M.",
"",
"DES Collaboration"
],
[
"Beas-Gonzalez",
"J.",
"",
"DES Collaboration"
],
[
"Yanny",
"B.",
"",
"DES Collaboration"
],
[
"Everett",
"S.",
"",
"DES Collaboration"
],
[
"Becker",
"M. R.",
"",
"DES Collaboration"
],
[
"Yamamoto",
"M.",
"",
"DES Collaboration"
],
[
"Legnani",
"E.",
"",
"DES Collaboration"
],
[
"De Vicente",
"J.",
"",
"DES Collaboration"
],
[
"Bechtol",
"K.",
"",
"DES Collaboration"
],
[
"Elvin-Poole",
"J.",
"",
"DES Collaboration"
],
[
"Bernstein",
"G. M.",
"",
"DES Collaboration"
],
[
"Choi",
"A.",
"",
"DES Collaboration"
],
[
"Gatti",
"M.",
"",
"DES Collaboration"
],
[
"Giannini",
"G.",
"",
"DES Collaboration"
],
[
"Gruendl",
"R. A.",
"",
"DES Collaboration"
],
[
"Jarvis",
"M.",
"",
"DES Collaboration"
],
[
"Lee",
"S.",
"",
"DES Collaboration"
],
[
"Mena-Fernández",
"J.",
"",
"DES Collaboration"
],
[
"Porredon",
"A.",
"",
"DES Collaboration"
],
[
"Rodriguez-Monroy",
"M.",
"",
"DES Collaboration"
],
[
"Rozo",
"E.",
"",
"DES Collaboration"
],
[
"Rykoff",
"E. S.",
"",
"DES Collaboration"
],
[
"Schutt",
"T.",
"",
"DES Collaboration"
],
[
"Sheldon",
"E.",
"",
"DES Collaboration"
],
[
"Troxel",
"M. A.",
"",
"DES Collaboration"
],
[
"Weaverdyck",
"N.",
"",
"DES Collaboration"
],
[
"Wetzell",
"V.",
"",
"DES Collaboration"
],
[
"Aguena",
"M.",
"",
"DES Collaboration"
],
[
"Alarcon",
"A.",
"",
"DES Collaboration"
],
[
"Allam",
"S.",
"",
"DES Collaboration"
],
[
"Amon",
"A.",
"",
"DES Collaboration"
],
[
"Andrade-Oliveira",
"F.",
"",
"DES Collaboration"
],
[
"Blazek",
"J.",
"",
"DES Collaboration"
],
[
"Brooks",
"D.",
"",
"DES Collaboration"
],
[
"Rosell",
"A. Carnero",
"",
"DES Collaboration"
],
[
"Carretero",
"J.",
"",
"DES Collaboration"
],
[
"Chang",
"C.",
"",
"DES Collaboration"
],
[
"Crocce",
"M.",
"",
"DES Collaboration"
],
[
"da Costa",
"L. N.",
"",
"DES Collaboration"
],
[
"Pereira",
"M. E. S.",
"",
"DES Collaboration"
],
[
"Davis",
"T. M.",
"",
"DES Collaboration"
],
[
"Desai",
"S.",
"",
"DES Collaboration"
],
[
"Diehl",
"H. T.",
"",
"DES Collaboration"
],
[
"Dodelson",
"S.",
"",
"DES Collaboration"
],
[
"Doel",
"P.",
"",
"DES Collaboration"
],
[
"Drlica-Wagner",
"A.",
"",
"DES Collaboration"
],
[
"Ferté",
"A.",
"",
"DES Collaboration"
],
[
"Frieman",
"J.",
"",
"DES Collaboration"
],
[
"García-Bellido",
"J.",
"",
"DES Collaboration"
],
[
"Gaztanaga",
"E.",
"",
"DES Collaboration"
],
[
"Gruen",
"D.",
"",
"DES Collaboration"
],
[
"Gutierrez",
"G.",
"",
"DES Collaboration"
],
[
"Hartley",
"W. G.",
"",
"DES Collaboration"
],
[
"Herner",
"K.",
"",
"DES Collaboration"
],
[
"Hinton",
"S. R.",
"",
"DES Collaboration"
],
[
"Hollowood",
"D. L.",
"",
"DES Collaboration"
],
[
"Honscheid",
"K.",
"",
"DES Collaboration"
],
[
"Huterer",
"D.",
"",
"DES Collaboration"
],
[
"James",
"D. J.",
"",
"DES Collaboration"
],
[
"Krause",
"E.",
"",
"DES Collaboration"
],
[
"Kuehn",
"K.",
"",
"DES Collaboration"
],
[
"Lahav",
"O.",
"",
"DES Collaboration"
],
[
"Marshall",
"J. L.",
"",
"DES Collaboration"
],
[
"Miquel",
"R.",
"",
"DES Collaboration"
],
[
"Muir",
"J.",
"",
"DES Collaboration"
],
[
"Myles",
"J.",
"",
"DES Collaboration"
],
[
"Pieres",
"A.",
"",
"DES Collaboration"
],
[
"Malagón",
"A. A. Plazas",
"",
"DES Collaboration"
],
[
"Prat",
"J.",
"",
"DES Collaboration"
],
[
"Raveri",
"M.",
"",
"DES Collaboration"
],
[
"Samuroff",
"S.",
"",
"DES Collaboration"
],
[
"Sanchez",
"E.",
"",
"DES Collaboration"
],
[
"Cid",
"D. Sanchez",
"",
"DES Collaboration"
],
[
"Sevilla-Noarbe",
"I.",
"",
"DES Collaboration"
],
[
"Smith",
"M.",
"",
"DES Collaboration"
],
[
"Suchyta",
"E.",
"",
"DES Collaboration"
],
[
"Tarle",
"G.",
"",
"DES Collaboration"
],
[
"Tucker",
"D. L.",
"",
"DES Collaboration"
],
[
"Walker",
"A. R.",
"",
"DES Collaboration"
],
[
"Wiseman",
"P.",
"",
"DES Collaboration"
],
[
"Zhang",
"Y.",
"",
"DES Collaboration"
]
] |
2501.05874 | Soyeong Jeong | Soyeong Jeong, Kangsan Kim, Jinheon Baek, Sung Ju Hwang | VideoRAG: Retrieval-Augmented Generation over Video Corpus | ACL Findings 2025 | null | null | null | cs.CV cs.AI cs.CL cs.IR cs.LG | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Retrieval-Augmented Generation (RAG) is a powerful strategy for improving the factual accuracy of models by retrieving external knowledge relevant to queries and incorporating it into the generation process. However, existing approaches primarily focus on text, with some recent advancements considering images, and they largely overlook videos, a rich source of multimodal knowledge capable of representing contextual details more effectively than any other modality. While very recent studies explore the use of videos in response generation, they either predefine query-associated videos without retrieval or convert videos into textual descriptions losing multimodal richness. To tackle these, we introduce VideoRAG, a framework that not only dynamically retrieves videos based on their relevance with queries but also utilizes both visual and textual information. The operation of VideoRAG is powered by recent Large Video Language Models (LVLMs), which enable the direct processing of video content to represent it for retrieval and the seamless integration of retrieved videos jointly with queries for response generation. Also, inspired by that the context size of LVLMs may not be sufficient to process all frames in extremely long videos and not all frames are equally important, we introduce a video frame selection mechanism to extract the most informative subset of frames, along with a strategy to extract textual information from videos (as it can aid the understanding of video content) when their subtitles are not available. We experimentally validate the effectiveness of VideoRAG, showcasing that it is superior to relevant baselines. Code is available at https://github.com/starsuzi/VideoRAG. | [
{
"version": "v1",
"created": "Fri, 10 Jan 2025 11:17:15 GMT"
},
{
"version": "v2",
"created": "Tue, 04 Mar 2025 07:29:52 GMT"
},
{
"version": "v3",
"created": "Wed, 28 May 2025 18:14:55 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Jeong",
"Soyeong",
""
],
[
"Kim",
"Kangsan",
""
],
[
"Baek",
"Jinheon",
""
],
[
"Hwang",
"Sung Ju",
""
]
] |
2501.06026 | Xiao-Feng Qian | G. Zhao, Y. Wang, and X.-F. Qian | Theory of Magnon Purcell Effect in Cavity Magnonic System | 6 pages, 4 figures | null | null | null | cond-mat.mes-hall quant-ph | http://creativecommons.org/licenses/by-nc-sa/4.0/ | We conduct a systematic analysis of cavity effects on the decay dynamics of an open magnonic system. The Purcell effect on the magnon oscillator decay is thoroughly examined for both driven and non-driven scenarios. Analytical conditions are determined to distinguish between strong and weak coupling regimes, corresponding to oscillatory and pure decay behaviors respectively. Additionally, our theory also predicts the decay of the photon mode within the cavity-magnonic open system, demonstrating excellent agreement with existing experimental data. Our findings and methodologies can provide valuable insights for advancing research in cavity magnonic quantum control, quantum information processing, and the development of magnonic quantum devices. | [
{
"version": "v1",
"created": "Fri, 10 Jan 2025 15:02:44 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Zhao",
"G.",
""
],
[
"Wang",
"Y.",
""
],
[
"Qian",
"X. -F.",
""
]
] |
2501.06425 | Yifan Zhang | Yifan Zhang, Yifeng Liu, Huizhuo Yuan, Zhen Qin, Yang Yuan, Quanquan Gu, Andrew C Yao | Tensor Product Attention Is All You Need | 52 pages, 11 figures | null | null | null | cs.CL cs.AI cs.LG | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Scaling language models to handle longer input sequences typically necessitates large key-value (KV) caches, resulting in substantial memory overhead during inference. In this paper, we propose Tensor Product Attention (TPA), a novel attention mechanism that uses tensor decompositions to represent queries, keys, and values compactly, substantially shrinking the KV cache size at inference time. By factorizing these representations into contextual low-rank components and seamlessly integrating with Rotary Position Embedding (RoPE), TPA achieves improved model quality alongside memory efficiency. Based on TPA, we introduce the Tensor Product Attention Transformer,(T6), a new model architecture for sequence modeling. Through extensive empirical evaluation on language modeling tasks, we demonstrate that T6 surpasses or matches the performance of standard Transformer baselines, including Multi-Head Attention (MHA), Multi-Query Attention (MQA), Grouped-Query Attention (GQA), and Multi-Head Latent Attention (MLA) across various metrics, including perplexity and a range of established evaluation benchmarks. Notably, TPA's memory efficiency and computational efficiency at the decoding stage enable processing longer sequences under fixed resource constraints, addressing a critical scalability challenge in modern language models. The code is available at https://github.com/tensorgi/T6. | [
{
"version": "v1",
"created": "Sat, 11 Jan 2025 03:37:10 GMT"
},
{
"version": "v2",
"created": "Fri, 07 Feb 2025 22:00:48 GMT"
},
{
"version": "v3",
"created": "Wed, 09 Apr 2025 20:51:08 GMT"
},
{
"version": "v4",
"created": "Thu, 29 May 2025 09:01:23 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Zhang",
"Yifan",
""
],
[
"Liu",
"Yifeng",
""
],
[
"Yuan",
"Huizhuo",
""
],
[
"Qin",
"Zhen",
""
],
[
"Yuan",
"Yang",
""
],
[
"Gu",
"Quanquan",
""
],
[
"Yao",
"Andrew C",
""
]
] |
2501.06922 | Ritesh Sharma | Woubishet Zewdu Taffese, Ritesh Sharma, Mohammad Hossein Afsharmovahed, Gunasekaran Manogaran and Genda Chen | Benchmarking YOLOv8 for Optimal Crack Detection in Civil Infrastructure | We would like to extend/modify this work and make changes for resubmission to a different place. Hence we would like to withdraw the paper | null | null | null | cs.CV | http://creativecommons.org/licenses/by-nc-nd/4.0/ | Ensuring the structural integrity and safety of bridges is crucial for the reliability of transportation networks and public safety. Traditional crack detection methods are increasingly being supplemented or replaced by advanced artificial intelligence (AI) techniques. However, most of the models rely on two-stage target detection algorithms, which pose concerns for real-time applications due to their lower speed. While models such as YOLO (You Only Look Once) have emerged as transformative tools due to their remarkable speed and accuracy. However, the potential of the latest YOLOv8 framework in this domain remains underexplored. This study bridges that gap by rigorously evaluating YOLOv8's performance across five model scales (nano, small, medium, large, and extra-large) using a high-quality Roboflow dataset. A comprehensive hyperparameter optimization was performed, testing six state-of-the-art optimizers-Stochastic Gradient Descent, Adaptive Moment Estimation, Adam with Decoupled Weight Decay, Root Mean Square Propagation, Rectified Adam, and Nesterov-accelerated Adam. Results revealed that YOLOv8, optimized with Stochastic Gradient Descent, delivered exceptional accuracy and speed, setting a new benchmark for real-time crack detection. Beyond its immediate application, this research positions YOLOv8 as a foundational approach for integrating advanced computer vision techniques into infrastructure monitoring. By enabling more reliable and proactive maintenance of aging bridge networks, this work paves the way for safer, more efficient transportation systems worldwide. | [
{
"version": "v1",
"created": "Sun, 12 Jan 2025 20:17:46 GMT"
},
{
"version": "v2",
"created": "Thu, 29 May 2025 02:38:24 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Taffese",
"Woubishet Zewdu",
""
],
[
"Sharma",
"Ritesh",
""
],
[
"Afsharmovahed",
"Mohammad Hossein",
""
],
[
"Manogaran",
"Gunasekaran",
""
],
[
"Chen",
"Genda",
""
]
] |
2501.07780 | A F M Anhar Uddin Bhuiyan | Stephen Margiotta, Binzhi Liu, Saleh Ahmed Khan, Gabriel Calderon Ortiz, Ahmed Ibreljic, Jinwoo Hwang and A F M Anhar Uddin Bhuiyan | Thermal Annealing and Radiation Effects on Structural and Electrical Properties of NbN/GaN Superconductor/Semiconductor Junction | null | J. Vac. Sci. Technol. A 43, 042701 (2025) | 10.1116/6.0004539 | null | cond-mat.supr-con quant-ph | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | In the rapidly evolving field of quantum computing, niobium nitride (NbN) superconductors have emerged as integral components due to their unique structural properties, including a high superconducting transition temperature (Tc), exceptional electrical conductivity, and compatibility with advanced device architectures. This study investigates the impact of high-temperature annealing and high-dose gamma irradiation on the structural and superconducting properties of NbN films grown on GaN via reactive DC magnetron sputtering. The as-deposited cubic {\delta}-NbN (111) films exhibited a high-intensity XRD peak, high Tc of 12.82K, and an atomically flat surface. Annealing at 500 and 950 {\deg}C for varying durations revealed notable structural and surface changes. High-resolution STEM indicated improved local ordering, while AFM showed reduced surface roughness after annealing. XPS revealed a gradual increase in the Nb/N ratio with higher annealing temperatures and durations. High-resolution XRD and STEM analyses showed lattice constant modifications in {\delta}-NbN films, attributed to residual stress changes following annealing. Additionally, XRD phi-scans revealed sixfold symmetry in NbN films due to rotational domains relative to GaN. While Tc remained stable after annealing at 500 {\deg}C, increasing the annealing temperature to 950 {\deg}C degraded Tc to ~8K and reduced the residual resistivity ratio from 0.85 in as-deposited films to 0.29 after 30 minutes. The effects of gamma radiation (5 Mrad (Si)) were also studied, demonstrating minimal changes to crystallinity and superconducting performance, indicating excellent radiation resilience. These findings highlight the potential of NbN superconductors for integration into advanced quantum devices and their suitability for applications in radiation-intensive environments such as space, satellites, and nuclear power plants. | [
{
"version": "v1",
"created": "Tue, 14 Jan 2025 01:37:21 GMT"
},
{
"version": "v2",
"created": "Fri, 23 May 2025 12:53:16 GMT"
},
{
"version": "v3",
"created": "Wed, 28 May 2025 21:46:57 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Margiotta",
"Stephen",
""
],
[
"Liu",
"Binzhi",
""
],
[
"Khan",
"Saleh Ahmed",
""
],
[
"Ortiz",
"Gabriel Calderon",
""
],
[
"Ibreljic",
"Ahmed",
""
],
[
"Hwang",
"Jinwoo",
""
],
[
"Bhuiyan",
"A F M Anhar Uddin",
""
]
] |
2501.08319 | Yoav Gur-Arieh | Yoav Gur-Arieh, Roy Mayan, Chen Agassy, Atticus Geiger, Mor Geva | Enhancing Automated Interpretability with Output-Centric Feature Descriptions | Accepted to ACL 2025 Main Conference | null | null | null | cs.CL | http://creativecommons.org/licenses/by/4.0/ | Automated interpretability pipelines generate natural language descriptions for the concepts represented by features in large language models (LLMs), such as plants or the first word in a sentence. These descriptions are derived using inputs that activate the feature, which may be a dimension or a direction in the model's representation space. However, identifying activating inputs is costly, and the mechanistic role of a feature in model behavior is determined both by how inputs cause a feature to activate and by how feature activation affects outputs. Using steering evaluations, we reveal that current pipelines provide descriptions that fail to capture the causal effect of the feature on outputs. To fix this, we propose efficient, output-centric methods for automatically generating feature descriptions. These methods use the tokens weighted higher after feature stimulation or the highest weight tokens after applying the vocabulary "unembedding" head directly to the feature. Our output-centric descriptions better capture the causal effect of a feature on model outputs than input-centric descriptions, but combining the two leads to the best performance on both input and output evaluations. Lastly, we show that output-centric descriptions can be used to find inputs that activate features previously thought to be "dead". | [
{
"version": "v1",
"created": "Tue, 14 Jan 2025 18:53:00 GMT"
},
{
"version": "v2",
"created": "Thu, 29 May 2025 15:26:06 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Gur-Arieh",
"Yoav",
""
],
[
"Mayan",
"Roy",
""
],
[
"Agassy",
"Chen",
""
],
[
"Geiger",
"Atticus",
""
],
[
"Geva",
"Mor",
""
]
] |
2501.08757 | Angela Monti | Fasma Diele, Andrew L. Krause, Deborah Lacitignola, Carmela Marangi, Angela Monti, Edgardo Villar-Sep\'ulveda | Transient Instability and Patterns of Reactivity in Diffusive-Chemotaxis Soil Carbon Dynamics | null | null | null | null | math.NA cs.NA | http://creativecommons.org/licenses/by/4.0/ | We study pattern formation in a chemotaxis model of bacteria and soil carbon dynamics as an example system where transient dynamics can give rise to pattern formation outside of Turing unstable regimes. We use a detailed analysis of the reactivity of the non-spatial and spatial dynamics, stability analyses, and numerical continuation to uncover detailed aspects of this system's pattern-forming potential. In addition to patterning in Turing unstable parameter regimes, reactivity of the spatial system can itself lead to a range of parameters where a spatially uniform state is asymptotically stable, but exhibits transient growth that can induce pattern formation. We show that this occurs in the bistable region of a subcritical Turing bifurcation. Intriguingly, such bistable regions appear in two spatial dimensions, but not in a one-dimensional domain, suggesting important interplays between geometry, transient growth, and the emergence of multistable patterns. We discuss the implications of our analysis for the bacterial soil organic carbon system, as well as for reaction-transport modelling more generally. | [
{
"version": "v1",
"created": "Wed, 15 Jan 2025 12:20:33 GMT"
},
{
"version": "v2",
"created": "Thu, 16 Jan 2025 06:36:05 GMT"
},
{
"version": "v3",
"created": "Thu, 29 May 2025 15:33:42 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Diele",
"Fasma",
""
],
[
"Krause",
"Andrew L.",
""
],
[
"Lacitignola",
"Deborah",
""
],
[
"Marangi",
"Carmela",
""
],
[
"Monti",
"Angela",
""
],
[
"Villar-Sepúlveda",
"Edgardo",
""
]
] |
2501.08797 | Giambattista Giacomin | Orph\'ee Collin, Giambattista Giacomin, Rafael L. Greenblatt, Yueyun Hu | On the Lyapunov exponent for the random field Ising transfer matrix, in the critical case | 35 pages, 2 figures | Published in Annales Henri Poincar\'e (2025) | 10.1007/s00023-025-01589-5 | null | math.PR math-ph math.MP | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We study the top Lyapunov exponent of a product of random $2 \times 2$ matrices appearing in the analysis of several statistical mechanical models with disorder, extending a previous treatment of the critical case (Giacomin and Greenblatt, ALEA 19 (2022), 701-728) by significantly weakening the assumptions on the disorder distribution. The argument we give completely revisits and improves the previous proof. As a key novelty we build a probability that is close to the Furstenberg probability, i.e. the invariant probability of the Markov chain corresponding to the evolution of the direction of a vector in ${\mathbb R}^2$ under the action of the random matrices, in terms of the ladder times of a centered random walk which is directly related to the random matrix sequence. We then show that sharp estimates on the ladder times (renewal) process lead to a sharp control on the probability measure we build and, in turn, to the control of its distance from the Furstenberg probability. | [
{
"version": "v1",
"created": "Wed, 15 Jan 2025 13:45:23 GMT"
},
{
"version": "v2",
"created": "Tue, 13 May 2025 11:43:42 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Collin",
"Orphée",
""
],
[
"Giacomin",
"Giambattista",
""
],
[
"Greenblatt",
"Rafael L.",
""
],
[
"Hu",
"Yueyun",
""
]
] |
2501.08967 | Vladimir Yurovsky | Vladimir A. Yurovsky (School of Chemistry, Tel Aviv University) and Amichay Vardi (Department of Chemistry, Ben-Gurion University and ITAMP, Harvard-Smithsonian Center for Astrophysics) | Eigenstate thermalization to non-monotonic distributions in strongly-interacting chaotic lattice gases | Submission to SciPost | null | null | null | cond-mat.quant-gas nlin.CD physics.atom-ph quant-ph | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We find non-monotonic equilibrium energy distributions, qualitatively different from the Fermi-Dirac and Bose-Einstein forms, in strongly-interacting many-body chaotic systems. The effect emerges in systems with finite energy spectra, supporting both positive and negative temperatures, in the regime of quantum ergodicity. The results are supported by exact diagonalization calculations for chaotic Fermi-Hubbard and Bose-Hubbard models, when they have Wigner-Dyson statistics of energy spectra and demonstrate eigenstate thermalization. The proposed effects may be observed in experiments with cold atoms in optical lattices. | [
{
"version": "v1",
"created": "Wed, 15 Jan 2025 17:23:38 GMT"
},
{
"version": "v2",
"created": "Thu, 29 May 2025 15:27:53 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Yurovsky",
"Vladimir A.",
"",
"School of Chemistry, Tel Aviv University"
],
[
"Vardi",
"Amichay",
"",
"Department of Chemistry, Ben-Gurion University and ITAMP, Harvard-Smithsonian Center for Astrophysics"
]
] |
2501.09417 | Alexander N. Poddubny | Jiaming Shi and Alexander N. Poddubny | Chiral Dissociation of Bound Photon Pairs for a Non-Hermitian Skin Effect | 6 pages, 4 figures + supplementary; version 2, to be published in PRL. Code is available at https://zenodo.org/records/15527091 | null | null | null | quant-ph cond-mat.mes-hall physics.optics | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We theoretically study the bound states of interacting photons propagating in a waveguide chirally coupled to an array of atoms. We demonstrate that the bound photon pairs can concentrate at the edge of the array and link this to the non-Hermitian skin effect. Unlike tight-binding non-Hermitian setups, the bound states in the waveguide-coupled array exhibit infinite radiative lifetimes when the array has an infinite size. However, in a finite array, non-Hermiticity and localization of bound pairs emerge due to their chiral dissociation into scattering states. Counterintuitively, when the photons are preferentially emitted to the right, the bound pairs are localized at the left edge of the array and vice versa. | [
{
"version": "v1",
"created": "Thu, 16 Jan 2025 09:44:00 GMT"
},
{
"version": "v2",
"created": "Thu, 29 May 2025 07:30:40 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Shi",
"Jiaming",
""
],
[
"Poddubny",
"Alexander N.",
""
]
] |
2501.09620 | Zhuokai Zhao | Chaoqi Wang, Zhuokai Zhao, Yibo Jiang, Zhaorun Chen, Chen Zhu, Yuxin Chen, Jiayi Liu, Lizhu Zhang, Xiangjun Fan, Hao Ma, Sinong Wang | Beyond Reward Hacking: Causal Rewards for Large Language Model Alignment | null | null | null | null | cs.LG cs.AI | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Recent advances in large language models (LLMs) have demonstrated significant progress in performing complex tasks. While Reinforcement Learning from Human Feedback (RLHF) has been effective in aligning LLMs with human preferences, it is susceptible to spurious correlations in reward modeling. Consequently, it often introduces biases-such as length bias, sycophancy, conceptual bias, and discrimination-that hinder the model's ability to capture true causal relationships. To address this, we propose a novel causal reward modeling approach that integrates causality to mitigate these spurious correlations. Our method enforces counterfactual invariance, ensuring reward predictions remain consistent when irrelevant variables are altered. Through experiments on both synthetic and real-world datasets, we show that our approach mitigates various types of spurious correlations effectively, resulting in more reliable and fair alignment of LLMs with human preferences. As a drop-in enhancement to the existing RLHF workflow, our causal reward modeling provides a practical way to improve the trustworthiness and fairness of LLM finetuning. | [
{
"version": "v1",
"created": "Thu, 16 Jan 2025 16:00:37 GMT"
},
{
"version": "v2",
"created": "Thu, 29 May 2025 02:21:03 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Wang",
"Chaoqi",
""
],
[
"Zhao",
"Zhuokai",
""
],
[
"Jiang",
"Yibo",
""
],
[
"Chen",
"Zhaorun",
""
],
[
"Zhu",
"Chen",
""
],
[
"Chen",
"Yuxin",
""
],
[
"Liu",
"Jiayi",
""
],
[
"Zhang",
"Lizhu",
""
],
[
"Fan",
"Xiangjun",
""
],
[
"Ma",
"Hao",
""
],
[
"Wang",
"Sinong",
""
]
] |
2501.09651 | Joaqu\'in Espinoza-Troni | Joaqu\'in Espinoza-Troni, Giuseppe Arr\`o, Felipe A Asenjo, Pablo S Moya | Electron scale magnetic holes generation driven by Whistler-to-Bernstein mode conversion in fully kinetic plasma turbulence | null | null | null | null | physics.space-ph | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Magnetic holes (MHs) are coherent structures characterized by a strong and localized magnetic field amplitude dip, commonly observed in the solar wind and planetary magnetosheaths. These structures come in different sizes, from magnetohydrodynamic to kinetic scales. Magnetospheric Multiscale (MMS) observations have revealed electron scale MHs to be ubiquitous in the turbulent Earth's magnetosheath, potentially playing an important role in the energy cascade and dissipation. Despite abundant observations, the origin of electron scale MHs is still unclear and debated. In this work, we use fully kinetic simulations to investigate the role of plasma turbulence in generating electron scale MHs. We perform a fully kinetic simulation of freely decaying plasma turbulence, initialized with typical Earth's magnetosheath parameters. We find that electron scale MHs can be generated by turbulence via the following mechanism: first, large-scale turbulent velocity shears produce regions with high electron temperature anisotropy; these localized regions become unstable, generating oblique electron scale whistler waves; as they propagate over the inhomogeneous turbulent background, whistler fluctuations develop an electrostatic component, turning into Bernstein-like modes; the strong electrostatic fluctuations produce current filaments that merge into an electron scale current vortex; the resulting electron vortex locally reduces the magnetic field amplitude, finally evolving into an electron scale MH. We show that MHs generated by this mechanism have properties consistent with MMS observations and nontrivial kinetic features. We provide numerical evidence of a new electron scale MH generation mechanism, driven by turbulence. Our results have potential implications for understanding the formation and occurrence of electron scale MHs in turbulent environments, such as the Earth's magnetosheath. | [
{
"version": "v1",
"created": "Thu, 16 Jan 2025 16:46:26 GMT"
},
{
"version": "v2",
"created": "Wed, 28 May 2025 23:34:29 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Espinoza-Troni",
"Joaquín",
""
],
[
"Arrò",
"Giuseppe",
""
],
[
"Asenjo",
"Felipe A",
""
],
[
"Moya",
"Pablo S",
""
]
] |
2501.10348 | Zizhou Zhang | Zizhou Zhang, Xinshi Li, Yu Cheng, Zhenrui Chen, Qianying Liu | Credit Risk Identification in Supply Chains Using Generative Adversarial Networks | The paper will be published and indexed by IEEE at 2025 8th International Conference on Advanced Algorithms and Control Engineering (ICAACE 2025) | null | null | null | cs.LG | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Credit risk management within supply chains has emerged as a critical research area due to its significant implications for operational stability and financial sustainability. The intricate interdependencies among supply chain participants mean that credit risks can propagate across networks, with impacts varying by industry. This study explores the application of Generative Adversarial Networks (GANs) to enhance credit risk identification in supply chains. GANs enable the generation of synthetic credit risk scenarios, addressing challenges related to data scarcity and imbalanced datasets. By leveraging GAN-generated data, the model improves predictive accuracy while effectively capturing dynamic and temporal dependencies in supply chain data. The research focuses on three representative industries-manufacturing (steel), distribution (pharmaceuticals), and services (e-commerce) to assess industry-specific credit risk contagion. Experimental results demonstrate that the GAN-based model outperforms traditional methods, including logistic regression, decision trees, and neural networks, achieving superior accuracy, recall, and F1 scores. The findings underscore the potential of GANs in proactive risk management, offering robust tools for mitigating financial disruptions in supply chains. Future research could expand the model by incorporating external market factors and supplier relationships to further enhance predictive capabilities. Keywords- Generative Adversarial Networks (GANs); Supply Chain Risk; Credit Risk Identification; Machine Learning; Data Augmentation | [
{
"version": "v1",
"created": "Fri, 17 Jan 2025 18:42:46 GMT"
},
{
"version": "v2",
"created": "Mon, 20 Jan 2025 08:54:34 GMT"
},
{
"version": "v3",
"created": "Fri, 24 Jan 2025 02:14:53 GMT"
},
{
"version": "v4",
"created": "Wed, 28 May 2025 19:41:01 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Zhang",
"Zizhou",
""
],
[
"Li",
"Xinshi",
""
],
[
"Cheng",
"Yu",
""
],
[
"Chen",
"Zhenrui",
""
],
[
"Liu",
"Qianying",
""
]
] |
2501.11106 | Zofia Grochulska | Zofia Grochulska | $L^{1}_{loc}$-convergence of Jacobians of Sobolev homeomorphisms via area formula | 17 pages | null | null | null | math.CA | http://creativecommons.org/licenses/by/4.0/ | We prove that given a sequence of homeomorphisms $f_k: \Omega \to \mathbb{R}^n$ convergent in $W^{1,p}(\Omega, \mathbb{R}^n)$, $p \geq 1$ for $n =2$ and $p > n-1$ for $n \geq 3$, to a homeomorphism $f$ which maps sets of measure zero onto sets of measure zero, Jacobians $Jf_k$ converge to $Jf$ in $L^1_{loc}(\Omega)$. We prove it via Federer's area formula and investigation of when $|f_k(E)| \to |f(E)|$ as $k \to \infty$ for Borel subsets $E \Subset \Omega$. | [
{
"version": "v1",
"created": "Sun, 19 Jan 2025 16:34:19 GMT"
},
{
"version": "v2",
"created": "Thu, 29 May 2025 12:15:03 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Grochulska",
"Zofia",
""
]
] |
2501.11437 | Masatake Hirao | Kenji Tanino, Tomoki Tamaru, Masatake Hirao, Masanori Sawa | More on the corner-vector construction for spherical designs | null | null | null | null | math.CO cs.NA math.NA | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | This paper explores a full generalization of the classical corner-vector method for constructing weighted spherical designs, which we call the {\it generalized corner-vector method}. First we establish a uniform upper bound for the degree of designs obtained from the proposed method. Our proof is a hybrid argument that employs techniques in analysis and combinatorics, especially a famous result by Xu(1998) on the interrelation between spherical designs and simplical designs, and the cross-ratio comparison method for Hilbert identities introduced by Nozaki and Sawa(2013). We extensively study conditions for the existence of designs obtained from our method, and present many curious examples of degree $7$ through $13$, some of which are, to our surprise, characterized in terms of integral lattices. | [
{
"version": "v1",
"created": "Mon, 20 Jan 2025 12:21:45 GMT"
},
{
"version": "v2",
"created": "Thu, 29 May 2025 00:35:42 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Tanino",
"Kenji",
""
],
[
"Tamaru",
"Tomoki",
""
],
[
"Hirao",
"Masatake",
""
],
[
"Sawa",
"Masanori",
""
]
] |
2501.11722 | Anton Vershovskii K. | Mikhail V. Petrenko and Anton K. Vershovskii | Anomalous suppression of spin-exchange relaxation in alignment signals in cesium in ultra-weak magnetic fields | 13 pages, 11 figures | null | null | null | physics.atom-ph | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | The results of a study of the dynamics of atomic moments alignment in cesium under optical pumping by linearly polarized resonant light in ultra-weak magnetic field are presented. It is shown that there are alignment components whose relaxation does not depend on spin-exchange broadening. The effect of suppression of spin-exchange relaxation in zero magnetic fields is detected, which is similar in its manifestations to the SERF (Spin-Exchange Relaxation Free) effect observed in orientation signals. This observation is interesting from the standpoint of general theory, since the law of conservation of angular momentum responsible for maintaining orientation in the SERF mode should not guarantee the preservation of alignment under the same conditions. A comparison with theoretically calculated parameters of SERF resonances in orientation is given. A qualitative explanation of the observed effect is presented. | [
{
"version": "v1",
"created": "Mon, 20 Jan 2025 20:09:23 GMT"
},
{
"version": "v2",
"created": "Fri, 18 Apr 2025 15:18:08 GMT"
},
{
"version": "v3",
"created": "Thu, 29 May 2025 09:51:32 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Petrenko",
"Mikhail V.",
""
],
[
"Vershovskii",
"Anton K.",
""
]
] |
2501.13114 | Purbita Jana | Purbita Jana and Prateek | Continuous Algebra: Algebraic Semantics for Continuous Propositional Logic | null | null | null | null | math.LO | http://creativecommons.org/publicdomain/zero/1.0/ | We have introduced continuous algebra as the algebraic semantics for Continuous Propositional Logic (CPL). A Continuous algebra is an MV-algebra together with an unary operator $\kappa$, analogous to the unary connective $\dfrac{1}{2}$ in CPL. We establish structural results, including the subdirect representation theorem. We also introduce $\ell u^*$-groups, which are lattice ordered groups with strong unit $u$, denoted by $\ell u$-groups, with a partial operator $^*$ that mimics the behavior of $\kappa$ over the interval $[id,u]$. This addition enables a natural correspondence between $\ell u^*$-groups and the continuous algebras, allowing us to prove the Chang's completeness theorem for the continuous algebras. | [
{
"version": "v1",
"created": "Sat, 18 Jan 2025 11:28:45 GMT"
},
{
"version": "v2",
"created": "Thu, 29 May 2025 07:49:39 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Jana",
"Purbita",
""
],
[
"Prateek",
"",
""
]
] |
2501.13210 | Andrei Lazanu | Philippe Brax and Andrei Lazanu | Primordial gravitational waves in DHOST inflation | 7 pages, 2 figures, published in MNRAS Letters | Mon.Not.Roy.Astron.Soc. 540 (2025) 1, L98-L103 | 10.1093/mnrasl/slaf040 | null | astro-ph.CO gr-qc | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We consider DHOST inflationary models with a shift symmetry leading to a de Sitter space-time at the background cosmological level. Deviations from scale invariance of the scalar and tensor perturbations follow from the breaking of the shift symmetry by quadratic and quartic operators. These models show a strong violation of the consistency relation of single-field inflationary models with a very flat spectrum of tensor perturbations. This opens up the prospects of future detection of primordial gravitational waves by mHz experiments. | [
{
"version": "v1",
"created": "Wed, 22 Jan 2025 20:49:46 GMT"
},
{
"version": "v2",
"created": "Thu, 29 May 2025 15:36:01 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Brax",
"Philippe",
""
],
[
"Lazanu",
"Andrei",
""
]
] |
2501.13679 | Jana Men\v{s}\'ikov\'a | Tom\'a\v{s} Hale, David Kubiz\v{n}\'ak, Jana Men\v{s}\'ikov\'a, Robert B. Mann, Jiayue Yang | On thermodynamics of charged and accelerating black holes | null | Phys. Rev. D 111, 104004 (2025) | 10.1103/PhysRevD.111.104004 | null | gr-qc hep-th | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We reconsider various C-metric spacetimes describing charged and (slowly) accelerating AdS black holes in different theories of (non-linear) electrodynamics and revisit their thermodynamic properties. Focusing first on the Maxwell theory, we find a parametrization of the metric where we can eliminate the non-trivial `normalization' of the boost Killing vector which was crucial for obtaining consistent thermodynamics in previous studies. We also calculate the Euclidean action using i) the standard holographic renormalization and ii) using the topological renormalization, showing that in the presence of overall cosmic string tension the two do not agree. These results are also extended to accelerating black holes in ModMax and RegMax non-linear electrodynamics. Interestingly, for the latter the electrostatic potential picks up a modification, that remains to be explained, but is consistent with the topological renormalization and the (generalized) Hawking--Ross prescription. Our study indicates that thermodynamics of charged accelerating black holes is far from being completely understood. | [
{
"version": "v1",
"created": "Thu, 23 Jan 2025 14:04:36 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Hale",
"Tomáš",
""
],
[
"Kubizňák",
"David",
""
],
[
"Menšíková",
"Jana",
""
],
[
"Mann",
"Robert B.",
""
],
[
"Yang",
"Jiayue",
""
]
] |
2501.13890 | Ayush Mohanty | Ayush Mohanty, Nazal Mohamed, Paritosh Ramanan and Nagi Gebraeel | Federated Granger Causality Learning for Interdependent Clients with State Space Representation | Published as a conference paper at International Conference on Learning Representations (ICLR) 2025 | null | null | null | cs.LG stat.ML | http://creativecommons.org/licenses/by/4.0/ | Advanced sensors and IoT devices have improved the monitoring and control of complex industrial enterprises. They have also created an interdependent fabric of geographically distributed process operations (clients) across these enterprises. Granger causality is an effective approach to detect and quantify interdependencies by examining how one client's state affects others over time. Understanding these interdependencies captures how localized events, such as faults and disruptions, can propagate throughout the system, possibly causing widespread operational impacts. However, the large volume and complexity of industrial data pose challenges in modeling these interdependencies. This paper develops a federated approach to learning Granger causality. We utilize a linear state space system framework that leverages low-dimensional state estimates to analyze interdependencies. This addresses bandwidth limitations and the computational burden commonly associated with centralized data processing. We propose augmenting the client models with the Granger causality information learned by the server through a Machine Learning (ML) function. We examine the co-dependence between the augmented client and server models and reformulate the framework as a standalone ML algorithm providing conditions for its sublinear and linear convergence rates. We also study the convergence of the framework to a centralized oracle model. Moreover, we include a differential privacy analysis to ensure data security while preserving causal insights. Using synthetic data, we conduct comprehensive experiments to demonstrate the robustness of our approach to perturbations in causality, the scalability to the size of communication, number of clients, and the dimensions of raw data. We also evaluate the performance on two real-world industrial control system datasets by reporting the volume of data saved by decentralization. | [
{
"version": "v1",
"created": "Thu, 23 Jan 2025 18:04:21 GMT"
},
{
"version": "v2",
"created": "Mon, 27 Jan 2025 16:49:19 GMT"
},
{
"version": "v3",
"created": "Mon, 03 Mar 2025 22:33:36 GMT"
},
{
"version": "v4",
"created": "Thu, 29 May 2025 14:27:43 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Mohanty",
"Ayush",
""
],
[
"Mohamed",
"Nazal",
""
],
[
"Ramanan",
"Paritosh",
""
],
[
"Gebraeel",
"Nagi",
""
]
] |
2501.13977 | Anshuman Chhabra | Rajvardhan Oak, Muhammad Haroon, Claire Jo, Magdalena Wojcieszak, Anshuman Chhabra | Re-ranking Using Large Language Models for Mitigating Exposure to Harmful Content on Social Media Platforms | Accepted to ACL 2025 Main Conference | null | null | null | cs.CL cs.AI cs.CY cs.SI | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Social media platforms utilize Machine Learning (ML) and Artificial Intelligence (AI) powered recommendation algorithms to maximize user engagement, which can result in inadvertent exposure to harmful content. Current moderation efforts, reliant on classifiers trained with extensive human-annotated data, struggle with scalability and adapting to new forms of harm. To address these challenges, we propose a novel re-ranking approach using Large Language Models (LLMs) in zero-shot and few-shot settings. Our method dynamically assesses and re-ranks content sequences, effectively mitigating harmful content exposure without requiring extensive labeled data. Alongside traditional ranking metrics, we also introduce two new metrics to evaluate the effectiveness of re-ranking in reducing exposure to harmful content. Through experiments on three datasets, three models and across three configurations, we demonstrate that our LLM-based approach significantly outperforms existing proprietary moderation approaches, offering a scalable and adaptable solution for harm mitigation. | [
{
"version": "v1",
"created": "Thu, 23 Jan 2025 00:26:32 GMT"
},
{
"version": "v2",
"created": "Fri, 16 May 2025 13:25:28 GMT"
},
{
"version": "v3",
"created": "Thu, 29 May 2025 12:42:55 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Oak",
"Rajvardhan",
""
],
[
"Haroon",
"Muhammad",
""
],
[
"Jo",
"Claire",
""
],
[
"Wojcieszak",
"Magdalena",
""
],
[
"Chhabra",
"Anshuman",
""
]
] |
2501.13978 | Sangyeop Yeo | Sangyeop Yeo, Seung-won Hwang, Yu-Seung Ma | Chain of Grounded Objectives: Bridging Process and Goal-oriented Prompting for Code Generation | Accepted by ECOOP 2025 main conference | null | null | null | cs.CL cs.AI cs.SE | http://creativecommons.org/licenses/by/4.0/ | The use of Large Language Models (LLMs) for code generation has gained significant attention in recent years. Existing methods often aim to improve the quality of generated code by incorporating additional contextual information or guidance into input prompts. Many of these approaches adopt sequential reasoning strategies, mimicking human-like step-by-step thinking. However, such strategies may constrain flexibility, as they do not always align with the structured characteristics of programming languages. This paper introduces the Chain of Grounded Objectives (CGO), a method that embeds functional objectives into input prompts to enhance code generation. By leveraging appropriately structured objectives as input and avoiding explicit sequential procedures, CGO adapts effectively to the structured nature of programming tasks. Empirical evaluations demonstrate that CGO effectively enhances code generation, addressing limitations of existing approaches. | [
{
"version": "v1",
"created": "Thu, 23 Jan 2025 01:45:09 GMT"
},
{
"version": "v2",
"created": "Thu, 29 May 2025 02:28:30 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Yeo",
"Sangyeop",
""
],
[
"Hwang",
"Seung-won",
""
],
[
"Ma",
"Yu-Seung",
""
]
] |
2501.14248 | Lanjing Zhang | Fei Deng (1), Catherine H Feng (1 and 2), Nan Gao (3 and 4), Lanjing Zhang (1, 4, 5 and 6) ((1) Department of Chemical Biology, Ernest Mario School of Pharmacy, Rutgers University, Piscataway, NJ, (2) Harvard University, Cambridge, MA, (3) Department of Biological Sciences, School of Arts & Sciences, Rutgers University, Newark, NJ, (4) Department of Pharmacology, Physiology, and Neuroscience, New Jersey Medical School, Rutgers University, Newark, NJ, (5) Department of Pathology, Princeton Medical Center, Plainsboro, NJ, (6) Rutgers Cancer Institute of New Jersey, New Brunswick, NJ.) | Normalization and selecting non-differentially expressed genes improve machine learning modelling of cross-platform transcriptomic data | 35 pages, 5 figures, 2 tables | Trans. Artif. Intell. 2025, 1(1), 5 | null | https://media.sciltp.com/articles/2505000683/2505000683.pdf | q-bio.QM q-bio.GN stat.CO stat.ME | http://creativecommons.org/licenses/by-nc-nd/4.0/ | Normalization is a critical step in quantitative analyses of biological processes. Recent works show that cross-platform integration and normalization enable machine learning (ML) training on RNA microarray and RNA-seq data, but no independent datasets were used in their studies. Therefore, it is unclear how to improve ML modelling performance on independent RNA array and RNA-seq based datasets. Inspired by the house-keeping genes that are commonly used in experimental biology, this study tests the hypothesis that non-differentially expressed genes (NDEG) may improve normalization of transcriptomic data and subsequently cross-platform modelling performance of ML models. Microarray and RNA-seq datasets of the TCGA breast cancer were used as independent training and test datasets, respectively, to classify the molecular subtypes of breast cancer. NDEG (p>0.85) and differentially expressed genes (DEG, p<0.05) were selected based on the p values of ANOVA analysis and used for subsequent data normalization and classification, respectively. Models trained based on data from one platform were used for testing on the other platform. Our data show that NDEG and DEG gene selection could effectively improve the model classification performance. Normalization methods based on parametric statistical analysis were inferior to those based on nonparametric statistics. In this study, the LOG_QN and LOG_QNZ normalization methods combined with the neural network classification model seem to achieve better performance. Therefore, NDEG-based normalization appears useful for cross-platform testing on completely independent datasets. However, more studies are required to examine whether NDEG-based normalization can improve ML classification performance in other datasets and other omic data types. | [
{
"version": "v1",
"created": "Fri, 24 Jan 2025 05:23:51 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Deng",
"Fei",
"",
"Department of Chemical Biology, Ernest Mario School of Pharmacy, Rutgers University, Piscataway, NJ"
],
[
"Feng",
"Catherine H",
"",
"1 and 2"
],
[
"Gao",
"Nan",
"",
"3 and 4"
],
[
"Zhang",
"Lanjing",
"",
"1, 4, 5 and 6"
]
] |
2501.14401 | Pratik Mazumder | Souvik Maji, Rhythm Baghel, Pratik Mazumder | CVOCSemRPL: Class-Variance Optimized Clustering, Semantic Information Injection and Restricted Pseudo Labeling based Improved Semi-Supervised Few-Shot Learning | null | null | null | null | cs.CV | http://creativecommons.org/licenses/by-nc-nd/4.0/ | Few-shot learning has been extensively explored to address problems where the amount of labeled samples is very limited for some classes. In the semi-supervised few-shot learning setting, substantial quantities of unlabeled samples are available. Such unlabeled samples are generally cheaper to obtain and can be used to improve the few-shot learning performance of the model. Some of the recent methods for this setting rely on clustering to generate pseudo-labels for the unlabeled samples. Since the effectiveness of clustering heavily influences the labeling of the unlabeled samples, it can significantly affect the few-shot learning performance. In this paper, we focus on improving the representation learned by the model in order to improve the clustering and, consequently, the model performance. We propose an approach for semi-supervised few-shot learning that performs a class-variance optimized clustering coupled with a cluster separation tuner in order to improve the effectiveness of clustering the labeled and unlabeled samples in this setting. It also optimizes the clustering-based pseudo-labeling process using a restricted pseudo-labeling approach and performs semantic information injection in order to improve the semi-supervised few-shot learning performance of the model. We experimentally demonstrate that our proposed approach significantly outperforms recent state-of-the-art methods on the benchmark datasets. | [
{
"version": "v1",
"created": "Fri, 24 Jan 2025 11:14:35 GMT"
},
{
"version": "v2",
"created": "Thu, 29 May 2025 17:23:17 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Maji",
"Souvik",
""
],
[
"Baghel",
"Rhythm",
""
],
[
"Mazumder",
"Pratik",
""
]
] |
2501.15370 | Chibuike Robinson Umeike | Robinson Umeike, Neil Getty, Fangfang Xia, and Rick Stevens | Scaling Large Vision-Language Models for Enhanced Multimodal Comprehension In Biomedical Image Analysis | 4 Pages, 4 Figures, 1 Table | null | 10.1109/ISBI60581.2025.10981159 | null | cs.CV cs.AI | http://creativecommons.org/licenses/by-nc-nd/4.0/ | Large language models (LLMs) have demonstrated immense capabilities in understanding textual data and are increasingly being adopted to help researchers accelerate scientific discovery through knowledge extraction (information retrieval), knowledge distillation (summarizing key findings and methodologies into concise forms), and knowledge synthesis (aggregating information from multiple scientific sources to address complex queries, generate hypothesis and formulate experimental plans). However, scientific data often exists in both visual and textual modalities. Vision language models (VLMs) address this by incorporating a pretrained vision backbone for processing images and a cross-modal projector that adapts image tokens into the LLM dimensional space, thereby providing richer multimodal comprehension. Nevertheless, off-the-shelf VLMs show limited capabilities in handling domain-specific data and are prone to hallucinations. We developed intelligent assistants finetuned from LLaVA models to enhance multimodal understanding in low-dose radiation therapy (LDRT)-a benign approach used in the treatment of cancer-related illnesses. Using multilingual data from 42,673 articles, we devise complex reasoning and detailed description tasks for visual question answering (VQA) benchmarks. Our assistants, trained on 50,882 image-text pairs, demonstrate superior performance over base models as evaluated using LLM-as-a-judge approach, particularly in reducing hallucination and improving domain-specific comprehension. | [
{
"version": "v1",
"created": "Sun, 26 Jan 2025 02:48:01 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Umeike",
"Robinson",
""
],
[
"Getty",
"Neil",
""
],
[
"Xia",
"Fangfang",
""
],
[
"Stevens",
"Rick",
""
]
] |
2501.15859 | Cl\'ement de Seguins Pazzis | Cl\'ement de Seguins Pazzis | The Free Hamilton Algebra | 167 pages (v2: extended bibliography and minor corrections) | null | null | null | math.RA | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Over an arbitrary field $\mathbb{F}$, let $p$ and $q$ be monic polynomials with degree $2$ in $\mathbb{F}[t]$. The free Hamilton algebra of the pair $(p,q)$ is the free noncommutative algebra in two generators $a$ and $b$ subject only to the relations $p(a)=0=q(b)$. Free Hamilton algebras are models of free products of two $2$-dimensional algebras over $\mathbb{F}$. They can be viewed as the most elementary nontrivial noncommutative algebras over fields.
It has been recently observed that the free Hamilton algebra has surprising connections with quaternion algebras. Here, we exploit these connections to investigate its zero divisors, group of units, maximal ideals, finite-dimensional subalgebras, and its automorphism group. | [
{
"version": "v1",
"created": "Mon, 27 Jan 2025 08:35:52 GMT"
},
{
"version": "v2",
"created": "Thu, 29 May 2025 09:11:09 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Pazzis",
"Clément de Seguins",
""
]
] |
2501.16143 | Marco Zambianco | Marco Zambianco, Silvio Cretti, Domenico Siracusa | Disruption-aware Microservice Re-orchestration for Cost-efficient Multi-cloud Deployments | null | null | null | null | cs.NI | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Multi-cloud environments enable a cost-efficient scaling of cloud-native applications across geographically distributed virtual nodes with different pricing models. In this context, the resource fragmentation caused by frequent changes in the resource demands of deployed microservices, along with the allocation or termination of new and existing microservices, increases the deployment cost. Therefore, re-orchestrating deployed microservices on a cheaper configuration of multi-cloud nodes offers a practical solution to restore the cost efficiency of deployment. However, the rescheduling procedure causes frequent service interruptions due to the continuous termination and rebooting of the containerized microservices. Moreover, it may potentially interfere with and delay other deployment operations, compromising the stability of the running applications. To address this issue, we formulate a multi-objective integer linear programming (ILP) problem that computes a microservice rescheduling solution capable of providing minimum deployment cost without significantly affecting the service continuity. At the same time, the proposed formulation also preserves the quality of service (QoS) requirements, including latency, expressed through microservice co-location constraints. Additionally, we present a heuristic algorithm to approximate the optimal solution, striking a balance between cost reduction and service disruption mitigation. We integrate the proposed approach as a custom plugin of the Kubernetes (K8s) scheduler. Results reveal that our approach significantly reduces multi-cloud deployment costs and service disruptions compared to the benchmark schemes, while ensuring QoS requirements are consistently met. | [
{
"version": "v1",
"created": "Mon, 27 Jan 2025 15:36:51 GMT"
},
{
"version": "v2",
"created": "Thu, 29 May 2025 14:52:07 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Zambianco",
"Marco",
""
],
[
"Cretti",
"Silvio",
""
],
[
"Siracusa",
"Domenico",
""
]
] |
2501.16472 | Edwan Pr\'eau | Edwan Pr\'eau | Revisiting color superconductors in bottom-up holography | 27 pages + Appendix; 14 figures. v3: added the last section with an example where the transition is first order | J. High Energ. Phys. 2025, 204 (2025) | 10.1007/JHEP05(2025)204 | null | hep-th hep-ph | http://creativecommons.org/licenses/by/4.0/ | We revisit the problem of constructing a color superconducting phase in bottom-up holography. We introduce a model that describes the five-dimensional dynamics of a scalar field dual to a chiral symmetry breaking condensate, which is coupled to the RR 4-form. The coupling is given by a topological WZ term, which is elegantly responsible both for scalar condensation at high density, and color Higgsing in the condensed phase. This construction realizes both properties at finite density and Higgsing fraction. | [
{
"version": "v1",
"created": "Mon, 27 Jan 2025 20:05:48 GMT"
},
{
"version": "v2",
"created": "Tue, 25 Feb 2025 09:58:22 GMT"
},
{
"version": "v3",
"created": "Wed, 28 May 2025 19:50:19 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Préau",
"Edwan",
""
]
] |
2501.17191 | Miao Li | Miao Li, Jey Han Lau, Eduard Hovy, Mirella Lapata | Decomposed Opinion Summarization with Verified Aspect-Aware Modules | 37 pages, long paper, present at ACL 2025 | null | null | null | cs.CL cs.IR | http://creativecommons.org/licenses/by/4.0/ | Opinion summarization plays a key role in deriving meaningful insights from large-scale online reviews. To make the process more explainable and grounded, we propose a domain-agnostic modular approach guided by review aspects (e.g., cleanliness for hotel reviews) which separates the tasks of aspect identification, opinion consolidation, and meta-review synthesis to enable greater transparency and ease of inspection. We conduct extensive experiments across datasets representing scientific research, business, and product domains. Results show that our approach generates more grounded summaries compared to strong baseline models, as verified through automated and human evaluations. Additionally, our modular approach, which incorporates reasoning based on review aspects, produces more informative intermediate outputs than other knowledge-agnostic decomposition approaches. Lastly, we provide empirical results to show that these intermediate outputs can support humans in summarizing opinions from large volumes of reviews. | [
{
"version": "v1",
"created": "Mon, 27 Jan 2025 09:29:55 GMT"
},
{
"version": "v2",
"created": "Tue, 18 Feb 2025 06:30:09 GMT"
},
{
"version": "v3",
"created": "Wed, 28 May 2025 18:31:34 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Li",
"Miao",
""
],
[
"Lau",
"Jey Han",
""
],
[
"Hovy",
"Eduard",
""
],
[
"Lapata",
"Mirella",
""
]
] |
2501.17432 | Vladimir Gavrichkov A. | V. A. Gavrichkov and S. I. Polukeev and S. G. Ovchinnikov | Diagonal superexchange in a simple square CuO$_2$ lattice | 12 pages, 4 figures | null | null | null | cond-mat.str-el | http://creativecommons.org/licenses/by/4.0/ | Many microscopic models with the interaction between the next-nearest neighbours as a key parameter for cuprate physics have inspired us to study the diagonal superexchange interaction in a CuO$_2$ layer. Our investigation shows that models with extended hopping provide a correct representation of magnetic interactions only in a hypothetical square CuO$_2$ layer, where the diagonal superexchange interaction with the next-nearest neighbors always has the AFM nature. The conclusions are based on the symmetry prohibition on FM contribution to the diagonal superexchange between the next-nearest neighbors for a simple square CuO$_2$ layer rather than for a real CuO$_2$ layer, where diagonal AFM superexchange may be overestimated. We also discuss the reasons for magnetic frustration effects and high sensitivity of spin nanoinhomogeneity to square symmetry breaking. | [
{
"version": "v1",
"created": "Wed, 29 Jan 2025 06:18:47 GMT"
},
{
"version": "v2",
"created": "Wed, 05 Feb 2025 05:09:00 GMT"
},
{
"version": "v3",
"created": "Thu, 29 May 2025 10:57:25 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Gavrichkov",
"V. A.",
""
],
[
"Polukeev",
"S. I.",
""
],
[
"Ovchinnikov",
"S. G.",
""
]
] |
2501.18221 | Antonio Irpino PhD | Andrea Diana, Elvira Romano, Antonio Irpino | Network Weighted Functional Regression: a method for modeling dependencies between functional data in a network | null | null | null | null | stat.ME | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | This paper focuses on predicting continuous signals in a sensor lab network, particularly studying microclimate changes. We propose two novel concepts: Network Functional Data (NFD), which represents time series signals as functions on network nodes, and the Network Weighted Functional Regression (NWFR) model, which analyzes relationships between functional responses and predictors in a weighted network. Additionally, we introduce a functional conformal method to provide prediction bands with guaranteed coverage probabilities, independent of data distribution.
Our statistical analysis on simulated and real-world data demonstrates that incorporating network structure enhances regression accuracy and improves the reliability of conformal prediction regions. These findings advance the analysis of complex network-structured data, offering a more precise and efficient approach. | [
{
"version": "v1",
"created": "Thu, 30 Jan 2025 09:23:31 GMT"
},
{
"version": "v2",
"created": "Thu, 29 May 2025 14:56:41 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Diana",
"Andrea",
""
],
[
"Romano",
"Elvira",
""
],
[
"Irpino",
"Antonio",
""
]
] |
2501.18411 | Nolan Koblischke | Nolan Koblischke, Hyunseok Jang, Kristen Menou, Mohamad Ali-Dib | Gravity-Bench-v1: A Benchmark on Gravitational Physics Discovery for Agents | Accepted at ICML 2025 | null | null | null | cs.AI astro-ph.IM physics.comp-ph | http://creativecommons.org/licenses/by/4.0/ | Modern science emerged from reasoning over repeatedly-observed planetary motions. We present Gravity-Bench-v1, an environment-based benchmark that challenges AI agents on tasks that parallel this historical development. Gravity-Bench-v1 evaluates agents on the discovery of physics concealed within a dynamic environment, using rigorous gravitational dynamics simulations. Gravity-Bench includes out-of-distribution cases, i.e. with physics that deviates from the real world, to evaluate true scientific generalization capabilities. Agents must plan to collect data within an experimental budget and must perform a dynamic form of data analysis and reasoning to solve tasks efficiently. Our benchmark admits an open-ended space of solutions. Reference solutions for each task are provided to calibrate AI performance against human expertise. Technically at an upper-undergraduate level, our benchmark proves challenging to baseline AI agents. Gravity-Bench-v1 and planned extensions should help map out AI progress towards scientific discovery capabilities. | [
{
"version": "v1",
"created": "Thu, 30 Jan 2025 15:06:34 GMT"
},
{
"version": "v2",
"created": "Thu, 29 May 2025 01:35:03 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Koblischke",
"Nolan",
""
],
[
"Jang",
"Hyunseok",
""
],
[
"Menou",
"Kristen",
""
],
[
"Ali-Dib",
"Mohamad",
""
]
] |
2501.18463 | Shiho Noda | Shiho Noda, Atsuyuki Miyai, Qing Yu, Go Irie, Kiyoharu Aizawa | A Benchmark and Evaluation for Real-World Out-of-Distribution Detection Using Vision-Language Models | Accepted at ICIP2025 Dataset and Benchmark Track | null | null | null | cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Out-of-distribution (OOD) detection is a task that detects OOD samples during inference to ensure the safety of deployed models. However, conventional benchmarks have reached performance saturation, making it difficult to compare recent OOD detection methods. To address this challenge, we introduce three novel OOD detection benchmarks that enable a deeper understanding of method characteristics and reflect real-world conditions. First, we present ImageNet-X, designed to evaluate performance under challenging semantic shifts. Second, we propose ImageNet-FS-X for full-spectrum OOD detection, assessing robustness to covariate shifts (feature distribution shifts). Finally, we propose Wilds-FS-X, which extends these evaluations to real-world datasets, offering a more comprehensive testbed. Our experiments reveal that recent CLIP-based OOD detection methods struggle to varying degrees across the three proposed benchmarks, and none of them consistently outperforms the others. We hope the community goes beyond specific benchmarks and includes more challenging conditions reflecting real-world scenarios. The code is https://github.com/hoshi23/OOD-X-Benchmarks. | [
{
"version": "v1",
"created": "Thu, 30 Jan 2025 16:30:20 GMT"
},
{
"version": "v2",
"created": "Mon, 03 Feb 2025 15:51:16 GMT"
},
{
"version": "v3",
"created": "Thu, 29 May 2025 15:56:25 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Noda",
"Shiho",
""
],
[
"Miyai",
"Atsuyuki",
""
],
[
"Yu",
"Qing",
""
],
[
"Irie",
"Go",
""
],
[
"Aizawa",
"Kiyoharu",
""
]
] |
2501.18584 | Kouichi Yasui | Kouichi Yasui | Corks, exotic 4-manifolds and genus functions | 51 pages, 5 figures, Corollary 1.18 added, exposition improved, typos fixed | null | null | null | math.GT math.SG | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We prove that every 4-dimensional oriented handlebody without 3- and 4-handles can be modified to admit infinitely many exotic smooth structures, and further that their genus functions are pairwise equivalent. We furthermore show that, for any 4-manifold admitting an embedding into a symplectic 4-manifold with weakly convex boundary, its genus function is algebraically realized as those of infinitely many pairwise exotic 4-manifolds. We also prove that, for any two 4-manifolds with non-degenerate intersection forms, algebraic inequivalences of genus functions are stable under connected sums and boundary sums with a certain type of 4-manifolds having arbitrarily large second Betti numbers. Furthermore, we introduce a notion of genus function type for diffeomorphism invariants and show that similar results hold for any such invariant. As an application of our exotic 4-manifolds, we also prove that for any (possibly non-orientable) 4-manifold, every submanifold of codimension less than two satisfying a mild condition can be modified to admit infinitely many exotically knotted copies. | [
{
"version": "v1",
"created": "Thu, 30 Jan 2025 18:55:55 GMT"
},
{
"version": "v2",
"created": "Thu, 06 Mar 2025 18:59:19 GMT"
},
{
"version": "v3",
"created": "Thu, 29 May 2025 17:19:13 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Yasui",
"Kouichi",
""
]
] |
2501.18887 | Shichang Zhang | Shichang Zhang, Tessa Han, Usha Bhalla, Himabindu Lakkaraju | Towards Unified Attribution in Explainable AI, Data-Centric AI, and Mechanistic Interpretability | null | null | null | null | cs.LG cs.AI | http://creativecommons.org/licenses/by/4.0/ | The increasing complexity of AI systems has made understanding their behavior critical. Numerous interpretability methods have been developed to attribute model behavior to three key aspects: input features, training data, and internal model components, which emerged from explainable AI, data-centric AI, and mechanistic interpretability, respectively. However, these attribution methods are studied and applied rather independently, resulting in a fragmented landscape of methods and terminology. This position paper argues that feature, data, and component attribution methods share fundamental similarities, and a unified view of them benefits both interpretability and broader AI research. To this end, we first analyze popular methods for these three types of attributions and present a unified view demonstrating that these seemingly distinct methods employ similar techniques (such as perturbations, gradients, and linear approximations) over different aspects and thus differ primarily in their perspectives rather than techniques. Then, we demonstrate how this unified view enhances understanding of existing attribution methods, highlights shared concepts and evaluation criteria among these methods, and leads to new research directions both in interpretability research, by addressing common challenges and facilitating cross-attribution innovation, and in AI more broadly, with applications in model editing, steering, and regulation. | [
{
"version": "v1",
"created": "Fri, 31 Jan 2025 04:42:45 GMT"
},
{
"version": "v2",
"created": "Thu, 13 Feb 2025 22:59:22 GMT"
},
{
"version": "v3",
"created": "Thu, 29 May 2025 16:49:00 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Zhang",
"Shichang",
""
],
[
"Han",
"Tessa",
""
],
[
"Bhalla",
"Usha",
""
],
[
"Lakkaraju",
"Himabindu",
""
]
] |
2501.18922 | Haoran Luo | Haoran Luo, Haihong E, Yikai Guo, Qika Lin, Xiaobao Wu, Xinyu Mu, Wenhao Liu, Meina Song, Yifan Zhu, Luu Anh Tuan | KBQA-o1: Agentic Knowledge Base Question Answering with Monte Carlo Tree Search | Accepted by ICML 2025 main conference | ICML 2025 | null | null | cs.CL cs.AI cs.DB | http://creativecommons.org/licenses/by-nc-sa/4.0/ | Knowledge Base Question Answering (KBQA) aims to answer natural language questions with a large-scale structured knowledge base (KB). Despite advancements with large language models (LLMs), KBQA still faces challenges in weak KB awareness, imbalance between effectiveness and efficiency, and high reliance on annotated data. To address these challenges, we propose KBQA-o1, a novel agentic KBQA method with Monte Carlo Tree Search (MCTS). It introduces a ReAct-based agent process for stepwise logical form generation with KB environment exploration. Moreover, it employs MCTS, a heuristic search method driven by policy and reward models, to balance agentic exploration's performance and search space. With heuristic exploration, KBQA-o1 generates high-quality annotations for further improvement by incremental fine-tuning. Experimental results show that KBQA-o1 outperforms previous low-resource KBQA methods with limited annotated data, boosting Llama-3.1-8B model's GrailQA F1 performance to 78.5% compared to 48.5% of the previous sota method with GPT-3.5-turbo. Our code is publicly available. | [
{
"version": "v1",
"created": "Fri, 31 Jan 2025 06:59:49 GMT"
},
{
"version": "v2",
"created": "Tue, 27 May 2025 08:36:37 GMT"
},
{
"version": "v3",
"created": "Thu, 29 May 2025 08:10:07 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Luo",
"Haoran",
""
],
[
"E",
"Haihong",
""
],
[
"Guo",
"Yikai",
""
],
[
"Lin",
"Qika",
""
],
[
"Wu",
"Xiaobao",
""
],
[
"Mu",
"Xinyu",
""
],
[
"Liu",
"Wenhao",
""
],
[
"Song",
"Meina",
""
],
[
"Zhu",
"Yifan",
""
],
[
"Tuan",
"Luu Anh",
""
]
] |
2501.19050 | Ruigang Wang | Ruigang Wang, Krishnamurthy Dvijotham, Ian R. Manchester | Norm-Bounded Low-Rank Adaptation | null | null | null | null | cs.LG | http://creativecommons.org/licenses/by/4.0/ | In this work, we propose norm-bounded low-rank adaptation (NB-LoRA) for parameter-efficient fine tuning. NB-LoRA is a novel parameterization of low-rank weight adaptations that admits explicit bounds on each singular value of the adaptation matrix, which can thereby satisfy any prescribed unitarily invariant norm bound, including the Schatten norms (e.g., nuclear, Frobenius, spectral norm). The proposed parameterization is unconstrained, smooth, and complete, i.e. it covers all matrices satisfying the prescribed rank and singular-value bounds. Comparative experiments on large language models show that NB-LoRA achieves superior adaptation performance and faster training over a range of models, tasks and ranks. Vision fine-tuning experiments show that NB-LoRA can achieve strong adaptation performance while avoiding model catastrophic forgetting, and compared to existing approaches it is substantially more robust to a hyper-parameters such as including adaptation rank, learning rate and number of training epochs. | [
{
"version": "v1",
"created": "Fri, 31 Jan 2025 11:24:57 GMT"
},
{
"version": "v2",
"created": "Mon, 26 May 2025 00:27:40 GMT"
},
{
"version": "v3",
"created": "Wed, 28 May 2025 22:04:11 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Wang",
"Ruigang",
""
],
[
"Dvijotham",
"Krishnamurthy",
""
],
[
"Manchester",
"Ian R.",
""
]
] |
2501.19300 | Xutong Liu | Xutong Liu, Xiangxiang Dai, Jinhang Zuo, Siwei Wang, Carlee Joe-Wong, John C.S. Lui, Wei Chen | Offline Learning for Combinatorial Multi-armed Bandits | null | null | null | null | cs.LG | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | The combinatorial multi-armed bandit (CMAB) is a fundamental sequential decision-making framework, extensively studied over the past decade. However, existing work primarily focuses on the online setting, overlooking the substantial costs of online interactions and the readily available offline datasets. To overcome these limitations, we introduce Off-CMAB, the first offline learning framework for CMAB. Central to our framework is the combinatorial lower confidence bound (CLCB) algorithm, which combines pessimistic reward estimations with combinatorial solvers. To characterize the quality of offline datasets, we propose two novel data coverage conditions and prove that, under these conditions, CLCB achieves a near-optimal suboptimality gap, matching the theoretical lower bound up to a logarithmic factor. We validate Off-CMAB through practical applications, including learning to rank, large language model (LLM) caching, and social influence maximization, showing its ability to handle nonlinear reward functions, general feedback models, and out-of-distribution action samples that excludes optimal or even feasible actions. Extensive experiments on synthetic and real-world datasets further highlight the superior performance of CLCB. | [
{
"version": "v1",
"created": "Fri, 31 Jan 2025 16:56:18 GMT"
},
{
"version": "v2",
"created": "Thu, 29 May 2025 03:11:10 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Liu",
"Xutong",
""
],
[
"Dai",
"Xiangxiang",
""
],
[
"Zuo",
"Jinhang",
""
],
[
"Wang",
"Siwei",
""
],
[
"Joe-Wong",
"Carlee",
""
],
[
"Lui",
"John C. S.",
""
],
[
"Chen",
"Wei",
""
]
] |
2501.19318 | Anirudh Chari | Anirudh Chari, Suraj Reddy, Aditya Tiwari, Richard Lian, Brian Zhou | MINDSTORES: Memory-Informed Neural Decision Synthesis for Task-Oriented Reinforcement in Embodied Systems | null | null | null | null | cs.AI | http://creativecommons.org/licenses/by/4.0/ | While large language models (LLMs) have shown promising capabilities as zero-shot planners for embodied agents, their inability to learn from experience and build persistent mental models limits their robustness in complex open-world environments like Minecraft. We introduce MINDSTORES, an experience-augmented planning framework that enables embodied agents to build and leverage mental models through natural interaction with their environment. Drawing inspiration from how humans construct and refine cognitive mental models, our approach extends existing zero-shot LLM planning by maintaining a database of past experiences that informs future planning iterations. The key innovation is representing accumulated experiences as natural language embeddings of (state, task, plan, outcome) tuples, which can then be efficiently retrieved and reasoned over by an LLM planner to generate insights and guide plan refinement for novel states and tasks. Through extensive experiments in the MineDojo environment, a simulation environment for agents in Minecraft that provides low-level controls for Minecraft, we find that MINDSTORES learns and applies its knowledge significantly better than existing memory-based LLM planners while maintaining the flexibility and generalization benefits of zero-shot approaches, representing an important step toward more capable embodied AI systems that can learn continuously through natural experience. | [
{
"version": "v1",
"created": "Fri, 31 Jan 2025 17:15:33 GMT"
},
{
"version": "v2",
"created": "Wed, 28 May 2025 16:59:03 GMT"
},
{
"version": "v3",
"created": "Thu, 29 May 2025 02:06:32 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Chari",
"Anirudh",
""
],
[
"Reddy",
"Suraj",
""
],
[
"Tiwari",
"Aditya",
""
],
[
"Lian",
"Richard",
""
],
[
"Zhou",
"Brian",
""
]
] |
2502.00264 | Binchi Zhang | Binchi Zhang, Zaiyi Zheng, Zhengzhang Chen, Jundong Li | Beyond the Permutation Symmetry of Transformers: The Role of Rotation for Model Fusion | ICML 2025 | null | null | null | cs.LG cs.CV | http://creativecommons.org/licenses/by/4.0/ | Symmetry in the parameter space of deep neural networks (DNNs) has proven beneficial for various deep learning applications. A well-known example is the permutation symmetry in Multi-Layer Perceptrons (MLPs), where permuting the rows of weight matrices in one layer and applying the inverse permutation to adjacent layers yields a functionally equivalent model. While permutation symmetry fully characterizes the equivalence set for MLPs, its discrete nature limits its utility for transformers. In this paper, we introduce rotation symmetry, a novel form of parameter space symmetry for transformers that generalizes permutation symmetry by rotating parameter matrices in self-attention layers. Unlike permutation symmetry, rotation symmetry operates in a continuous domain, thereby significantly expanding the equivalence set for transformers. Based on this property, we propose a theoretically optimal parameter matching algorithm as a plug-and-play module to enhance model fusion. We evaluate our approach using pre-trained transformers across diverse natural language and vision tasks. Experimental results demonstrate that our rotation symmetry-based matching algorithm substantially improves model fusion, highlighting the potential of parameter space symmetry to facilitate model fusion. Our code is available on https://github.com/zhengzaiyi/RotationSymmetry. | [
{
"version": "v1",
"created": "Sat, 01 Feb 2025 01:44:55 GMT"
},
{
"version": "v2",
"created": "Wed, 28 May 2025 20:23:12 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Zhang",
"Binchi",
""
],
[
"Zheng",
"Zaiyi",
""
],
[
"Chen",
"Zhengzhang",
""
],
[
"Li",
"Jundong",
""
]
] |
2502.00288 | Jijia Liu | Jijia Liu, Feng Gao, Qingmin Liao, Chao Yu, Yu Wang | Learning from Suboptimal Data in Continuous Control via Auto-Regressive Soft Q-Network | Accepted by ICML 2025 | null | null | null | cs.LG cs.RO | http://creativecommons.org/licenses/by/4.0/ | Reinforcement learning (RL) for continuous control often requires large amounts of online interaction data. Value-based RL methods can mitigate this burden by offering relatively high sample efficiency. Some studies further enhance sample efficiency by incorporating offline demonstration data to "kick-start" training, achieving promising results in continuous control. However, they typically compute the Q-function independently for each action dimension, neglecting interdependencies and making it harder to identify optimal actions when learning from suboptimal data, such as non-expert demonstration and online-collected data during the training process. To address these issues, we propose Auto-Regressive Soft Q-learning (ARSQ), a value-based RL algorithm that models Q-values in a coarse-to-fine, auto-regressive manner. First, ARSQ decomposes the continuous action space into discrete spaces in a coarse-to-fine hierarchy, enhancing sample efficiency for fine-grained continuous control tasks. Next, it auto-regressively predicts dimensional action advantages within each decision step, enabling more effective decision-making in continuous control tasks. We evaluate ARSQ on two continuous control benchmarks, RLBench and D4RL, integrating demonstration data into online training. On D4RL, which includes non-expert demonstrations, ARSQ achieves an average $1.62\times$ performance improvement over SOTA value-based baseline. On RLBench, which incorporates expert demonstrations, ARSQ surpasses various baselines, demonstrating its effectiveness in learning from suboptimal online-collected data. Project page is at https://sites.google.com/view/ar-soft-q | [
{
"version": "v1",
"created": "Sat, 01 Feb 2025 03:04:53 GMT"
},
{
"version": "v2",
"created": "Thu, 29 May 2025 03:49:50 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Liu",
"Jijia",
""
],
[
"Gao",
"Feng",
""
],
[
"Liao",
"Qingmin",
""
],
[
"Yu",
"Chao",
""
],
[
"Wang",
"Yu",
""
]
] |
2502.00488 | Chuqi Chen | Chuqi Chen, Yahong Yang, Yang Xiang, Wenrui Hao | Learn Singularly Perturbed Solutions via Homotopy Dynamics | null | null | null | null | cs.LG cs.NA math.NA | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Solving partial differential equations (PDEs) using neural networks has become a central focus in scientific machine learning. Training neural networks for singularly perturbed problems is particularly challenging due to certain parameters in the PDEs that introduce near-singularities in the loss function. In this study, we overcome this challenge by introducing a novel method based on homotopy dynamics to effectively manipulate these parameters. From a theoretical perspective, we analyze the effects of these parameters on training difficulty in these singularly perturbed problems and establish the convergence of the proposed homotopy dynamics method. Experimentally, we demonstrate that our approach significantly accelerates convergence and improves the accuracy of these singularly perturbed problems. These findings present an efficient optimization strategy leveraging homotopy dynamics, offering a robust framework to extend the applicability of neural networks for solving singularly perturbed differential equations. | [
{
"version": "v1",
"created": "Sat, 01 Feb 2025 16:26:53 GMT"
},
{
"version": "v2",
"created": "Sun, 09 Feb 2025 01:54:47 GMT"
},
{
"version": "v3",
"created": "Thu, 29 May 2025 08:37:00 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Chen",
"Chuqi",
""
],
[
"Yang",
"Yahong",
""
],
[
"Xiang",
"Yang",
""
],
[
"Hao",
"Wenrui",
""
]
] |
2502.01179 | Wen Lai | Wen Lai, Alexander Fraser, Ivan Titov | Joint Localization and Activation Editing for Low-Resource Fine-Tuning | Accepted by ICML 2025 (camera-ready version). The code is released at https://github.com/wenlai-lavine/jola | null | null | null | cs.CL cs.AI cs.LG | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Parameter-efficient fine-tuning (PEFT) methods, such as LoRA, are commonly used to adapt LLMs. However, the effectiveness of standard PEFT methods is limited in low-resource scenarios with only a few hundred examples. Recent advances in interpretability research have inspired the emergence of activation editing (or steering) techniques, which modify the activations of specific model components. Due to their extremely small parameter counts, these methods show promise for small datasets. However, their performance is highly dependent on identifying the correct modules to edit and often lacks stability across different datasets. In this paper, we propose Joint Localization and Activation Editing (JoLA), a method that jointly learns (1) which heads in the Transformer to edit (2) whether the intervention should be additive, multiplicative, or both and (3) the intervention parameters themselves - the vectors applied as additive offsets or multiplicative scalings to the head output. Through evaluations on three benchmarks spanning commonsense reasoning, natural language understanding, and natural language generation, we demonstrate that JoLA consistently outperforms existing methods. The code for the method is released at https://github.com/wenlai-lavine/jola. | [
{
"version": "v1",
"created": "Mon, 03 Feb 2025 09:13:09 GMT"
},
{
"version": "v2",
"created": "Mon, 24 Feb 2025 20:52:08 GMT"
},
{
"version": "v3",
"created": "Mon, 19 May 2025 11:36:10 GMT"
},
{
"version": "v4",
"created": "Thu, 29 May 2025 14:57:31 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Lai",
"Wen",
""
],
[
"Fraser",
"Alexander",
""
],
[
"Titov",
"Ivan",
""
]
] |
2502.01310 | Milena Gazdieva | Roman Tarasov, Petr Mokrov, Milena Gazdieva, Evgeny Burnaev, Alexander Korotin | A Statistical Learning Perspective on Semi-dual Adversarial Neural Optimal Transport Solvers | null | null | null | null | cs.LG cs.AI | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Neural network-based optimal transport (OT) is a recent and fruitful direction in the generative modeling community. It finds its applications in various fields such as domain translation, image super-resolution, computational biology and others. Among the existing OT approaches, of considerable interest are adversarial minimax solvers based on semi-dual formulations of OT problems. While promising, these methods lack theoretical investigation from a statistical learning perspective. Our work fills this gap by establishing upper bounds on the generalization error of an approximate OT map recovered by the minimax quadratic OT solver. Importantly, the bounds we derive depend solely on some standard statistical and mathematical properties of the considered functional classes (neural nets). While our analysis focuses on the quadratic OT, we believe that similar bounds could be derived for general OT case, paving the promising direction for future research. | [
{
"version": "v1",
"created": "Mon, 03 Feb 2025 12:37:20 GMT"
},
{
"version": "v2",
"created": "Thu, 29 May 2025 12:15:10 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Tarasov",
"Roman",
""
],
[
"Mokrov",
"Petr",
""
],
[
"Gazdieva",
"Milena",
""
],
[
"Burnaev",
"Evgeny",
""
],
[
"Korotin",
"Alexander",
""
]
] |
2502.01662 | Yuchu Jiang | Jiale Fu, Yuchu Jiang, Junkai Chen, Jiaming Fan, Xin Geng, Xu Yang | Fast Large Language Model Collaborative Decoding via Speculation | null | null | null | null | cs.CL cs.AI cs.LG | http://creativecommons.org/licenses/by/4.0/ | Large Language Model (LLM) collaborative decoding techniques improve output quality by combining the outputs of multiple models at each generation step, but they incur high computational costs. In this paper, we introduce Collaborative decoding via Speculation (CoS), a novel framework that accelerates collaborative decoding without compromising performance. Inspired by Speculative Decoding--where a small proposal model generates tokens sequentially, and a larger target model verifies them in parallel, our approach builds on two key insights: (1) the verification distribution can be the combined distribution of both the proposal and target models, and (2) alternating each model as the proposer and verifier can further enhance efficiency. We generalize this method to collaboration among n models and theoretically prove that CoS is never slower than standard collaborative decoding, typically achieving faster speed. Extensive experiments demonstrate CoS is 1.11x-2.23x faster than standard collaborative decoding without compromising generation quality. Our code is available at https://github.com/Kamichanw/CoS/. | [
{
"version": "v1",
"created": "Sat, 01 Feb 2025 05:22:11 GMT"
},
{
"version": "v2",
"created": "Thu, 29 May 2025 15:20:23 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Fu",
"Jiale",
""
],
[
"Jiang",
"Yuchu",
""
],
[
"Chen",
"Junkai",
""
],
[
"Fan",
"Jiaming",
""
],
[
"Geng",
"Xin",
""
],
[
"Yang",
"Xu",
""
]
] |
2502.01699 | Tianlin Zhang | Tianlin Zhang, En Yu, Yi Shao, Jiande Sun | Multimodal Inverse Attention Network with Intrinsic Discriminant Feature Exploitation for Fake News Detection | 13 pages, 6 figures | null | null | null | cs.LG cs.CL cs.CV cs.IR cs.MM | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Multimodal fake news detection has garnered significant attention due to its profound implications for social security. While existing approaches have contributed to understanding cross-modal consistency, they often fail to leverage modal-specific representations and explicit discrepant features. To address these limitations, we propose a Multimodal Inverse Attention Network (MIAN), a novel framework that explores intrinsic discriminative features based on news content to advance fake news detection. Specifically, MIAN introduces a hierarchical learning module that captures diverse intra-modal relationships through local-to-global and local-to-local interactions, thereby generating enhanced unimodal representations to improve the identification of fake news at the intra-modal level. Additionally, a cross-modal interaction module employs a co-attention mechanism to establish and model dependencies between the refined unimodal representations, facilitating seamless semantic integration across modalities. To explicitly extract inconsistency features, we propose an inverse attention mechanism that effectively highlights the conflicting patterns and semantic deviations introduced by fake news in both intra- and inter-modality. Extensive experiments on benchmark datasets demonstrate that MIAN significantly outperforms state-of-the-art methods, underscoring its pivotal contribution to advancing social security through enhanced multimodal fake news detection. | [
{
"version": "v1",
"created": "Mon, 03 Feb 2025 07:58:22 GMT"
},
{
"version": "v2",
"created": "Thu, 29 May 2025 07:33:39 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Zhang",
"Tianlin",
""
],
[
"Yu",
"En",
""
],
[
"Shao",
"Yi",
""
],
[
"Sun",
"Jiande",
""
]
] |
2502.01753 | Eliot Quataert | Eliot Quataert and Philip F. Hopkins | Cosmic Ray Feedback in Massive Halos: Implications for the Distribution of Baryons | 12 pages, 5 figures. final version accepted by the Open Journal of Astrophysics | null | 10.33232/001c.138772 | null | astro-ph.CO astro-ph.HE | http://creativecommons.org/licenses/by/4.0/ | We use order of magnitude estimates and observational constraints to argue that feedback from relativistic cosmic rays (CRs) produced by massive black holes is likely to have a particularly large effect at radii of order the virial radius and larger in group-mass halos. We show that for a range of plausible (but uncertain) CR transport parameters and energetics, the pressure produced by CRs generated by the central massive black hole over its lifetime can be of order the thermal gas pressure in the outskirts of $\sim 10^{13-14} M_\odot$ halos (but not in more massive clusters). The properties of this CR feedback at low redshift are not well predicted by the radiative cooling rate of hot gas at smaller radii, which is often used as a proxy for `current' black hole feedback. This is because most black hole growth happens early in massive halos, and CR transport timescales in halo outskirts are Gyr or more; the accumulated CR energy thus depends on the full history of black hole activity in the halo. The large CR pressure in group-mass systems likely leads to CR-driven outflows that move gas from large halo radii to outside the virial radius. Such feedback would not be captured by current cosmological simulations that focus on mechanical black hole feedback; in particular, CR feedback remains active even long after the mechanical feedback sourcing the CRs has turned off. We speculate that this CR feedback may be important for explaining the weak lensing $S_8$ tension and the evidence for strong feedback at large halo radii from kinetic Sunyaev-Zeldovich measurements. Prospects for testing this mechanism observationally and implementing the necessary physics in cosmological simulations are discussed. | [
{
"version": "v1",
"created": "Mon, 03 Feb 2025 19:01:11 GMT"
},
{
"version": "v2",
"created": "Wed, 28 May 2025 01:36:28 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Quataert",
"Eliot",
""
],
[
"Hopkins",
"Philip F.",
""
]
] |
2502.01918 | Zachary Cooper-Baldock | Zachary Cooper-Baldock and Stephen Turnock and Karl Sammut | Wake-Informed 3D Path Planning for Autonomous Underwater Vehicles Using A* and Neural Network Approximations | 11 pages, 6 figures, preprint of journal paper | null | null | null | cs.RO cs.AI cs.LG | http://creativecommons.org/licenses/by-nc-sa/4.0/ | Autonomous Underwater Vehicles (AUVs) encounter significant energy, control and navigation challenges in complex underwater environments, particularly during close-proximity operations, such as launch and recovery (LAR), where fluid interactions and wake effects present additional navigational and energy challenges. Traditional path planning methods fail to incorporate these detailed wake structures, resulting in increased energy consumption, reduced control stability, and heightened safety risks. This paper presents a novel wake-informed, 3D path planning approach that fully integrates localized wake effects and global currents into the planning algorithm. Two variants of the A* algorithm - a current-informed planner and a wake-informed planner - are created to assess its validity and two neural network models are then trained to approximate these planners for real-time applications. Both the A* planners and NN models are evaluated using important metrics such as energy expenditure, path length, and encounters with high-velocity and turbulent regions. The results demonstrate a wake-informed A* planner consistently achieves the lowest energy expenditure and minimizes encounters with high-velocity regions, reducing energy consumption by up to 11.3%. The neural network models are observed to offer computational speedup of 6 orders of magnitude, but exhibit 4.51 - 19.79% higher energy expenditures and 9.81 - 24.38% less optimal paths. These findings underscore the importance of incorporating detailed wake structures into traditional path planning algorithms and the benefits of neural network approximations to enhance energy efficiency and operational safety for AUVs in complex 3D domains. | [
{
"version": "v1",
"created": "Tue, 04 Feb 2025 01:23:23 GMT"
},
{
"version": "v2",
"created": "Thu, 29 May 2025 15:53:13 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Cooper-Baldock",
"Zachary",
""
],
[
"Turnock",
"Stephen",
""
],
[
"Sammut",
"Karl",
""
]
] |
2502.02389 | Pau Colomer | Pau Colomer, Christian Deppe, Holger Boche, Andreas Winter | Rate-reliability tradeoff for deterministic identification | 16 pages, 2 figures. A preliminary version of this work has been accepted for presentation at the 2025 IEEE International Conference on Communications, Montreal (Canada) 8-12 June 2025 | null | null | null | cs.IT math.IT quant-ph | http://creativecommons.org/licenses/by/4.0/ | We investigate deterministic identification over arbitrary memoryless channels under the constraint that the error probabilities of first and second kind are exponentially small in the block length $n$, controlled by reliability exponents $E_1,E_2 \geq 0$. In contrast to the regime of slowly vanishing errors, where the identifiable message length scales linearithmically as $\Theta(n\log n)$, here we find that for positive exponents linear scaling is restored, now with a rate that is a function of the reliability exponents. We give upper and lower bounds on the ensuing rate-reliability function in terms of (the logarithm of) the packing and covering numbers of the channel output set, which for small error exponents $E_1,E_2>0$ can be expanded in leading order as the product of the Minkowski dimension of a certain parametrisation the channel output set and $\log\min\{E_1,E_2\}$. These allow us to recover the previously observed slightly superlinear identification rates, and offer a different perspective for understanding them in more traditional information theory terms. We also show that even if only one of the two errors is required to be exponentially small, the linearithmic scaling is lost. We further illustrate our results with a discussion of the case of dimension zero, and extend them to classical-quantum channels and quantum channels with tensor product input restriction. | [
{
"version": "v1",
"created": "Tue, 04 Feb 2025 15:09:14 GMT"
},
{
"version": "v2",
"created": "Thu, 27 Mar 2025 12:42:43 GMT"
},
{
"version": "v3",
"created": "Thu, 29 May 2025 11:19:35 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Colomer",
"Pau",
""
],
[
"Deppe",
"Christian",
""
],
[
"Boche",
"Holger",
""
],
[
"Winter",
"Andreas",
""
]
] |
2502.02410 | Jan Schuchardt | Jan Schuchardt, Mina Dalirrooyfard, Jed Guzelkabaagac, Anderson Schneider, Yuriy Nevmyvaka, Stephan G\"unnemann | Privacy Amplification by Structured Subsampling for Deep Differentially Private Time Series Forecasting | Accepted as ICML 2025 Spotlight | null | null | null | cs.LG cs.CR stat.ML | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Many forms of sensitive data, such as web traffic, mobility data, or hospital occupancy, are inherently sequential. The standard method for training machine learning models while ensuring privacy for units of sensitive information, such as individual hospital visits, is differentially private stochastic gradient descent (DP-SGD). However, we observe in this work that the formal guarantees of DP-SGD are incompatible with time-series-specific tasks like forecasting, since they rely on the privacy amplification attained by training on small, unstructured batches sampled from an unstructured dataset. In contrast, batches for forecasting are generated by (1) sampling sequentially structured time series from a dataset, (2) sampling contiguous subsequences from these series, and (3) partitioning them into context and ground-truth forecast windows. We theoretically analyze the privacy amplification attained by this structured subsampling to enable the training of forecasting models with sound and tight event- and user-level privacy guarantees. Towards more private models, we additionally prove how data augmentation amplifies privacy in self-supervised training of sequence models. Our empirical evaluation demonstrates that amplification by structured subsampling enables the training of forecasting models with strong formal privacy guarantees. | [
{
"version": "v1",
"created": "Tue, 04 Feb 2025 15:29:00 GMT"
},
{
"version": "v2",
"created": "Thu, 29 May 2025 13:11:57 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Schuchardt",
"Jan",
""
],
[
"Dalirrooyfard",
"Mina",
""
],
[
"Guzelkabaagac",
"Jed",
""
],
[
"Schneider",
"Anderson",
""
],
[
"Nevmyvaka",
"Yuriy",
""
],
[
"Günnemann",
"Stephan",
""
]
] |
2502.02432 | Michael Griffin | Tewodros Amdeberhan, Michael Griffin, Ken Ono | Some topological genera and Jacobi forms | Minor revision to address the comments by the two referees | null | null | null | math.NT math.AT | http://creativecommons.org/licenses/by/4.0/ | We revisit and elucidate the $\widehat{A}$-genus, Hirzebruch's $L$-genus and Witten's $W$-genus, cobordism invariants of special classes of manifolds. After slight modification, involving Hecke's trick, we find that the $\widehat{A}$-genus and $L$-genus arise directly from Jacobi's theta function. For every $k\geq 0,$ we obtain exact formulas for the quasimodular expressions of $\widehat{A}_k$ and $L_k$ as ``traces'' of partition Eisenstein series \[ \widehat{\mathcal{A}}_k(\tau)= \operatorname{Tr}_k(\phi_{\widehat{A}};\tau)\ \ \ \ \ \ {\text {and}}\ \ \ \ \ \ \mathcal{L}_k(\tau)= \operatorname{Tr}_k(\phi_L;\tau), \] which are easily converted to the original topological expressions. Surprisingly, Ramanujan defined twists of the $\widehat{\mathcal{A}}_k(\tau)$ in his ``lost notebook'' in his study of derivatives of theta functions, decades before Borel and Hirzebruch rediscovered them in the context of spin manifolds. In addition, we show that the nonholomorphic $G_2^{\star}$-completion of the characteristic series of the Witten genus is the Jacobi theta function avatar of the $\widehat{A}$-genus. | [
{
"version": "v1",
"created": "Tue, 04 Feb 2025 15:56:57 GMT"
},
{
"version": "v2",
"created": "Wed, 05 Feb 2025 02:15:46 GMT"
},
{
"version": "v3",
"created": "Wed, 28 May 2025 20:44:13 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Amdeberhan",
"Tewodros",
""
],
[
"Griffin",
"Michael",
""
],
[
"Ono",
"Ken",
""
]
] |
2502.02516 | Alessio Russo | Alessio Russo, Aldo Pacchiano | Adaptive Exploration for Multi-Reward Multi-Policy Evaluation | Accepted at the International Conference on Machine Learning, 2025 | null | null | null | cs.LG cs.AI stat.ML | http://creativecommons.org/licenses/by/4.0/ | We study the policy evaluation problem in an online multi-reward multi-policy discounted setting, where multiple reward functions must be evaluated simultaneously for different policies. We adopt an $(\epsilon,\delta)$-PAC perspective to achieve $\epsilon$-accurate estimates with high confidence across finite or convex sets of rewards, a setting that has not been investigated in the literature. Building on prior work on Multi-Reward Best Policy Identification, we adapt the MR-NaS exploration scheme to jointly minimize sample complexity for evaluating different policies across different reward sets. Our approach leverages an instance-specific lower bound revealing how the sample complexity scales with a measure of value deviation, guiding the design of an efficient exploration policy. Although computing this bound entails a hard non-convex optimization, we propose an efficient convex approximation that holds for both finite and convex reward sets. Experiments in tabular domains demonstrate the effectiveness of this adaptive exploration scheme. | [
{
"version": "v1",
"created": "Tue, 04 Feb 2025 17:35:51 GMT"
},
{
"version": "v2",
"created": "Wed, 28 May 2025 18:52:23 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Russo",
"Alessio",
""
],
[
"Pacchiano",
"Aldo",
""
]
] |
2502.02673 | Adibvafa Fallahpour | Adibvafa Fallahpour, Jun Ma, Alif Munim, Hongwei Lyu, Bo Wang | MedRAX: Medical Reasoning Agent for Chest X-ray | 16 pages, 4 figures, 5 Tables | null | null | null | cs.LG cs.AI cs.MA | http://creativecommons.org/licenses/by/4.0/ | Chest X-rays (CXRs) play an integral role in driving critical decisions in disease management and patient care. While recent innovations have led to specialized models for various CXR interpretation tasks, these solutions often operate in isolation, limiting their practical utility in clinical practice. We present MedRAX, the first versatile AI agent that seamlessly integrates state-of-the-art CXR analysis tools and multimodal large language models into a unified framework. MedRAX dynamically leverages these models to address complex medical queries without requiring additional training. To rigorously evaluate its capabilities, we introduce ChestAgentBench, a comprehensive benchmark containing 2,500 complex medical queries across 7 diverse categories. Our experiments demonstrate that MedRAX achieves state-of-the-art performance compared to both open-source and proprietary models, representing a significant step toward the practical deployment of automated CXR interpretation systems. Data and code have been publicly available at https://github.com/bowang-lab/MedRAX | [
{
"version": "v1",
"created": "Tue, 04 Feb 2025 19:31:00 GMT"
},
{
"version": "v2",
"created": "Thu, 29 May 2025 01:45:45 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Fallahpour",
"Adibvafa",
""
],
[
"Ma",
"Jun",
""
],
[
"Munim",
"Alif",
""
],
[
"Lyu",
"Hongwei",
""
],
[
"Wang",
"Bo",
""
]
] |
2502.03025 | Matteo Fornoni | Elena Beretta, Cecilia Cavaterra, Matteo Fornoni, Maurizio Grasselli | Optimal control of the fidelity coefficient in a Cahn-Hilliard image inpainting model | 40 pages, revised version, to appear in ESAIM: Control Optim. Calc. Var | null | null | null | math.OC math.AP | http://creativecommons.org/licenses/by/4.0/ | We consider an inpainting model proposed by A. Bertozzi et al., which is based on a Cahn-Hilliard-type equation. This equation describes the evolution of an order parameter that represents an approximation of the original image occupying a bounded two-dimensional domain. The given image is assumed to be damaged in a fixed subdomain, and the equation is characterised by a linear reaction term. This term is multiplied by the so-called fidelity coefficient, which is a strictly positive bounded function defined in the undamaged region. The idea is that, given an initial image, the order parameter evolves towards the given image, and this process properly diffuses through the boundary of the damaged region, restoring the damaged image, provided that the fidelity coefficient is large enough. Here, we formulate an optimal control problem based on this fact, namely, our cost functional accounts for the magnitude of the fidelity coefficient. Assuming a singular potential to ensure that the order parameter takes its values in between 0 and 1, we first analyse the control-to-state operator and prove the existence of at least one optimal control, establishing the validity of first-order optimality conditions. Then, under suitable assumptions, we demonstrate second-order optimality conditions. | [
{
"version": "v1",
"created": "Wed, 05 Feb 2025 09:30:29 GMT"
},
{
"version": "v2",
"created": "Thu, 29 May 2025 08:27:28 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Beretta",
"Elena",
""
],
[
"Cavaterra",
"Cecilia",
""
],
[
"Fornoni",
"Matteo",
""
],
[
"Grasselli",
"Maurizio",
""
]
] |
2502.03085 | Fiona Sawyer | Fiona Sawyer, James S. Bolton, George D. Becker, Luke Conaboy, Martin G. Haehnelt, Laura Keating, Girish Kulkarni and Ewald Puchwein | How probable is the Lyman-$\alpha$ damping wing in the spectrum of the redshift z = 5.9896 quasar ULAS J0148+0600? | 16 pages, 9 figures. Accepted for publication in MNRAS | null | null | null | astro-ph.CO | http://creativecommons.org/licenses/by/4.0/ | The shape of the Ly-$\alpha$ transmission in the near zone of the redshift $z=5.9896$ quasar ULAS J0148$+$0600 (hereafter J0148) is consistent with a damping wing arising from an extended neutral hydrogen island in the diffuse intergalactic medium (IGM). Here we use simulations of late-ending reionisation from Sherwood-Relics to assess the expected incidence of quasars with Ly-$\alpha$ and Ly-$\beta$ absorption similar to the observed J0148 spectrum. We find a late end to reionisation at $z=5.3$ is a necessary requirement for reproducing a Ly-$\alpha$ damping wing consistent with J0148. This occurs in $\sim3$ per cent of our simulated spectra for an IGM neutral fraction $\langle x_{\rm HI}\rangle=0.14$ at $z=6$. However, using standard assumptions for the ionising photon output of J0148, the a priori probability of drawing a simulated quasar spectrum with a Ly-$\alpha$ damping wing profile \emph{and} Ly-$\alpha$ near zone size that simultaneously match J0148 is low, $p<10^{-2}$. This may indicate that the ionising emission from J0148 is variable on timescales $t<10^{5}\rm\,yr$, or alternatively that the Ly-$\alpha$ transmission in the J0148 near zone is impacted by the transverse proximity effect from nearby star-forming galaxies or undetected quasars. We also predict the IGM temperature should be $T\sim 4\times 10^{4}\rm\,K$ within a few proper Mpc of the Ly-$\alpha$ near zone edge due to recent H$\,\rm \scriptstyle I$ and He$\,\rm \scriptstyle II$ photo-heating. Evidence for enhanced thermal broadening in the Ly-$\alpha$ absorption near the damping wing edge would provide further evidence that the final stages of reionisation are occurring at $z<6$. | [
{
"version": "v1",
"created": "Wed, 05 Feb 2025 11:24:33 GMT"
},
{
"version": "v2",
"created": "Wed, 28 May 2025 11:58:34 GMT"
},
{
"version": "v3",
"created": "Thu, 29 May 2025 11:14:41 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Sawyer",
"Fiona",
""
],
[
"Bolton",
"James S.",
""
],
[
"Becker",
"George D.",
""
],
[
"Conaboy",
"Luke",
""
],
[
"Haehnelt",
"Martin G.",
""
],
[
"Keating",
"Laura",
""
],
[
"Kulkarni",
"Girish",
""
],
[
"Puchwein",
"Ewald",
""
]
] |
2502.03397 | Hongli Zhan | Hongli Zhan, Muneeza Azmat, Raya Horesh, Junyi Jessy Li, Mikhail Yurochkin | SPRI: Aligning Large Language Models with Context-Situated Principles | Forty-Second International Conference on Machine Learning (ICML 2025) Camera-Ready Version | null | null | null | cs.CL cs.AI | http://creativecommons.org/licenses/by/4.0/ | Aligning Large Language Models to integrate and reflect human values, especially for tasks that demand intricate human oversight, is arduous since it is resource-intensive and time-consuming to depend on human expertise for context-specific guidance. Prior work has utilized predefined sets of rules or principles to steer the behavior of models (Bai et al., 2022; Sun et al., 2023). However, these principles tend to be generic, making it challenging to adapt them to each individual input query or context. In this work, we present Situated-PRInciples (SPRI), a framework requiring minimal or no human effort that is designed to automatically generate guiding principles in real-time for each input query and utilize them to align each response. We evaluate SPRI on three tasks, and show that 1) SPRI can derive principles in a complex domain-specific task that leads to on-par performance as expert-crafted ones; 2) SPRI-generated principles lead to instance-specific rubrics that outperform prior LLM-as-a-judge frameworks; 3) using SPRI to generate synthetic SFT data leads to substantial improvement on truthfulness. We release our code and model generations at https://github.com/honglizhan/SPRI-public. | [
{
"version": "v1",
"created": "Wed, 05 Feb 2025 17:32:29 GMT"
},
{
"version": "v2",
"created": "Thu, 29 May 2025 13:20:34 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Zhan",
"Hongli",
""
],
[
"Azmat",
"Muneeza",
""
],
[
"Horesh",
"Raya",
""
],
[
"Li",
"Junyi Jessy",
""
],
[
"Yurochkin",
"Mikhail",
""
]
] |
2502.03543 | Jeffrey Kuntz | Jisuke Kubo and Jeffrey Kuntz | Primordial Gravitational Waves in Quadratic Gravity | 19+4 pages, v3: matches published version | null | 10.1088/1475-7516/2025/05/093 | null | gr-qc astro-ph.CO hep-ph | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Quadratic gravity is a fourth-order (in derivatives) theory that can serve as an attractive upgrade to the standard description of gravity provided by General Relativity, thanks to its renormalizability and its built-in description of primordial inflation. We bring quadratic gravity into a second-order form by introducing an auxiliary tensor field and we consider the primordial tensor fluctuations (gravitational waves) in the theory around a Friedmann-Lema\^itre-Robertson-Walker background. After a canonical quantization of the perturbations, we calculate the tensor power spectrum in quasi de Sitter spacetime. We find that the spectral index $n_t$ and the amplitude $A_t$ of the tensor power spectrum are both suppressed by the factor $(1 + 2{\bf H}^2_*/m_\text{gh}^2)^{-1}$, where ${\bf H}_*$ is the Hubble rate at horizon exit and $m_\text{gh}$ is the mass of the spin-two ghost. This restores the slow-roll consistency condition familiar from single-field inflation models, where the tensor-to-scalar ratio $r$ is equal to $-8n_t$ in the lowest nontrivial order in the slow-roll approximation. We also discuss the well-known issue of the ghost problem in fourth-order theories and how it pertains to the results at hand. | [
{
"version": "v1",
"created": "Wed, 05 Feb 2025 19:01:13 GMT"
},
{
"version": "v2",
"created": "Wed, 12 Feb 2025 14:27:27 GMT"
},
{
"version": "v3",
"created": "Thu, 29 May 2025 10:11:59 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Kubo",
"Jisuke",
""
],
[
"Kuntz",
"Jeffrey",
""
]
] |
2502.03708 | Adityanarayanan Radhakrishnan | Daniel Beaglehole, Adityanarayanan Radhakrishnan, Enric Boix-Adser\`a, Mikhail Belkin | Toward universal steering and monitoring of AI models | null | null | null | null | cs.CL cs.AI stat.ML | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Modern AI models contain much of human knowledge, yet understanding of their internal representation of this knowledge remains elusive. Characterizing the structure and properties of this representation will lead to improvements in model capabilities and development of effective safeguards. Building on recent advances in feature learning, we develop an effective, scalable approach for extracting linear representations of general concepts in large-scale AI models (language models, vision-language models, and reasoning models). We show how these representations enable model steering, through which we expose vulnerabilities, mitigate misaligned behaviors, and improve model capabilities. Additionally, we demonstrate that concept representations are remarkably transferable across human languages and combinable to enable multi-concept steering. Through quantitative analysis across hundreds of concepts, we find that newer, larger models are more steerable and steering can improve model capabilities beyond standard prompting. We show how concept representations are effective for monitoring misaligned content (hallucinations, toxic content). We demonstrate that predictive models built using concept representations are more accurate for monitoring misaligned content than using models that judge outputs directly. Together, our results illustrate the power of using internal representations to map the knowledge in AI models, advance AI safety, and improve model capabilities. | [
{
"version": "v1",
"created": "Thu, 06 Feb 2025 01:41:48 GMT"
},
{
"version": "v2",
"created": "Wed, 28 May 2025 19:52:55 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Beaglehole",
"Daniel",
""
],
[
"Radhakrishnan",
"Adityanarayanan",
""
],
[
"Boix-Adserà",
"Enric",
""
],
[
"Belkin",
"Mikhail",
""
]
] |
2502.04080 | Peter Clark | Peter Clark, Joseph Callow, Or Graur, Claire Greenwell, Lei Hu, Jessica Aguilar, Steven Ahlen, Davide Bianchi, David Brooks, Todd Claybaugh, Kyle Dawson, Axel de la Macorra, Peter Doel, Satya Gontcho A Gontcho, Gaston Gutierrez, Klaus Honscheid, Stephanie Juneau, Robert Kehoe, Theodore Kisner, Anthony Kremin, Martin Landriau, Laurent Le Guillou, Aaron Meisner, Ramon Miquel, John Moustakas, Ignasi P\'erez-R\`afols, Eusebio Sanchez, Michael Schubnell, David Sprayberry, Gregory Tarl\'e, Benjamin A.Weaver and Hu Zou | AT 2018dyk: tidal disruption event or active galactic nucleus? Follow-up observations of an extreme coronal line emitter with the Dark Energy Spectroscopic Instrument | 44 pages, 17 total figures. Author compiled version of the paper now published in MNRAS | null | 10.1093/mnras/staf724 | null | astro-ph.HE | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We present fresh insights into the nature of the tidal disruption event (TDE) candidate AT 2018dyk. AT 2018dyk has sparked a debate in the literature around its classification as either a bona-fide TDE or as an active galactic nucleus (AGN) turn-on state change. A new follow-up spectrum taken with the Dark Energy Spectroscopic Instrument, in combination with host-galaxy analysis using archival SDSS-MaNGA data, supports the identification of AT 2018dyk as a TDE. Specifically, we classify this object as a TDE that occurred within a gas-rich environment, which was responsible for both its mid-infrared (MIR) outburst and development of Fe coronal emission lines. Comparison with the known sample of TDE-linked extreme coronal line emitters (TDE-ECLEs) and other TDEs displaying coronal emission lines (CrL-TDEs) reveals similar characteristics and shared properties. For example, the MIR properties of both groups appear to form a continuum with links to the content and density of the material in their local environments. This includes evidence for a MIR colour-luminosity relationship in TDEs occurring within such gas-rich environments, with those with larger MIR outbursts also exhibiting redder peaks. | [
{
"version": "v1",
"created": "Thu, 06 Feb 2025 13:43:59 GMT"
},
{
"version": "v2",
"created": "Thu, 29 May 2025 12:41:52 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Clark",
"Peter",
""
],
[
"Callow",
"Joseph",
""
],
[
"Graur",
"Or",
""
],
[
"Greenwell",
"Claire",
""
],
[
"Hu",
"Lei",
""
],
[
"Aguilar",
"Jessica",
""
],
[
"Ahlen",
"Steven",
""
],
[
"Bianchi",
"Davide",
""
],
[
"Brooks",
"David",
""
],
[
"Claybaugh",
"Todd",
""
],
[
"Dawson",
"Kyle",
""
],
[
"de la Macorra",
"Axel",
""
],
[
"Doel",
"Peter",
""
],
[
"Gontcho",
"Satya Gontcho A",
""
],
[
"Gutierrez",
"Gaston",
""
],
[
"Honscheid",
"Klaus",
""
],
[
"Juneau",
"Stephanie",
""
],
[
"Kehoe",
"Robert",
""
],
[
"Kisner",
"Theodore",
""
],
[
"Kremin",
"Anthony",
""
],
[
"Landriau",
"Martin",
""
],
[
"Guillou",
"Laurent Le",
""
],
[
"Meisner",
"Aaron",
""
],
[
"Miquel",
"Ramon",
""
],
[
"Moustakas",
"John",
""
],
[
"Pérez-Ràfols",
"Ignasi",
""
],
[
"Sanchez",
"Eusebio",
""
],
[
"Schubnell",
"Michael",
""
],
[
"Sprayberry",
"David",
""
],
[
"Tarlé",
"Gregory",
""
],
[
"Weaver",
"Benjamin A.",
""
],
[
"Zou",
"Hu",
""
]
] |
2502.04350 | Yongchao Chen | Yongchao Chen, Yilun Hao, Yueying Liu, Yang Zhang, Chuchu Fan | CodeSteer: Symbolic-Augmented Language Models via Code/Text Guidance | 28 pages, 12 figures | International Conference on Machine Learning (ICML'2025) | null | null | cs.CL cs.AI cs.LG cs.SC cs.SE | http://creativecommons.org/publicdomain/zero/1.0/ | Existing methods fail to effectively steer Large Language Models (LLMs) between textual reasoning and code generation, leaving symbolic computing capabilities underutilized. We introduce CodeSteer, an effective method for guiding LLM code/text generation. We construct a comprehensive benchmark SymBench comprising 37 symbolic tasks with adjustable complexity and also synthesize datasets of 12k multi-turn guidance/generation trajectories and 5.5k guidance comparison pairs. We fine-tune the Llama-3-8B model with a newly designed multi-turn supervised fine-tuning (SFT) and direct preference optimization (DPO). The resulting model, CodeSteerLLM, augmented with the proposed symbolic and self-answer checkers, effectively guides the code/text generation of larger models. Augmenting GPT-4o with CodeSteer raises its average performance score from 53.3 to 86.4, even outperforming the existing best LLM OpenAI o1 (82.7), o1-preview (74.8), and DeepSeek R1 (76.8) across all 37 tasks (28 seen, 9 unseen). Trained for GPT-4o, CodeSteer demonstrates superior generalizability, providing an average 41.8 performance boost on Claude, Mistral, and GPT-3.5. CodeSteer-guided LLMs fully harness symbolic computing to maintain strong performance on highly complex tasks. Models, Datasets, and Codes are available at https://github.com/yongchao98/CodeSteer-v1.0 and https://huggingface.co/yongchao98. | [
{
"version": "v1",
"created": "Tue, 04 Feb 2025 15:53:59 GMT"
},
{
"version": "v2",
"created": "Thu, 29 May 2025 00:38:10 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Chen",
"Yongchao",
""
],
[
"Hao",
"Yilun",
""
],
[
"Liu",
"Yueying",
""
],
[
"Zhang",
"Yang",
""
],
[
"Fan",
"Chuchu",
""
]
] |
2502.04358 | Elliot Meyerson | Elliot Meyerson and Xin Qiu | Position: Scaling LLM Agents Requires Asymptotic Analysis with LLM Primitives | In Proceedings of the 42nd International Conference on Machine Learning (ICML 2025); 13 pages including references | null | null | null | cs.CL cs.AI cs.CC cs.LG cs.NE | http://creativecommons.org/licenses/by/4.0/ | Decomposing hard problems into subproblems often makes them easier and more efficient to solve. With large language models (LLMs) crossing critical reliability thresholds for a growing slate of capabilities, there is an increasing effort to decompose systems into sets of LLM-based agents, each of whom can be delegated sub-tasks. However, this decomposition (even when automated) is often intuitive, e.g., based on how a human might assign roles to members of a human team. How close are these role decompositions to optimal? This position paper argues that asymptotic analysis with LLM primitives is needed to reason about the efficiency of such decomposed systems, and that insights from such analysis will unlock opportunities for scaling them. By treating the LLM forward pass as the atomic unit of computational cost, one can separate out the (often opaque) inner workings of a particular LLM from the inherent efficiency of how a set of LLMs are orchestrated to solve hard problems. In other words, if we want to scale the deployment of LLMs to the limit, instead of anthropomorphizing LLMs, asymptotic analysis with LLM primitives should be used to reason about and develop more powerful decompositions of large problems into LLM agents. | [
{
"version": "v1",
"created": "Tue, 04 Feb 2025 20:47:43 GMT"
},
{
"version": "v2",
"created": "Thu, 29 May 2025 16:46:00 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Meyerson",
"Elliot",
""
],
[
"Qiu",
"Xin",
""
]
] |
2502.04434 | Oliver Philcox | Oliver H. E. Philcox | Searching for Inflationary Physics with the CMB Trispectrum: 1. Primordial Theory & Optimal Estimators | 56 pages, 2 tables, 226 equations, accepted by Phys. Rev. D | null | null | null | astro-ph.CO gr-qc hep-ph hep-th | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | The primordial four-point function encodes a wealth of information about the inflationary Universe. Despite extensive theoretical work, most models of four-point physics have never been compared to data. In this series, we conduct a detailed analysis of Cosmic Microwave Background temperature and polarization trispectra, searching for a wide variety of phenomena including local effects, self-interactions, curvatons, DBI inflation, gauge fields, solid inflation, scalar field exchange, spinning massive field exchange, chiral physics, point sources, and gravitational lensing. After presenting a suite of separable primordial templates, we derive thirteen quasi-optimal estimators that directly estimate the underlying template amplitudes. These are unbiased, minimum variance, mask-deconvolved, and account for correlations between templates (including with lensing). Each estimator can be efficiently implemented using spherical harmonic transforms, Monte Carlo methods, and optimization techniques, and asymptotes to standard forms in certain limits. In Paper 2, we implement these estimators in public code, and in Paper 3, use them to constrain primordial trispectra with Planck data. This enables a wide variety of tests of inflation, including some of the first direct constraints on cosmological collider physics. | [
{
"version": "v1",
"created": "Thu, 06 Feb 2025 19:00:00 GMT"
},
{
"version": "v2",
"created": "Wed, 12 Feb 2025 15:44:13 GMT"
},
{
"version": "v3",
"created": "Wed, 28 May 2025 18:15:24 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Philcox",
"Oliver H. E.",
""
]
] |
2502.04778 | Chen-Xiao Gao | Chen-Xiao Gao, Chenyang Wu, Mingjun Cao, Chenjun Xiao, Yang Yu, Zongzhang Zhang | Behavior-Regularized Diffusion Policy Optimization for Offline Reinforcement Learning | Accepted by ICML 2025 | null | null | null | cs.LG cs.AI | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Behavior regularization, which constrains the policy to stay close to some behavior policy, is widely used in offline reinforcement learning (RL) to manage the risk of hazardous exploitation of unseen actions. Nevertheless, existing literature on behavior-regularized RL primarily focuses on explicit policy parameterizations, such as Gaussian policies. Consequently, it remains unclear how to extend this framework to more advanced policy parameterizations, such as diffusion models. In this paper, we introduce BDPO, a principled behavior-regularized RL framework tailored for diffusion-based policies, thereby combining the expressive power of diffusion policies and the robustness provided by regularization. The key ingredient of our method is to calculate the Kullback-Leibler (KL) regularization analytically as the accumulated discrepancies in reverse-time transition kernels along the diffusion trajectory. By integrating the regularization, we develop an efficient two-time-scale actor-critic RL algorithm that produces the optimal policy while respecting the behavior constraint. Comprehensive evaluations conducted on synthetic 2D tasks and continuous control tasks from the D4RL benchmark validate its effectiveness and superior performance. | [
{
"version": "v1",
"created": "Fri, 07 Feb 2025 09:30:35 GMT"
},
{
"version": "v2",
"created": "Thu, 29 May 2025 09:14:17 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Gao",
"Chen-Xiao",
""
],
[
"Wu",
"Chenyang",
""
],
[
"Cao",
"Mingjun",
""
],
[
"Xiao",
"Chenjun",
""
],
[
"Yu",
"Yang",
""
],
[
"Zhang",
"Zongzhang",
""
]
] |
2502.04964 | Roman Vashurin | Roman Vashurin, Maiya Goloburda, Albina Ilina, Aleksandr Rubashevskii, Preslav Nakov, Artem Shelmanov, Maxim Panov | Uncertainty Quantification for LLMs through Minimum Bayes Risk: Bridging Confidence and Consistency | null | null | null | null | cs.CL | http://creativecommons.org/licenses/by/4.0/ | Uncertainty quantification (UQ) methods for Large Language Models (LLMs) encompass a variety of approaches, with two major types being particularly prominent: information-based, which focus on model confidence expressed as token probabilities, and consistency-based, which assess the semantic relationship between multiple outputs generated using repeated sampling. Several recent methods have combined these two approaches to boost UQ performance. However, they sometimes fail to outperform much simpler baseline methods. Our work discusses the fundamental approach to constructing uncertainty measures that directly links uncertainty with the minimum Bayes risks achieved by LLM decoding. Building on these findings, we propose a novel approach to integrating model confidence with output consistency, resulting in a family of efficient and robust UQ methods. Our investigation reveals distinctive characteristics of LLMs as probabilistic models, which help to explain why these UQ methods underperform in certain tasks. Based on these findings, we propose a new way of synthesizing model confidence and output consistency, leading to a family of efficient and robust UQ methods. We evaluate our approach across various tasks such as question answering, abstractive summarization, and machine translation, demonstrating sizable improvements over state-of-the-art UQ approaches. | [
{
"version": "v1",
"created": "Fri, 07 Feb 2025 14:30:12 GMT"
},
{
"version": "v2",
"created": "Tue, 11 Feb 2025 14:32:15 GMT"
},
{
"version": "v3",
"created": "Wed, 21 May 2025 10:24:49 GMT"
},
{
"version": "v4",
"created": "Thu, 29 May 2025 09:39:51 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Vashurin",
"Roman",
""
],
[
"Goloburda",
"Maiya",
""
],
[
"Ilina",
"Albina",
""
],
[
"Rubashevskii",
"Aleksandr",
""
],
[
"Nakov",
"Preslav",
""
],
[
"Shelmanov",
"Artem",
""
],
[
"Panov",
"Maxim",
""
]
] |
2502.05210 | Zizhou Zhang | Shicheng Zhou, Zizhou Zhang, Rong Zhang, Yuchen Yin, Chia Hong Chang, Qinyan Shen | Regression and Forecasting of U.S. Stock Returns Based on LSTM | 5pages | null | null | null | q-fin.ST cs.LG | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | This paper analyses the investment returns of three stock sectors, Manuf, Hitec, and Other, in the U.S. stock market, based on the Fama-French three-factor model, the Carhart four-factor model, and the Fama-French five-factor model, in order to test the validity of the Fama-French three-factor model, the Carhart four-factor model, and the Fama-French five-factor model for the three sectors of the market. French five-factor model for the three sectors of the market. Also, the LSTM model is used to explore the additional factors affecting stock returns. The empirical results show that the Fama-French five-factor model has better validity for the three segments of the market under study, and the LSTM model has the ability to capture the factors affecting the returns of certain industries, and can better regress and predict the stock returns of the relevant industries. Keywords- Fama-French model; Carhart model; Factor model; LSTM model. | [
{
"version": "v1",
"created": "Mon, 03 Feb 2025 19:26:44 GMT"
},
{
"version": "v2",
"created": "Tue, 04 Mar 2025 15:25:14 GMT"
},
{
"version": "v3",
"created": "Wed, 28 May 2025 19:42:17 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Zhou",
"Shicheng",
""
],
[
"Zhang",
"Zizhou",
""
],
[
"Zhang",
"Rong",
""
],
[
"Yin",
"Yuchen",
""
],
[
"Chang",
"Chia Hong",
""
],
[
"Shen",
"Qinyan",
""
]
] |
2502.05258 | Oliver Philcox | Oliver H. E. Philcox | Searching for Inflationary Physics with the CMB Trispectrum: 2. Code & Validation | 35 pages, 27 figures, 10263 lines of code. PolySpec is available at https://github.com/oliverphilcox/PolySpec. Accepted by Phys. Rev. D | null | null | null | astro-ph.CO gr-qc hep-ph hep-th | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | To unlock the vast potential of the CMB trispectrum, we require both robust estimators and efficient computational tools. In this work, we introduce the public code PolySpec: a suite of quartic estimators designed to measure the amplitudes of a wide variety of inflationary templates, including local non-Gaussianity, effective field theory models, direction-dependent trispectra, spinning massive particle exchange, and weak gravitational lensing. PolySpec includes a python/cython implementation of each estimator derived in Paper 1 and has been carefully optimized to ensure efficient use of computational resources. We perform a broad range of validation tests, which demonstrate that the estimator is unbiased and minimum-variance, both in Gaussian and non-Gaussian regimes. In addition, we forecast constraints on various types of trispectra; this highlights the utility of CMB polarization and demonstrates that many models of primordial physics are poorly correlated with the simple templates considered in previous studies. This work lays the foundation for the Planck trispectrum analyses performed in Paper 3. | [
{
"version": "v1",
"created": "Fri, 07 Feb 2025 19:00:00 GMT"
},
{
"version": "v2",
"created": "Wed, 12 Feb 2025 15:47:51 GMT"
},
{
"version": "v3",
"created": "Wed, 28 May 2025 20:04:21 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Philcox",
"Oliver H. E.",
""
]
] |
2502.05527 | Hyeon-Woo Jeong | Hyeon-Woo Jeong, Jiho Kim, Boknam Chae, Kenji Watanabe, Takashi Taniguchi and Gil-Ho Lee | Preserving Twist-Angle in Marginally Twisted Double-Bilayer Graphene Devices During Fabrication | 15 pages, 4 figures | null | 10.1021/acsanm.5c00459 | null | cond-mat.mes-hall cond-mat.mtrl-sci | http://creativecommons.org/licenses/by/4.0/ | Twisted van der Waals heterostructures provide a platform for studying a wide range of electron correlation phenomena, including unconventional superconductivity and correlated insulating states. However, fabricating such devices is challenging due to the difficulty in achieving and maintaining homogeneous twist-angles. Here, we present a fabrication method to preserve the twist-angle with minimal deformation. We fabricated marginally twisted double-bilayer graphene (mTDBG) stacks and directly imaged the resulting triangular superlattice periodicity via scattering-type scanning near-field optical microscopy (s-SNOM). This technique enabled us to monitor twist-angle deformation at each fabrication step, paving the way for more reliable device fabrication and facilitating the exploration of twist-angle-dependent physics. | [
{
"version": "v1",
"created": "Sat, 08 Feb 2025 11:17:44 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Jeong",
"Hyeon-Woo",
""
],
[
"Kim",
"Jiho",
""
],
[
"Chae",
"Boknam",
""
],
[
"Watanabe",
"Kenji",
""
],
[
"Taniguchi",
"Takashi",
""
],
[
"Lee",
"Gil-Ho",
""
]
] |
2502.05540 | Qirui Wu | Qirui Wu, Shizhou Zhang, De Cheng, Yinghui Xing, Di Xu, Peng Wang, Yanning Zhang | Demystifying Catastrophic Forgetting in Two-Stage Incremental Object Detector | Accepted in ICML2025 | null | null | null | cs.CV | http://creativecommons.org/licenses/by/4.0/ | Catastrophic forgetting is a critical chanllenge for incremental object detection (IOD). Most existing methods treat the detector monolithically, relying on instance replay or knowledge distillation without analyzing component-specific forgetting. Through dissection of Faster R-CNN, we reveal a key insight: Catastrophic forgetting is predominantly localized to the RoI Head classifier, while regressors retain robustness across incremental stages. This finding challenges conventional assumptions, motivating us to develop a framework termed NSGP-RePRE. Regional Prototype Replay (RePRE) mitigates classifier forgetting via replay of two types of prototypes: coarse prototypes represent class-wise semantic centers of RoI features, while fine-grained prototypes model intra-class variations. Null Space Gradient Projection (NSGP) is further introduced to eliminate prototype-feature misalignment by updating the feature extractor in directions orthogonal to subspace of old inputs via gradient projection, aligning RePRE with incremental learning dynamics. Our simple yet effective design allows NSGP-RePRE to achieve state-of-the-art performance on the Pascal VOC and MS COCO datasets under various settings. Our work not only advances IOD methodology but also provide pivotal insights for catastrophic forgetting mitigation in IOD. Code is available at \href{https://github.com/fanrena/NSGP-RePRE}{https://github.com/fanrena/NSGP-RePRE} . | [
{
"version": "v1",
"created": "Sat, 08 Feb 2025 12:10:02 GMT"
},
{
"version": "v2",
"created": "Mon, 17 Feb 2025 12:36:11 GMT"
},
{
"version": "v3",
"created": "Thu, 29 May 2025 07:12:52 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Wu",
"Qirui",
""
],
[
"Zhang",
"Shizhou",
""
],
[
"Cheng",
"De",
""
],
[
"Xing",
"Yinghui",
""
],
[
"Xu",
"Di",
""
],
[
"Wang",
"Peng",
""
],
[
"Zhang",
"Yanning",
""
]
] |
2502.05689 | Oleksandr Stashko | R. A. Konoplya, O. S. Stashko | Transition from Regular Black Holes to Wormholes in Covariant Effective Quantum Gravity: Scattering, Quasinormal Modes, and Hawking Radiation | null | Phys. Rev. D 111, 084031, (2025) | 10.1103/PhysRevD.111.084031 | null | gr-qc | http://creativecommons.org/licenses/by/4.0/ | Utilizing the Hamiltonian constraints approach, a quantum-corrected solution has been derived \cite{Zhang:2024ney}, which describes either a regular black hole or a traversable wormhole, contingent upon the value of the quantum parameter. In this work, we compute the quasinormal modes associated with axial gravitational and test fields' perturbations of these objects. We see that due to quantum corrections near the event horizon, the first several overtones deviate from their Schwarzschild values at an increasing rate. The transition between the black hole and wormhole states is marked by modifications in the late-time signal. Our findings reveal that the fundamental quasinormal modes of quantum-corrected black holes exhibit only slight deviations from those of the classical Schwarzschild solution. However, at the transition, the spectrum undergoes significant changes, with the wormhole state characterized by exceptionally long-lived quasinormal modes. In addition, we calculate absorption cross-sections of partial waves, grey-body factors and energy emission rates of Hawking radiation. | [
{
"version": "v1",
"created": "Sat, 08 Feb 2025 20:44:31 GMT"
},
{
"version": "v2",
"created": "Wed, 28 May 2025 21:22:04 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Konoplya",
"R. A.",
""
],
[
"Stashko",
"O. S.",
""
]
] |
2502.05743 | Xiao Li | Xiao Li, Zekai Zhang, Xiang Li, Siyi Chen, Zhihui Zhu, Peng Wang, Qing Qu | Understanding Representation Dynamics of Diffusion Models via Low-Dimensional Modeling | First two authors contributed equally | null | null | null | cs.LG cs.CV | http://creativecommons.org/licenses/by/4.0/ | Diffusion models, though originally designed for generative tasks, have demonstrated impressive self-supervised representation learning capabilities. A particularly intriguing phenomenon in these models is the emergence of unimodal representation dynamics, where the quality of learned features peaks at an intermediate noise level. In this work, we conduct a comprehensive theoretical and empirical investigation of this phenomenon. Leveraging the inherent low-dimensionality structure of image data, we theoretically demonstrate that the unimodal dynamic emerges when the diffusion model successfully captures the underlying data distribution. The unimodality arises from an interplay between denoising strength and class confidence across noise scales. Empirically, we further show that, in classification tasks, the presence of unimodal dynamics reliably indicates generalization: it emerges when the model generalizes and gradually transitions to a monotonically decreasing curve as the model begins to memorize the training data. | [
{
"version": "v1",
"created": "Sun, 09 Feb 2025 01:58:28 GMT"
},
{
"version": "v2",
"created": "Wed, 28 May 2025 18:17:58 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Li",
"Xiao",
""
],
[
"Zhang",
"Zekai",
""
],
[
"Li",
"Xiang",
""
],
[
"Chen",
"Siyi",
""
],
[
"Zhu",
"Zhihui",
""
],
[
"Wang",
"Peng",
""
],
[
"Qu",
"Qing",
""
]
] |
2502.06044 | Juraj Marusic | Getoar Sopa, Juraj Marusic, Marco Avella-Medina and John P. Cunningham | Scalable Differentially Private Bayesian Optimization | 25 pages, 5 figures | null | null | null | stat.ML cs.LG | http://creativecommons.org/licenses/by/4.0/ | In recent years, there has been much work on scaling Bayesian Optimization to high-dimensional problems, for example hyperparameter tuning in large machine learning models. These scalable methods have been successful, finding high objective values much more quickly than traditional global Bayesian Optimization or random search-based methods. At the same time, these large models often use sensitive data, but preservation of Differential Privacy has not scaled alongside these modern Bayesian Optimization procedures. Here we develop a method to privately optimize potentially high-dimensional parameter spaces using privatized Gradient Informative Bayesian Optimization. Our theoretical results show that under suitable conditions, our method converges exponentially fast to a locally optimal parameter configuration, up to a natural privacy error. Moreover, regardless of whether the assumptions are satisfied, we prove that our algorithm maintains privacy and empirically display superior performance to existing methods in the high-dimensional hyperparameter setting. | [
{
"version": "v1",
"created": "Sun, 09 Feb 2025 21:49:50 GMT"
},
{
"version": "v2",
"created": "Wed, 28 May 2025 20:06:18 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Sopa",
"Getoar",
""
],
[
"Marusic",
"Juraj",
""
],
[
"Avella-Medina",
"Marco",
""
],
[
"Cunningham",
"John P.",
""
]
] |
2502.06152 | Ziyang Guo | Ziyang Guo, Yifan Wu, Jason Hartline and Jessica Hullman | The Value of Information in Human-AI Decision-making | null | null | null | null | cs.AI cs.LG | http://creativecommons.org/licenses/by/4.0/ | Multiple agents -- including humans and AI models -- are increasingly combined to make decisions with the expectation of achieving complementary performance, where the decisions they make together outperform those made individually. However, knowing how to improve the performance of collaborating agents is often difficult without knowing more about what particular information and strategies each agent employs. With a focus on human-AI pairings, we contribute a decision-theoretic framework for characterizing the value of information -- and consequently, opportunities for agents to better exploit available information -- in AI-assisted decision workflows. We present a novel explanation technique (ILIV-SHAP) that adapts SHAP explanations to highlight human-complementing information. We validate the effectiveness of the framework and ILIV-SHAP through a study of human-AI decision-making. We show that our measure of complementary information can be used to identify which AI model will best complement human decisions. We also find that presenting ILIV-SHAP with AI predictions leads to reliably greater reductions in error over non-AI assisted decisions more than vanilla SHAP. | [
{
"version": "v1",
"created": "Mon, 10 Feb 2025 04:50:42 GMT"
},
{
"version": "v2",
"created": "Thu, 03 Apr 2025 14:57:23 GMT"
},
{
"version": "v3",
"created": "Tue, 15 Apr 2025 19:26:06 GMT"
},
{
"version": "v4",
"created": "Wed, 28 May 2025 18:44:35 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Guo",
"Ziyang",
""
],
[
"Wu",
"Yifan",
""
],
[
"Hartline",
"Jason",
""
],
[
"Hullman",
"Jessica",
""
]
] |
2502.06381 | Lukas Pin | Lukas Pin, Sof\'ia S. Villar, William F. Rosenberger | Revisiting Optimal Allocations for Binary Responses: Insights from Considering Type-I Error Rate Control | 14 pages, 2 figures, 5 tables | null | null | null | stat.ME math.ST stat.AP stat.TH | http://creativecommons.org/licenses/by/4.0/ | This work revisits optimal response-adaptive designs from a type-I error rate perspective, highlighting when and how much these allocations exacerbate type-I error rate inflation - an issue previously undocumented. We explore a range of approaches from the literature that can be applied to reduce type-I error rate inflation. However, we found that all of these approaches fail to give a robust solution to the problem. To address this, we derive two optimal allocation proportions, incorporating the more robust score test (instead of the Wald test) with finite sample estimators (instead of the unknown true values) in the formulation of the optimization problem. One proportion optimizes statistical power and the other minimizes the total number failures in a trial while maintaining a fixed variance level. Through simulations based on an early-phase and a confirmatory trial we provide crucial practical insight into how these new optimal proportion designs can offer substantial patient outcomes advantages while controlling type-I error rate. While we focused on binary outcomes, the framework offers valuable insights that naturally extend to other outcome types, multi-armed trials and alternative measures of interest. | [
{
"version": "v1",
"created": "Mon, 10 Feb 2025 12:06:07 GMT"
},
{
"version": "v2",
"created": "Fri, 28 Feb 2025 09:31:38 GMT"
},
{
"version": "v3",
"created": "Thu, 29 May 2025 14:44:15 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Pin",
"Lukas",
""
],
[
"Villar",
"Sofía S.",
""
],
[
"Rosenberger",
"William F.",
""
]
] |
2502.06545 | Annie Marsden | Annie Marsden and Elad Hazan | Universal Sequence Preconditioning | 35 pages, 7 figures | null | null | null | cs.LG stat.ML | http://creativecommons.org/licenses/by/4.0/ | We study the problem of preconditioning in the setting of sequential prediction. From the theoretical lens of linear dynamical systems, we show that applying a convolution to the input sequence translates to applying a polynomial to the unknown transition matrix in the hidden space. With this insight, we develop a novel preconditioning method that convolves the input sequence with the coefficients of the Chebyshev or Legendre polynomials. We formally prove that this improves the regret of two distinct prediction methods. Moreover, using this preconditioning technique on either method gives the first sublinear regret bounds that are also hidden dimension independent (up to logarithmic factors) even when the hidden transition matrix is asymmetric. From rigorous experiments on synthetic data we show that our simple preconditioning method generalizes to both 1) settings where the data is not from a linear dynamical system and 2) a broad range of learning algorithms, including recurrent neural networks. | [
{
"version": "v1",
"created": "Mon, 10 Feb 2025 15:10:06 GMT"
},
{
"version": "v2",
"created": "Thu, 29 May 2025 02:26:29 GMT"
}
] | 2025-05-30T00:00:00 | [
[
"Marsden",
"Annie",
""
],
[
"Hazan",
"Elad",
""
]
] |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.