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 |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
2410.01405 | Kevin Xu | Kevin Xu and Issei Sato | On Expressive Power of Looped Transformers: Theoretical Analysis and Enhancement via Timestep Encoding | Camera-ready version for ICML 2025 | null | null | null | cs.LG | http://creativecommons.org/licenses/by/4.0/ | Looped Transformers provide advantages in parameter efficiency, computational capabilities, and generalization for reasoning tasks. However, their expressive power regarding function approximation remains underexplored. In this paper, we establish the approximation rate of Looped Transformers by defining the modulus of continuity for sequence-to-sequence functions. This reveals a limitation specific to the looped architecture. That is, the analysis prompts the incorporation of scaling parameters for each loop, conditioned on timestep encoding. Experiments validate the theoretical results, showing that increasing the number of loops enhances performance, with further gains achieved through the timestep encoding. | [
{
"version": "v1",
"created": "Wed, 02 Oct 2024 10:31:17 GMT"
},
{
"version": "v2",
"created": "Tue, 08 Oct 2024 16:41:40 GMT"
},
{
"version": "v3",
"created": "Mon, 25 Nov 2024 08:17:14 GMT"
},
{
"version": "v4",
"created": "Mon, 03 Feb 2025 07:30:21 GMT"
},
{
"version": "v5",
"created": "Tue, 11 Mar 2025 15:51:21 GMT"
},
{
"version": "v6",
"created": "Fri, 02 May 2025 17:13:48 GMT"
},
{
"version": "v7",
"created": "Thu, 05 Jun 2025 13:30:23 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Xu",
"Kevin",
""
],
[
"Sato",
"Issei",
""
]
] |
2410.01482 | Gabriel Kasmi | Gabriel Kasmi and Amandine Brunetto and Thomas Fel and Jayneel Parekh | One Wave To Explain Them All: A Unifying Perspective On Feature Attribution | Accepted to ICML 2025 | null | null | null | stat.ML cs.AI cs.LG | http://creativecommons.org/licenses/by/4.0/ | Feature attribution methods aim to improve the transparency of deep neural networks by identifying the input features that influence a model's decision. Pixel-based heatmaps have become the standard for attributing features to high-dimensional inputs, such as images, audio representations, and volumes. While intuitive and convenient, these pixel-based attributions fail to capture the underlying structure of the data. Moreover, the choice of domain for computing attributions has often been overlooked. This work demonstrates that the wavelet domain allows for informative and meaningful attributions. It handles any input dimension and offers a unified approach to feature attribution. Our method, the Wavelet Attribution Method (WAM), leverages the spatial and scale-localized properties of wavelet coefficients to provide explanations that capture both the where and what of a model's decision-making process. We show that WAM quantitatively matches or outperforms existing gradient-based methods across multiple modalities, including audio, images, and volumes. Additionally, we discuss how WAM bridges attribution with broader aspects of model robustness and transparency. Project page: https://gabrielkasmi.github.io/wam/ | [
{
"version": "v1",
"created": "Wed, 02 Oct 2024 12:34:04 GMT"
},
{
"version": "v2",
"created": "Thu, 05 Jun 2025 16:15:38 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Kasmi",
"Gabriel",
""
],
[
"Brunetto",
"Amandine",
""
],
[
"Fel",
"Thomas",
""
],
[
"Parekh",
"Jayneel",
""
]
] |
2410.01704 | Soham Govande | Soham V. Govande | An Exploration of Self-Supervised Mutual Information Alignment for Multi-Task Settings | null | null | null | null | cs.CL | http://creativecommons.org/publicdomain/zero/1.0/ | There is a growing need for pluralistic alignment methods that can steer language models towards individual attributes and preferences. One such method, Self-Supervised Alignment with Mutual Information (SAMI), uses conditional mutual information to encourage the connection between behavioral preferences and model responses. We conduct two experiments exploring SAMI in multi-task settings. First, we compare SAMI to Direct Preference Optimization (DPO) on a multi-task benchmark (MT-Bench), using a stronger model to generate training data for a weaker one across diverse categories (humanities, STEM, extraction, coding, math, reasoning, and roleplay). Our results indicate that one iteration of SAMI has a 57% win rate against DPO, with significant variation in performance between task categories. Second, we examine SAMI's impact on mathematical accuracy (GSM-8K) relative to supervised fine-tuning (SFT). While SAMI increases zero-shot performance by 1.1%, SFT is more effective with a 3.2% boost. However, SAMI shows interesting scaling trends. When given 10 attempts, SAMI improves accuracy by 3.9%, while SFT achieves a 10.1% increase. Combining SAMI with SFT yields an additional improvement of 1.3% in multi-attempt settings, though single-attempt accuracy remains unchanged. | [
{
"version": "v1",
"created": "Wed, 02 Oct 2024 16:15:04 GMT"
},
{
"version": "v2",
"created": "Thu, 05 Jun 2025 00:32:25 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Govande",
"Soham V.",
""
]
] |
2410.02292 | Haiyuan Wang Dr. | Haiyuan Wang, Nicolas Stenger, Peder Lyngby, Mikael Kuisma, Jakob Schi{\o}tz, and Kristian Sommer Thygesen | Two-dimensional materials as ideal substrates for molecular quantum emitters | null | null | null | null | cond-mat.mtrl-sci physics.app-ph physics.med-ph | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | The generation and manipulation of non-classical states of light is central to quantum technologies. Color centers in insulators have been extensively studied for single-photon generation, but organic molecules immobilized on substrates have gained attention due to their superior scalability, large oscillator strengths, and tunable emission frequency. Here, we use first-principles calculations to investigate the photoemission from organic molecules adsorbed on 2D materials. Focusing on terrylene on hexagonal boron nitride (hBN), we obtain zero phonon line (ZPL) energies and emission lineshapes in excellent agreement with experiments. Notably, antisite defects in hBN can immobilize the molecule without influencing its key emission features. We further show that the main effect of the 2D substrate is to introduce sharp sidebands near the ZPL as a fingerprint of hindered rotational, translational, or bending modes of the molecule. Our findings provide insight into how substrate interactions shape the optical properties of molecular systems for quantum applications. | [
{
"version": "v1",
"created": "Thu, 03 Oct 2024 08:22:57 GMT"
},
{
"version": "v2",
"created": "Thu, 05 Jun 2025 12:33:54 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Wang",
"Haiyuan",
""
],
[
"Stenger",
"Nicolas",
""
],
[
"Lyngby",
"Peder",
""
],
[
"Kuisma",
"Mikael",
""
],
[
"Schiøtz",
"Jakob",
""
],
[
"Thygesen",
"Kristian Sommer",
""
]
] |
2410.03208 | Iulia Duta | Iulia Duta and Pietro Li\`o | SPHINX: Structural Prediction using Hypergraph Inference Network | Accepted at International Conference on Machine Learning (ICML 2025) | null | null | null | cs.LG | http://creativecommons.org/licenses/by/4.0/ | The importance of higher-order relations is widely recognized in a large number of real-world systems. However, annotating them is a tedious and sometimes impossible task. Consequently, current approaches for data modelling either ignore the higher-order interactions altogether or simplify them into pairwise connections. In order to facilitate higher-order processing, even when a hypergraph structure is not available, we introduce Structural Prediction using Hypergraph Inference Network (SPHINX), a model that learns to infer a latent hypergraph structure in an unsupervised way, solely from the final node-level signal. The model consists of a soft, differentiable clustering method used to sequentially predict, for each hyperedge, the probability distribution over the nodes and a sampling algorithm that converts them into an explicit hypergraph structure. We show that the recent advancement in $k$-subset sampling represents a suitable tool for producing discrete hypergraph structures, addressing some of the training instabilities exhibited by prior works. The resulting model can generate the higher-order structure necessary for any modern hypergraph neural network, facilitating the capture of higher-order interaction in domains where annotating them is difficult. Through extensive ablation studies and experiments conducted on two challenging datasets for trajectory prediction, we demonstrate that our model is capable of inferring suitable latent hypergraphs, that are interpretable and enhance the final performance. | [
{
"version": "v1",
"created": "Fri, 04 Oct 2024 07:49:57 GMT"
},
{
"version": "v2",
"created": "Thu, 05 Jun 2025 08:13:02 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Duta",
"Iulia",
""
],
[
"Liò",
"Pietro",
""
]
] |
2410.03562 | Alexander Barg | Arda Aydin and Alexander Barg | Class of codes correcting absorptions and emissions | null | Physical Review A, vol.111, no.2, 2025, 022415 | 10.1103/PhysRevA.111.022415 | null | quant-ph cs.IT math.IT | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We construct a general family of quantum codes that protect against all emission, absorption, dephasing, and raising/lowering errors up to an arbitrary fixed order. Such codes are known in the literature as absorption-emission (AE) codes. We derive simplified error correction conditions for a general AE code and show that any permutation-invariant code that corrects $\le t$ errors can be mapped to an AE code that corrects up to order-$t$ transitions. Carefully tuning the parameters of permutationally invariant codes, we construct several examples of efficient AE codes, hosted in systems with low total angular momentum. Our results also imply that spin codes can be mapped to AE codes, enabling us to characterize logical operators for certain subclasses of such codes. | [
{
"version": "v1",
"created": "Fri, 04 Oct 2024 16:14:03 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Aydin",
"Arda",
""
],
[
"Barg",
"Alexander",
""
]
] |
2410.04055 | Jiayi He | Jiayi He, Hehai Lin, Qingyun Wang, Yi Fung, Heng Ji | Self-Correction is More than Refinement: A Learning Framework for Visual and Language Reasoning Tasks | Accepted by ACL 2025 Findings | null | null | null | cs.CL | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | While Vision-Language Models (VLMs) have shown remarkable abilities in visual and language reasoning tasks, they invariably generate flawed responses. Self-correction that instructs models to refine their outputs presents a promising solution to this issue. Previous studies have mainly concentrated on Large Language Models (LLMs), while the self-correction abilities of VLMs, particularly concerning both visual and linguistic information, remain largely unexamined. This study investigates the self-correction capabilities of VLMs during both inference and fine-tuning stages. We introduce a Self-Correction Learning (SCL) approach that enables VLMs to learn from their self-generated self-correction data through Direct Preference Optimization (DPO) without relying on external feedback, facilitating self-improvement. Specifically, we collect preferred and disfavored samples based on the correctness of initial and refined responses, which are obtained by two-turn self-correction with VLMs during the inference stage. Experimental results demonstrate that although VLMs struggle to self-correct effectively during iterative inference without additional fine-tuning and external feedback, they can enhance their performance and avoid previous mistakes through preference fine-tuning when their self-generated self-correction data are categorized into preferred and disfavored samples. This study emphasizes that self-correction is not merely a refinement process; rather, it should enhance the reasoning abilities of models through additional training, enabling them to generate high-quality responses directly without further refinement. | [
{
"version": "v1",
"created": "Sat, 05 Oct 2024 06:28:54 GMT"
},
{
"version": "v2",
"created": "Thu, 08 May 2025 02:10:43 GMT"
},
{
"version": "v3",
"created": "Thu, 05 Jun 2025 13:25:43 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"He",
"Jiayi",
""
],
[
"Lin",
"Hehai",
""
],
[
"Wang",
"Qingyun",
""
],
[
"Fung",
"Yi",
""
],
[
"Ji",
"Heng",
""
]
] |
2410.04264 | Yoonsoo Nam | Yoonsoo Nam, Nayara Fonseca, Seok Hyeong Lee, Chris Mingard, Niclas Goring, Ouns El Harzli, Abdurrahman Hadi Erturk, Soufiane Hayou, and Ard A. Louis | Disentangling Rich Dynamics from Feature Learning: A Framework for Independent Measurements | null | null | null | null | stat.ML cs.LG | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | In machine learning, it is widely believed that dynamic feature transformation (the rich regime) enhances predictive performance. However, this link does not always hold, and existing richness measures rely on correlated factors - such as performance or parameter norms - which can complicate the analysis of feature learning. We introduce (1) a measure that quantifies the rich regime independently of performance, and (2) interpretable feature metrics for visualization. Leveraging low-rank bias, our approach generalizes neural collapse metrics and captures lazy-to-rich transitions (e.g., grokking) without relying on performance as a proxy. We reveal how batch normalization and training set size influence lazy/rich dynamics for VGG16 and ResNet18 on CIFAR-10/100, opening avenues for better understanding feature learning. | [
{
"version": "v1",
"created": "Sat, 05 Oct 2024 18:53:48 GMT"
},
{
"version": "v2",
"created": "Wed, 04 Jun 2025 21:53:51 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Nam",
"Yoonsoo",
""
],
[
"Fonseca",
"Nayara",
""
],
[
"Lee",
"Seok Hyeong",
""
],
[
"Mingard",
"Chris",
""
],
[
"Goring",
"Niclas",
""
],
[
"Harzli",
"Ouns El",
""
],
[
"Erturk",
"Abdurrahman Hadi",
""
],
[
"Hayou",
"Soufiane",
""
],
[
"Louis",
"Ard A.",
""
]
] |
2410.04522 | Frans Van Die | Frans van Die, Ivan Rapoport, Yonadav Barry Ginat, Vincent Desjacques | Detection prospects for the GW background of Galactic (sub)solar mass primordial black holes | 44 pages, 17 figures, fixed typos and unclear phrasing | JCAP 05 (2025) 036 | 10.1088/1475-7516/2025/05/036 | null | astro-ph.CO astro-ph.GA | http://creativecommons.org/licenses/by/4.0/ | In multi-component dark matter models, a fraction $f_\text{pbh}$ of the dark matter could be in the form of primordial black holes (PBHs) with (sub)solar masses. Some would have formed binaries that presently trace the Milky Way halo of particle dark matter. We explore the gravitational wave (GW) signal produced by such a hypothetical population of Galactic PBH binaries and assess its detectability by the LISA experiment. For this purpose, we model the formation and evolution of early-type PBH binaries accounting for GW hardening and binary disruption in the Milky Way. Our analysis reveals that the present-day Galactic population of PBH binaries is characterized by very high orbital eccentricities $|1-e|\ll 1$. For a PBH mass $M_{\rm pbh} \sim 0.1 - 1 M_\odot$, this yields a GW background that peaks in the millihertz frequency range where the LISA instrumental noise is minimum. While this signal remains below the LISA detection threshold for viable $f_\text{pbh}\lesssim 0.01$, future GW observatories such as DECIGO and BBO could detect it if $0.01\lesssim M_{\rm pbh} \lesssim 0.1 M_\odot$. Furthermore, we anticipate that, after 5 years of observations, LISA should be able to detect $\mathcal{O}(100)$ (resp. $\mathcal{O}(1)$) loud Galactic PBH binaries of mass $M_{\rm pbh} \sim 0.1 - 1 M_\odot$ with a SNR $\geq 5$ if $f_{\rm pbh}=0.01$ (resp. $f_{\rm pbh}=0.001$). Nonlinear effects not considered here such as mass accretion and dynamical capture could alter these predictions. | [
{
"version": "v1",
"created": "Sun, 06 Oct 2024 15:30:12 GMT"
},
{
"version": "v2",
"created": "Thu, 05 Jun 2025 12:53:31 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"van Die",
"Frans",
""
],
[
"Rapoport",
"Ivan",
""
],
[
"Ginat",
"Yonadav Barry",
""
],
[
"Desjacques",
"Vincent",
""
]
] |
2410.06307 | Dheeraj Narasimha | Nicolas Gast, Dheeraj Narasimha | Model Predictive Control is Almost Optimal for Restless Bandit | Reviewed and accepted to COLT 2025 | null | null | null | math.OC cs.LG math.PR stat.ML | http://creativecommons.org/licenses/by-sa/4.0/ | We consider the discrete time infinite horizon average reward restless markovian bandit (RMAB) problem. We propose a \emph{model predictive control} based non-stationary policy with a rolling computational horizon $\tau$. At each time-slot, this policy solves a $\tau$ horizon linear program whose first control value is kept as a control for the RMAB. Our solution requires minimal assumptions and quantifies the loss in optimality in terms of $\tau$ and the number of arms, $N$. We show that its sub-optimality gap is $O(1/\sqrt{N})$ in general, and $\exp(-\Omega(N))$ under a local-stability condition. Our proof is based on a framework from dynamic control known as \emph{dissipativity}. Our solution easy to implement and performs very well in practice when compared to the state of the art. Further, both our solution and our proof methodology can easily be generalized to more general constrained MDP settings and should thus, be of great interest to the burgeoning RMAB community. | [
{
"version": "v1",
"created": "Tue, 08 Oct 2024 19:34:23 GMT"
},
{
"version": "v2",
"created": "Thu, 05 Jun 2025 07:08:44 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Gast",
"Nicolas",
""
],
[
"Narasimha",
"Dheeraj",
""
]
] |
2410.06415 | Jillian Fisher | Jillian Fisher, Shangbin Feng, Robert Aron, Thomas Richardson, Yejin Choi, Daniel W. Fisher, Jennifer Pan, Yulia Tsvetkov, Katharina Reinecke | Biased AI can Influence Political Decision-Making | null | null | null | null | cs.HC cs.AI | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | As modern large language models (LLMs) become integral to everyday tasks, concerns about their inherent biases and their potential impact on human decision-making have emerged. While bias in models are well-documented, less is known about how these biases influence human decisions. This paper presents two interactive experiments investigating the effects of partisan bias in LLMs on political opinions and decision-making. Participants interacted freely with either a biased liberal, biased conservative, or unbiased control model while completing these tasks. We found that participants exposed to partisan biased models were significantly more likely to adopt opinions and make decisions which matched the LLM's bias. Even more surprising, this influence was seen when the model bias and personal political partisanship of the participant were opposite. However, we also discovered that prior knowledge of AI was weakly correlated with a reduction of the impact of the bias, highlighting the possible importance of AI education for robust mitigation of bias effects. Our findings not only highlight the critical effects of interacting with biased LLMs and its ability to impact public discourse and political conduct, but also highlights potential techniques for mitigating these risks in the future. | [
{
"version": "v1",
"created": "Tue, 08 Oct 2024 22:56:00 GMT"
},
{
"version": "v2",
"created": "Mon, 04 Nov 2024 20:12:07 GMT"
},
{
"version": "v3",
"created": "Thu, 05 Jun 2025 15:55:15 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Fisher",
"Jillian",
""
],
[
"Feng",
"Shangbin",
""
],
[
"Aron",
"Robert",
""
],
[
"Richardson",
"Thomas",
""
],
[
"Choi",
"Yejin",
""
],
[
"Fisher",
"Daniel W.",
""
],
[
"Pan",
"Jennifer",
""
],
[
"Tsvetkov",
"Yulia",
""
],
[
"Reinecke",
"Katharina",
""
]
] |
2410.06895 | Chenhao Sun | Chenhao Sun, Yuhao Mao, Mark Niklas M\"uller, Martin Vechev | Average Certified Radius is a Poor Metric for Randomized Smoothing | ICML'25 | null | null | null | cs.LG | http://creativecommons.org/licenses/by/4.0/ | Randomized smoothing (RS) is popular for providing certified robustness guarantees against adversarial attacks. The average certified radius (ACR) has emerged as a widely used metric for tracking progress in RS. However, in this work, for the first time we show that ACR is a poor metric for evaluating robustness guarantees provided by RS. We theoretically prove not only that a trivial classifier can have arbitrarily large ACR, but also that ACR is extremely sensitive to improvements on easy samples. In addition, the comparison using ACR has a strong dependence on the certification budget. Empirically, we confirm that existing training strategies, though improving ACR, reduce the model's robustness on hard samples consistently. To strengthen our findings, we propose strategies, including explicitly discarding hard samples, reweighing the dataset with approximate certified radius, and extreme optimization for easy samples, to replicate the progress in RS training and even achieve the state-of-the-art ACR on CIFAR-10, without training for robustness on the full data distribution. Overall, our results suggest that ACR has introduced a strong undesired bias to the field, and its application should be discontinued in RS. Finally, we suggest using the empirical distribution of $p_A$, the accuracy of the base model on noisy data, as an alternative metric for RS. | [
{
"version": "v1",
"created": "Wed, 09 Oct 2024 13:58:41 GMT"
},
{
"version": "v2",
"created": "Fri, 31 Jan 2025 13:07:32 GMT"
},
{
"version": "v3",
"created": "Tue, 03 Jun 2025 14:57:28 GMT"
},
{
"version": "v4",
"created": "Wed, 04 Jun 2025 21:15:56 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Sun",
"Chenhao",
""
],
[
"Mao",
"Yuhao",
""
],
[
"Müller",
"Mark Niklas",
""
],
[
"Vechev",
"Martin",
""
]
] |
2410.06952 | Mikhail Zubkov Dr | Praveen D. Xavier, M.A.Zubkov | Generalized Wigner-Weyl calculus for gauge theory and non-dissipative transport | Latex, 46 pages | null | null | null | hep-th | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We consider a theory of fermions interacting with a (in general, non-Abelian) gauge field. The theory is assumed to be essentially inhomogeneous, which might be provided by non-trivial background fields interacting with both fermions and gauge bosons. For this theory, a version of Wigner-Weyl calculus is developed, in which the Wigner transformation of the fermion Green function belongs to a matrix representation of gauge group. We demonstrate the power of the proposed formalism through the representation of responses of vector and axial currents to the gauge field strength through the topological invariants composed of the Wigner transformed two-point Green functions. This way a new family of non-dissipative transport phenomena is introduced. In particular, we discuss the non-Abelian versions of the chiral separation effect and of the quantum Hall effect. | [
{
"version": "v1",
"created": "Wed, 09 Oct 2024 14:50:27 GMT"
},
{
"version": "v2",
"created": "Wed, 05 Feb 2025 22:18:03 GMT"
},
{
"version": "v3",
"created": "Wed, 04 Jun 2025 23:50:31 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Xavier",
"Praveen D.",
""
],
[
"Zubkov",
"M. A.",
""
]
] |
2410.07096 | Mingde Zhao | Mingde Zhao, Tristan Sylvain, Romain Laroche, Doina Precup, Yoshua Bengio | Rejecting Hallucinated State Targets during Planning | [20250604]: ICML 2025 Camera Ready, https://github.com/mila-iqia/delusions | null | null | null | cs.AI | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Generative models can be used in planning to propose targets corresponding to states that agents deem either likely or advantageous to experience. However, imperfections, common in learned models, lead to infeasible hallucinated targets, which can cause delusional behaviors and thus safety concerns. This work first categorizes and investigates the properties of several kinds of infeasible targets. Then, we devise a strategy to reject infeasible targets with a generic target evaluator, which trains alongside planning agents as an add-on without the need to change the behavior nor the architectures of the agent (and the generative model) it is attached to. We highlight that, without proper design, the evaluator can produce delusional estimates, rendering the strategy futile. Thus, to learn correct evaluations of infeasible targets, we propose to use a combination of learning rule, architecture, and two assistive hindsight relabeling strategies. Our experiments validate significant reductions in delusional behaviors and performance improvements for several kinds of existing planning agents. | [
{
"version": "v1",
"created": "Wed, 09 Oct 2024 17:35:25 GMT"
},
{
"version": "v2",
"created": "Thu, 10 Oct 2024 16:55:33 GMT"
},
{
"version": "v3",
"created": "Sun, 13 Oct 2024 02:54:20 GMT"
},
{
"version": "v4",
"created": "Wed, 16 Oct 2024 18:46:35 GMT"
},
{
"version": "v5",
"created": "Mon, 18 Nov 2024 17:40:57 GMT"
},
{
"version": "v6",
"created": "Fri, 07 Feb 2025 18:10:13 GMT"
},
{
"version": "v7",
"created": "Sun, 11 May 2025 20:13:14 GMT"
},
{
"version": "v8",
"created": "Wed, 04 Jun 2025 20:43:13 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Zhao",
"Mingde",
""
],
[
"Sylvain",
"Tristan",
""
],
[
"Laroche",
"Romain",
""
],
[
"Precup",
"Doina",
""
],
[
"Bengio",
"Yoshua",
""
]
] |
2410.07984 | Ke Li | Shi-Bing Li, Ke Li, Lei Yu | Large Deviation Analysis for the Reverse Shannon Theorem | See also concurrent and independent works arXiv:2410.07051 and arXiv:2410.10770. V1: prelimilary version. V2: presentation significantly improved, errors and typos fixed | null | null | null | cs.IT math.IT | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Channel simulation is to simulate a noisy channel using noiseless channels with unlimited shared randomness. This can be interpreted as the reverse problem to Shannon's noisy coding theorem. In contrast to previous works, our approach employs R\'enyi divergence (with the parameter $\alpha\in(0,\infty)$) to measure the level of approximation. Specifically, we obtain the reverse Shannon theorem under the R\'enyi divergence, which characterizes the R\'enyi simulation rate, the minimum communication cost rate required for the R\'enyi divergence vanishing asymptotically. We also investigate the behaviors of the R\'enyi divergence when the communication cost rate is above or below the R\'enyi simulation rate. When the communication cost rate is above the R\'enyi simulation rate, we provide a complete characterization of the convergence exponent, called the reliability function. When the communication cost rate is below the R\'enyi simulation rate, we determine the linear increasing rate for the R\'enyi divergence with parameter $\alpha\in(0,\infty]$, which implies the strong converse exponent for the $\alpha$-order fidelity. | [
{
"version": "v1",
"created": "Thu, 10 Oct 2024 14:39:05 GMT"
},
{
"version": "v2",
"created": "Thu, 05 Jun 2025 09:21:26 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Li",
"Shi-Bing",
""
],
[
"Li",
"Ke",
""
],
[
"Yu",
"Lei",
""
]
] |
2410.09425 | Adam Kasperski | Adam Kasperski, Pawel Zielinski | Computational complexity of the recoverable robust shortest path problem in acyclic digraphs | null | null | null | null | cs.DS math.OC | http://creativecommons.org/licenses/by/4.0/ | In this paper, the recoverable robust shortest path problem in acyclic digraphs is considered. The interval budgeted uncertainty representation is used to model the uncertain second-stage costs. The computational complexity of this problem has been open to date. In this paper, we prove that the problem is strongly NP-hard even for the case of layered acyclic digraphs. We also show that for the discrete budgeted uncertainty, the problem is not approximable unless P=NP. | [
{
"version": "v1",
"created": "Sat, 12 Oct 2024 08:02:25 GMT"
},
{
"version": "v2",
"created": "Thu, 05 Jun 2025 05:31:53 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Kasperski",
"Adam",
""
],
[
"Zielinski",
"Pawel",
""
]
] |
2410.09586 | Qiang Zhu | Kevin Parrish, Qingyang Hu and Qiang Zhu | PyLRO: A Python Calculator for Analyzing Long Range Structural Order | 8 pages, 5 figures | J. Appl. Phys. 137, 025101, 2025 | 10.1063/5.0244012 | null | cond-mat.mtrl-sci | http://creativecommons.org/licenses/by/4.0/ | We present PyLRO, an open-source Python calculator designed to detect, quantify, and display long-range order in periodic structures. The program's design methodology, workflow, and approach to order quantification are described and demonstrated using a simple toy model. Additionally, we apply PyLRO to a series of metastable AlPO4 structural intermediates from a prior high-pressure study, demonstrating how to compute and visualize structural order in all directions on a Miller sphere. We further highlight the program's capabilities through a high-throughput analysis of structural patterns in the pressure-induced amorphization of AlPO4, revealing atomistic insights within specific energy regions of massive amorphous structures. These results suggest that PyLRO can be a valuable tool for investigating crystal-amorphous transition in materials research. | [
{
"version": "v1",
"created": "Sat, 12 Oct 2024 16:39:44 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Parrish",
"Kevin",
""
],
[
"Hu",
"Qingyang",
""
],
[
"Zhu",
"Qiang",
""
]
] |
2410.10390 | Cornelius V. Braun | Cornelius V. Braun, Robert T. Lange, Marc Toussaint | Stein Variational Evolution Strategies | null | null | null | null | cs.LG cs.AI cs.NE | http://creativecommons.org/licenses/by/4.0/ | Stein Variational Gradient Descent (SVGD) is a highly efficient method to sample from an unnormalized probability distribution. However, the SVGD update relies on gradients of the log-density, which may not always be available. Existing gradient-free versions of SVGD make use of simple Monte Carlo approximations or gradients from surrogate distributions, both with limitations. To improve gradient-free Stein variational inference, we combine SVGD steps with evolution strategy (ES) updates. Our results demonstrate that the resulting algorithm generates high-quality samples from unnormalized target densities without requiring gradient information. Compared to prior gradient-free SVGD methods, we find that the integration of the ES update in SVGD significantly improves the performance on multiple challenging benchmark problems. | [
{
"version": "v1",
"created": "Mon, 14 Oct 2024 11:24:41 GMT"
},
{
"version": "v2",
"created": "Thu, 05 Jun 2025 13:25:39 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Braun",
"Cornelius V.",
""
],
[
"Lange",
"Robert T.",
""
],
[
"Toussaint",
"Marc",
""
]
] |
2410.11283 | Pankayaraj Pathmanathan | Pankayaraj Pathmanathan, Udari Madhushani Sehwag, Michael-Andrei Panaitescu-Liess, Furong Huang | AdvBDGen: Adversarially Fortified Prompt-Specific Fuzzy Backdoor Generator Against LLM Alignment | Published at the Neurips Safe Generative AI Workshop 2024 | null | null | null | cs.LG | http://creativecommons.org/licenses/by/4.0/ | With the growing adoption of reinforcement learning with human feedback (RLHF) for aligning large language models (LLMs), the risk of backdoor installation during alignment has increased, leading to unintended and harmful behaviors. Existing backdoor triggers are typically limited to fixed word patterns, making them detectable during data cleaning and easily removable post-poisoning. In this work, we explore the use of prompt-specific paraphrases as backdoor triggers, enhancing their stealth and resistance to removal during LLM alignment. We propose AdvBDGen, an adversarially fortified generative fine-tuning framework that automatically generates prompt-specific backdoors that are effective, stealthy, and transferable across models. AdvBDGen employs a generator-discriminator pair, fortified by an adversary, to ensure the installability and stealthiness of backdoors. It enables the crafting and successful installation of complex triggers using as little as 3% of the fine-tuning data. Once installed, these backdoors can jailbreak LLMs during inference, demonstrate improved stability against perturbations compared to traditional constant triggers, and are more challenging to remove. These findings underscore an urgent need for the research community to develop more robust defenses against adversarial backdoor threats in LLM alignment. | [
{
"version": "v1",
"created": "Tue, 15 Oct 2024 05:05:56 GMT"
},
{
"version": "v2",
"created": "Wed, 09 Apr 2025 16:09:27 GMT"
},
{
"version": "v3",
"created": "Thu, 05 Jun 2025 03:02:29 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Pathmanathan",
"Pankayaraj",
""
],
[
"Sehwag",
"Udari Madhushani",
""
],
[
"Panaitescu-Liess",
"Michael-Andrei",
""
],
[
"Huang",
"Furong",
""
]
] |
2410.11295 | Qin Wang | Minfeng Qi, Qin Wang, Zhipeng Wang, Lin Zhong, Zhixiong Gao, Tianqing Zhu, Shiping Chen, William Knottenbelt | BRC20 Pinning Attack | null | null | null | null | cs.CR cs.CE cs.ET | http://creativecommons.org/licenses/by/4.0/ | BRC20 tokens are a type of non-fungible asset on the Bitcoin network. They allow users to embed customised content within Bitcoin's satoshis. The token frenzy reached a market size of US\$2.811\,b (2023Q3--2025Q1). However, this intuitive design has not undergone serious security scrutiny.
We present the first analysis of BRC20's \emph{transfer} mechanism and identify a new attack vector. A typical BRC20 transfer involves two "bundled" on-chain transactions with different fee levels: the first (i.e., \textbf{Tx1}) with a lower fee inscribes the \textsf{transfer} request, while the second (i.e., \textbf{Tx2}) with a higher fee finalizes the actual transfer. An adversary can send a manipulated fee transaction (falling between the two fee levels), which causes \textbf{Tx1} to be processed while \textbf{Tx2} is pinned in the mempool. This locks BRC20 liquidity and disrupts normal withdrawal requests from users. We term this the \emph{BRC20 pinning attack}.
We validated the attack in real-world settings in collaboration with Binance researchers. With their knowledge and permission, we conducted a controlled test against Binance's ORDI hot wallet, resulting in a temporary suspension of ORDI withdrawals for 3.5 hours. Recovery was performed shortly after. Further analysis confirms that the attack can be applied to over \textbf{90\%} of inscription-based tokens within the Bitcoin ecosystem. | [
{
"version": "v1",
"created": "Tue, 15 Oct 2024 05:33:16 GMT"
},
{
"version": "v2",
"created": "Sun, 20 Oct 2024 23:57:21 GMT"
},
{
"version": "v3",
"created": "Thu, 05 Jun 2025 16:45:17 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Qi",
"Minfeng",
""
],
[
"Wang",
"Qin",
""
],
[
"Wang",
"Zhipeng",
""
],
[
"Zhong",
"Lin",
""
],
[
"Gao",
"Zhixiong",
""
],
[
"Zhu",
"Tianqing",
""
],
[
"Chen",
"Shiping",
""
],
[
"Knottenbelt",
"William",
""
]
] |
2410.11802 | Zhe Li | Zhe Li, Xiangfei Qiu, Peng Chen, Yihang Wang, Hanyin Cheng, Yang Shu, Jilin Hu, Chenjuan Guo, Aoying Zhou, Christian S. Jensen, Bin Yang | TSFM-Bench: A Comprehensive and Unified Benchmarking of Foundation Models for Time Series Forecasting | null | null | null | null | cs.LG | http://creativecommons.org/licenses/by-nc-nd/4.0/ | Time Series Forecasting (TSF) is key functionality in numerous fields, such as financial investment, weather services, and energy management. Although increasingly capable TSF methods occur, many of them require domain-specific data collection and model training and do not generalize well when applied in other domains. Time Series Foundation Models (TSFMs) that are pre-trained on massive heterogeneous time series data aim to overcome these limitations. The prospects for generalizability have spurred the development of a new generation of TSFMs. This study proposes a benchmark, TSFM-Bench, to facilitate comprehensive and unified evaluation of TSFMs. TSFM-Bench covers a wide range of TSFMs, including those based on large language models and those pre-trained on time series data. TSFM-Bench supports multiple forecasting scenarios, including zero-shot, few-shot, and full-shot, enabling assessment across the full range of adaptation strategies. TSFM-Bench also provides a standardized experimental protocols for critical evaluation processes such as dataset splitting, loading, normalization, and few-shot sampling, facilitating consistency and fairness. We report on an extensive evaluation of TSFMs across a diverse range of datasets spanning multiple domains and exhibiting varied statistical characteristics. Specifically, we identify pros and cons and inherent limitations of existing TSFMs, and we propose potential directions for new model designs. | [
{
"version": "v1",
"created": "Tue, 15 Oct 2024 17:23:49 GMT"
},
{
"version": "v2",
"created": "Mon, 21 Oct 2024 07:11:07 GMT"
},
{
"version": "v3",
"created": "Fri, 01 Nov 2024 05:14:40 GMT"
},
{
"version": "v4",
"created": "Wed, 27 Nov 2024 00:58:59 GMT"
},
{
"version": "v5",
"created": "Thu, 05 Jun 2025 11:28:35 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Li",
"Zhe",
""
],
[
"Qiu",
"Xiangfei",
""
],
[
"Chen",
"Peng",
""
],
[
"Wang",
"Yihang",
""
],
[
"Cheng",
"Hanyin",
""
],
[
"Shu",
"Yang",
""
],
[
"Hu",
"Jilin",
""
],
[
"Guo",
"Chenjuan",
""
],
[
"Zhou",
"Aoying",
""
],
[
"Jensen",
"Christian S.",
""
],
[
"Yang",
"Bin",
""
]
] |
2410.12436 | Dr. Goutam Manna | Samit Ganguly, Narayan Banerjee, Abhijit Bhattacharyya and Goutam Manna | Particle production rate for a dynamical system using the path integral approach | 16 pages, 6 figures, Accepted from Physical Review D | Phys. Rev. D 111, 085014, (2025) | 10.1103/PhysRevD.111.085014 | null | gr-qc hep-th | http://creativecommons.org/licenses/by-sa/4.0/ | In this work, we investigate the particle creation rate in a dynamical (Vaidya) spacetime using Feynman's path integral formalism within the framework of the effective action approach. We examine three distinct cases involving the following mass functions, each representing dynamical geometries: (i) $m(v,r)=\mu v$, (ii) $m(v,r)=\mu v +\nu r$, and (iii) $m(v,r)=\mu v -\frac{\mu^2 v^2}{2r}$, where $\mu$ and $\nu$ are positive constants that satisfy all known energy conditions. We analyze particle production rates in the region of dynamical horizons, revealing an initial high rate followed by a rapid decline in all cases. Additionally, we explore the thermodynamic properties by calculating the surface gravity and corresponding Hayward-Kodama temperatures for each scenario. Graphical representations show the variation of surface gravity over time for the three cases, offering insights into the system's thermodynamic evolution. Our research investigates the connection between background geometry and the particle creation process, placing it within the broader context of quantum field theory in curved spacetime. The non-stationary nature of Vaidya geometry is highlighted as a valuable framework for examining the dynamic aspects of particle creation. This in-depth analysis enhances our understanding of quantum processes in curved spacetime and may offer insights relevant to thermodynamics and studies of gravitational collapse. | [
{
"version": "v1",
"created": "Wed, 16 Oct 2024 10:28:27 GMT"
},
{
"version": "v2",
"created": "Fri, 21 Mar 2025 04:49:05 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Ganguly",
"Samit",
""
],
[
"Banerjee",
"Narayan",
""
],
[
"Bhattacharyya",
"Abhijit",
""
],
[
"Manna",
"Goutam",
""
]
] |
2410.12656 | Mete Ismayilzada | Mete Ismayilzada, Defne Circi, Jonne S\"alev\"a, Hale Sirin, Abdullatif K\"oksal, Bhuwan Dhingra, Antoine Bosselut, Duygu Ataman, Lonneke van der Plas | Evaluating Morphological Compositional Generalization in Large Language Models | Accepted to NAACL 2025 | null | null | null | cs.CL cs.AI | http://creativecommons.org/licenses/by/4.0/ | Large language models (LLMs) have demonstrated significant progress in various natural language generation and understanding tasks. However, their linguistic generalization capabilities remain questionable, raising doubts about whether these models learn language similarly to humans. While humans exhibit compositional generalization and linguistic creativity in language use, the extent to which LLMs replicate these abilities, particularly in morphology, is under-explored. In this work, we systematically investigate the morphological generalization abilities of LLMs through the lens of compositionality. We define morphemes as compositional primitives and design a novel suite of generative and discriminative tasks to assess morphological productivity and systematicity. Focusing on agglutinative languages such as Turkish and Finnish, we evaluate several state-of-the-art instruction-finetuned multilingual models, including GPT-4 and Gemini. Our analysis shows that LLMs struggle with morphological compositional generalization particularly when applied to novel word roots, with performance declining sharply as morphological complexity increases. While models can identify individual morphological combinations better than chance, their performance lacks systematicity, leading to significant accuracy gaps compared to humans. | [
{
"version": "v1",
"created": "Wed, 16 Oct 2024 15:17:20 GMT"
},
{
"version": "v2",
"created": "Wed, 06 Nov 2024 14:14:58 GMT"
},
{
"version": "v3",
"created": "Sun, 09 Feb 2025 22:08:16 GMT"
},
{
"version": "v4",
"created": "Thu, 05 Jun 2025 10:08:39 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Ismayilzada",
"Mete",
""
],
[
"Circi",
"Defne",
""
],
[
"Sälevä",
"Jonne",
""
],
[
"Sirin",
"Hale",
""
],
[
"Köksal",
"Abdullatif",
""
],
[
"Dhingra",
"Bhuwan",
""
],
[
"Bosselut",
"Antoine",
""
],
[
"Ataman",
"Duygu",
""
],
[
"van der Plas",
"Lonneke",
""
]
] |
2410.12872 | JongWoo Kim | JongWoo Kim, SeongYeub Chu, Bryan Wong, Mun Yi | Not All Options Are Created Equal: Textual Option Weighting for Token-Efficient LLM-Based Knowledge Tracing | 11 pages | null | null | null | cs.CL cs.AI cs.CY cs.LG | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Large Language Models (LLMs) have recently emerged as promising tools for knowledge tracing (KT) due to their strong reasoning and generalization abilities. While recent LLM-based KT methods have proposed new prompt formats, they struggle to represent the full interaction histories of example learners within a single prompt during in-context learning (ICL), resulting in limited scalability and high computational cost under token constraints. In this work, we present \textit{LLM-based Option-weighted Knowledge Tracing (LOKT)}, a simple yet effective framework that encodes the interaction histories of example learners in context as \textit{textual categorical option weights (TCOW)}. TCOW are semantic labels (e.g., ``inadequate'') assigned to the options selected by learners when answering questions, enhancing the interpretability of LLMs. Experiments on multiple-choice datasets show that LOKT outperforms existing non-LLM and LLM-based KT models in both cold-start and warm-start settings. Moreover, LOKT enables scalable and cost-efficient inference, achieving strong performance even under strict token constraints. Our code is available at \href{https://anonymous.4open.science/r/LOKT_model-3233}{https://anonymous.4open.science/r/LOKT\_model-3233}. | [
{
"version": "v1",
"created": "Mon, 14 Oct 2024 16:25:48 GMT"
},
{
"version": "v2",
"created": "Thu, 05 Jun 2025 05:41:21 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Kim",
"JongWoo",
""
],
[
"Chu",
"SeongYeub",
""
],
[
"Wong",
"Bryan",
""
],
[
"Yi",
"Mun",
""
]
] |
2410.13135 | Shun Hashiyada | Shun Hashiyada and Yoshito Y. Tanaka | Conservation law for angular momentum based on optical field derivatives: Analysis of optical spin-orbit conversion | 17 pages, 2 figures | null | 10.1103/PhysRevResearch.7.L022052 | null | physics.optics | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We present a theoretical framework for analyzing the loss of optical angular momentum (AM), including spin (SAM) and orbital (OAM) components, in light-matter interactions. Conventional SAM and OAM conservation laws rely on transverse field components, neglecting longitudinal fields and limiting applicability to vacuum. Our approach defines optical AM using time derivatives of the electric and magnetic fields, yielding a gauge-invariant formulation that includes both transverse and longitudinal components and explicitly incorporates charge and current densities. This enables a more complete description of AM dissipation in materials. We apply this framework to analyze spin-orbit conversion (SOC) in two scenarios: scattering of circularly polarized (CP) beams by a gold nanoparticle and focusing of CP and linearly polarized optical vortex beams by a lens. The results show that SOC depends on particle size and polarization, with notable OAM loss in larger particles and CP beam focusing. This framework enables the evaluation of previously overlooked SAM and OAM losses, providing a powerful tool for studying systems in which the analysis of AM losses is intrinsically important, such as chiral materials, as well as for designing photonic devices and exploring light-matter interactions at the nanoscale. | [
{
"version": "v1",
"created": "Thu, 17 Oct 2024 01:44:17 GMT"
},
{
"version": "v2",
"created": "Fri, 18 Oct 2024 06:07:01 GMT"
},
{
"version": "v3",
"created": "Thu, 10 Apr 2025 02:51:29 GMT"
},
{
"version": "v4",
"created": "Mon, 12 May 2025 01:09:52 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Hashiyada",
"Shun",
""
],
[
"Tanaka",
"Yoshito Y.",
""
]
] |
2410.13577 | Benjamin Leblanc | Benjamin Leblanc, Mathieu Bazinet, Nathaniel D'Amours, Alexandre Drouin and Pascal Germain | Generalization Bounds via Meta-Learned Model Representations: PAC-Bayes and Sample Compression Hypernetworks | Accepted at ICML 2025 (also at the NeurIPS 2024 Workshop on Compression in Machine Learning) | null | null | null | cs.LG | http://creativecommons.org/licenses/by/4.0/ | Both PAC-Bayesian and Sample Compress learning frameworks are instrumental for deriving tight (non-vacuous) generalization bounds for neural networks. We leverage these results in a meta-learning scheme, relying on a hypernetwork that outputs the parameters of a downstream predictor from a dataset input. The originality of our approach lies in the investigated hypernetwork architectures that encode the dataset before decoding the parameters: (1) a PAC-Bayesian encoder that expresses a posterior distribution over a latent space, (2) a Sample Compress encoder that selects a small sample of the dataset input along with a message from a discrete set, and (3) a hybrid between both approaches motivated by a new Sample Compress theorem handling continuous messages. The latter theorem exploits the pivotal information transiting at the encoder-decoder junction in order to compute generalization guarantees for each downstream predictor obtained by our meta-learning scheme. | [
{
"version": "v1",
"created": "Thu, 17 Oct 2024 14:12:35 GMT"
},
{
"version": "v2",
"created": "Fri, 21 Feb 2025 14:23:17 GMT"
},
{
"version": "v3",
"created": "Thu, 05 Jun 2025 15:32:11 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Leblanc",
"Benjamin",
""
],
[
"Bazinet",
"Mathieu",
""
],
[
"D'Amours",
"Nathaniel",
""
],
[
"Drouin",
"Alexandre",
""
],
[
"Germain",
"Pascal",
""
]
] |
2410.13634 | Victor Alexandrov | Victor Alexandrov | Additional first order equation for infinitesimal bendings of smooth surfaces in the isothermal coordinates | 11 pages; this v3 is an English translation of v2 which was in Russian | Sib. Math. J. 66, No. 3, 618-628 (2025) | 10.1134/S0037446625030024 | null | math.DG | http://creativecommons.org/licenses/by/4.0/ | The article contributes to the theory of infinitesimal bendings of smooth surfaces in Euclidean 3-space. We derive a linear differential equation of the first order, which previously did not appear in the literature and which is satisfied by any Darboux rotation field of a smooth surface. We show that, for some surfaces, this additional equation is functionally independent of the three standard equations that the Darboux rotation field satisfies (and by which it is determined). As a consequence of this additional equation, we prove the maximum principle for the components of the Darboux rotation field for a class of disk-homeomorphic surfaces containing not only surfaces of positive Gaussian curvature. | [
{
"version": "v1",
"created": "Thu, 17 Oct 2024 15:05:35 GMT"
},
{
"version": "v2",
"created": "Fri, 28 Mar 2025 05:11:49 GMT"
},
{
"version": "v3",
"created": "Thu, 05 Jun 2025 14:50:57 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Alexandrov",
"Victor",
""
]
] |
2410.13831 | Kajetan Schweighofer | Kajetan Schweighofer, Adrian Arnaiz-Rodriguez, Sepp Hochreiter, Nuria Oliver | The Disparate Benefits of Deep Ensembles | ICML 2025 | null | null | null | cs.LG cs.AI | http://creativecommons.org/licenses/by/4.0/ | Ensembles of Deep Neural Networks, Deep Ensembles, are widely used as a simple way to boost predictive performance. However, their impact on algorithmic fairness is not well understood yet. Algorithmic fairness examines how a model's performance varies across socially relevant groups defined by protected attributes such as age, gender, or race. In this work, we explore the interplay between the performance gains from Deep Ensembles and fairness. Our analysis reveals that they unevenly favor different groups, a phenomenon that we term the disparate benefits effect. We empirically investigate this effect using popular facial analysis and medical imaging datasets with protected group attributes and find that it affects multiple established group fairness metrics, including statistical parity and equal opportunity. Furthermore, we identify that the per-group differences in predictive diversity of ensemble members can explain this effect. Finally, we demonstrate that the classical Hardt post-processing method is particularly effective at mitigating the disparate benefits effect of Deep Ensembles by leveraging their better-calibrated predictive distributions. | [
{
"version": "v1",
"created": "Thu, 17 Oct 2024 17:53:01 GMT"
},
{
"version": "v2",
"created": "Wed, 04 Jun 2025 18:48:31 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Schweighofer",
"Kajetan",
""
],
[
"Arnaiz-Rodriguez",
"Adrian",
""
],
[
"Hochreiter",
"Sepp",
""
],
[
"Oliver",
"Nuria",
""
]
] |
2410.14812 | Victoria Lin | Victoria Lin, Louis-Philippe Morency, Eli Ben-Michael | Isolated Causal Effects of Natural Language | ICML 2025 | null | null | null | cs.CL stat.ME | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | As language technologies become widespread, it is important to understand how changes in language affect reader perceptions and behaviors. These relationships may be formalized as the isolated causal effect of some focal language-encoded intervention (e.g., factual inaccuracies) on an external outcome (e.g., readers' beliefs). In this paper, we introduce a formal estimation framework for isolated causal effects of language. We show that a core challenge of estimating isolated effects is the need to approximate all non-focal language outside of the intervention. Drawing on the principle of omitted variable bias, we provide measures for evaluating the quality of both non-focal language approximations and isolated effect estimates themselves. We find that poor approximation of non-focal language can lead to bias in the corresponding isolated effect estimates due to omission of relevant variables, and we show how to assess the sensitivity of effect estimates to such bias along the two key axes of fidelity and overlap. In experiments on semi-synthetic and real-world data, we validate the ability of our framework to correctly recover isolated effects and demonstrate the utility of our proposed measures. | [
{
"version": "v1",
"created": "Fri, 18 Oct 2024 18:32:38 GMT"
},
{
"version": "v2",
"created": "Fri, 09 May 2025 23:26:32 GMT"
},
{
"version": "v3",
"created": "Wed, 04 Jun 2025 18:20:45 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Lin",
"Victoria",
""
],
[
"Morency",
"Louis-Philippe",
""
],
[
"Ben-Michael",
"Eli",
""
]
] |
2410.14899 | Zhongze Cai | Zhongze Cai, Hansheng Jiang, Xiaocheng Li | Out-of-distribution Robust Optimization | null | null | null | null | math.OC | http://creativecommons.org/licenses/by/4.0/ | In this paper, we consider the contextual robust optimization problem under an out-of-distribution setting. The contextual robust optimization problem considers a risk-sensitive objective function for an optimization problem with the presence of a context vector (also known as covariates or side information) capturing related information. While the existing works mainly consider the in-distribution setting, and the resultant robustness achieved is in an out-of-sample sense, our paper studies an out-of-distribution setting where there can be a difference between the test environment and the training environment where the data are collected. We propose methods that handle this out-of-distribution setting, and the key relies on a density ratio estimation for the distribution shift. We show that additional structures such as covariate shift and label shift are not only helpful in defending distribution shift but also necessary in avoiding non-trivial solutions compared to other principled methods such as distributionally robust optimization. We also illustrate how the covariates can be useful in this procedure. Numerical experiments generate more intuitions and demonstrate that the proposed methods can help avoid over-conservative solutions. | [
{
"version": "v1",
"created": "Fri, 18 Oct 2024 22:59:01 GMT"
},
{
"version": "v2",
"created": "Tue, 22 Oct 2024 18:37:40 GMT"
},
{
"version": "v3",
"created": "Thu, 05 Jun 2025 09:44:53 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Cai",
"Zhongze",
""
],
[
"Jiang",
"Hansheng",
""
],
[
"Li",
"Xiaocheng",
""
]
] |
2410.17940 | Xu-Yang Hou | Jia-Chen Tang, Xu-Yang Hou, and Hao Guo | Geometry effect of the dynamical quantum phase transitions at finite temperatures | 14 pages, 5 figures | Phys. Rev. B 111, 174310(2025) | 10.1103/PhysRevB.111.174310 | null | quant-ph cond-mat.other | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Dynamical quantum phase transitions (DQPTs) probe the nonequilibrium evolution of quantum systems, unveiling their geometric and topological characteristics. In this study, we introduce the concepts of parallel quench and dynamic geometrical order parameter (DGOP) for non-band models, where these quantities capture the geometric shifts associated with DQPTs. At zero temperature, the DGOP corresponds to the Pancharatnam geometric phase, while at finite temperatures, it extends to the interferometric geometric phase. We further generalize the dynamic topological order parameter (DTOP) to finite-temperature band models, examining its behavior in the Su-Schrieffer-Heeger (SSH) model. Our analysis shows that thermal fluctuations and boundary effects at finite temperatures disrupt the quantization of the DTOP, yet it retains signatures of topological transitions. These findings deepen the understanding of geometric and topological properties in quantum dynamics, illuminating DQPTs across both non-band and band frameworks. | [
{
"version": "v1",
"created": "Wed, 23 Oct 2024 15:07:08 GMT"
},
{
"version": "v2",
"created": "Thu, 27 Mar 2025 16:19:50 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Tang",
"Jia-Chen",
""
],
[
"Hou",
"Xu-Yang",
""
],
[
"Guo",
"Hao",
""
]
] |
2410.18067 | Valeria Ruscio | Valeria Ruscio, Umberto Nanni, Fabrizio Silvestri | Beyond Position: the emergence of wavelet-like properties in Transformers | null | null | null | null | cs.LG cs.AI | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | This paper studies how Transformer models with Rotary Position Embeddings (RoPE) develop emergent, wavelet-like properties that compensate for the positional encoding's theoretical limitations. Through an analysis spanning model scales, architectures, and training checkpoints, we show that attention heads evolve to implement multi-resolution processing analogous to wavelet transforms. We demonstrate that this scale-invariant behavior is unique to RoPE, emerges through distinct evolutionary phases during training, and statistically adheres to the fundamental uncertainty principle. Our findings suggest that the effectiveness of modern Transformers stems from their remarkable ability to spontaneously develop optimal, multi-resolution decompositions to address inherent architectural constraints. | [
{
"version": "v1",
"created": "Wed, 23 Oct 2024 17:48:28 GMT"
},
{
"version": "v2",
"created": "Fri, 25 Oct 2024 20:27:05 GMT"
},
{
"version": "v3",
"created": "Tue, 21 Jan 2025 17:50:47 GMT"
},
{
"version": "v4",
"created": "Wed, 04 Jun 2025 21:06:45 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Ruscio",
"Valeria",
""
],
[
"Nanni",
"Umberto",
""
],
[
"Silvestri",
"Fabrizio",
""
]
] |
2410.18881 | Weijian Luo | Weijian Luo | Diff-Instruct++: Training One-step Text-to-image Generator Model to Align with Human Preferences | Revision: The paper was accepted by Transactions of Machine Learning Research (TMLR) | null | null | null | cs.CV cs.AI cs.LG | http://creativecommons.org/licenses/by/4.0/ | One-step text-to-image generator models offer advantages such as swift inference efficiency, flexible architectures, and state-of-the-art generation performance. In this paper, we study the problem of aligning one-step generator models with human preferences for the first time. Inspired by the success of reinforcement learning using human feedback (RLHF), we formulate the alignment problem as maximizing expected human reward functions while adding an Integral Kullback-Leibler divergence term to prevent the generator from diverging. By overcoming technical challenges, we introduce Diff-Instruct++ (DI++), the first, fast-converging and image data-free human preference alignment method for one-step text-to-image generators. We also introduce novel theoretical insights, showing that using CFG for diffusion distillation is secretly doing RLHF with DI++. Such an interesting finding brings understanding and potential contributions to future research involving CFG. In the experiment sections, we align both UNet-based and DiT-based one-step generators using DI++, which use the Stable Diffusion 1.5 and the PixelArt-$\alpha$ as the reference diffusion processes. The resulting DiT-based one-step text-to-image model achieves a strong Aesthetic Score of 6.19 and an Image Reward of 1.24 on the COCO validation prompt dataset. It also achieves a leading Human preference Score (HPSv2.0) of 28.48, outperforming other open-sourced models such as Stable Diffusion XL, DMD2, SD-Turbo, as well as PixelArt-$\alpha$. Both theoretical contributions and empirical evidence indicate that DI++ is a strong human-preference alignment approach for one-step text-to-image models. The homepage of the paper is https://github.com/pkulwj1994/diff_instruct_pp. | [
{
"version": "v1",
"created": "Thu, 24 Oct 2024 16:17:18 GMT"
},
{
"version": "v2",
"created": "Thu, 05 Jun 2025 12:44:07 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Luo",
"Weijian",
""
]
] |
2410.18959 | Arjun Ashok | Andrew Robert Williams, Arjun Ashok, \'Etienne Marcotte, Valentina Zantedeschi, Jithendaraa Subramanian, Roland Riachi, James Requeima, Alexandre Lacoste, Irina Rish, Nicolas Chapados, Alexandre Drouin | Context is Key: A Benchmark for Forecasting with Essential Textual Information | ICML 2025. First two authors contributed equally | null | null | null | cs.LG cs.AI stat.ML | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Forecasting is a critical task in decision-making across numerous domains. While historical numerical data provide a start, they fail to convey the complete context for reliable and accurate predictions. Human forecasters frequently rely on additional information, such as background knowledge and constraints, which can efficiently be communicated through natural language. However, in spite of recent progress with LLM-based forecasters, their ability to effectively integrate this textual information remains an open question. To address this, we introduce "Context is Key" (CiK), a time-series forecasting benchmark that pairs numerical data with diverse types of carefully crafted textual context, requiring models to integrate both modalities; crucially, every task in CiK requires understanding textual context to be solved successfully. We evaluate a range of approaches, including statistical models, time series foundation models, and LLM-based forecasters, and propose a simple yet effective LLM prompting method that outperforms all other tested methods on our benchmark. Our experiments highlight the importance of incorporating contextual information, demonstrate surprising performance when using LLM-based forecasting models, and also reveal some of their critical shortcomings. This benchmark aims to advance multimodal forecasting by promoting models that are both accurate and accessible to decision-makers with varied technical expertise. The benchmark can be visualized at https://servicenow.github.io/context-is-key-forecasting/v0/. | [
{
"version": "v1",
"created": "Thu, 24 Oct 2024 17:56:08 GMT"
},
{
"version": "v2",
"created": "Tue, 04 Feb 2025 23:12:13 GMT"
},
{
"version": "v3",
"created": "Thu, 06 Feb 2025 19:05:41 GMT"
},
{
"version": "v4",
"created": "Thu, 05 Jun 2025 17:25:49 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Williams",
"Andrew Robert",
""
],
[
"Ashok",
"Arjun",
""
],
[
"Marcotte",
"Étienne",
""
],
[
"Zantedeschi",
"Valentina",
""
],
[
"Subramanian",
"Jithendaraa",
""
],
[
"Riachi",
"Roland",
""
],
[
"Requeima",
"James",
""
],
[
"Lacoste",
"Alexandre",
""
],
[
"Rish",
"Irina",
""
],
[
"Chapados",
"Nicolas",
""
],
[
"Drouin",
"Alexandre",
""
]
] |
2410.20898 | Weijian Luo | Weijian Luo and Colin Zhang and Debing Zhang and Zhengyang Geng | David and Goliath: Small One-step Model Beats Large Diffusion with Score Post-training | Revision: paper accepted by the ICML2025 main conference | null | null | null | cs.CV cs.AI cs.LG cs.MM | http://creativecommons.org/licenses/by/4.0/ | We propose Diff-Instruct* (DI*), a data-efficient post-training approach for one-step text-to-image generative models to improve its human preferences without requiring image data. Our method frames alignment as online reinforcement learning from human feedback (RLHF), which optimizes the one-step model to maximize human reward functions while being regularized to be kept close to a reference diffusion process. Unlike traditional RLHF approaches, which rely on the Kullback-Leibler divergence as the regularization, we introduce a novel general score-based divergence regularization that substantially improves performance as well as post-training stability. Although the general score-based RLHF objective is intractable to optimize, we derive a strictly equivalent tractable loss function in theory that can efficiently compute its \emph{gradient} for optimizations. We introduce \emph{DI*-SDXL-1step}, which is a 2.6B one-step text-to-image model at a resolution of $1024\times 1024$, post-trained from DMD2 w.r.t SDXL. \textbf{Our 2.6B \emph{DI*-SDXL-1step} model outperforms the 50-step 12B FLUX-dev model} in ImageReward, PickScore, and CLIP score on the Parti prompts benchmark while using only 1.88\% of the inference time. This result clearly shows that with proper post-training, the small one-step model is capable of beating huge multi-step diffusion models. Our model is open-sourced at this link: https://github.com/pkulwj1994/diff_instruct_star. We hope our findings can contribute to human-centric machine learning techniques. | [
{
"version": "v1",
"created": "Mon, 28 Oct 2024 10:26:19 GMT"
},
{
"version": "v2",
"created": "Tue, 24 Dec 2024 05:22:40 GMT"
},
{
"version": "v3",
"created": "Thu, 05 Jun 2025 12:11:49 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Luo",
"Weijian",
""
],
[
"Zhang",
"Colin",
""
],
[
"Zhang",
"Debing",
""
],
[
"Geng",
"Zhengyang",
""
]
] |
2410.21862 | Massimo Bilancia | Massimo Bilancia and Samuele Magro | Hierarchical mixtures of Unigram models for short text clustering: The role of Beta-Liouville priors | 28 pages, 6 figures. Submitted | Annals of Operation Research, 2025 | 10.1007/s10479-025-06693-7 | null | stat.ML cs.LG stat.CO | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | This paper presents a variant of the Multinomial mixture model tailored to the unsupervised classification of short text data. While the Multinomial probability vector is traditionally assigned a Dirichlet prior distribution, this work explores an alternative formulation based on the Beta-Liouville distribution, which offers a more flexible correlation structure than the Dirichlet. We examine the theoretical properties of the Beta-Liouville distribution, with particular focus on its conjugacy with the Multinomial likelihood. This property enables the derivation of update equations for a CAVI (Coordinate Ascent Variational Inference) algorithm, facilitating approximate posterior inference of the model parameters. In addition, we introduce a stochastic variant of the CAVI algorithm to enhance scalability. The paper concludes with empirical examples demonstrating effective strategies for selecting the Beta-Liouville hyperparameters. | [
{
"version": "v1",
"created": "Tue, 29 Oct 2024 08:56:29 GMT"
},
{
"version": "v2",
"created": "Thu, 14 Nov 2024 09:17:31 GMT"
},
{
"version": "v3",
"created": "Mon, 28 Apr 2025 10:27:31 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Bilancia",
"Massimo",
""
],
[
"Magro",
"Samuele",
""
]
] |
2410.22038 | Thomas Ransford | Ricardo Fraiman, Leonardo Moreno, Thomas Ransford | A Cram\'er-Wold theorem for mixtures | 12 pages | null | null | null | math.PR | http://creativecommons.org/licenses/by/4.0/ | We show how a Cram\'er-Wold theorem for a family of multivariate probability distributions can be used to generate a similar theorem for mixtures (convex combinations) of distributions drawn from the same family.
Using this abstract result, we establish a Cram\'er-Wold theorem for mixtures of multivariate Gaussian distributions. According to this theorem, two such mixtures can be distinguished by projecting them onto a certain predetermined finite set of lines, the number of lines depending only on the total number Gaussian distributions involved and on the ambient dimension. A similar result is also obtained for mixtures of multivariate $t$-distributions. | [
{
"version": "v1",
"created": "Tue, 29 Oct 2024 13:45:11 GMT"
},
{
"version": "v2",
"created": "Thu, 05 Jun 2025 15:59:00 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Fraiman",
"Ricardo",
""
],
[
"Moreno",
"Leonardo",
""
],
[
"Ransford",
"Thomas",
""
]
] |
2410.22391 | Thomas Schmied | Thomas Schmied, Thomas Adler, Vihang Patil, Maximilian Beck, Korbinian P\"oppel, Johannes Brandstetter, G\"unter Klambauer, Razvan Pascanu, Sepp Hochreiter | A Large Recurrent Action Model: xLSTM enables Fast Inference for Robotics Tasks | null | null | null | null | cs.LG cs.AI | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | In recent years, there has been a trend in the field of Reinforcement Learning (RL) towards large action models trained offline on large-scale datasets via sequence modeling. Existing models are primarily based on the Transformer architecture, which result in powerful agents. However, due to slow inference times, Transformer-based approaches are impractical for real-time applications, such as robotics. Recently, modern recurrent architectures, such as xLSTM and Mamba, have been proposed that exhibit parallelization benefits during training similar to the Transformer architecture while offering fast inference. In this work, we study the aptitude of these modern recurrent architectures for large action models. Consequently, we propose a Large Recurrent Action Model (LRAM) with an xLSTM at its core that comes with linear-time inference complexity and natural sequence length extrapolation abilities. Experiments on 432 tasks from 6 domains show that LRAM compares favorably to Transformers in terms of performance and speed. | [
{
"version": "v1",
"created": "Tue, 29 Oct 2024 17:55:47 GMT"
},
{
"version": "v2",
"created": "Thu, 20 Feb 2025 09:29:58 GMT"
},
{
"version": "v3",
"created": "Wed, 04 Jun 2025 17:58:20 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Schmied",
"Thomas",
""
],
[
"Adler",
"Thomas",
""
],
[
"Patil",
"Vihang",
""
],
[
"Beck",
"Maximilian",
""
],
[
"Pöppel",
"Korbinian",
""
],
[
"Brandstetter",
"Johannes",
""
],
[
"Klambauer",
"Günter",
""
],
[
"Pascanu",
"Razvan",
""
],
[
"Hochreiter",
"Sepp",
""
]
] |
2410.22882 | Adrian Signer | Riccardo Aliberti, Paolo Beltrame, Ettore Budassi, Carlo M. Carloni Calame, Gilberto Colangelo, Lorenzo Cotrozzi, Achim Denig, Anna Driutti, Tim Engel, Lois Flower, Andrea Gurgone, Martin Hoferichter, Fedor Ignatov, Sophie Kollatzsch, Bastian Kubis, Andrzej Kup\'s\'c, Fabian Lange, Alberto Lusiani, Stefan E. M\"uller, J\'er\'emy Paltrinieri, Pau Petit Ros\`as, Fulvio Piccinini, Alan Price, Lorenzo Punzi, Marco Rocco, Olga Shekhovtsova, Andrzej Si\'odmok, Adrian Signer, Giovanni Stagnitto, Peter Stoffer, Thomas Teubner, William J. Torres Bobadilla, Francesco P. Ucci, Yannick Ulrich, Graziano Venanzoni | Radiative corrections and Monte Carlo tools for low-energy hadronic cross sections in $e^+ e^-$ collisions | RadioMonteCarLow 2 Working Group report Phase I, 67 pages, 34 figures; version accepted for publication in SciPost Physics Community Reports | null | null | null | hep-ph hep-ex | http://creativecommons.org/licenses/by/4.0/ | We present the results of Phase I of an ongoing review of Monte Carlo tools relevant for low-energy hadronic cross sections. This includes a detailed comparison of Monte Carlo codes for electron-positron scattering into a muon pair, pion pair, and electron pair, for scan and radiative-return experiments. After discussing the various approaches that are used and effects that are included, we show differential cross sections obtained with AfkQed, BabaYaga@NLO, KKMC, MCGPJ, McMule, Phokhara, and Sherpa, for scenarios that are inspired by experiments providing input for the dispersive evaluation of the hadronic vacuum polarisation. | [
{
"version": "v1",
"created": "Wed, 30 Oct 2024 10:23:29 GMT"
},
{
"version": "v2",
"created": "Wed, 04 Dec 2024 11:07:25 GMT"
},
{
"version": "v3",
"created": "Thu, 05 Jun 2025 09:42:39 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Aliberti",
"Riccardo",
""
],
[
"Beltrame",
"Paolo",
""
],
[
"Budassi",
"Ettore",
""
],
[
"Calame",
"Carlo M. Carloni",
""
],
[
"Colangelo",
"Gilberto",
""
],
[
"Cotrozzi",
"Lorenzo",
""
],
[
"Denig",
"Achim",
""
],
[
"Driutti",
"Anna",
""
],
[
"Engel",
"Tim",
""
],
[
"Flower",
"Lois",
""
],
[
"Gurgone",
"Andrea",
""
],
[
"Hoferichter",
"Martin",
""
],
[
"Ignatov",
"Fedor",
""
],
[
"Kollatzsch",
"Sophie",
""
],
[
"Kubis",
"Bastian",
""
],
[
"Kupść",
"Andrzej",
""
],
[
"Lange",
"Fabian",
""
],
[
"Lusiani",
"Alberto",
""
],
[
"Müller",
"Stefan E.",
""
],
[
"Paltrinieri",
"Jérémy",
""
],
[
"Rosàs",
"Pau Petit",
""
],
[
"Piccinini",
"Fulvio",
""
],
[
"Price",
"Alan",
""
],
[
"Punzi",
"Lorenzo",
""
],
[
"Rocco",
"Marco",
""
],
[
"Shekhovtsova",
"Olga",
""
],
[
"Siódmok",
"Andrzej",
""
],
[
"Signer",
"Adrian",
""
],
[
"Stagnitto",
"Giovanni",
""
],
[
"Stoffer",
"Peter",
""
],
[
"Teubner",
"Thomas",
""
],
[
"Bobadilla",
"William J. Torres",
""
],
[
"Ucci",
"Francesco P.",
""
],
[
"Ulrich",
"Yannick",
""
],
[
"Venanzoni",
"Graziano",
""
]
] |
2410.22944 | Tom Lamb | Tom A. Lamb, Adam Davies, Alasdair Paren, Philip H.S. Torr and Francesco Pinto | Focus On This, Not That! Steering LLMs with Adaptive Feature Specification | 36pages, 19 figures | null | null | null | cs.LG cs.AI cs.CL | http://creativecommons.org/licenses/by/4.0/ | Despite the success of Instruction Tuning (IT) in training large language models (LLMs), such models often leverage spurious or biased features learnt from their training data and can become misaligned, leading to undesired behaviours. While existing techniques can steer model behaviour at inference-time, they are often post-hoc and do not embed steering as an intrinsic model feature. In this work, we introduce Focus Instruction Tuning (FIT), which trains LLMs to condition their responses by focusing on specific features whilst ignoring others, leading to different behaviours based on what features are specified. Across diverse benchmarks, we demonstrate that FIT: (i) successfully steers behaviour at inference time; (ii) increases robustness by amplifying core task signals and down-weighting spurious cues; (iii) mitigates social bias by suppressing demographic attributes; and (iv) generalises under distribution shifts and to previously unseen focus features. FIT therefore offers a lightweight, intrinsic mechanism for building more robust, fair, and easily controllable LLMs. | [
{
"version": "v1",
"created": "Wed, 30 Oct 2024 12:01:48 GMT"
},
{
"version": "v2",
"created": "Thu, 16 Jan 2025 11:26:02 GMT"
},
{
"version": "v3",
"created": "Mon, 10 Feb 2025 23:03:19 GMT"
},
{
"version": "v4",
"created": "Thu, 05 Jun 2025 08:25:12 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Lamb",
"Tom A.",
""
],
[
"Davies",
"Adam",
""
],
[
"Paren",
"Alasdair",
""
],
[
"Torr",
"Philip H. S.",
""
],
[
"Pinto",
"Francesco",
""
]
] |
2410.23121 | Pauli Virtanen | P. Virtanen, R. P. S. Penttil\"a, P. T\"orm\"a, A. D\'iez-Carl\'on, D. K. Efetov, T. T. Heikkil\"a | Superconducting junctions with flat bands | 17 pages | null | null | null | cond-mat.supr-con | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We analyze the properties of flat-band superconductor junctions that behave differently from ordinary junctions containing only metals with Fermi surfaces. In particular, we show how in the tunneling limit the critical Josephson current between flat-band superconductors is inversely proportional to the pair potential, how the quantum geometric contribution to the supercurrent contributes even in the normal state of a flat-band weak link, and how Andreev reflection is strongly affected by the presence of bound states. Our results are relevant for analyzing the superconducting properties of junctions involving electronic systems with flat bands. | [
{
"version": "v1",
"created": "Wed, 30 Oct 2024 15:29:59 GMT"
},
{
"version": "v2",
"created": "Thu, 05 Jun 2025 07:46:15 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Virtanen",
"P.",
""
],
[
"Penttilä",
"R. P. S.",
""
],
[
"Törmä",
"P.",
""
],
[
"Díez-Carlón",
"A.",
""
],
[
"Efetov",
"D. K.",
""
],
[
"Heikkilä",
"T. T.",
""
]
] |
2410.23158 | Oliver Urs Lenz | Oliver Urs Lenz, Matthijs van Leeuwen | Monotonic anomaly detection | null | null | null | null | cs.LG | http://creativecommons.org/licenses/by-nc-nd/4.0/ | Semi-supervised anomaly detection is based on the principle that potential anomalies are those records that look different from normal training data. However, in some cases we are specifically interested in anomalies that correspond to high attribute values (or low, but not both). We present two asymmetrical distance measures that take this monotonicity into account: ramp distance and signed distance. Through experiments on synthetic and real-life datasets, we show that ramp distance increases anomaly detection performance over the traditional absolute distance. While signed distance also performs well on synthetic data, it performs substantially poorer on real-life datasets. We argue that this is a consequence of the fact that when using signed distance, low values of certain attributes automatically compensate for high values of other attributes, such that anomaly detection is reduced to counting the total attribute value sum, which is too simplistic in practice. | [
{
"version": "v1",
"created": "Wed, 30 Oct 2024 16:11:40 GMT"
},
{
"version": "v2",
"created": "Thu, 05 Jun 2025 13:14:42 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Lenz",
"Oliver Urs",
""
],
[
"van Leeuwen",
"Matthijs",
""
]
] |
2410.23518 | Helio Huet | H. Huet, P. R. Ramesh, S. C. Wein, N. Coste, P. Hilaire, N. Somaschi, M. Morassi, A. Lema\^itre, I. Sagnes, M. F. Doty, O. Krebs, L. Lanco, D. A. Fioretto and P. Senellart | Deterministic and reconfigurable graph state generation with a single solid-state quantum emitter | null | Nat Commun 16, 4337 (2025) | 10.1038/s41467-025-59693-3 | null | quant-ph | http://creativecommons.org/licenses/by/4.0/ | Measurement-based quantum computing offers a promising route towards scalable, universal photonic quantum computation. This approach relies on the deterministic and efficient generation of photonic graph states in which many photons are mutually entangled with various topologies. Recently, deterministic sources of graph states have been demonstrated with quantum emitters in both the optical and microwave domains. In this work, we demonstrate deterministic and reconfigurable graph state generation with optical solid-state integrated quantum emitters. Specifically, we use a single semiconductor quantum dot in a cavity to generate caterpillar graph states, the most general type of graph state that can be produced with a single emitter. By using fast detuned optical pulses, we achieve full control over the spin state, enabling us to vary the entanglement topology at will. We perform quantum state tomography of two successive photons, measuring Bell state fidelities up to 0.80$\pm$0.04 and concurrences up to 0.69$\pm$0.09, while maintaining high photon indistinguishability. This simple optical scheme, compatible with commercially available quantum dot-based single photon sources, brings us a step closer to fault-tolerant quantum computing with spins and photons. | [
{
"version": "v1",
"created": "Wed, 30 Oct 2024 23:59:54 GMT"
},
{
"version": "v2",
"created": "Sat, 19 Apr 2025 15:16:30 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Huet",
"H.",
""
],
[
"Ramesh",
"P. R.",
""
],
[
"Wein",
"S. C.",
""
],
[
"Coste",
"N.",
""
],
[
"Hilaire",
"P.",
""
],
[
"Somaschi",
"N.",
""
],
[
"Morassi",
"M.",
""
],
[
"Lemaître",
"A.",
""
],
[
"Sagnes",
"I.",
""
],
[
"Doty",
"M. F.",
""
],
[
"Krebs",
"O.",
""
],
[
"Lanco",
"L.",
""
],
[
"Fioretto",
"D. A.",
""
],
[
"Senellart",
"P.",
""
]
] |
2410.23743 | Ming Li | Ming Li, Yanhong Li, Tianyi Zhou | What Happened in LLMs Layers when Trained for Fast vs. Slow Thinking: A Gradient Perspective | ACL2025 main, Camera-ready | null | null | null | cs.CL cs.AI cs.LG | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | What makes a difference in the post-training of LLMs? We investigate the training patterns of different layers in large language models (LLMs) through the lens of the gradient. We are specifically interested in how fast vs. slow thinking affects the layer-wise gradients, given the recent popularity of training LLMs on reasoning paths such as chain-of-thoughts (CoT) and process rewards. In our study, fast thinking without CoT leads to larger gradients and larger differences of gradients across layers than slow thinking (Detailed CoT), indicating the learning stability brought by the latter. Additionally, we study whether the gradient patterns can reflect the correctness of responses when training different LLMs using slow vs. fast thinking paths. The results show that the gradients of slow thinking can distinguish correct and irrelevant reasoning paths. As a comparison, we conduct similar gradient analyses on non-reasoning knowledge learning tasks, on which, however, trivially increasing the response length does not lead to similar behaviors of slow thinking. Our study strengthens fundamental understandings of LLM training and sheds novel insights on its efficiency and stability, which pave the way towards building a generalizable System-2 agent. Our code, data, and gradient statistics can be found in: https://github.com/MingLiiii/Layer_Gradient. | [
{
"version": "v1",
"created": "Thu, 31 Oct 2024 08:58:06 GMT"
},
{
"version": "v2",
"created": "Thu, 05 Jun 2025 02:01:49 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Li",
"Ming",
""
],
[
"Li",
"Yanhong",
""
],
[
"Zhou",
"Tianyi",
""
]
] |
2410.23818 | Thomas Bland | A. Chaika, A. O. Oliinyk, I. V. Yatsuta, N. P. Proukakis, M. Edwards, A. I. Yakimenko, T. Bland | Acceleration-induced transport of quantum vortices in joined atomtronic circuits | null | null | null | null | cond-mat.quant-gas quant-ph | http://creativecommons.org/licenses/by/4.0/ | Persistent currents--inviscid quantized flow around an atomic circuit--are a crucial building block of atomtronic devices. We investigate how acceleration influences the transfer of persistent currents between two density-connected, ring-shaped atomic Bose-Einstein condensates, joined by a tunable weak link that controls system topology. We find that the acceleration of this system modifies both the density and phase dynamics between the rings, leading to a bias in the periodic vortex oscillations studied in T. Bland et al., Phys. Rev. Research 4, 043171 (2022). Accounting for dissipation suppressing such vortex oscillations, the acceleration facilitates a unilateral vortex transfer to the leading ring. We analyze how this transfer depends on the weak-link amplitude, the initial persistent current configuration, and the acceleration strength and direction. Characterization of the sensitivity to these parameters paves the way for a new platform for acceleration measurements, for which we outline a proof-of-concept ultracold double-ring accelerometer. | [
{
"version": "v1",
"created": "Thu, 31 Oct 2024 11:02:10 GMT"
},
{
"version": "v2",
"created": "Thu, 05 Jun 2025 13:24:36 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Chaika",
"A.",
""
],
[
"Oliinyk",
"A. O.",
""
],
[
"Yatsuta",
"I. V.",
""
],
[
"Proukakis",
"N. P.",
""
],
[
"Edwards",
"M.",
""
],
[
"Yakimenko",
"A. I.",
""
],
[
"Bland",
"T.",
""
]
] |
2410.23884 | Khurram Yamin | Khurram Yamin, Shantanu Gupta, Gaurav R. Ghosal, Zachary C. Lipton, Bryan Wilder | Failure Modes of LLMs for Causal Reasoning on Narratives | null | null | null | null | cs.LG cs.CL | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | In this work, we investigate the causal reasoning abilities of large language models (LLMs) through the representative problem of inferring causal relationships from narratives. We find that even state-of-the-art language models rely on unreliable shortcuts, both in terms of the narrative presentation and their parametric knowledge. For example, LLMs tend to determine causal relationships based on the topological ordering of events (i.e., earlier events cause later ones), resulting in lower performance whenever events are not narrated in their exact causal order. Similarly, we demonstrate that LLMs struggle with long-term causal reasoning and often fail when the narratives are long and contain many events. Additionally, we show LLMs appear to rely heavily on their parametric knowledge at the expense of reasoning over the provided narrative. This degrades their abilities whenever the narrative opposes parametric knowledge. We extensively validate these failure modes through carefully controlled synthetic experiments, as well as evaluations on real-world narratives. Finally, we observe that explicitly generating a causal graph generally improves performance while naive chain-of-thought is ineffective. Collectively, our results distill precise failure modes of current state-of-the-art models and can pave the way for future techniques to enhance causal reasoning in LLMs. | [
{
"version": "v1",
"created": "Thu, 31 Oct 2024 12:48:58 GMT"
},
{
"version": "v2",
"created": "Tue, 24 Dec 2024 23:07:13 GMT"
},
{
"version": "v3",
"created": "Wed, 04 Jun 2025 18:26:30 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Yamin",
"Khurram",
""
],
[
"Gupta",
"Shantanu",
""
],
[
"Ghosal",
"Gaurav R.",
""
],
[
"Lipton",
"Zachary C.",
""
],
[
"Wilder",
"Bryan",
""
]
] |
2411.00155 | Hans Ke{\ss}ler | Jayson G. Cosme, Phatthamon Kongkhambut, Anton B\"olian, Richelle Jade L. Tuquero, Jim Skulte, Ludwig Mathey, Andreas Hemmerich, Hans Ke{\ss}ler | Torus bifurcation of a dissipative time crystal | null | Phys. Rev. Lett. 134, 223601 (2025) | 10.1103/PhysRevLett.134.223601 | null | cond-mat.quant-gas nlin.AO quant-ph | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Using a quantum gas setup consisting of a Bose-Einstein condensate strongly coupled to a high-finesse optical cavity by a transverse pump laser, we experimentally observe an instability of a dissipative continuous time crystal (CTC) towards a time crystalline state exhibiting two prominent oscillation frequencies. Applying a mean-field approximation model and a Floquet analysis, we theoretically confirm that this transition is a manifestation in a many-body system of a torus bifurcation between a limit cycle (LC) and a limit torus (LT). We theoretically illustrate the LC and LT attractors using the minimal model and experimentally reconstruct them using Takens' embedding theorem applied to the non-destructively measured intracavity photon dynamics. | [
{
"version": "v1",
"created": "Thu, 31 Oct 2024 18:59:55 GMT"
},
{
"version": "v2",
"created": "Thu, 08 May 2025 13:08:16 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Cosme",
"Jayson G.",
""
],
[
"Kongkhambut",
"Phatthamon",
""
],
[
"Bölian",
"Anton",
""
],
[
"Tuquero",
"Richelle Jade L.",
""
],
[
"Skulte",
"Jim",
""
],
[
"Mathey",
"Ludwig",
""
],
[
"Hemmerich",
"Andreas",
""
],
[
"Keßler",
"Hans",
""
]
] |
2411.00160 | Bernat Espigule | Bernat Espigule, David Juher, Joan Salda\~na | Collinear Fractals and Bandt's Conjecture | 15 pages, 10 figures | Fractal and Fractional 2024 | 10.3390/fractalfract8120725 | 8, no. 12: 725 | math.DS | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | For a complex parameter $c$ outside the unit disk and an integer $n\ge2$, we examine the $n$-ary collinear fractal $E(c,n)$, defined as the attractor of the iterated function system $\{\mbox{$f_k \colon \mathbb{C} \longrightarrow \mathbb{C}$}\}_{k=1}^n$, where $f_k(z):=1+n-2k+c^{-1}z$. We investigate some topological features of the connectedness locus $\mathcal{M}_n$, similar to the Mandelbrot set, defined as the set of those $c$ for which $E(c,n)$ is connected. In particular, we provide a detailed answer to an open question posed by Calegari, Koch, and Walker in 2017. We also extend and refine the technique of the covering property by Solomyak and Xu to any $n\ge2$. We use it to show that a nontrivial portion of $\mathcal{M}_n$ is regular-closed. When $n\ge21$, we enhance this result by showing that, in fact, the whole $\mathcal{M}_n\setminus\mathbb{R}$ lies within the closure of its interior, thus proving that the generalized Bandt's conjecture is true. | [
{
"version": "v1",
"created": "Thu, 31 Oct 2024 19:06:48 GMT"
},
{
"version": "v2",
"created": "Fri, 06 Dec 2024 00:29:39 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Espigule",
"Bernat",
""
],
[
"Juher",
"David",
""
],
[
"Saldaña",
"Joan",
""
]
] |
2411.00872 | Hrishabh Bharadwaj | Hrishabh Bharadwaj | Exploring vector dark matter via effective interactions | 13 pages, 7 figures, To be published in Chinese Physics C | Chinese Physics C Vol. 49, No. 7 (2025) 073104 | 10.1088/1674-1137/add5c5 | null | hep-ph | http://creativecommons.org/licenses/by/4.0/ | We explore the properties of self-conjugate dark matter (DM) particles that predominantly interact with Standard Model electroweak gauge bosons, using an effective field theory approach. The study emphasizes effective contact interactions, invariant under the Standard Model gauge group, between vector DM and SM-neutral electroweak gauge bosons. Focusing on interaction terms up to dimension-8, we establish constraints on the model parameters based on the observed DM relic density and indirect detection signals. We also examine the prospects for dark matter-nucleon scattering in direct detection experiments. In addition, we analyze the sensitivity of low-energy LEP data to the pair production of light DM particles (with masses up to 80 GeV). Finally, we assess the potential of the proposed International Linear Collider (ILC) to probe these effective operators through the detection of DM particles produced in association with mono-photons. | [
{
"version": "v1",
"created": "Thu, 31 Oct 2024 04:17:52 GMT"
},
{
"version": "v2",
"created": "Fri, 02 May 2025 05:42:55 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Bharadwaj",
"Hrishabh",
""
]
] |
2411.01503 | Xinchi Han | Xinchi Han, Yongxi Lv, Shuyuan Zhang, Yingming Mao, Shizhen Zhao, ZhuoRan Liu, Zhuotao Liu, Peirui Cao, Ximeng Liu, Xinbing Wang, Wu Dongchao, Yang Jian, Zhang zhanbang | A Highly Scalable LLM Clusters with Optical Interconnect | null | null | null | null | cs.NI | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | The rapid development of large-scale GPU clusters for LLM training has driven enterprises to replace core-layer electrical switches with optical circuit switches (OCS) to meet escalating bandwidth demands. However, current physical topology design of OCS-based clusters faces two critical challenges. First, there exist unrealizable logical topologies, leading to underutilization of bandwidth resource. Second, calculating OCS reconfiguration constitutes an NP-Complete problem and is time-consuming for multi-tenant GPU clusters which need real-time scheduling. In this paper, we propose \emph{Cross Wiring}, a new physical topology design that resolves both limitations. Our physical topology guarantees full compatibility with all logical topologies under L2-compatibility constraints. Through a proposed \emph{Symmetric Integer Matrix Decomposition Theorem}, we design a polynomial-time OCS reconfiguration algorithm that satisfies arbitrary logical topology requirements. Evaluations show a up to 39.5\% higher training throughput versus prior architectures such as \emph{Gemini} in 128-NPU testbed and a 12.6\% reduction in average job completion time through real-workload based multi-tenant large-scale simulations. | [
{
"version": "v1",
"created": "Sun, 03 Nov 2024 09:49:12 GMT"
},
{
"version": "v2",
"created": "Thu, 20 Mar 2025 09:49:03 GMT"
},
{
"version": "v3",
"created": "Tue, 01 Apr 2025 10:22:26 GMT"
},
{
"version": "v4",
"created": "Wed, 23 Apr 2025 07:27:35 GMT"
},
{
"version": "v5",
"created": "Fri, 16 May 2025 15:06:56 GMT"
},
{
"version": "v6",
"created": "Thu, 05 Jun 2025 15:26:44 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Han",
"Xinchi",
""
],
[
"Lv",
"Yongxi",
""
],
[
"Zhang",
"Shuyuan",
""
],
[
"Mao",
"Yingming",
""
],
[
"Zhao",
"Shizhen",
""
],
[
"Liu",
"ZhuoRan",
""
],
[
"Liu",
"Zhuotao",
""
],
[
"Cao",
"Peirui",
""
],
[
"Liu",
"Ximeng",
""
],
[
"Wang",
"Xinbing",
""
],
[
"Dongchao",
"Wu",
""
],
[
"Jian",
"Yang",
""
],
[
"zhanbang",
"Zhang",
""
]
] |
2411.01574 | Olga Mashkova | Olga Mashkova, Fernando Zhapa-Camacho, Robert Hoehndorf | DELE: Deductive $\mathcal{EL}^{++}$ Embeddings for Knowledge Base Completion | Extended version of the paper "Enhancing Geometric Ontology Embeddings for $\mathcal{EL}^{++}$ with Negative Sampling and Deductive Closure Filtering" presented at NeSy 2024 conference, revised version | null | null | null | cs.AI | http://creativecommons.org/licenses/by/4.0/ | Ontology embeddings map classes, relations, and individuals in ontologies into $\mathbb{R}^n$, and within $\mathbb{R}^n$ similarity between entities can be computed or new axioms inferred. For ontologies in the Description Logic $\mathcal{EL}^{++}$, several optimization-based embedding methods have been developed that explicitly generate models of an ontology. However, these methods suffer from some limitations; they do not distinguish between statements that are unprovable and provably false, and therefore they may use entailed statements as negatives. Furthermore, they do not utilize the deductive closure of an ontology to identify statements that are inferred but not asserted. We evaluated a set of embedding methods for $\mathcal{EL}^{++}$ ontologies, incorporating several modifications that aim to make use of the ontology deductive closure. In particular, we designed novel negative losses that account both for the deductive closure and different types of negatives and formulated evaluation methods for knowledge base completion. We demonstrate that our embedding methods improve over the baseline ontology embedding in the task of knowledge base or ontology completion. | [
{
"version": "v1",
"created": "Sun, 03 Nov 2024 14:00:04 GMT"
},
{
"version": "v2",
"created": "Thu, 05 Jun 2025 08:51:13 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Mashkova",
"Olga",
""
],
[
"Zhapa-Camacho",
"Fernando",
""
],
[
"Hoehndorf",
"Robert",
""
]
] |
2411.01679 | Tennison Liu | Nicol\'as Astorga, Tennison Liu, Yuanzhang Xiao, Mihaela van der Schaar | Autoformulation of Mathematical Optimization Models Using LLMs | *Astorga and Liu contributed equally. Published as a conference paper at ICML 2025 | null | null | null | cs.LG | http://creativecommons.org/licenses/by/4.0/ | Mathematical optimization is fundamental to decision-making across diverse domains, from operations research to healthcare. Yet, translating real-world problems into optimization models remains a difficult task, often demanding specialized expertise. This paper approaches the problem of $\textit{autoformulation}$: the automated creation of solver-ready optimization models from natural language problem descriptions. We identify three core challenges of autoformulation: $\textit{(1)}$ the vast, problem-dependent hypothesis space, $\textit{(2)}$ efficient and diverse exploration of this space under uncertainty, and $\textit{(3)}$ evaluation of formulation correctness against problem description. To address these challenges, we present a novel method leveraging $\textit{Large Language Models}$ (LLMs) with $\textit{Monte-Carlo Tree Search}$, exploiting the hierarchical nature of optimization modeling to generate and systematically explore possible formulations. To enhance search efficiency, we introduce symbolic pruning to eliminate trivially equivalent search paths (branches), and employ LLM-based evaluation of partial formulations to guide search. Empirical analysis on linear and mixed-integer programming benchmarks demonstrates our method's effectiveness, with significant performance gains from both LLM-based value estimation and symbolic pruning techniques. | [
{
"version": "v1",
"created": "Sun, 03 Nov 2024 20:41:38 GMT"
},
{
"version": "v2",
"created": "Thu, 05 Jun 2025 14:26:44 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Astorga",
"Nicolás",
""
],
[
"Liu",
"Tennison",
""
],
[
"Xiao",
"Yuanzhang",
""
],
[
"van der Schaar",
"Mihaela",
""
]
] |
2411.02582 | Shaoming He | Juanqin Liu, Leonardo Plotegher, Eloy Roura, Cristino de Souza Junior, Shaoming He | Real-Time Detection for Small UAVs: Combining YOLO and Multi-frame Motion Analysis | null | null | 10.1109/TAES.2025.3577592 | null | cs.CV | http://creativecommons.org/licenses/by/4.0/ | Unmanned Aerial Vehicle (UAV) detection technology plays a critical role in mitigating security risks and safeguarding privacy in both military and civilian applications. However, traditional detection methods face significant challenges in identifying UAV targets with extremely small pixels at long distances. To address this issue, we propose the Global-Local YOLO-Motion (GL-YOMO) detection algorithm, which combines You Only Look Once (YOLO) object detection with multi-frame motion detection techniques, markedly enhancing the accuracy and stability of small UAV target detection. The YOLO detection algorithm is optimized through multi-scale feature fusion and attention mechanisms, while the integration of the Ghost module further improves efficiency. Additionally, a motion detection approach based on template matching is being developed to augment detection capabilities for minute UAV targets. The system utilizes a global-local collaborative detection strategy to achieve high precision and efficiency. Experimental results on a self-constructed fixed-wing UAV dataset demonstrate that the GL-YOMO algorithm significantly enhances detection accuracy and stability, underscoring its potential in UAV detection applications. | [
{
"version": "v1",
"created": "Thu, 10 Oct 2024 14:30:50 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Liu",
"Juanqin",
""
],
[
"Plotegher",
"Leonardo",
""
],
[
"Roura",
"Eloy",
""
],
[
"Junior",
"Cristino de Souza",
""
],
[
"He",
"Shaoming",
""
]
] |
2411.02896 | Khamdam Rakhimov | M.M. Suleimanov, M.U. Nosirov, H.T. Yusupov, A. Chaves, G.R. Berdiyorov, and Kh.Yu. Rakhimov | Wave-packet dynamics in a graphene with periodic potentials | null | null | null | null | cond-mat.mes-hall | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We use the Dirac continuum model to study the propagation of electronic wave packets in graphene with periodically arranged circular potential steps. The time propagation of the wave packets are calculated using the split-operator method for different size, height and separation of the barriers. The time propagation of the wave packets is calculated using the split-operator method for various barrier sizes, heights, and separations. We found that, despite the pronounced Klein tunneling effect in graphene, the presence of a lattice of defects significantly impacts the propagation properties of the wave packets. For example, depending on the height and size of the incident wave packet, the transmission probability can decrease by more than 20\%. The alteration of the polarity of the potential barriers also contributes to the transmission probabilities of the wave packet in graphene. The obtained results could provide valuable insights into the fundamental understanding of charge carrier dynamics in graphene-based nanodevices. | [
{
"version": "v1",
"created": "Tue, 05 Nov 2024 08:17:19 GMT"
},
{
"version": "v2",
"created": "Thu, 05 Jun 2025 05:29:37 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Suleimanov",
"M. M.",
""
],
[
"Nosirov",
"M. U.",
""
],
[
"Yusupov",
"H. T.",
""
],
[
"Chaves",
"A.",
""
],
[
"Berdiyorov",
"G. R.",
""
],
[
"Rakhimov",
"Kh. Yu.",
""
]
] |
2411.04846 | Faisal Abu-Khzam | Faisal N. Abu-Khzam, Tom Davot, Lucas Isenmann and Sergio Thoumi | On the Complexity of 2-club Cluster Editing with Vertex Splitting | null | null | null | null | cs.DS cs.CC | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Editing a graph to obtain a disjoint union of s-clubs is one of the models for correlation clustering, which seeks a partition of the vertex set of a graph so that elements of each resulting set are close enough according to some given criterion. For example, in the case of editing into s-clubs, the criterion is proximity since any pair of vertices (in an s-club) are within a distance of s from each other. In this work we consider the vertex splitting operation, which allows a vertex to belong to more than one cluster. This operation was studied as one of the parameters associated with the Cluster Editing problem. We study the complexity and parameterized complexity of the s-Club Cluster Edge Deletion with Vertex Splitting and s-Club Cluster Vertex Splitting problems. Both problems are shown to be NP-Complete and APX-hard. On the positive side, we show that both problems are Fixed-Parameter Tractable with respect to the number of allowed editing operations and that s-Club Cluster Vertex Splitting is solvable in polynomial-time on the class of forests. | [
{
"version": "v1",
"created": "Thu, 07 Nov 2024 16:33:45 GMT"
},
{
"version": "v2",
"created": "Thu, 05 Jun 2025 17:25:21 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Abu-Khzam",
"Faisal N.",
""
],
[
"Davot",
"Tom",
""
],
[
"Isenmann",
"Lucas",
""
],
[
"Thoumi",
"Sergio",
""
]
] |
2411.05685 | Yilin Bi | Junhap Bian, Tao Zhou, Yilin Bi | Beyond Pairwise Interactions: Unveiling the Role of Higher-Order Interactions via Stepwise Reduction | 14 pages, 6 figures | Communications Physics, vol. 8, article number 228, 2025 | 10.1038/s42005-025-02157-3 | null | physics.soc-ph cs.SI | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Complex systems, such as economic, social, biological, and ecological systems, usually feature interactions not only between pairwise entities but also among three or more entities. These multi-entity interactions are known as higher-order interactions. Hypergraph, as a mathematical tool, can effectively characterize higher-order interactions, where nodes denote entities and hyperedges represent interactions among multiple entities. Meanwhile, all higher-order interactions can also be projected into a number of lower-order interactions or even some pairwise interactions. Whether it is necessary to consider all higher-order interactions, and whether it is with little loss to replace them by lower-order or even pairwise interactions, remain a controversial issue. If the role of higher-order interactions is insignificant, the complexity of computation and the difficulty of analysis can be drastically reduced by projecting higher-order interactions into lower-order or pairwise interactions. We use link prediction, a fundamental problem in network science, as the entry point. Specifically, we evaluate the impact of higher-order interactions on link predictive accuracy to explore the necessity of these structures. We propose a method to decompose the higher-order structures in a stepwise way, thereby allowing to systematically explore the impacts of structures at different orders on link prediction. The results indicate that in some networks, incorporating higher-order interactions significantly enhances the accuracy of link prediction, while in others, the effect is insignificant. Therefore, we think that the role of higher-order interactions varies in different types of networks. Overall, since the improvement in predictive accuracy provided by higher-order interactions is significant in some networks, we believe that the study of higher-order interactions is both necessary and valuable. | [
{
"version": "v1",
"created": "Fri, 08 Nov 2024 16:33:17 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Bian",
"Junhap",
""
],
[
"Zhou",
"Tao",
""
],
[
"Bi",
"Yilin",
""
]
] |
2411.06842 | Vladyslav Zalevskyi | Vladyslav Zalevskyi, Thomas Sanchez, Margaux Roulet, H\'el\`ene Lajous, Jordina Aviles Verdera, Roxane Licandro, Georg Langs, Gregor Kasprian, Jana Hutter, Hamza Kebiri and Meritxell Bach Cuadra | DRIFTS: Optimizing Domain Randomization with Synthetic Data and Weight Interpolation for Fetal Brain Tissue Segmentation | null | null | null | null | eess.IV cs.CV | http://creativecommons.org/licenses/by/4.0/ | Fetal brain tissue segmentation in magnetic resonance imaging (MRI) is a crucial tool that supports understanding of neurodevelopment, yet it faces challenges due to the heterogeneity of data coming from different scanners and settings, as well as data scarcity. Recent approaches based on domain randomization, like SynthSeg, have shown great potential for single-source domain generalization by simulating images with randomized contrast and image resolution from the label maps. In this work, we investigate how to maximize the out-of-domain (OOD) generalization potential of SynthSegbased methods in fetal brain MRI. Specifically, we demonstrate that the simple Gaussian mixture models employed in FetalSynthSeg outperform physics-informed generation methods in terms of OOD generalization. We further show that incorporating intensity clustering significantly enhances generalization in settings with limited label classes by producing more realistic synthetic data. By combining synthetic pretraining with fine-tuning on real images and applying weight-space interpolation between the two models, we propose DRIFTS as an effective and practical solution for single-source domain generalization. DRIFTS consistently outperforms current state-of-the-art models across multiple benchmarks and is, to our knowledge, the first method to achieve accurate brain tissue segmentation on fetal T1-weighted images. We validate our approach on 308 subjects from four datasets acquired at three different sites, covering a range of scanner field strengths (0.55T to 3T) and both T1w and T2w modalities. We conclude with five practical recommendations to guide the development of SynthSeg-based methods for other organs and imaging modalities. | [
{
"version": "v1",
"created": "Mon, 11 Nov 2024 10:17:44 GMT"
},
{
"version": "v2",
"created": "Wed, 04 Jun 2025 18:56:48 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Zalevskyi",
"Vladyslav",
""
],
[
"Sanchez",
"Thomas",
""
],
[
"Roulet",
"Margaux",
""
],
[
"Lajous",
"Hélène",
""
],
[
"Verdera",
"Jordina Aviles",
""
],
[
"Licandro",
"Roxane",
""
],
[
"Langs",
"Georg",
""
],
[
"Kasprian",
"Gregor",
""
],
[
"Hutter",
"Jana",
""
],
[
"Kebiri",
"Hamza",
""
],
[
"Cuadra",
"Meritxell Bach",
""
]
] |
2411.07077 | Yuxin Ma | Erin Carson and Yuxin Ma | A stable one-synchronization variant of reorthogonalized block classical Gram--Schmidt | 25 pages, 8. figures | null | null | null | math.NA cs.NA | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | The block classical Gram--Schmidt (BCGS) algorithm and its reorthogonalized variant are widely-used methods for computing the economic QR factorization of block columns $X$ due to their lower communication cost compared to other approaches such as modified Gram--Schmidt and Householder QR. To further reduce communication, i.e., synchronization, there has been a long ongoing search for a variant of reorthogonalized BCGS variant that achieves $O(u)$ loss of orthogonality while requiring only \emph{one} synchronization point per block column, where $u$ represents the unit roundoff. Utilizing Pythagorean inner products and delayed normalization techniques, we propose the first provably stable one-synchronization reorthogonalized BCGS variant, demonstrating that it has $O(u)$ loss of orthogonality under the condition $O(u) \kappa^2(X) \leq 1/2$, where $\kappa(\cdot)$ represents the condition number.
By incorporating one additional synchronization point, we develop a two-synchronization reorthogonalized BCGS variant which maintains $O(u)$ loss of orthogonality under the improved condition $O(u) \kappa(X) \leq 1/2$. An adaptive strategy is then proposed to combine these two variants, ensuring $O(u)$ loss of orthogonality while using as few synchronization points as possible under the less restrictive condition $O(u) \kappa(X) \leq 1/2$. As an example of where this adaptive approach is beneficial, we show that using the adaptive orthogonalization variant, $s$-step GMRES achieves a backward error comparable to $s$-step GMRES with BCGSI+, also known as BCGS2, both theoretically and numerically, but requires fewer synchronization points. | [
{
"version": "v1",
"created": "Mon, 11 Nov 2024 15:54:43 GMT"
},
{
"version": "v2",
"created": "Thu, 05 Jun 2025 13:10:42 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Carson",
"Erin",
""
],
[
"Ma",
"Yuxin",
""
]
] |
2411.07103 | Zakaria Derbazi | Zakaria Derbazi | On a connection between total positivity and Bernoulli stopping problems | 13 pages, 1 figure. Fixing some errors and removing sections 3.4 and 4.3 | null | null | null | math.PR | http://creativecommons.org/licenses/by/4.0/ | Consider the optimal stopping problem of maximising the expected payoff in a game where a nonnegative reward is granted upon stopping on a success in a sequence of independent Bernoulli trials. These Bernoulli stopping problems are characterised by a recurrence relation connecting the stopping and continuation payoffs. This recurrence is used to establish optimality of the myopic strategy under unimodal continuation payoffs. Further, by embedding the success epochs of the trials into a Markov chain, the transition matrix of the chain maps the stopping payoffs into continuation payoffs. A total positivity argument shows that the expectation of a unimodal function of the chain is shown to be unimodal in the initial state. This property establishes the unimodality of the stopping payoffs as a sufficient condition for the optimality of the myopic rule. Illustrative applications of this result include the $m$th last-success problem and its generalisation. | [
{
"version": "v1",
"created": "Mon, 11 Nov 2024 16:27:22 GMT"
},
{
"version": "v2",
"created": "Wed, 04 Jun 2025 20:19:19 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Derbazi",
"Zakaria",
""
]
] |
2411.07350 | Sanghyuk Moon | Sanghyuk Moon and Eve C. Ostriker | Prestellar Cores in Turbulent Clouds: Properties of Critical Cores | 28 pages, 13 figures, accepted for publication in ApJ; replaced with the accepted version | null | null | null | astro-ph.GA astro-ph.SR | http://creativecommons.org/licenses/by/4.0/ | A fraction of the dense cores within a turbulent molecular cloud will eventually collapse to form stars. Identifying the physical criteria for instability and analyzing critical core properties is therefore necessary to star formation theory. Here we quantify the characteristics of an ensemble of "critical cores" on the verge of collapse. This critical epoch was identified in a companion paper, which followed the dynamical evolution of prestellar cores in numerical simulations of turbulent, self-gravitating clouds. We find that radial profiles of density and turbulent velocity dispersion in individual critical cores are consistent with our new model for turbulent equilibrium spheres (TESs). While a global linewidth-size relation exists for a cloud with given size and Mach number, the turbulent scaling relations around each core exhibit significant variations. As a result, there is no single density threshold for collapse; instead, cores collapse at a range of densities determined by the local sonic scale and gravitational potential environment. The critical cores in our simulations are mostly transonic; we do not find either purely thermal or highly turbulent cores. In our low Mach number model which better resolves the characteristic mass and sonic scales, we find marginal evidence that the core mass function (CMF) of critical cores peaks around a characteristic mass scale associated with the large-scale cloud properties. We highlight the importance of constructing the CMF at the critical time for the purpose of testing gravoturbulent fragmentation theories, and derive the resolution requirements to unambiguously identify the peak of the CMF. | [
{
"version": "v1",
"created": "Mon, 11 Nov 2024 20:13:26 GMT"
},
{
"version": "v2",
"created": "Wed, 13 Nov 2024 18:55:10 GMT"
},
{
"version": "v3",
"created": "Wed, 04 Jun 2025 18:05:40 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Moon",
"Sanghyuk",
""
],
[
"Ostriker",
"Eve C.",
""
]
] |
2411.07970 | Cheng Zhao | Cheng Zhao, Song Huang, Mengfan He, Paulo Montero-Camacho, Yu Liu, Pablo Renard, Yunyi Tang, Aurelien Verdier, Wenshuo Xu, Xiaorui Yang, Jiaxi Yu, Yao Zhang, Siyi Zhao, Xingchen Zhou, Shengyu He, Jean-Paul Kneib, Jiayi Li, Zhuoyang Li, Wen-Ting Wang, Zhong-Zhi Xianyu, Yidian Zhang, Rafaela Gsponer, Xiao-Dong Li, Antoine Rocher, Siwei Zou, Ting Tan, Zhiqi Huang, Zhuoxiao Wang, Pei Li, Maxime Rombach, Chenxing Dong, Daniel Forero-Sanchez, Yuanhang Ning, Huanyuan Shan, Tao Wang, Yin Li, Zhongxu Zhai, Yuting Wang, Gong-Bo Zhao, Yong Shi, Shude Mao, Lei Huang, Liquan Guo, and Zheng Cai | MUltiplexed Survey Telescope: Perspectives for Large-Scale Structure Cosmology in the Era of Stage-V Spectroscopic Survey | Submitted to SCPMA | null | null | null | astro-ph.CO astro-ph.IM | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | The MUltiplexed Survey Telescope (MUST) is a 6.5-meter telescope under development. Dedicated to highly-multiplexed, wide-field spectroscopic surveys, MUST observes over 20,000 targets simultaneously using 6.2-mm pitch positioning robots within a ~5 deg$^2$ field of view. MUST aims to conduct the first Stage-V spectroscopic survey in the 2030s, mapping the 3D Universe with over 100 million galaxies and quasars, spanning from the nearby Universe to a redshift of z ~ 5.5, corresponding to approximately 1 billion years after the Big Bang. To cover this extensive redshift range, we present an initial conceptual target selection algorithm for different types of galaxies, ranging from local bright galaxies and luminous red galaxies to emission-line galaxies, and high-redshift (2 < z < 5.5) Lyman-break galaxies. Using Fisher forecasts, we demonstrate that MUST can address fundamental questions in cosmology, including the nature of dark energy, tests of gravity theories, and investigations into primordial physics. This is the first paper in the series of science white papers for MUST, with subsequent developments focusing on additional scientific cases such as galaxy and quasar evolution, Milky Way physics, and dynamic phenomena in the time-domain Universe. | [
{
"version": "v1",
"created": "Tue, 12 Nov 2024 17:51:21 GMT"
},
{
"version": "v2",
"created": "Wed, 13 Nov 2024 17:29:11 GMT"
},
{
"version": "v3",
"created": "Thu, 05 Jun 2025 16:29:38 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Zhao",
"Cheng",
""
],
[
"Huang",
"Song",
""
],
[
"He",
"Mengfan",
""
],
[
"Montero-Camacho",
"Paulo",
""
],
[
"Liu",
"Yu",
""
],
[
"Renard",
"Pablo",
""
],
[
"Tang",
"Yunyi",
""
],
[
"Verdier",
"Aurelien",
""
],
[
"Xu",
"Wenshuo",
""
],
[
"Yang",
"Xiaorui",
""
],
[
"Yu",
"Jiaxi",
""
],
[
"Zhang",
"Yao",
""
],
[
"Zhao",
"Siyi",
""
],
[
"Zhou",
"Xingchen",
""
],
[
"He",
"Shengyu",
""
],
[
"Kneib",
"Jean-Paul",
""
],
[
"Li",
"Jiayi",
""
],
[
"Li",
"Zhuoyang",
""
],
[
"Wang",
"Wen-Ting",
""
],
[
"Xianyu",
"Zhong-Zhi",
""
],
[
"Zhang",
"Yidian",
""
],
[
"Gsponer",
"Rafaela",
""
],
[
"Li",
"Xiao-Dong",
""
],
[
"Rocher",
"Antoine",
""
],
[
"Zou",
"Siwei",
""
],
[
"Tan",
"Ting",
""
],
[
"Huang",
"Zhiqi",
""
],
[
"Wang",
"Zhuoxiao",
""
],
[
"Li",
"Pei",
""
],
[
"Rombach",
"Maxime",
""
],
[
"Dong",
"Chenxing",
""
],
[
"Forero-Sanchez",
"Daniel",
""
],
[
"Ning",
"Yuanhang",
""
],
[
"Shan",
"Huanyuan",
""
],
[
"Wang",
"Tao",
""
],
[
"Li",
"Yin",
""
],
[
"Zhai",
"Zhongxu",
""
],
[
"Wang",
"Yuting",
""
],
[
"Zhao",
"Gong-Bo",
""
],
[
"Shi",
"Yong",
""
],
[
"Mao",
"Shude",
""
],
[
"Huang",
"Lei",
""
],
[
"Guo",
"Liquan",
""
],
[
"Cai",
"Zheng",
""
]
] |
2411.08570 | Shuai S. A. Yuan | Shuai S. A. Yuan, Xinyi Y. I. Xu, Jinpeng Yuan, Guoda Xie, Chongwen Huang, Xiaoming Chen, Zhixiang Huang, and Wei E. I. Sha | Electromagnetic Modeling and Capacity Analysis of Rydberg Atom-Based MIMO System | in IEEE Antennas and Wireless Propagation Letters, 2025 | null | null | null | eess.SP | http://creativecommons.org/licenses/by-nc-nd/4.0/ | Rydberg atom-based antennas exploit the quantum properties of highly excited Rydberg atoms, providing unique advantages over classical antennas, such as high sensitivity, broad frequency range, and compact size. Despite the increasing interests in their applications in antenna and communication engineering, two key properties, involving the lack of polarization multiplexing and isotropic reception without mutual coupling, remain unexplored in the analysis of Rydberg atom-based spatial multiplexing, i.e., multiple-input and multiple-output (MIMO), communications. Generally, the design considerations for any antenna, even for atomic ones, can be extracted to factors such as radiation patterns, efficiency, and polarization, allowing them to be seamlessly integrated into existing system models. In this letter, we extract the antenna properties from relevant quantum characteristics, enabling electromagnetic modeling and capacity analysis of Rydberg MIMO systems in both far-field and near-field scenarios. By employing ray-based method for far-field analysis and dyadic Green's function for near-field calculation, our results indicate that Rydberg atom-based antenna arrays offer specific advantages over classical dipole-type arrays in single-polarization MIMO communications. | [
{
"version": "v1",
"created": "Wed, 13 Nov 2024 12:32:20 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Yuan",
"Shuai S. A.",
""
],
[
"Xu",
"Xinyi Y. I.",
""
],
[
"Yuan",
"Jinpeng",
""
],
[
"Xie",
"Guoda",
""
],
[
"Huang",
"Chongwen",
""
],
[
"Chen",
"Xiaoming",
""
],
[
"Huang",
"Zhixiang",
""
],
[
"Sha",
"Wei E. I.",
""
]
] |
2411.09414 | Tejhas Kapoor | Bhubanjyoti Bhattacharya, Thomas E. Browder, Alakabha Datta, Tejhas Kapoor, Emi Kou, and Lopamudra Mukherjee | New physics searches via angular distributions of $ \bar{B} \to D^* (\to D \pi) \tau (\to \ell \nu_\tau \bar{\nu}_\ell) \bar{\nu}_\tau$ decays | 38 pages, 8 figures; new version: added references, corrected typos | JHEP 04 (2025) 135 | 10.1007/JHEP04(2025)135 | null | hep-ph hep-ex | http://creativecommons.org/licenses/by/4.0/ | The study of $\bar{B} \to D^* \tau {\bar{\nu}}_\tau$ angular distribution can be used to obtain information about new physics (or beyond the Standard Model) couplings, which are motivated by various $B$ anomalies. However, the inability to measure precisely the three-momentum of the $\tau$ lepton hinders such measurements, as the tau decay contains one or more undetected neutrinos. Here, we present a measurable angular distribution of $\bar{B} \to D^* \tau {\bar{\nu}}_\tau$ by considering the additional decay $\tau \to \ell \nu_\tau \bar{\nu}_\ell$, where $\ell \in \{ e , \mu \}$. The full process used is $\bar{B} \to D^* (\to D \pi) \tau (\to \ell \nu_\tau \bar{\nu}_\ell) \bar{\nu}_\tau$, in which only the $\ell$ and $D^*$ are reconstructed. A fit to the experimental angular distribution of this process can be used to extract information on new physics parameters. To demonstrate the feasibility of this approach, we generate simulated data for this process and perform a sensitivity study to obtain the expected statistical errors on new physics parameters from experiments in the near future. We obtain a sensitivity of the order of 5% for the right-handed current and around 6% for the tensor current. In addition, we use the recent lattice QCD data on $B \to D^*$ form factors and obtain correlations between form factors and new physics parameters. | [
{
"version": "v1",
"created": "Thu, 14 Nov 2024 13:07:36 GMT"
},
{
"version": "v2",
"created": "Wed, 15 Jan 2025 02:16:46 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Bhattacharya",
"Bhubanjyoti",
""
],
[
"Browder",
"Thomas E.",
""
],
[
"Datta",
"Alakabha",
""
],
[
"Kapoor",
"Tejhas",
""
],
[
"Kou",
"Emi",
""
],
[
"Mukherjee",
"Lopamudra",
""
]
] |
2411.09740 | St\'ephane Blondin | St\'ephane Blondin | Type Ia supernovae | This is a pre-print of a chapter for the Encyclopedia of Astrophysics (edited by Ilya Mandel, section editor Jeffrey Andrews) to be published by Elsevier as a Reference Module. 18 pages, 7 figures. Edited to match the final revised version | null | null | null | astro-ph.HE astro-ph.SR | http://creativecommons.org/licenses/by/4.0/ | Type Ia supernovae (SNe Ia) correspond to the thermonuclear explosion of a carbon-oxygen white dwarf (C-O WD) star in a binary system, triggered by the accretion of material from another star, or the merger/collision with a secondary WD. Their phenomenal luminosity -- several billion times that of the sun -- has motivated their use as cosmological distance indicators and led to the discovery of the accelerated expansion of the universe. SNe Ia are also the main producers of iron and hence play a fundamental role in the chemical evolution of galaxies. While recent observations have confirmed the basic theoretical picture of an exploding C-O WD star whose luminosity is powered by the radioactive decay of $^{56}$Ni, a number of uncertainties remain concerning the nature of the binary companion and the explosion mechanism. Several lines of evidence point towards the existence of multiple progenitor channels in order to explain the full range of the observed diversity. A complete physical understanding of these energetic stellar explosions remains a long-lasting goal of modern astrophysics. | [
{
"version": "v1",
"created": "Thu, 14 Nov 2024 19:00:02 GMT"
},
{
"version": "v2",
"created": "Thu, 05 Jun 2025 13:48:28 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Blondin",
"Stéphane",
""
]
] |
2411.10380 | Kirill Kuzanyan | Nathan Kleeorin, Kirill Kuzanyan, Nikolai Safiullin, Igor Rogachevskii, Vladimir Obridko, Sergey Porshnev, and Rodion Stepanov | Forecast of solar activity based on mean-field dynamo model and neural network | 12 pages, 4 figures, 1 table | Astronomical and Astrophysical Transactions 34, No. 4, 371-388 (2024) | null | null | astro-ph.SR physics.space-ph | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We discuss a prediction of the solar activity on a short time-scale applying the method based on a combination of a nonlinear mean-field dynamo model and the artificial neural network. The artificial neural network which serves as a correction scheme for the forecast, uses the currently available observational data (e.g., the 13 month running average of the observed solar sunspot numbers) and the dynamo model output. The nonlinear mean-field $\alpha\,\Omega$ dynamo produces the large-scale magnetic flux which is redistributed by negative effective magnetic pressure instability (NEMPI) producing sunspots and active regions. The nonlinear mean-field dynamo model includes algebraic nonlinearity (caused by the feedback of the growing magnetic field on the plasma motion) and dynamic nonlinearities (related to the dynamics of the magnetic helicity of small-scale magnetic field). We compare the forecast errors with a horizon of 1, 6, 12 and 18 months, for different forecast methods, with the same corrections on the current monthly observations. Our forecast is in good agreement with the observed solar activity, the forecast error is almost stably small over short-medium ranges of forecasting windows. Despite a strong level of chaotic component in the solar magnetic activity we present quantitative evidence that the solar activity on a short range can be stably well predicted, by the joint use of the physically based model with the neural network. This result may have an immediate practical implementation for predictions of various phenomena of solar activity and other astrophysical processes, so may be of interest to a broad community. | [
{
"version": "v1",
"created": "Fri, 15 Nov 2024 17:38:22 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Kleeorin",
"Nathan",
""
],
[
"Kuzanyan",
"Kirill",
""
],
[
"Safiullin",
"Nikolai",
""
],
[
"Rogachevskii",
"Igor",
""
],
[
"Obridko",
"Vladimir",
""
],
[
"Porshnev",
"Sergey",
""
],
[
"Stepanov",
"Rodion",
""
]
] |
2411.10778 | Kaiji Motoyama | Kaiji Motoyama, Yoshimichi Teratani, Kazuhiko Tsutsumi, Kohei Wake, Ryosuke Kobayashi, Rui Sakano, Akira Oguri | Three-body Fermi liquid corrections for an infinite-$U$ SU($N$) Anderson impurity model | 20 pages, 13 figures, expanded explanation of the formulation | Phys. Rev. B 111, 235301 (2025) | 10.1103/PhysRevB.111.235301 | NITEP 229 | cond-mat.mes-hall cond-mat.str-el | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We study the three-body Fermi liquid effects in the SU($N$) Anderson impurity model in the strong interaction limit where the occupation number $N_d^{}$ of the impurity levels varies over the range of $0<N_d^{}<1$. The three-body correlation of impurity electrons contributes to the next-to-leading order terms of transport coefficients at low energies when the electron-hole symmetry, the time-reversal symmetries, or both are broken by external fields or potentials. Using the numerical renormalization group approach, we calculate the differential conductance and the nonlinear current noise through quantum dots, as well as the thermal conductivity of both quantum dots and magnetic alloys. Specifically, we focus on the SU(2) and SU(4) cases and demonstrate how the three-body contributions evolve in the limit of $U\to\infty$, across the $1/N$-filling Kondo regime and the valence fluctuation regime. Our results clarify how the three-body correlation affects low-energy transport, with a crucial dependence on the occupation number $N_d^{}$. We also show that the three-body correlation strongly couples with asymmetries in the tunnel couplings between quantum dots and reservoirs; in particular it significantly affects the nonlinear current through the quarter-filling Kondo state for $N=4$. | [
{
"version": "v1",
"created": "Sat, 16 Nov 2024 11:33:51 GMT"
},
{
"version": "v2",
"created": "Thu, 05 Jun 2025 04:49:05 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Motoyama",
"Kaiji",
""
],
[
"Teratani",
"Yoshimichi",
""
],
[
"Tsutsumi",
"Kazuhiko",
""
],
[
"Wake",
"Kohei",
""
],
[
"Kobayashi",
"Ryosuke",
""
],
[
"Sakano",
"Rui",
""
],
[
"Oguri",
"Akira",
""
]
] |
2411.11055 | Ofir Zafrir | Ofir Zafrir, Igor Margulis, Dorin Shteyman, Shira Guskin, Guy Boudoukh | FastDraft: How to Train Your Draft | Accepted at ACL 2025 | null | null | null | cs.CL | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Speculative Decoding has gained popularity as an effective technique for accelerating the auto-regressive inference process of Large Language Models. However, Speculative Decoding entirely relies on the availability of efficient draft models, which are often lacking for many existing language models due to a stringent constraint of vocabulary compatibility. In this work we introduce FastDraft, a novel and efficient approach for pre-training and aligning a draft model to any large language model by incorporating efficient pre-training, followed by fine-tuning over synthetic datasets generated by the target model. We demonstrate FastDraft by training two highly parameter efficient drafts for the popular Phi-3-mini and Llama-3.1-8B models. Using FastDraft, we were able to produce a draft model with approximately 10 billion tokens on a single server with 8 Intel$^\circledR$ Gaudi$^\circledR$ 2 accelerators in under 24 hours. Our results show that the draft model achieves impressive results in key metrics of acceptance rate, block efficiency and up to 3x memory bound speed up when evaluated on code completion and up to 2x in summarization, text completion and instruction tasks. We validate our theoretical findings through benchmarking on the latest Intel$^\circledR$ Core$^{\tiny \text{TM}}$ Ultra, achieving a wall-clock time speedup of up to 2x, indicating a significant reduction in runtime. Due to its high quality, FastDraft unlocks large language models inference on AI-PC and other edge-devices. | [
{
"version": "v1",
"created": "Sun, 17 Nov 2024 12:32:44 GMT"
},
{
"version": "v2",
"created": "Wed, 21 May 2025 08:28:22 GMT"
},
{
"version": "v3",
"created": "Thu, 05 Jun 2025 08:09:22 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Zafrir",
"Ofir",
""
],
[
"Margulis",
"Igor",
""
],
[
"Shteyman",
"Dorin",
""
],
[
"Guskin",
"Shira",
""
],
[
"Boudoukh",
"Guy",
""
]
] |
2411.12930 | Dimple Vijay Kochar | Dimple Vijay Kochar, Hanrui Wang, Anantha Chandrakasan, Xin Zhang | LEDRO: LLM-Enhanced Design Space Reduction and Optimization for Analog Circuits | Accepted at IEEE International Conference on LLM-Aided Design (ICLAD), 2025; Accepted as a WIP poster at Design Automation Conference (DAC), 2025 | null | null | null | cs.LG cs.SY eess.SY | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Traditional approaches for designing analog circuits are time-consuming and require significant human expertise. Existing automation efforts using methods like Bayesian Optimization (BO) and Reinforcement Learning (RL) are sub-optimal and costly to generalize across different topologies and technology nodes. In our work, we introduce a novel approach, LEDRO, utilizing Large Language Models (LLMs) in conjunction with optimization techniques to iteratively refine the design space for analog circuit sizing. LEDRO is highly generalizable compared to other RL and BO baselines, eliminating the need for design annotation or model training for different topologies or technology nodes. We conduct a comprehensive evaluation of our proposed framework and baseline on 22 different Op-Amp topologies across four FinFET technology nodes. Results demonstrate the superior performance of LEDRO as it outperforms our best baseline by an average of 13% FoM improvement with 2.15x speed-up on low complexity Op-Amps and 48% FoM improvement with 1.7x speed-up on high complexity Op-Amps. This highlights LEDRO's effective performance, efficiency, and generalizability. | [
{
"version": "v1",
"created": "Tue, 19 Nov 2024 23:43:25 GMT"
},
{
"version": "v2",
"created": "Thu, 05 Jun 2025 02:02:32 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Kochar",
"Dimple Vijay",
""
],
[
"Wang",
"Hanrui",
""
],
[
"Chandrakasan",
"Anantha",
""
],
[
"Zhang",
"Xin",
""
]
] |
2411.13280 | Keyue Qiu | Keyue Qiu, Yuxuan Song, Jie Yu, Hongbo Ma, Ziyao Cao, Zhilong Zhang, Yushuai Wu, Mingyue Zheng, Hao Zhou, Wei-Ying Ma | Empower Structure-Based Molecule Optimization with Gradient Guided Bayesian Flow Networks | Accepted to ICML 2025 | null | null | null | q-bio.BM cs.AI | http://creativecommons.org/licenses/by-nc-sa/4.0/ | Structure-Based molecule optimization (SBMO) aims to optimize molecules with both continuous coordinates and discrete types against protein targets. A promising direction is to exert gradient guidance on generative models given its remarkable success in images, but it is challenging to guide discrete data and risks inconsistencies between modalities. To this end, we leverage a continuous and differentiable space derived through Bayesian inference, presenting Molecule Joint Optimization (MolJO), the gradient-based SBMO framework that facilitates joint guidance signals across different modalities while preserving SE(3)-equivariance. We introduce a novel backward correction strategy that optimizes within a sliding window of the past histories, allowing for a seamless trade-off between explore-and-exploit during optimization. MolJO achieves state-of-the-art performance on CrossDocked2020 benchmark (Success Rate 51.3%, Vina Dock -9.05 and SA 0.78), more than 4x improvement in Success Rate compared to the gradient-based counterpart, and 2x "Me-Better" Ratio as much as 3D baselines. Furthermore, we extend MolJO to a wide range of optimization settings, including multi-objective optimization and challenging tasks in drug design such as R-group optimization and scaffold hopping, further underscoring its versatility. Code is available at https://github.com/AlgoMole/MolCRAFT. | [
{
"version": "v1",
"created": "Wed, 20 Nov 2024 12:48:29 GMT"
},
{
"version": "v2",
"created": "Thu, 21 Nov 2024 07:23:37 GMT"
},
{
"version": "v3",
"created": "Mon, 12 May 2025 07:13:56 GMT"
},
{
"version": "v4",
"created": "Thu, 05 Jun 2025 12:35:30 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Qiu",
"Keyue",
""
],
[
"Song",
"Yuxuan",
""
],
[
"Yu",
"Jie",
""
],
[
"Ma",
"Hongbo",
""
],
[
"Cao",
"Ziyao",
""
],
[
"Zhang",
"Zhilong",
""
],
[
"Wu",
"Yushuai",
""
],
[
"Zheng",
"Mingyue",
""
],
[
"Zhou",
"Hao",
""
],
[
"Ma",
"Wei-Ying",
""
]
] |
2411.13669 | Robert A. Lang Mr. | Danial Motlagh, Robert A. Lang, Paarth Jain, Jorge A. Campos-Gonzalez-Angulo, Tao Zeng, Alan Aspuru-Guzik, and Juan Miguel Arrazola | Quantum Algorithm for Vibronic Dynamics: Case Study on Singlet Fission Solar Cell Design | null | null | null | null | quant-ph | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Vibronic interactions between nuclear motion and electronic states are critical for the accurate modeling of photochemistry. However, accurate simulations of fully quantum non-adiabatic dynamics are often prohibitively expensive for classical methods beyond small systems. In this work, we present a quantum algorithm based on product formulas for simulating time evolution under a general vibronic Hamiltonian in real space, capable of handling an arbitrary number of electronic states and vibrational modes. We develop the first trotterization scheme for vibronic Hamiltonians beyond two electronic states and introduce an array of optimization techniques for the exponentiation of each fragment in the product formula, resulting in a remarkably low cost of implementation. To demonstrate practical relevance, we outline a proof-of-principle integration of our algorithm into a materials discovery pipeline for designing more efficient singlet fission-based organic solar cells. We estimate that $100$ femtoseconds of propagation using a second-order Trotter product formula for a $6$-state, $21$-mode model of exciton transport at an anthracene dimer requires $154$ qubits and $2.76 \times 10^6$ Toffoli gates. While a $4$-state, $246$-mode model describing charge transfer at an anthracene-fullerene interface requires $1053$ qubits and $2.66 \times 10^7$ Toffoli gates. | [
{
"version": "v1",
"created": "Wed, 20 Nov 2024 19:22:10 GMT"
},
{
"version": "v2",
"created": "Wed, 04 Jun 2025 21:42:33 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Motlagh",
"Danial",
""
],
[
"Lang",
"Robert A.",
""
],
[
"Jain",
"Paarth",
""
],
[
"Campos-Gonzalez-Angulo",
"Jorge A.",
""
],
[
"Zeng",
"Tao",
""
],
[
"Aspuru-Guzik",
"Alan",
""
],
[
"Arrazola",
"Juan Miguel",
""
]
] |
2411.13841 | Hsuan Lo | Hsuan Lo, You Wang, Rimi Banerjee, Baile Zhang, and Y. D. Chong | Switchable Non-Hermitian Skin Effect in Bogoliubov Modes | null | null | 10.1103/PhysRevB.111.L241401 | null | cond-mat.mes-hall | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Interacting or nonlinear lattices can host emergent particle-like modes, such as Bogoliubov quasiparticles, whose band topology and other properties are potentially highly tunable. Despite originating in the study of superconducting materials, Bogoliubov quasiparticles can also occur in synthetic metamaterials. Here, we implement a nonlinear driven-dissipative circuit whose fluctuations are Bogoliubov modes possessing nontrivial non-Hermitian band topology. We show experimentally that the system exhibits a switchable non-Hermitian skin effect (NHSE), which abruptly appears when the on-site driving voltage amplitude exceeds a threshold. In contrast to earlier realizations of the NHSE and related phenomena in circuit models, the switchable NHSE in our system occurs in Bogoliubov modes, which are strongly affected by how the system is driven. Moreover, unlike other experimental platforms hosting non-Hermitian Bogoliubov modes, our system does not contain unconventional asymmetric hopping nonlinearities, only a local Kerr-type nonlinearity. | [
{
"version": "v1",
"created": "Thu, 21 Nov 2024 05:02:04 GMT"
},
{
"version": "v2",
"created": "Mon, 24 Mar 2025 06:32:45 GMT"
},
{
"version": "v3",
"created": "Wed, 14 May 2025 03:22:22 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Lo",
"Hsuan",
""
],
[
"Wang",
"You",
""
],
[
"Banerjee",
"Rimi",
""
],
[
"Zhang",
"Baile",
""
],
[
"Chong",
"Y. D.",
""
]
] |
2411.13983 | Hansung Kim | Hansung Kim, Edward L. Zhu, Chang Seok Lim, Francesco Borrelli | Learning Two-agent Motion Planning Strategies from Generalized Nash Equilibrium for Model Predictive Control | Accepted Proceeding at 2025 Learning for Dynamics and Control Conference (L4DC) | null | null | null | cs.MA cs.RO cs.SY eess.SY | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We introduce an Implicit Game-Theoretic MPC (IGT-MPC), a decentralized algorithm for two-agent motion planning that uses a learned value function that predicts the game-theoretic interaction outcomes as the terminal cost-to-go function in a model predictive control (MPC) framework, guiding agents to implicitly account for interactions with other agents and maximize their reward. This approach applies to competitive and cooperative multi-agent motion planning problems which we formulate as constrained dynamic games. Given a constrained dynamic game, we randomly sample initial conditions and solve for the generalized Nash equilibrium (GNE) to generate a dataset of GNE solutions, computing the reward outcome of each game-theoretic interaction from the GNE. The data is used to train a simple neural network to predict the reward outcome, which we use as the terminal cost-to-go function in an MPC scheme. We showcase emerging competitive and coordinated behaviors using IGT-MPC in scenarios such as two-vehicle head-to-head racing and un-signalized intersection navigation. IGT-MPC offers a novel method integrating machine learning and game-theoretic reasoning into model-based decentralized multi-agent motion planning. | [
{
"version": "v1",
"created": "Thu, 21 Nov 2024 09:47:15 GMT"
},
{
"version": "v2",
"created": "Sat, 23 Nov 2024 02:42:55 GMT"
},
{
"version": "v3",
"created": "Sun, 02 Mar 2025 23:56:38 GMT"
},
{
"version": "v4",
"created": "Thu, 05 Jun 2025 04:38:52 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Kim",
"Hansung",
""
],
[
"Zhu",
"Edward L.",
""
],
[
"Lim",
"Chang Seok",
""
],
[
"Borrelli",
"Francesco",
""
]
] |
2411.14293 | Gaurav Kumar | Gaurav Kumar, Arezoo M. Ardekani | Machine learning framework to predict the performance of lipid nanoparticles for nucleic acid delivery | This is a preprint of a manuscript under review | ACS Appl. Bio Mater. 2025, 8, 5, 3717-3727 | 10.1021/acsabm.4c01716 | null | q-bio.BM | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Lipid nanoparticles (LNPs) are highly effective carriers for gene therapies, including mRNA and siRNA delivery, due to their ability to transport nucleic acids across biological membranes, low cytotoxicity, improved pharmacokinetics, and scalability. A typical approach to formulate LNPs is to establish a quantitative structure-activity relationship (QSAR) between their compositions and in vitro/in vivo activities which allows for the prediction of activity based on molecular structure. However, developing QSAR for LNPs can be challenging due to the complexity of multi-component formulations, interactions with biological membranes, and stability in physiological environments. To address these challenges, we developed a machine learning framework to predict the activity and cell viability of LNPs for nucleic acid delivery. We curated data from 6,398 LNP formulations in the literature, applied nine featurization techniques to extract chemical information, and trained five machine learning models for binary and multiclass classification. Our binary models achieved over 90% accuracy, while the multiclass models reached over 95% accuracy. Our results demonstrated that molecular descriptors, particularly when used with random forest and gradient boosting models, provided the most accurate predictions. Our findings also emphasized the need for large training datasets and comprehensive LNP composition details, such as constituent structures, molar ratios, nucleic acid types, and dosages, to enhance predictive performance. | [
{
"version": "v1",
"created": "Thu, 21 Nov 2024 16:40:00 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Kumar",
"Gaurav",
""
],
[
"Ardekani",
"Arezoo M.",
""
]
] |
2411.14608 | Lorena Alzate-Vargas | Lorena Alzate-Vargas, Kashi N. Subedi, Nicholas Lubbers, Michael W.D Cooper, Roxanne M. Tutchton, Tammie Gibson, Richard A. Messerly | Toward machine learning interatomic potentials for modeling uranium mononitride | null | null | null | null | cond-mat.mtrl-sci | http://creativecommons.org/licenses/by-nc-nd/4.0/ | Uranium mononitride (UN) is a promising accident-tolerant fuel because of its high fissile density and high thermal conductivity. In this study, we developed the first machine learning interatomic potentials for reliable atomic-scale modeling of UN at finite temperatures. We constructed a training set using density functional theory (DFT) calculations that was enriched through an active learning procedure, and two neural network potentials were generated. Both potentials successfully reproduce key thermophysical properties of interest, such as temperature-dependent lattice parameter, specific heat capacity, and bulk modulus. We also evaluated the energy of stoichiometric defect reactions and defect migration barriers and found close agreement with DFT predictions, demonstrating that our potentials can be used for modeling defects in UN. Additional tests provide evidence that our potentials are reliable for simulating diffusion, noble gas impurities, and radiation damage. | [
{
"version": "v1",
"created": "Thu, 21 Nov 2024 22:12:27 GMT"
},
{
"version": "v2",
"created": "Thu, 05 Jun 2025 17:44:17 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Alzate-Vargas",
"Lorena",
""
],
[
"Subedi",
"Kashi N.",
""
],
[
"Lubbers",
"Nicholas",
""
],
[
"Cooper",
"Michael W. D",
""
],
[
"Tutchton",
"Roxanne M.",
""
],
[
"Gibson",
"Tammie",
""
],
[
"Messerly",
"Richard A.",
""
]
] |
2411.14831 | Anton Zlotnikov | A. D. Fedoseev, A. O. Zlotnikov | Coexistence of gapless and gapped vortex modes with Majorana corner states in a 2D second-order topological superconductor | 9 pages, 3 figures | Phys. Rev. B 111, 214507 (2025) | 10.1103/PhysRevB.111.214507 | null | cond-mat.supr-con | http://creativecommons.org/licenses/by/4.0/ | Although the appearance of vortex-localized states with zero energy in first-order topological superconductors is well known, their possibility to form in the higher-order topological phase of 2D systems has not been completely uncovered yet. Here we demonstrate the coexistence of zero-energy vortex modes and Majorana corner modes in the model of a 2D second-order topological superconductor. The model describes an interface between a normal layer supporting the topological insulating phase and a superconducting layer, for which different symmetries of the spin-singlet superconducting order parameter are considered. We show that the gapless vortex modes can appear under certain conditions in the superconducting state with a vortex if the bulk energy spectrum of the normal (non-superconducting) state is gapless and has Dirac cones. The number of pairs of such vortex modes corresponds to the number of Dirac cones. It is essential that if the normal bulk spectrum becomes gapped and the system is in the state of a topological insulator, then the zero-energy vortex modes can not be realized, while Majorana corner modes hold in the superconducting state. The interaction of the vortex modes with the edge and topological corner modes is studied when the vortex appears near the boundaries. | [
{
"version": "v1",
"created": "Fri, 22 Nov 2024 10:10:41 GMT"
},
{
"version": "v2",
"created": "Thu, 05 Jun 2025 02:23:18 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Fedoseev",
"A. D.",
""
],
[
"Zlotnikov",
"A. O.",
""
]
] |
2411.15479 | Jie Wang | Jared Miller, Jie Wang, Feng Guo | Sparse Polynomial Matrix Optimization | 31 pages, 9 tables, 3 figures | null | null | null | math.OC | http://creativecommons.org/licenses/by/4.0/ | A polynomial matrix inequality is a formula asserting that a polynomial matrix is positive semidefinite. Polynomial matrix optimization concerns minimizing the smallest eigenvalue of a symmetric polynomial matrix subject to a tuple of polynomial matrix inequalities. This work explores the use of sparsity methods in reducing the complexity of sum-of-squares based methods in verifying polynomial matrix inequalities or solving polynomial matrix optimization. In the unconstrained setting, Newton polytopes can be employed to sparsify the monomial basis, resulting in smaller semidefinite programs. In the general setting, we show how to exploit different types of sparsity (term sparsity, correlative sparsity, matrix sparsity) encoded in polynomial matrices to derive sparse semidefinite programming relaxations for polynomial matrix optimization. For term sparsity, we show that the block structures of the term sparsity iterations with maximal chordal extensions converge to the one determined by PMI sign symmetries. For correlative sparsity, unlike the scalar case, we provide a counterexample showing that asymptotic convergence does not hold under the Archimedean condition and the running intersection property. By employing the theory of matrix-valued measures, we establish several results on detecting global optimality and retrieving optimal solutions under correlative sparsity. The effectiveness of sparsity methods on reducing computational complexity is demonstrated on various examples of polynomial matrix optimization. | [
{
"version": "v1",
"created": "Sat, 23 Nov 2024 07:23:17 GMT"
},
{
"version": "v2",
"created": "Sun, 22 Dec 2024 12:27:14 GMT"
},
{
"version": "v3",
"created": "Thu, 05 Jun 2025 15:14:37 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Miller",
"Jared",
""
],
[
"Wang",
"Jie",
""
],
[
"Guo",
"Feng",
""
]
] |
2411.15850 | Alvaro De Rujula | A. De R\'ujula | A revamped understanding of Cosmic Rays and Gamma-Ray Bursts | 8 pages, 14 figures. Accepted by PLB after usual referees' torture | null | null | null | hep-ph astro-ph.HE | http://creativecommons.org/licenses/by/4.0/ | Interesting data on Gamma Ray Burts (GRBs) and Cosmic Rays (CRs) have recently been made public. GRB221009A has a record ``peak energy". The CR electron spectrum has been measured to unprecedented high energies and exhibits a ``knee" akin to the ones in all-particle or individual-element CR nuclei. IceCube has not seen high-energy neutrinos associated with GRBs. AMS has published a CR positron spectrum conducive to much speculation. We examine these data in the light of the ``CannonBall Model" of GRBs and CRs, in which they are intimately related and which they do strongly validate. | [
{
"version": "v1",
"created": "Sun, 24 Nov 2024 14:10:57 GMT"
},
{
"version": "v2",
"created": "Sun, 08 Dec 2024 16:55:40 GMT"
},
{
"version": "v3",
"created": "Tue, 03 Jun 2025 12:42:38 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"De Rújula",
"A.",
""
]
] |
2411.16000 | Wayne Lewis | Wayne Lewis | Global Primitive Roots of Unity | 20 pages | null | null | null | math.NT | http://creativecommons.org/licenses/by-nc-nd/4.0/ | A valuation ring defined as an ultraproduct with respect to a particular algebraic nonprincipal ultrafilter on $\wp(\mathbb{P})$ provides a natural setting to exhibit an infinite set of primes $q$ relative to which a non-perfect-power integer $m \neq -1$ is a primitive root of unity $(\!\!\!\!\mod q)$. This application of ultraproducts demonstrates their potential in addressing foundational problems in number theory requiring verification of denumerable conforming sets. | [
{
"version": "v1",
"created": "Sun, 24 Nov 2024 22:44:50 GMT"
},
{
"version": "v2",
"created": "Tue, 26 Nov 2024 21:19:42 GMT"
},
{
"version": "v3",
"created": "Fri, 06 Dec 2024 04:11:00 GMT"
},
{
"version": "v4",
"created": "Fri, 27 Dec 2024 06:06:23 GMT"
},
{
"version": "v5",
"created": "Sun, 05 Jan 2025 08:17:33 GMT"
},
{
"version": "v6",
"created": "Tue, 18 Feb 2025 02:10:40 GMT"
},
{
"version": "v7",
"created": "Mon, 24 Feb 2025 11:03:27 GMT"
},
{
"version": "v8",
"created": "Wed, 26 Mar 2025 13:41:32 GMT"
},
{
"version": "v9",
"created": "Wed, 09 Apr 2025 06:44:30 GMT"
},
{
"version": "v10",
"created": "Wed, 16 Apr 2025 02:07:35 GMT"
},
{
"version": "v11",
"created": "Sun, 27 Apr 2025 01:24:07 GMT"
},
{
"version": "v12",
"created": "Mon, 05 May 2025 08:39:40 GMT"
},
{
"version": "v13",
"created": "Tue, 13 May 2025 07:37:26 GMT"
},
{
"version": "v14",
"created": "Tue, 20 May 2025 11:28:08 GMT"
},
{
"version": "v15",
"created": "Thu, 29 May 2025 10:40:33 GMT"
},
{
"version": "v16",
"created": "Wed, 04 Jun 2025 18:28:20 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Lewis",
"Wayne",
""
]
] |
2411.16331 | Xiaobin Hu | Xiaozhong Ji, Xiaobin Hu, Zhihong Xu, Junwei Zhu, Chuming Lin, Qingdong He, Jiangning Zhang, Donghao Luo, Yi Chen, Qin Lin, Qinglin Lu, Chengjie Wang | Sonic: Shifting Focus to Global Audio Perception in Portrait Animation | refer to our main-page \url{https://jixiaozhong.github.io/Sonic/} | null | null | null | cs.MM cs.CV cs.GR cs.SD eess.AS | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | The study of talking face generation mainly explores the intricacies of synchronizing facial movements and crafting visually appealing, temporally-coherent animations. However, due to the limited exploration of global audio perception, current approaches predominantly employ auxiliary visual and spatial knowledge to stabilize the movements, which often results in the deterioration of the naturalness and temporal inconsistencies.Considering the essence of audio-driven animation, the audio signal serves as the ideal and unique priors to adjust facial expressions and lip movements, without resorting to interference of any visual signals. Based on this motivation, we propose a novel paradigm, dubbed as Sonic, to {s}hift f{o}cus on the exploration of global audio per{c}ept{i}o{n}.To effectively leverage global audio knowledge, we disentangle it into intra- and inter-clip audio perception and collaborate with both aspects to enhance overall perception.For the intra-clip audio perception, 1). \textbf{Context-enhanced audio learning}, in which long-range intra-clip temporal audio knowledge is extracted to provide facial expression and lip motion priors implicitly expressed as the tone and speed of speech. 2). \textbf{Motion-decoupled controller}, in which the motion of the head and expression movement are disentangled and independently controlled by intra-audio clips. Most importantly, for inter-clip audio perception, as a bridge to connect the intra-clips to achieve the global perception, \textbf{Time-aware position shift fusion}, in which the global inter-clip audio information is considered and fused for long-audio inference via through consecutively time-aware shifted windows. Extensive experiments demonstrate that the novel audio-driven paradigm outperform existing SOTA methodologies in terms of video quality, temporally consistency, lip synchronization precision, and motion diversity. | [
{
"version": "v1",
"created": "Mon, 25 Nov 2024 12:24:52 GMT"
},
{
"version": "v2",
"created": "Wed, 04 Jun 2025 07:50:44 GMT"
},
{
"version": "v3",
"created": "Thu, 05 Jun 2025 11:49:59 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Ji",
"Xiaozhong",
""
],
[
"Hu",
"Xiaobin",
""
],
[
"Xu",
"Zhihong",
""
],
[
"Zhu",
"Junwei",
""
],
[
"Lin",
"Chuming",
""
],
[
"He",
"Qingdong",
""
],
[
"Zhang",
"Jiangning",
""
],
[
"Luo",
"Donghao",
""
],
[
"Chen",
"Yi",
""
],
[
"Lin",
"Qin",
""
],
[
"Lu",
"Qinglin",
""
],
[
"Wang",
"Chengjie",
""
]
] |
2411.16699 | Pawe{\l} Pilarczyk | Miko{\l}aj Rosman, Micha{\l} Palczewski, Pawe{\l} Pilarczyk, Agnieszka Bart{\l}omiejczyk | Bistability and chaos in the discrete two-gene Andrecut-Kauffman model | 20 pages, 11 figures | Discrete and Continuous Dynamical Systems - Series B, Vol. 30, No. 11, November 2025, pp. 4442-4461 | 10.3934/dcdsb.2025028 | null | nlin.CD math.DS | http://creativecommons.org/licenses/by-nc-nd/4.0/ | We conduct numerical analysis of the 2-dimensional discrete-time gene expression model originally introduced by Andrecut and Kauffman (Phys. Lett. A 367: 281-287, 2007). In contrast to the previous studies, we analyze the dynamics with different reaction rates $\alpha_1$ and $\alpha_2$ for each of the two genes under consideration. We explore bifurcation diagrams for the model with $\alpha_1$ varying in a wide range and $\alpha_2$ fixed. We detect chaotic dynamics by means of the positive maximum Lyapunov exponent and we scan through selected parameters to detect those combinations for which chaotic dynamics can be found in the model. Moreover, we find bistability in the model, that is, the existence of two disjoint attractors. Both situations are interesting from the point of view of applications, as they imply unpredictability of the dynamics encountered. Finally, we show some specific values of parameters of the model in which the two attractors are of different kind (a periodic orbit and a chaotic attractor) or of the same kind (two periodic orbits or two chaotic attractors). | [
{
"version": "v1",
"created": "Sun, 10 Nov 2024 20:10:40 GMT"
},
{
"version": "v2",
"created": "Thu, 05 Jun 2025 07:53:11 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Rosman",
"Mikołaj",
""
],
[
"Palczewski",
"Michał",
""
],
[
"Pilarczyk",
"Paweł",
""
],
[
"Bartłomiejczyk",
"Agnieszka",
""
]
] |
2411.18575 | Cristian Pach\'on-Garc\'ia | Pedro Delicado and Cristian Pach\'on-Garc\'ia | Functional relevance based on the continuous Shapley value | 36 pages, 13 figures | null | null | null | stat.ML cs.AI cs.LG stat.AP | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | The presence of artificial intelligence (AI) in our society is increasing, which brings with it the need to understand the behavior of AI mechanisms, including machine learning predictive algorithms fed with tabular data, text or images, among others. This work focuses on interpretability of predictive models based on functional data. Designing interpretability methods for functional data models implies working with a set of features whose size is infinite. In the context of scalar on function regression, we propose an interpretability method based on the Shapley value for continuous games, a mathematical formulation that allows for the fair distribution of a global payoff among a continuous set of players. The method is illustrated through a set of experiments with simulated and real data sets. The open source Python package ShapleyFDA is also presented. | [
{
"version": "v1",
"created": "Wed, 27 Nov 2024 18:20:00 GMT"
},
{
"version": "v2",
"created": "Wed, 04 Jun 2025 18:42:38 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Delicado",
"Pedro",
""
],
[
"Pachón-García",
"Cristian",
""
]
] |
2412.00193 | Amir-Reza Negari | Amir-Reza Negari and Tyler D. Ellison and Timothy H. Hsieh | Spacetime Markov length: a diagnostic for fault tolerance via mixed-state phases | null | null | null | null | quant-ph cond-mat.stat-mech cond-mat.str-el | http://creativecommons.org/licenses/by/4.0/ | We establish a correspondence between the fault-tolerance of local stabilizer codes experiencing measurement and physical errors and the mixed-state phases of decohered resource states in one higher dimension. Drawing from recent developments in mixed-state phases of matter, this motivates a diagnostic of fault-tolerance, which we refer to as the spacetime Markov length. This is a length scale determined by the decay of the (classical) conditional mutual information of repeated syndrome measurement outcomes in spacetime. The diagnostic is independent of the decoder, and its divergence signals the intrinsic breakdown of fault tolerance. As a byproduct, we find that decoherence may be useful for exposing transitions from higher-form symmetry-protected topological phases driven by both incoherent and coherent perturbations. | [
{
"version": "v1",
"created": "Fri, 29 Nov 2024 19:00:03 GMT"
},
{
"version": "v2",
"created": "Thu, 05 Jun 2025 14:50:47 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Negari",
"Amir-Reza",
""
],
[
"Ellison",
"Tyler D.",
""
],
[
"Hsieh",
"Timothy H.",
""
]
] |
2412.00908 | A. P. Saiko | A.P. Saiko, G.A. Rusetsky, S.A. Markevich, R. Fedaruk | Optomechanical systems with nonlinear interactions: photon blockade, collapse-revival effect and Fano-like resonance | 10 pages, 9 figures | null | null | null | quant-ph cond-mat.mes-hall | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Closed-form expressions for the average amplitude of the optical field in optomechanical systems are obtained, in which, in addition to the linear interaction, quadratic and cubic interactions of the vibrational mode of the mechanical resonator with the mode of the optical resonator are considered. In the framework of the non-secular perturbation theory, using the Bogoliubov averaging method, it is shown that the effects of photon blockade, collapse and revival of optical oscillations in such systems can be realized. The main contribution to the formation of revivals is provided by the Kerr self-action of the optical mode and the cross-Kerr interaction of the fourth degree in optical and mechanical amplitudes. The cross-Kerr interactions of the sixth- and eighth-order in amplitudes destroy the regular structure of revivals. The influence of these cross-Kerr nonlinearities disappears with an increase in the decay rate of the optical mode and is also completely suppressed at zero temperature. It is shown that the asymmetry of the spectral line of the optical field intensity in the cavity is most pronounced with an increase in the degree of nonlinearity and is explained by Fano interference. | [
{
"version": "v1",
"created": "Sun, 01 Dec 2024 17:50:04 GMT"
},
{
"version": "v2",
"created": "Thu, 12 Dec 2024 13:59:29 GMT"
},
{
"version": "v3",
"created": "Thu, 05 Jun 2025 13:51:58 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Saiko",
"A. P.",
""
],
[
"Rusetsky",
"G. A.",
""
],
[
"Markevich",
"S. A.",
""
],
[
"Fedaruk",
"R.",
""
]
] |
2412.01271 | Sen Xing | Sen Xing, Muyan Zhong, Zeqiang Lai, Liangchen Li, Jiawen Liu, Yaohui Wang, Jifeng Dai, Wenhai Wang | MuLan: Adapting Multilingual Diffusion Models for Hundreds of Languages with Negligible Cost | null | null | null | null | cs.CL cs.AI | http://creativecommons.org/licenses/by/4.0/ | In this work, we explore a cost-effective framework for multilingual image generation. We find that, unlike models tuned on high-quality images with multilingual annotations, leveraging text encoders pre-trained on widely available, noisy Internet image-text pairs significantly enhances data efficiency in text-to-image (T2I) generation across multiple languages.Based on this insight, we introduce MuLan, Multi-Language adapter, a lightweight language adapter with fewer than 20M parameters, trained alongside a frozen text encoder and image diffusion model. Compared to previous multilingual T2I models, this framework offers: (1) Cost efficiency. Using readily accessible English data and off-the-shelf multilingual text encoders minimizes the training cost; (2) High performance. Achieving comparable generation capabilities in over 110 languages with CLIP similarity scores nearly matching those in English (39.57 for English vs. 39.61 for other languages); and (3) Broad applicability. Seamlessly integrating with compatible community tools like LoRA, LCM, ControlNet, and IP-Adapter, expanding its potential use cases. | [
{
"version": "v1",
"created": "Mon, 02 Dec 2024 08:38:19 GMT"
},
{
"version": "v2",
"created": "Thu, 05 Jun 2025 10:34:56 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Xing",
"Sen",
""
],
[
"Zhong",
"Muyan",
""
],
[
"Lai",
"Zeqiang",
""
],
[
"Li",
"Liangchen",
""
],
[
"Liu",
"Jiawen",
""
],
[
"Wang",
"Yaohui",
""
],
[
"Dai",
"Jifeng",
""
],
[
"Wang",
"Wenhai",
""
]
] |
2412.01612 | Kosuke Mizuno | Taiga Adachi, Kosuke Mizuno, Sohei Tateno | Iwasawa theory for weighted graphs | 33 pages, 7 figures | null | null | null | math.NT math.CO | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Let $p$ be a prime number and let $d$ be a positive integer. In this paper, we generalize Iwasawa theory for graphs initiated by Gonet and Valli\`{e}res to weighted graphs. In particular, we prove an analogue of Iwasawa's class number formula and that of Kida's formula for compatible systems of $(\mathbb{Z}/p^n\mathbb{Z})^d$-covers of weighted graphs. We also provide numerical examples of characteristic elements and Iwasawa invariants. At the end of this paper, we give an application of the ideas of Iwasawa theory to the theory of discrete-time quantum walks in graphs. | [
{
"version": "v1",
"created": "Mon, 02 Dec 2024 15:32:08 GMT"
},
{
"version": "v2",
"created": "Fri, 27 Dec 2024 15:59:11 GMT"
},
{
"version": "v3",
"created": "Thu, 05 Jun 2025 14:58:52 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Adachi",
"Taiga",
""
],
[
"Mizuno",
"Kosuke",
""
],
[
"Tateno",
"Sohei",
""
]
] |
2412.03782 | Andrew Lampinen | Andrew Kyle Lampinen, Stephanie C. Y. Chan, Aaditya K. Singh, Murray Shanahan | The broader spectrum of in-context learning | null | null | null | null | cs.CL cs.LG | http://creativecommons.org/licenses/by/4.0/ | The ability of language models to learn a task from a few examples in context has generated substantial interest. Here, we provide a perspective that situates this type of supervised few-shot learning within a much broader spectrum of meta-learned in-context learning. Indeed, we suggest that any distribution of sequences in which context non-trivially decreases loss on subsequent predictions can be interpreted as eliciting a kind of in-context learning. We suggest that this perspective helps to unify the broad set of in-context abilities that language models exhibit -- such as adapting to tasks from instructions or role play, or extrapolating time series. This perspective also sheds light on potential roots of in-context learning in lower-level processing of linguistic dependencies (e.g. coreference or parallel structures). Finally, taking this perspective highlights the importance of generalization, which we suggest can be studied along several dimensions: not only the ability to learn something novel, but also flexibility in learning from different presentations, and in applying what is learned. We discuss broader connections to past literature in meta-learning and goal-conditioned agents, and other perspectives on learning and adaptation. We close by suggesting that research on in-context learning should consider this broader spectrum of in-context capabilities and types of generalization. | [
{
"version": "v1",
"created": "Thu, 05 Dec 2024 00:05:11 GMT"
},
{
"version": "v2",
"created": "Mon, 09 Dec 2024 18:28:06 GMT"
},
{
"version": "v3",
"created": "Thu, 05 Jun 2025 17:58:57 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Lampinen",
"Andrew Kyle",
""
],
[
"Chan",
"Stephanie C. Y.",
""
],
[
"Singh",
"Aaditya K.",
""
],
[
"Shanahan",
"Murray",
""
]
] |
2412.04119 | Cristian-George Craciun | Cristian-George Cr\u{a}ciun, R\u{a}zvan-Alexandru Sm\u{a}du, Dumitru-Clementin Cercel, Mihaela-Claudia Cercel | GRAF: Graph Retrieval Augmented by Facts for Romanian Legal Multi-Choice Question Answering | Accepted to ACL 2025 Findings | null | null | null | cs.CL | http://creativecommons.org/licenses/by-nc-sa/4.0/ | Pre-trained Language Models (PLMs) have shown remarkable performances in recent years, setting a new paradigm for NLP research and industry. The legal domain has received some attention from the NLP community partly due to its textual nature. Some tasks from this domain are represented by question-answering (QA) tasks. This work explores the legal domain Multiple-Choice QA (MCQA) for a low-resource language. The contribution of this work is multi-fold. We first introduce JuRO, the first openly available Romanian legal MCQA dataset, comprising three different examinations and a number of 10,836 total questions. Along with this dataset, we introduce CROL, an organized corpus of laws that has a total of 93 distinct documents with their modifications from 763 time spans, that we leveraged in this work for Information Retrieval (IR) techniques. Moreover, we are the first to propose Law-RoG, a Knowledge Graph (KG) for the Romanian language, and this KG is derived from the aforementioned corpus. Lastly, we propose a novel approach for MCQA, Graph Retrieval Augmented by Facts (GRAF), which achieves competitive results with generally accepted SOTA methods and even exceeds them in most settings. | [
{
"version": "v1",
"created": "Thu, 05 Dec 2024 12:37:27 GMT"
},
{
"version": "v2",
"created": "Thu, 19 Dec 2024 21:11:00 GMT"
},
{
"version": "v3",
"created": "Thu, 05 Jun 2025 17:37:25 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Crăciun",
"Cristian-George",
""
],
[
"Smădu",
"Răzvan-Alexandru",
""
],
[
"Cercel",
"Dumitru-Clementin",
""
],
[
"Cercel",
"Mihaela-Claudia",
""
]
] |
2412.04338 | Alexander K. Hartmann | Alexander K. Hartmann | Numerical Aspects of Large Deviations | Lectures notes for lectures given at 2024 Les Houches Summer School on "Large deviations and applications". For C Codes see DARE (Oldenburg University research data repository) at doi.org/10.57782/HXEGVF. Revised version V02 contains now detailed overview over paper, section on needed numerical resources and overview over other approaches, plus small changes | null | null | null | physics.comp-ph physics.data-an | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | An introduction to numerical large-deviation sampling is provided. First, direct biasing with a known distribution is explained. As simple example, the Bernoulli experiment is used throughout the text. Next, Markov chain Monte Carlo (MCMC) simulations are introduced. In particular, the Metropolis-Hastings algorithm is explained. As first implementation of MCMC, sampling of the plain Bernoulli model is shown. Next, an exponential bias is used for the same model, which allows one to obtain the tails of the distribution of a measurable quantity. This approach is generalized to MCMC simulations, where the states are vectors of $U(0,1)$ random entries. This allows one to use the exponential or any other bias to access the large-deviation properties of rather arbitrary random processes. Finally, some recent research applications to study more complex models are discussed. | [
{
"version": "v1",
"created": "Thu, 05 Dec 2024 16:56:51 GMT"
},
{
"version": "v2",
"created": "Thu, 05 Jun 2025 16:29:17 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Hartmann",
"Alexander K.",
""
]
] |
2412.04954 | Xi Zhang | Xi Zhang, Zaiqiao Meng, Jake Lever, Edmond S. L. Ho | Gla-AI4BioMed at RRG24: Visual Instruction-tuned Adaptation for Radiology Report Generation | Accepted by BioNLP@ACL 2024 | Proceedings of the 23rd Workshop on Biomedical Natural Language Processing (BioNLP 2024) 2024 | 10.18653/v1/2024.bionlp-1.54 | null | cs.CV cs.CL cs.LG | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We introduce a radiology-focused visual language model designed to generate radiology reports from chest X-rays. Building on previous findings that large language models (LLMs) can acquire multimodal capabilities when aligned with pretrained vision encoders, we demonstrate similar potential with chest X-ray images. This integration enhances the ability of model to understand and describe chest X-ray images. Our model combines an image encoder with a fine-tuned LLM based on the Vicuna-7B architecture, enabling it to generate different sections of a radiology report with notable accuracy. The training process involves a two-stage approach: (i) initial alignment of chest X-ray features with the LLM (ii) followed by fine-tuning for radiology report generation. | [
{
"version": "v1",
"created": "Fri, 06 Dec 2024 11:14:03 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Zhang",
"Xi",
""
],
[
"Meng",
"Zaiqiao",
""
],
[
"Lever",
"Jake",
""
],
[
"Ho",
"Edmond S. L.",
""
]
] |
2412.05219 | Liam Scanlon | Liam J. Scanlon, Santosh Bhusal, Christina M. Hoffmann, Junhong He, Sean R. Parkin, Brennan J. Arnold, William J. Gannon | Structural characterization of the candidate Weyl semimetal CeGaGe | null | Physical Review B. 111 (2025) 184102 | 10.1103/PhysRevB.111.184102 | null | cond-mat.str-el cond-mat.mtrl-sci | http://creativecommons.org/licenses/by-nc-sa/4.0/ | Weyl semimetals have a variety of intriguing physical properties, including topologically protected electronic states that coexist with conducting states. Possible exploitation of topologically protected states in a conducting material is promising for technological applications. Weyl semimetals that form in a noncentrosymmetric structure that also contain magnetic moments may host a variety of emergent phenomena that cannot be seen in magnetic, centrosymmetric Weyl materials. It can be difficult to distinguish definitively between a centrosymmetric structure and one of its noncentrosymmetric subgroups with standard powder X-ray diffractometers in cases where two atoms in the compound have nearly the same atomic number, as is the case for the candidate Weyl semimetal CeGaGe. In these cases, a careful single-crystal neutron diffraction experiment with high-angle reflections provides complimentary information to X-ray diffraction and definitively resolves any ambiguity between centrosymmetric and noncentrosymmetric crystal structures. Single-crystal neutron diffraction measurements on the candidate Weyl semimetal CeGaGe confirm that its structure is noncentrosymmetric, described by space group 109 $\left(I4_1md\right)$ rather than the centrosymmetric space group 141 $\left(I4_1/amd\right)$. There are many high-angle reflections in the data set that give clear, physically intuitive evidence that CeGaGe forms with $I4_1md$ symmetry since Bragg planes of these reflections can contain Ga with no Ge or vice versa, whereas the Bragg planes for a structure with $I4_1/amd$ symmetry would have a mix of Ga and Ge. Further, in some crystals we have studied, there is clear evidence for a structural transition from body-centered $I4_1md$ symmetry to primitive $P4_3$ and/or $P4_1$ symmetry. | [
{
"version": "v1",
"created": "Fri, 06 Dec 2024 17:50:52 GMT"
},
{
"version": "v2",
"created": "Tue, 27 May 2025 17:41:29 GMT"
},
{
"version": "v3",
"created": "Wed, 04 Jun 2025 18:05:37 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Scanlon",
"Liam J.",
""
],
[
"Bhusal",
"Santosh",
""
],
[
"Hoffmann",
"Christina M.",
""
],
[
"He",
"Junhong",
""
],
[
"Parkin",
"Sean R.",
""
],
[
"Arnold",
"Brennan J.",
""
],
[
"Gannon",
"William J.",
""
]
] |
2412.08151 | Junsup Shim | Junsup Shim, Teppei Okumura, Atsushi Taruya | Improving cosmological constraints via galaxy intrinsic alignment in full-shape analysis | Accepted for publication in PRD. 21 pages, 10 figures, 3 tables | null | 10.1103/2k9z-299n | YITP-24-170 | astro-ph.CO | http://creativecommons.org/licenses/by-nc-nd/4.0/ | The intrinsic alignment (IA) of galaxy shapes probes the underlying gravitational tidal field, thus offering cosmological information complementary to galaxy clustering. In this paper, we perform a Fisher forecast to assess the benefit of IA in improving cosmological parameter constraints, for the first time, leveraging the full-shape (FS) information of IA statistics. Our forecast is based on PFS-like and Euclid-like surveys as examples of deep and wide galaxy surveys, respectively. We explore various cosmological models, with the most comprehensive one simultaneously including dynamical dark energy, curvature, massive neutrinos, and modified gravity (MG). We find that adding FS IA information significantly tightens cosmological constraints relative to the FS clustering-only cases, particularly for dynamical dark energy and nonflat-MG models. For a deep galaxy survey, the Figure-of-Merit for the dark energy equation of state parameters is improved by at least more than $40\%$ in all dynamical dark energy models investigated. For nonflat-MG models, parameter constraints are tightened by $6-28\%$, except for the dark matter density and spectral index parameters. For a wide galaxy survey, improvements with IA become milder, although its joint constraints are tighter than those from the deep survey. Our findings highlight the efficacy of the galaxy IA as a complementary cosmological probe to galaxy clustering. | [
{
"version": "v1",
"created": "Wed, 11 Dec 2024 07:11:41 GMT"
},
{
"version": "v2",
"created": "Thu, 05 Jun 2025 09:13:24 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Shim",
"Junsup",
""
],
[
"Okumura",
"Teppei",
""
],
[
"Taruya",
"Atsushi",
""
]
] |
2412.08435 | Lifan Zhao | Lifan Zhao, Yanyan Shen | Proactive Model Adaptation Against Concept Drift for Online Time Series Forecasting | Accepted by KDD 2025. This version fixed typos in Eq. (3) | null | 10.1145/3690624.3709210 | null | cs.LG cs.AI cs.CE stat.ML | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Time series forecasting always faces the challenge of concept drift, where data distributions evolve over time, leading to a decline in forecast model performance. Existing solutions are based on online learning, which continually organize recent time series observations as new training samples and update model parameters according to the forecasting feedback on recent data. However, they overlook a critical issue: obtaining ground-truth future values of each sample should be delayed until after the forecast horizon. This delay creates a temporal gap between the training samples and the test sample. Our empirical analysis reveals that the gap can introduce concept drift, causing forecast models to adapt to outdated concepts. In this paper, we present Proceed, a novel proactive model adaptation framework for online time series forecasting. Proceed first estimates the concept drift between the recently used training samples and the current test sample. It then employs an adaptation generator to efficiently translate the estimated drift into parameter adjustments, proactively adapting the model to the test sample. To enhance the generalization capability of the framework, Proceed is trained on synthetic diverse concept drifts. Extensive experiments on five real-world datasets across various forecast models demonstrate that Proceed brings more performance improvements than the state-of-the-art online learning methods, significantly facilitating forecast models' resilience against concept drifts. Code is available at https://github.com/SJTU-DMTai/OnlineTSF. | [
{
"version": "v1",
"created": "Wed, 11 Dec 2024 14:57:10 GMT"
},
{
"version": "v2",
"created": "Mon, 16 Dec 2024 15:02:52 GMT"
},
{
"version": "v3",
"created": "Fri, 07 Feb 2025 13:54:59 GMT"
},
{
"version": "v4",
"created": "Thu, 05 Jun 2025 04:25:32 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Zhao",
"Lifan",
""
],
[
"Shen",
"Yanyan",
""
]
] |
2412.08693 | Martin Sasieta | Javier M. Magan, Martin Sasieta and Brian Swingle | Random Circuits in the Black Hole Interior | 56 pages + appendices. v3: journal version, minor changes | null | null | null | hep-th quant-ph | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | In this paper, we present a quantitative holographic relation between a microscopic measure of randomness and the geometric length of the wormhole in the black hole interior. To this end, we perturb an AdS black hole with Brownian semiclassical sources, implementing the continuous version of a random quantum circuit for the black hole. We use the random circuit to prepare ensembles of states of the black hole whose semiclassical duals contain Einstein-Rosen (ER) caterpillars: long cylindrical wormholes with large numbers of matter inhomogeneities, of linearly growing length with the circuit time. In this setup, we show semiclassically that the ensemble of ER caterpillars of average length $k\ell_{\Delta}$ and matter correlation scale $\ell_{\Delta}$ forms an approximate quantum state $k$-design of the black hole. At exponentially long circuit times, the ensemble of ER caterpillars becomes polynomial-copy indistinguishable from a collection of random states of the black hole. We comment on the implications of these results for holographic circuit complexity and for the holographic description of the black hole interior. | [
{
"version": "v1",
"created": "Wed, 11 Dec 2024 19:00:00 GMT"
},
{
"version": "v2",
"created": "Mon, 23 Dec 2024 14:24:51 GMT"
},
{
"version": "v3",
"created": "Thu, 05 Jun 2025 13:14:42 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Magan",
"Javier M.",
""
],
[
"Sasieta",
"Martin",
""
],
[
"Swingle",
"Brian",
""
]
] |
2412.08794 | Prajwal Koirala | Prajwal Koirala, Zhanhong Jiang, Soumik Sarkar, Cody Fleming | Latent Safety-Constrained Policy Approach for Safe Offline Reinforcement Learning | null | null | null | null | cs.LG stat.ML | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | In safe offline reinforcement learning (RL), the objective is to develop a policy that maximizes cumulative rewards while strictly adhering to safety constraints, utilizing only offline data. Traditional methods often face difficulties in balancing these constraints, leading to either diminished performance or increased safety risks. We address these issues with a novel approach that begins by learning a conservatively safe policy through the use of Conditional Variational Autoencoders, which model the latent safety constraints. Subsequently, we frame this as a Constrained Reward-Return Maximization problem, wherein the policy aims to optimize rewards while complying with the inferred latent safety constraints. This is achieved by training an encoder with a reward-Advantage Weighted Regression objective within the latent constraint space. Our methodology is supported by theoretical analysis, including bounds on policy performance and sample complexity. Extensive empirical evaluation on benchmark datasets, including challenging autonomous driving scenarios, demonstrates that our approach not only maintains safety compliance but also excels in cumulative reward optimization, surpassing existing methods. Additional visualizations provide further insights into the effectiveness and underlying mechanisms of our approach. | [
{
"version": "v1",
"created": "Wed, 11 Dec 2024 22:00:07 GMT"
},
{
"version": "v2",
"created": "Thu, 05 Jun 2025 06:04:43 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Koirala",
"Prajwal",
""
],
[
"Jiang",
"Zhanhong",
""
],
[
"Sarkar",
"Soumik",
""
],
[
"Fleming",
"Cody",
""
]
] |
2412.09429 | Yi Luo | Yi Luo, Linghang Shi, Yihao Li, Aobo Zhuang, Yeyun Gong, Ling Liu, Chen Lin | From Intention To Implementation: Automating Biomedical Research via LLMs | To appear in SCIENCE CHINA Information Sciences. If you find our work useful, please cite us as: @article{ BioResearcher, author = "Yi Luo and Linghang Shi and Yihao Li and Aobo Zhuang and Yeyun Gong and Ling Liu and Chen Lin", title = "From Intention To Implementation: Automating Biomedical Research via LLMs", journal = "SCIENCE CHINA Information Sciences", year = "2025" } | null | null | null | cs.MA cs.AI cs.CL | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Conventional biomedical research is increasingly labor-intensive due to the exponential growth of scientific literature and datasets. Artificial intelligence (AI), particularly Large Language Models (LLMs), has the potential to revolutionize this process by automating various steps. Still, significant challenges remain, including the need for multidisciplinary expertise, logicality of experimental design, and performance measurements. This paper introduces BioResearcher, the first end-to-end automated system designed to streamline the entire biomedical research process involving dry lab experiments. BioResearcher employs a modular multi-agent architecture, integrating specialized agents for search, literature processing, experimental design, and programming. By decomposing complex tasks into logically related sub-tasks and utilizing a hierarchical learning approach, BioResearcher effectively addresses the challenges of multidisciplinary requirements and logical complexity. Furthermore, BioResearcher incorporates an LLM-based reviewer for in-process quality control and introduces novel evaluation metrics to assess the quality and automation of experimental protocols. BioResearcher successfully achieves an average execution success rate of 63.07% across eight previously unmet research objectives. The generated protocols, on average, outperform typical agent systems by 22.0% on five quality metrics. The system demonstrates significant potential to reduce researchers' workloads and accelerate biomedical discoveries, paving the way for future innovations in automated research systems. | [
{
"version": "v1",
"created": "Thu, 12 Dec 2024 16:35:05 GMT"
},
{
"version": "v2",
"created": "Sun, 22 Dec 2024 05:34:46 GMT"
},
{
"version": "v3",
"created": "Wed, 04 Jun 2025 06:48:06 GMT"
},
{
"version": "v4",
"created": "Thu, 05 Jun 2025 05:44:18 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Luo",
"Yi",
""
],
[
"Shi",
"Linghang",
""
],
[
"Li",
"Yihao",
""
],
[
"Zhuang",
"Aobo",
""
],
[
"Gong",
"Yeyun",
""
],
[
"Liu",
"Ling",
""
],
[
"Lin",
"Chen",
""
]
] |
2412.10510 | Tobias Braun | Tobias Braun, Mark Rothermel, Marcus Rohrbach, Anna Rohrbach | DEFAME: Dynamic Evidence-based FAct-checking with Multimodal Experts | null | null | null | null | cs.CV cs.CL | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | The proliferation of disinformation demands reliable and scalable fact-checking solutions. We present Dynamic Evidence-based FAct-checking with Multimodal Experts (DEFAME), a modular, zero-shot MLLM pipeline for open-domain, text-image claim verification. DEFAME operates in a six-stage process, dynamically selecting the tools and search depth to extract and evaluate textual and visual evidence. Unlike prior approaches that are text-only, lack explainability, or rely solely on parametric knowledge, DEFAME performs end-to-end verification, accounting for images in claims and evidence while generating structured, multimodal reports. Evaluation on the popular benchmarks VERITE, AVerITeC, and MOCHEG shows that DEFAME surpasses all previous methods, establishing itself as the new state-of-the-art fact-checking system for uni- and multimodal fact-checking. Moreover, we introduce a new multimodal benchmark, ClaimReview2024+, featuring claims after the knowledge cutoff of GPT-4o, avoiding data leakage. Here, DEFAME drastically outperforms the GPT-4o baselines, showing temporal generalizability and the potential for real-time fact-checking. | [
{
"version": "v1",
"created": "Fri, 13 Dec 2024 19:11:18 GMT"
},
{
"version": "v2",
"created": "Thu, 06 Feb 2025 13:27:38 GMT"
},
{
"version": "v3",
"created": "Thu, 05 Jun 2025 17:10:20 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Braun",
"Tobias",
""
],
[
"Rothermel",
"Mark",
""
],
[
"Rohrbach",
"Marcus",
""
],
[
"Rohrbach",
"Anna",
""
]
] |
2412.11224 | Shrisha Bharadwaj | Shrisha Bharadwaj, Haiwen Feng, Giorgio Becherini, Victoria Abrevaya, Michael J. Black | GenLit: Reformulating Single-Image Relighting as Video Generation | null | null | null | null | cs.CV cs.GR | http://creativecommons.org/licenses/by/4.0/ | Manipulating the illumination of a 3D scene within a single image represents a fundamental challenge in computer vision and graphics. This problem has traditionally been addressed using inverse rendering techniques, which involve explicit 3D asset reconstruction and costly ray-tracing simulations. Meanwhile, recent advancements in visual foundation models suggest that a new paradigm could soon be possible -- one that replaces explicit physical models with networks that are trained on large amounts of image and video data. In this paper, we exploit the physical world understanding of a video diffusion model, particularly Stable Video Diffusion, to relight a single image. We introduce GenLit, a framework that distills the ability of a graphics engine to perform light manipulation into a video-generation model, enabling users to directly insert and manipulate a point light in the 3D world within a given image, and generate results directly as a video sequence. We find that a model fine-tuned on only a small synthetic dataset generalizes to real-world scenes, enabling single-image relighting with plausible and convincing shadows. Our results highlight the ability of video foundation models to capture rich information about lighting, material, and, shape and our findings indicate that such models, with minimal training, can be used to perform relighting without explicit asset reconstruction or complex ray tracing. | [
{
"version": "v1",
"created": "Sun, 15 Dec 2024 15:40:40 GMT"
},
{
"version": "v2",
"created": "Thu, 05 Jun 2025 12:36:04 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Bharadwaj",
"Shrisha",
""
],
[
"Feng",
"Haiwen",
""
],
[
"Becherini",
"Giorgio",
""
],
[
"Abrevaya",
"Victoria",
""
],
[
"Black",
"Michael J.",
""
]
] |
2412.11901 | Zixiang Xu | Gennian Ge and Zixiang Xu and Chi Hoi Yip and Shengtong Zhang and Xiaochen Zhao | The Frankl-Pach upper bound is not tight for any uniformity | 7 pages | J. Combin. Theory Ser. A 217 (2026), Paper No. 106078, 9pp | 10.1016/j.jcta.2025.106078 | null | math.CO | http://creativecommons.org/licenses/by/4.0/ | For any positive integers $n\ge d+1\ge 3$, what is the maximum size of a $(d+1)$-uniform set system in $[n]$ with VC-dimension at most $d$? In 1984, Frankl and Pach initiated the study of this fundamental problem and provided an upper bound $\binom{n}{d}$ via an elegant algebraic proof. Surprisingly, in 2007, Mubayi and Zhao showed that when $n$ is sufficiently large and $d$ is a prime power, the Frankl-Pach upper bound is not tight. They also remarked that their method requires $d$ to be a prime power, and asked for new ideas to improve the Frankl-Pach upper bound without extra assumptions on $n$ and $d$.
In this paper, we provide an improvement for any $d\ge 2$ and $n\ge 2d+2$, which demonstrates that the long-standing Frankl-Pach upper bound $\binom{n}{d}$ is not tight for any uniformity. Our proof combines a simple yet powerful polynomial method and structural analysis. | [
{
"version": "v1",
"created": "Mon, 16 Dec 2024 15:46:31 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Ge",
"Gennian",
""
],
[
"Xu",
"Zixiang",
""
],
[
"Yip",
"Chi Hoi",
""
],
[
"Zhang",
"Shengtong",
""
],
[
"Zhao",
"Xiaochen",
""
]
] |
2412.12106 | Andreu Ball\'us Santacana | Andreu Ballus Santacana | The Syncytial Mesh Model: A Biophysical Framework for Scale-Dependent Coherence in the Brain | This manuscript was prepared with assistance from AI tools for language refinement, drafting support, and formula verification. The preprint was originally posted on bioRxiv on 25 November 2024. Available at https://www.biorxiv.org/content/10.1101/2024.11.22.624908v1 | null | null | null | q-bio.NC | http://creativecommons.org/licenses/by/4.0/ | The brain-mesh model introduces a novel three-layered architecture that integrates local and macro-regional connectivity with an underlying, mesh-inspired network layer. This foundational mesh layer, based on metallic mesh structures, spans the entire brain and generates interference patterns, noise, and resonance effects that modulate both local and global neural dynamics. The fused model goes beyond traditional connectivity frameworks by providing a unified explanation for phenomena such as brain-wide phase gradients, stable low-frequency resonance frequencies, and long-range plasticity effects, which are often difficult to explain cohesively within existing models.
In addition to accounting for classical neurobiological observations, such as phase synchrony, functional connectivity fluctuations, and local Hebbian plasticity, the model offers novel insights into less understood phenomena. Specifically, it predicts connectivity-independent phase gradients across non-synaptic regions, harmonic resonance peaks consistent across individuals, and diffuse plasticity driven by global interference patterns, all of which are challenging to explain under current frameworks.
These unique predictions align with partial empirical observations, such as traveling wave dynamics, consistent low-frequency oscillations, and task-induced connectivity shifts, underscoring the model's relevance. Additionally, the brain-mesh model generates testable hypotheses that distinguish it from traditional approaches. This provides a promising framework for future experimental validation and opens new avenues for understanding global brain function. | [
{
"version": "v1",
"created": "Fri, 29 Nov 2024 10:52:48 GMT"
},
{
"version": "v2",
"created": "Thu, 05 Jun 2025 01:56:40 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Santacana",
"Andreu Ballus",
""
]
] |
2412.12186 | Michael Faulkner | Michael F. Faulkner | Emergent electrostatics in planar XY spin models: the bridge connecting topological order with broken $U(1)$ symmetry | 35 pages, 17 figures | null | null | null | cond-mat.stat-mech | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Topological phases have been a central focus of condensed-matter physics for over 50 years. Along with many experimental applications, they have provided much intellectual interest due to their characterization via some form of topological ordering, as opposed to the symmetry-breaking ordering of conventional continuous phase transitions. This distinction is most subtle in the case of the Berezinskii-Kosterlitz-Thouless (BKT) transition as its experimental realizations appear to break U(1) symmetry at low temperature. It also presents two further paradoxes: i) its prototypical short-range interacting planar XY spin model behaves as an emergent long-range interacting electrolyte; ii) its topological ordering is not accompanied by a topological nonergodicity within the BKT picture. This review paper addresses these three interconnected questions. We review a series of papers that demonstrate that U(1) symmetry is indeed broken, but within a broader framework than that traditionally used to characterize broken symmetry. We discuss recovery of this symmetry by breaking velocity-symmetry in a deterministic Markov process. We then expand on a modern field theory of the emergent electrolyte that maps directly from the spin field to an emergent lattice electric field governed by an augmented electrostatic Boltzmann distribution. This local model of electrolyte physics resolves both the short-range-long-range paradox and the question of topological nonergodicity - as in contrast with the BKT picture, it describes global topological defects and their nonergodic freezing by the topological ordering. It also connects the broken U(1) symmetry with the topological ordering, providing a comprehensive framework for broken symmetry at the transition. We introduce long-time topological stability as a measure of topological nonergodicity - within a general framework for weakly broken ergodicity. | [
{
"version": "v1",
"created": "Fri, 13 Dec 2024 14:23:02 GMT"
},
{
"version": "v2",
"created": "Thu, 05 Jun 2025 15:42:35 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Faulkner",
"Michael F.",
""
]
] |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.