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 |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
2502.10897 | Xunchuan Liu | Xunchuan Liu, Tie Liu, Pak-Shing Li, Xiaofeng Mai, Christian Henkel, Paul F. Goldsmith, Sheng-Li Qin, Yan Gong, Xing Lu, Fengwei Xu, Qiuyi Luo, Hong-Li Liu, Tianwei Zhang, Yu Cheng, Yihuan Di, Yuefang Wu, Qilao Gu, Ningyu Tang, Aiyuan Yang, Zhiqiang Shen | Network of velocity-coherent filaments formed by supersonic turbulence in a very-high-velocity HI cloud | Accepted for publication in Nature Astronomy | null | null | null | astro-ph.GA astro-ph.SR | http://creativecommons.org/licenses/by/4.0/ | The warm neutral medium (WNM) was thought to be subsonically/transonically turbulent, and it lacks a network of intertwined filaments that are commonly seen in both molecular clouds and cold neutral medium (CNM). Here, we report HI~21 cm line observations of a very-high-velocity (-330 km s$^{-1}$ $<V_{\rm LSR}<$ -250 km s$^{-1}$) cloud (VHVC), using the Five-hundred-meter Aperture Spherical radio Telescope (FAST) with unprecedented resolution and sensitivity. For the first time, such a VHVC is clearly revealed to be a supersonic WNM system consisting of a network of velocity-coherent HI~filaments. The filaments are in the forms of slim curves, hubs, and webs, distributed in different layers within the position-position-velocity ({\it ppv}) data cube. The entire cloud has skewed log-normal probability distribution of column density and the filaments themselves show asymmetrical radial density profiles, indicating shock compression by supersonic magnetohydrodynamic (MHD) turbulence, as is also confirmed by our MHD simulation (sonic Mach number $M_{\rm s}=3$ and Alfv\'en Mach number $M_{\rm A}=1$). This work suggests that hierarchical filaments can be established by shocks in a low-density WNM, where gravity is negligible, offering a viable pathway to structure formation in the earliest evolutionary phases of the interstellar medium (ISM). | [
{
"version": "v1",
"created": "Sat, 15 Feb 2025 20:26:26 GMT"
},
{
"version": "v2",
"created": "Thu, 05 Jun 2025 15:31:27 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Liu",
"Xunchuan",
""
],
[
"Liu",
"Tie",
""
],
[
"Li",
"Pak-Shing",
""
],
[
"Mai",
"Xiaofeng",
""
],
[
"Henkel",
"Christian",
""
],
[
"Goldsmith",
"Paul F.",
""
],
[
"Qin",
"Sheng-Li",
""
],
[
"Gong",
"Yan",
""
],
[
"Lu",
"Xing",
""
],
[
"Xu",
"Fengwei",
""
],
[
"Luo",
"Qiuyi",
""
],
[
"Liu",
"Hong-Li",
""
],
[
"Zhang",
"Tianwei",
""
],
[
"Cheng",
"Yu",
""
],
[
"Di",
"Yihuan",
""
],
[
"Wu",
"Yuefang",
""
],
[
"Gu",
"Qilao",
""
],
[
"Tang",
"Ningyu",
""
],
[
"Yang",
"Aiyuan",
""
],
[
"Shen",
"Zhiqiang",
""
]
] |
2502.11028 | Prateek Chhikara | Prateek Chhikara | Mind the Confidence Gap: Overconfidence, Calibration, and Distractor Effects in Large Language Models | null | null | null | null | cs.CL cs.AI | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Large Language Models (LLMs) show remarkable proficiency in natural language tasks, yet their frequent overconfidence-misalignment between predicted confidence and true correctness-poses significant risks in critical decision-making applications. We present a comprehensive analysis on calibration in LLMs across nine LLMs and three factual Question-Answering (QA) datasets, systematically comparing standard free-generation settings against structured distractor-augmented prompts. Our evaluation reveals that explicitly incorporating distractors can substantially mitigate miscalibration, achieving relative accuracy improvements up to 460% and ECE reductions up to 90%. Despite general trends, we uncover nuanced findings: large RLHF-tuned models display inherent calibration strengths but can paradoxically suffer increased miscalibration on easier queries, whereas smaller models benefit disproportionately from distractor prompts but remain significantly miscalibrated. Through detailed analyses across question types, we identify persistent calibration failures, particularly in person-based queries. We conclude with concrete recommendations-targeted fine-tuning, structured prompting, and strategic model choice-to ensure reliable, trustworthy LLM deployments. | [
{
"version": "v1",
"created": "Sun, 16 Feb 2025 07:46:09 GMT"
},
{
"version": "v2",
"created": "Thu, 05 Jun 2025 07:14:36 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Chhikara",
"Prateek",
""
]
] |
2502.11260 | Riccardo Zamboni | Riccardo Zamboni, Enrico Brunetti, Marcello Restelli | Scalable Multi-Agent Offline Reinforcement Learning and the Role of Information | null | null | null | null | cs.LG | http://creativecommons.org/licenses/by/4.0/ | Offline Reinforcement Learning (RL) focuses on learning policies solely from a batch of previously collected data. offering the potential to leverage such datasets effectively without the need for costly or risky active exploration. While recent advances in Offline Multi-Agent RL (MARL) have shown promise, most existing methods either rely on large datasets jointly collected by all agents or agent-specific datasets collected independently. The former approach ensures strong performance but raises scalability concerns, while the latter emphasizes scalability at the expense of performance guarantees. In this work, we propose a novel scalable routine for both dataset collection and offline learning. Agents first collect diverse datasets coherently with a pre-specified information-sharing network and subsequently learn coherent localized policies without requiring either full observability or falling back to complete decentralization. We theoretically demonstrate that this structured approach allows a multi-agent extension of the seminal Fitted Q-Iteration (FQI) algorithm to globally converge, in high probability, to near-optimal policies. The convergence is subject to error terms that depend on the informativeness of the shared information. Furthermore, we show how this approach allows to bound the inherent error of the supervised-learning phase of FQI with the mutual information between shared and unshared information. Our algorithm, SCAlable Multi-agent FQI (SCAM-FQI), is then evaluated on a distributed decision-making problem. The empirical results align with our theoretical findings, supporting the effectiveness of SCAM-FQI in achieving a balance between scalability and policy performance. | [
{
"version": "v1",
"created": "Sun, 16 Feb 2025 20:28:42 GMT"
},
{
"version": "v2",
"created": "Thu, 05 Jun 2025 13:43:08 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Zamboni",
"Riccardo",
""
],
[
"Brunetti",
"Enrico",
""
],
[
"Restelli",
"Marcello",
""
]
] |
2502.12171 | Haonan He | Haonan He, Peng Ye, Yuchen Ren, Yuan Yuan, Luyang Zhou, Shucun Ju, Lei Chen | GoRA: Gradient-driven Adaptive Low Rank Adaptation | null | null | null | null | cs.LG cs.AI cs.CL | http://creativecommons.org/licenses/by/4.0/ | Low-Rank Adaptation (LoRA) is a crucial method for efficiently fine-tuning large language models (LLMs), with its effectiveness influenced by two key factors: rank selection and weight initialization. While numerous LoRA variants have been proposed to improve performance by addressing one of these aspects, they often compromise usability or computational efficiency. In this paper, we analyze and identify the core limitations of existing approaches and propose a novel framework -- GoRA (Gradient-driven Adaptive Low Rank Adaptation) -- that simultaneously adapts both the rank and initialization strategy within a unified framework. GoRA leverages gradient information during training to dynamically assign optimal ranks and initialize low-rank adapter weights in an adaptive manner. To our knowledge, GoRA is the first method that not only addresses the limitations of prior approaches -- which often focus on either rank selection or initialization in isolation -- but also unifies both aspects within a single framework, enabling more effective and efficient adaptation. Extensive experiments across various architectures and modalities show that GoRA consistently outperforms existing LoRA-based methods while preserving the efficiency of vanilla LoRA. For example, when fine-tuning Llama3.1-8B-Base for mathematical reasoning, GoRA achieves a 5.13-point improvement over standard LoRA and even outperforms full fine-tuning by 2.05 points under high-rank settings. | [
{
"version": "v1",
"created": "Thu, 13 Feb 2025 10:33:58 GMT"
},
{
"version": "v2",
"created": "Thu, 05 Jun 2025 14:16:46 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"He",
"Haonan",
""
],
[
"Ye",
"Peng",
""
],
[
"Ren",
"Yuchen",
""
],
[
"Yuan",
"Yuan",
""
],
[
"Zhou",
"Luyang",
""
],
[
"Ju",
"Shucun",
""
],
[
"Chen",
"Lei",
""
]
] |
2502.12316 | Igor Rogachevskii | Nathan Kleeorin and Igor Rogachevskii | Large-scale clustering of inertial particles in a rotating, stratified and inhomogeneous turbulence | 18 pages, revtex4-2, revised, final version | null | null | NORDITA-2025-013 | physics.flu-dyn astro-ph.EP astro-ph.SR physics.geo-ph | http://creativecommons.org/licenses/by/4.0/ | We develop a theory of various kinds of large-scale clustering of inertial particles in a rotating density stratified or inhomogeneous turbulent fluid flows. The large-scale particle clustering occurs in scales which are much larger than the integral scale of turbulence, and it is described in terms of the effective pumping velocity in a turbulent flux of particles. We show that for a fast rotating strongly anisotropic turbulence, the large-scale clustering occurs in the plane perpendicular to rotation axis in the direction of the fluid density stratification. We apply the theory of the large-scale particle clustering for explanation of the formation of planetesimals (progenitors of planets) in accretion protoplanetary discs. We determine the radial profiles of the radial and azimuthal components of the effective pumping velocity of particles which have two maxima corresponding to different regimes of the particle--fluid interactions: at the small radius it is the Stokes regime, while at the larger radius it is the Epstein regime. With the decrease the particle radius, the distance between the maxima increases. This implies that smaller-size particles are concentrated nearby the central body of the accretion disk, while larger-size particles are accumulated far from the central body. The dynamic time of the particle clustering is about $\tau_{\rm dyn} \sim 10^5$--$10^6$ years, while the turbulent diffusion time is about $10^7$ years, that is much larger than the characteristic formation time of large-scale particle clusters ($\sim \tau_{\rm dyn}$). | [
{
"version": "v1",
"created": "Mon, 17 Feb 2025 20:34:55 GMT"
},
{
"version": "v2",
"created": "Wed, 28 May 2025 15:36:34 GMT"
},
{
"version": "v3",
"created": "Thu, 05 Jun 2025 14:58:43 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Kleeorin",
"Nathan",
""
],
[
"Rogachevskii",
"Igor",
""
]
] |
2502.12408 | Abdul Waheed | Abdul Waheed, Hanin Atwany, Rita Singh, Bhiksha Raj | On the Robust Approximation of ASR Metrics | ACL 2025 camera-ready | null | null | null | cs.CL | http://creativecommons.org/licenses/by/4.0/ | Recent advances in speech foundation models are largely driven by scaling both model size and data, enabling them to perform a wide range of tasks, including speech recognition. Traditionally, ASR models are evaluated using metrics like Word Error Rate (WER) and Character Error Rate (CER), which depend on ground truth labels. As a result of limited labeled data from diverse domains and testing conditions, the true generalization capabilities of these models beyond standard benchmarks remain unclear. Moreover, labeling data is both costly and time-consuming. To address this, we propose a novel label-free approach for approximating ASR performance metrics, eliminating the need for ground truth labels. Our method utilizes multimodal embeddings in a unified space for speech and transcription representations, combined with a high-quality proxy model to compute proxy metrics. These features are used to train a regression model to predict key ASR metrics like Word Error Rate (WER) and Character Error Rate (CER). We experiment with over 40 models across 14 datasets representing both standard and in-the-wild testing conditions. Our results show that we approximate the metrics within a single-digit absolute difference across all experimental configurations, outperforming the most recent baseline by more than 50\%. | [
{
"version": "v1",
"created": "Tue, 18 Feb 2025 01:10:17 GMT"
},
{
"version": "v2",
"created": "Wed, 04 Jun 2025 22:29:37 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Waheed",
"Abdul",
""
],
[
"Atwany",
"Hanin",
""
],
[
"Singh",
"Rita",
""
],
[
"Raj",
"Bhiksha",
""
]
] |
2502.12414 | Abdul Waheed | Hanin Atwany, Abdul Waheed, Rita Singh, Monojit Choudhury, Bhiksha Raj | Lost in Transcription, Found in Distribution Shift: Demystifying Hallucination in Speech Foundation Models | ACL2025 camera-ready | null | null | null | cs.CL | http://creativecommons.org/licenses/by/4.0/ | Speech foundation models trained at a massive scale, both in terms of model and data size, result in robust systems capable of performing multiple speech tasks, including automatic speech recognition (ASR). These models transcend language and domain barriers, yet effectively measuring their performance remains a challenge. Traditional metrics like word error rate (WER) and character error rate (CER) are commonly used to evaluate ASR performance but often fail to reflect transcription quality in critical contexts, particularly when detecting fabricated outputs. This phenomenon, known as hallucination, is especially concerning in high-stakes domains such as healthcare, legal, and aviation, where errors can have severe consequences. In our work, we address this gap by investigating hallucination in ASR models. We examine how factors such as distribution shifts, model size, and model architecture influence the hallucination error rate (HER), a metric we introduce to quantify hallucinations. Our analysis of over 20 ASR models reveals \numinsights~key insights: (1) High WERs can mask low hallucination rates, while low WERs may conceal dangerous hallucinations. (2) Synthetic noise, both adversarial and common perturbations like white noise, pitch shift, and time stretching, increase HER. (3) Distribution shift correlates strongly with HER ($\alpha = 0.91$). Our findings highlight the importance of incorporating HER alongside traditional metrics like WER to better assess ASR model performance, particularly in high-stakes domains. | [
{
"version": "v1",
"created": "Tue, 18 Feb 2025 01:25:39 GMT"
},
{
"version": "v2",
"created": "Wed, 04 Jun 2025 23:04:18 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Atwany",
"Hanin",
""
],
[
"Waheed",
"Abdul",
""
],
[
"Singh",
"Rita",
""
],
[
"Choudhury",
"Monojit",
""
],
[
"Raj",
"Bhiksha",
""
]
] |
2502.12436 | Wichayaporn Wongkamjan | Wichayaporn Wongkamjan, Yanze Wang, Feng Gu, Denis Peskoff, Jonathan K. Kummerfeld, Jonathan May, Jordan Lee Boyd-Graber | Should I Trust You? Detecting Deception in Negotiations using Counterfactual RL | ACL Findings 2025 | null | null | null | cs.CL | http://creativecommons.org/licenses/by/4.0/ | An increasingly common socio-technical problem is people being taken in by offers that sound ``too good to be true'', where persuasion and trust shape decision-making. This paper investigates how \abr{ai} can help detect these deceptive scenarios. We analyze how humans strategically deceive each other in \textit{Diplomacy}, a board game that requires both natural language communication and strategic reasoning. This requires extracting logical forms of proposed agreements in player communications and computing the relative rewards of the proposal using agents' value functions. Combined with text-based features, this can improve our deception detection. Our method detects human deception with a high precision when compared to a Large Language Model approach that flags many true messages as deceptive. Future human-\abr{ai} interaction tools can build on our methods for deception detection by triggering \textit{friction} to give users a chance of interrogating suspicious proposals. | [
{
"version": "v1",
"created": "Tue, 18 Feb 2025 02:11:41 GMT"
},
{
"version": "v2",
"created": "Fri, 21 Feb 2025 19:52:58 GMT"
},
{
"version": "v3",
"created": "Thu, 05 Jun 2025 06:27:54 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Wongkamjan",
"Wichayaporn",
""
],
[
"Wang",
"Yanze",
""
],
[
"Gu",
"Feng",
""
],
[
"Peskoff",
"Denis",
""
],
[
"Kummerfeld",
"Jonathan K.",
""
],
[
"May",
"Jonathan",
""
],
[
"Boyd-Graber",
"Jordan Lee",
""
]
] |
2502.12687 | Taehyeon Song | Kyung Kiu Kim, Jeongwon Ho, Seungjoon Hyun, Taehyeon Song | Black Hole Chemistry Knows Extra Dimensions | 31 pages, 11 figures, 1 Table, Some claims were deleted | null | null | null | hep-th gr-qc | http://creativecommons.org/licenses/by/4.0/ | : In this note, we study an extra dimension effect on the black hole chemistry in the 8-dimensional Einstein-Yang-Mills-Maxwell theory. The base spacetime contains 4- dimensional compact manifolds and an instanton on top of those. We demonstrate how the extra dimensions affect the phase transition and viable black hole sizes in the 4-dimensional Einstein frame through the black hole chemistry. We focus on asymptotically anti-de Sitter spacetimes for the effective 4-dimensional model obtained by a dimensional reduction. The extra-dimension size determines thermodynamic pressure, and the thermodynamic volume is roughly the horizon size of black holes. Thus, the extra dimension and black hole sizes are related as a conjugate pair of thermodynamic variables. | [
{
"version": "v1",
"created": "Tue, 18 Feb 2025 09:48:19 GMT"
},
{
"version": "v2",
"created": "Thu, 13 Mar 2025 04:18:15 GMT"
},
{
"version": "v3",
"created": "Thu, 05 Jun 2025 07:39:47 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Kim",
"Kyung Kiu",
""
],
[
"Ho",
"Jeongwon",
""
],
[
"Hyun",
"Seungjoon",
""
],
[
"Song",
"Taehyeon",
""
]
] |
2502.13033 | Guillaume Munch-Maccagnoni | \'El\'eonore Mangel, Paul-Andr\'e Melli\`es and Guillaume Munch-Maccagnoni | Classical notions of computation and the Hasegawa-Thielecke theorem | 24 pages with appendix | null | null | null | cs.LO cs.PL math.CT | http://creativecommons.org/licenses/by-nc-sa/4.0/ | In the spirit of the Curry-Howard correspondence between proofs and programs, we define and study a syntax and semantics for classical logic equipped with a computationally involutive negation, using a polarised effect calculus. A main challenge in designing a denotational semantics is to accommodate both call-by-value and call-by-name evaluation strategies, which leads to a failure of associativity of composition. Building on the work of the third author, we devise the notion of dialogue duploid, which provides a non-associative and effectful counterpart to the notion of dialogue category introduced by the second author in his 2-categorical account, based on adjunctions, of logical polarities and continuations. We show that the syntax of the polarised calculus can be interpreted in any dialogue duploid, and that it defines in fact a syntactic dialogue duploid. As an application, we establish, by semantic as well as syntactic means, the Hasegawa-Thielecke theorem, which states that the notions of central map and of thunkable map coincide in any dialogue duploid (in particular, for any double negation monad on a symmetric monoidal category). | [
{
"version": "v1",
"created": "Tue, 18 Feb 2025 16:48:06 GMT"
},
{
"version": "v2",
"created": "Thu, 05 Jun 2025 17:49:54 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Mangel",
"Éléonore",
""
],
[
"Melliès",
"Paul-André",
""
],
[
"Munch-Maccagnoni",
"Guillaume",
""
]
] |
2502.13063 | Yuri Kuratov | Yuri Kuratov, Mikhail Arkhipov, Aydar Bulatov, Mikhail Burtsev | Cramming 1568 Tokens into a Single Vector and Back Again: Exploring the Limits of Embedding Space Capacity | ACL 2025 (main conference) | null | null | null | cs.CL cs.LG | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | A range of recent works addresses the problem of compression of sequence of tokens into a shorter sequence of real-valued vectors to be used as inputs instead of token embeddings or key-value cache. These approaches are focused on reduction of the amount of compute in existing language models rather than minimization of number of bits needed to store text. Despite relying on powerful models as encoders, the maximum attainable lossless compression ratio is typically not higher than x10. This fact is highly intriguing because, in theory, the maximum information capacity of large real-valued vectors is far beyond the presented rates even for 16-bit precision and a modest vector size. In this work, we explore the limits of compression by replacing the encoder with a per-sample optimization procedure. We show that vectors with compression ratios up to x1500 exist, which highlights two orders of magnitude gap between existing and practically attainable solutions. Furthermore, we empirically show that the compression limits are determined not by the length of the input but by the amount of uncertainty to be reduced, namely, the cross-entropy loss on this sequence without any conditioning. The obtained limits highlight the substantial gap between the theoretical capacity of input embeddings and their practical utilization, suggesting significant room for optimization in model design. | [
{
"version": "v1",
"created": "Tue, 18 Feb 2025 17:08:45 GMT"
},
{
"version": "v2",
"created": "Thu, 05 Jun 2025 13:20:09 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Kuratov",
"Yuri",
""
],
[
"Arkhipov",
"Mikhail",
""
],
[
"Bulatov",
"Aydar",
""
],
[
"Burtsev",
"Mikhail",
""
]
] |
2502.13264 | Kaveh Ahadi | Samuel J. Poage, Xueshi Gao, Merve Baksi, Salva Salmani-Rezaie, David A. Muller, Divine P. Kumah, Chun Ning Lau, Jose Lorenzana, Maria N. Gastiasoro and Kaveh Ahadi | Violation of Pauli Limit at KTaO3(110) Interfaces | null | null | 10.1103/PhysRevB.111.214506 | null | cond-mat.supr-con | http://creativecommons.org/licenses/by/4.0/ | The superconducting order parameter at the KTaO3 interfaces and its dependence on interface orientation remains a subject of debate. The superconductivity at these interfaces exhibits strong resilience against in-plane magnetic field and violates Pauli limit. The interface orientation dependence of critical field and violation of Pauli limit, however, have not been investigated. To address this problem, we grew epitaxial LaMnO3/KTaO3 heterostructures using molecular beam epitaxy. We show that superconductivity is extremely robust against the in-plane magnetic field. Our results indicate that the interface orientation, despite impacting the critical temperature, does not affect the ratio of critical field to the Pauli limiting field. These results offer opportunities to engineer superconductors which are resilient against magnetic field. | [
{
"version": "v1",
"created": "Tue, 18 Feb 2025 20:18:44 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Poage",
"Samuel J.",
""
],
[
"Gao",
"Xueshi",
""
],
[
"Baksi",
"Merve",
""
],
[
"Salmani-Rezaie",
"Salva",
""
],
[
"Muller",
"David A.",
""
],
[
"Kumah",
"Divine P.",
""
],
[
"Lau",
"Chun Ning",
""
],
[
"Lorenzana",
"Jose",
""
],
[
"Gastiasoro",
"Maria N.",
""
],
[
"Ahadi",
"Kaveh",
""
]
] |
2502.13276 | Armando Capasso | Armando Capasso | CW-complexes and minimal Hilbert vector of graded Artinian Gorenstein algebras | New sections and results | null | null | null | math.AC math.CO math.RA | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | I introduce a geometric interpretation of the set of standard graded Artinian Gorenstein algebras of codimension $n$ and degree $d$: the standard locus, which is a subset of the projective space of degree $d$ polynomials in $n$ variables, and I characterize it. Under opportune hypothesis, I prove that the locus of full Perazzo polynomials is the union of the minimal dimensional irreducible components of the standard locus and it is pure dimensional subset. On the other hand, I associate to any homogeneous polynomial a topological space, which is a CW-complex. Using all these sets, I prove that the Hilbert function restricted to the standard locus has minimal values on any irreducible component of the domain. I apply all this to the Full Perazzo Conjecture and I prove it. | [
{
"version": "v1",
"created": "Tue, 18 Feb 2025 20:48:16 GMT"
},
{
"version": "v2",
"created": "Wed, 04 Jun 2025 20:29:28 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Capasso",
"Armando",
""
]
] |
2502.13854 | Pedro Montealegre | Augusto Modanese, Pedro Montealegre, and Mart\'in R\'ios-Wilson | Strong and Hiding Distributed Certification of Bipartiteness | 32 pages, 8 figures. Abstract shortened to meet arXiv's requirements | null | null | null | cs.DC | http://creativecommons.org/licenses/by-nc-nd/4.0/ | Distributed certification is a framework in distributed computing where nodes in a network jointly verify whether the whole graph satisfies a given property. A locally checkable proof (LCP) is a non-deterministic distributed algorithm used to verify global properties of a graph $G$, involving a prover and a verifier. The prover is a powerful entity that assigns certificates to nodes, which are then locally checked by the verifier.
An LCP is correct if it satisfies completeness and soundness. Completeness means that, for any graph $G$ satisfying a property $\Pi$, there exists a certificate assignment accepted by all nodes. Soundness ensures that for every graph not satisfying $\Pi$, at least one node rejects any certificate assignment.
We study how to certify that a graph is bipartite (i.e., $2$-colorable) with an LCP that hides the $2$-coloring from the verifier. An LCP is hiding if no local algorithm can reconstruct the coloring from a valid certificate. Motivated by promise-free separations in the LOCAL model and its variants, we also require strong soundness: in a no-instance, the subgraph induced by accepting nodes must be $2$-colorable. An LCP with completeness, soundness, hiding, and strong soundness is called strong and hiding.
We show that such LCPs for $2$-coloring exist in specific graph classes, using only $O(\log n)$-size certificates. If the input is a cycle or has a node of degree 1, these LCPs also work in anonymous networks with constant-size certificates.
We also prove that no strong and hiding LCP exists for general graphs unless node identifiers are available and certificates are of size $\omega(1)$. In anonymous networks, this lower bound holds regardless of the certificate size. We also present a characterization of the hiding property for $k$-coloring, which plays a key role in future investigations. | [
{
"version": "v1",
"created": "Wed, 19 Feb 2025 16:18:50 GMT"
},
{
"version": "v2",
"created": "Thu, 05 Jun 2025 16:00:44 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Modanese",
"Augusto",
""
],
[
"Montealegre",
"Pedro",
""
],
[
"Ríos-Wilson",
"Martín",
""
]
] |
2502.13941 | Andreas Aigner | Enrico Bau, Andreas Aigner, Jonas Biechteler, Connor Heimig, Thorsten Goelz, Stefan A. Maier, and Andreas Tittl | Spatially Encoded Polaritonic Ultra-Strong Coupling in Gradient Metasurfaces with Epsilon-Near-Zero Modes | null | null | null | null | physics.optics cond-mat.mtrl-sci physics.app-ph | http://creativecommons.org/licenses/by/4.0/ | We introduce a platform to achieve ultra-strong coupling (USC) between light and matter using widely available materials. USC is a light-matter interaction regime characterized by coupling strengths exceeding 10% of the ground state energy. It gives rise to novel physical phenomena, such as efficient single-photon coupling and quantum gates, with applications in quantum sensing, nonlinear optics, and low-threshold lasing. Although early demonstrations in plasmonic systems have been realized, achieving USC in dielectric platforms, which offer lower losses and high Q-factors, remains challenging due to typically low mode overlap between the photonic field and the material resonance. Here we leverage dielectric dual gradient metasurfaces supporting quasi-bound states in the continuum to spatially encode both the spectral and coupling parameter space and demonstrate USC to an epsilon-near-zero (ENZ) mode in an ultra-thin SiO2 layer. The strong out-of-plane electric fields in our tapered bar structure overlap exceptionally well with those of the ENZ mode, resulting in a normalized coupling strength of 0.101 and a mode splitting equivalent to 20% of the ENZ mode energy; a four- to five-fold increase compared to previous approaches. The strong field confinement of our approach opens new possibilities for compact and scalable polaritonic devices, such as tunable frequency converters and low-energy optical modulators. | [
{
"version": "v1",
"created": "Wed, 19 Feb 2025 18:34:58 GMT"
},
{
"version": "v2",
"created": "Thu, 05 Jun 2025 13:06:13 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Bau",
"Enrico",
""
],
[
"Aigner",
"Andreas",
""
],
[
"Biechteler",
"Jonas",
""
],
[
"Heimig",
"Connor",
""
],
[
"Goelz",
"Thorsten",
""
],
[
"Maier",
"Stefan A.",
""
],
[
"Tittl",
"Andreas",
""
]
] |
2502.14032 | Ryan Spieler | Andreas Karch, Ryan C. Spieler | Critical theories connecting gapped phases with $\mathbb{Z}_2\times\mathbb{Z}_2$ symmetry from the duality web | v2. Typos corrected and references added. Matches version in JHEP | J. High Energ. Phys. 2025, 28 (2025) | 10.1007/JHEP06(2025)028 | null | cond-mat.str-el cond-mat.stat-mech hep-th | http://creativecommons.org/licenses/by/4.0/ | We use the ideas behind the duality web to construct numerous conformal field theories mediating the phase transitions between various symmetry broken and topological phases. In particular we obtain the full field theory version of the Kennedy Tasaki transformation, mapping a gapless theory mediating a topological phase transition of symmetry protected topological orders to a standard symmetry breaking one in a 1+1 dimensional $\mathbb{Z}_2 \times \mathbb{Z}_2$ gauge theory. When we consider all possible discrete gauging operations, we obtain bosonic and fermionic webs with 9 critical theories per web, each connecting 4 separate gapped phases, some of them topological. Bosonization maps the two webs into each other. In addition to discussing the multi-critical theory connecting the four gapped phases in each phase diagram, we discuss the partially gapped theories connecting two of those four. Some of these are gapless symmetry protected topological phases. | [
{
"version": "v1",
"created": "Wed, 19 Feb 2025 19:00:01 GMT"
},
{
"version": "v2",
"created": "Thu, 05 Jun 2025 02:57:54 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Karch",
"Andreas",
""
],
[
"Spieler",
"Ryan C.",
""
]
] |
2502.14184 | Marjolein Oostrom | Marjolein Oostrom, Alex Hagen, Nicole LaHaye, Karl Pazdernik | Bayesian SegNet for Semantic Segmentation with Improved Interpretation of Microstructural Evolution During Irradiation of Materials | null | null | 10.1016/j.commatsci.2025.113943 | null | cs.CV cs.LG | http://creativecommons.org/licenses/by/4.0/ | Understanding the relationship between the evolution of microstructures of irradiated LiAlO2 pellets and tritium diffusion, retention and release could improve predictions of tritium-producing burnable absorber rod performance. Given expert-labeled segmented images of irradiated and unirradiated pellets, we trained Deep Convolutional Neural Networks to segment images into defect, grain, and boundary classes. Qualitative microstructural information was calculated from these segmented images to facilitate the comparison of unirradiated and irradiated pellets. We tested modifications to improve the sensitivity of the model, including incorporating meta-data into the model and utilizing uncertainty quantification. The predicted segmentation was similar to the expert-labeled segmentation for most methods of microstructural qualification, including pixel proportion, defect area, and defect density. Overall, the high performance metrics for the best models for both irradiated and unirradiated images shows that utilizing neural network models is a viable alternative to expert-labeled images. | [
{
"version": "v1",
"created": "Thu, 20 Feb 2025 01:26:05 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Oostrom",
"Marjolein",
""
],
[
"Hagen",
"Alex",
""
],
[
"LaHaye",
"Nicole",
""
],
[
"Pazdernik",
"Karl",
""
]
] |
2502.14425 | Yuxing Cheng | Yuxing Cheng, Yi Chang, Yuan Wu | A Survey on Data Contamination for Large Language Models | null | null | null | null | cs.CL | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Recent advancements in Large Language Models (LLMs) have demonstrated significant progress in various areas, such as text generation and code synthesis. However, the reliability of performance evaluation has come under scrutiny due to data contamination-the unintended overlap between training and test datasets. This overlap has the potential to artificially inflate model performance, as LLMs are typically trained on extensive datasets scraped from publicly available sources. These datasets often inadvertently overlap with the benchmarks used for evaluation, leading to an overestimation of the models' true generalization capabilities. In this paper, we first examine the definition and impacts of data contamination. Secondly, we review methods for contamination-free evaluation, focusing on three strategies: data updating-based methods, data rewriting-based methods, and prevention-based methods. Specifically, we highlight dynamic benchmarks and LLM-driven evaluation methods. Finally, we categorize contamination detecting methods based on model information dependency: white-Box, gray-Box, and black-Box detection approaches. Our survey highlights the requirements for more rigorous evaluation protocols and proposes future directions for addressing data contamination challenges. | [
{
"version": "v1",
"created": "Thu, 20 Feb 2025 10:23:27 GMT"
},
{
"version": "v2",
"created": "Thu, 05 Jun 2025 00:49:05 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Cheng",
"Yuxing",
""
],
[
"Chang",
"Yi",
""
],
[
"Wu",
"Yuan",
""
]
] |
2502.14470 | Antonio Franco | A. Franco, A. A. Nucita, F. De Paolis and F. Strafella | COSMIC-S: a photometric Catalogue of Observed Stars in the Small MagellanIc Cloud | 7 pages, 7 figures, 3 tables, Accepted for publication on ApJS | The Astrophysical Journal Supplement Series, 277:33 (6pp), 2025 April | 10.3847/1538-4365/adb62a | null | astro-ph.SR astro-ph.GA astro-ph.IM | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | The Dark Energy Camera (DECam) is a wide-field instrument mounted on the 4m V. Blanco Telescope (CTIO). Its impressive technical characteristics makes it one of the most suitable ground-based telescope for the production of accurate stellar photometry even towards crowded regions such as the Magellanic Clouds. We analysed DECam images acquired from February 2018 to January 2020 towards the Small Magellanic Cloud. We performed a PSF photometry by using the SExtractor and PSFEx tools and producing a comprehensive photometric catalogue in the SDSS system, considering the $gri$ filters. Then, we present COSMIC-S, a photometric catalogue consisting of 10 971 906 sources, including $gri$ magnitudes with a mean error $<\sigma_m> \simeq 0.04$ mag. A total of 2 456 434 sources have good photometry in all three bands. The catalogue appears virtually complete to $m \simeq 22$, with a limiting magnitude $m\simeq 25$. We derived the colour-magnitude and colour-colour diagrams in order to prove the goodness of the catalogue. | [
{
"version": "v1",
"created": "Thu, 20 Feb 2025 11:49:42 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Franco",
"A.",
""
],
[
"Nucita",
"A. A.",
""
],
[
"De Paolis",
"F.",
""
],
[
"Strafella",
"F.",
""
]
] |
2502.14578 | Alicia Castro | T. Mursheed Amith and Alicia Castro | The massive BMS character in 3D quantum gravity | 31 pages, 5 figures | null | 10.1088/1361-6382/ade110 | null | gr-qc hep-th | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We derive the one-loop partition function for three-dimensional quantum gravity in a finite-radius thermal twisted flat space with a conical defect, reproducing the massive BMS$_3$ character. We perform the computation in both discrete and continuum geometry formulations, showing consistency between them. In the discrete case, we integrate out bulk degrees of freedom in a Regge gravity framework, while in the continuum, we construct a dual non-local boundary field theory encoding geodesic length fluctuations. Our study shows that the additional modes of the massive character, compared to the vacuum case, originate from the explicit breaking of radial diffeomorphism symmetry by the defect. This provides a concrete geometric mechanism in Regge gravity, tracing the appearance of massive BMS$_3$ particles to diffeomorphism breaking by conical defects, and highlights the broader relevance of discrete geometry approaches to quantum gravity with matter. | [
{
"version": "v1",
"created": "Thu, 20 Feb 2025 14:01:33 GMT"
},
{
"version": "v2",
"created": "Thu, 05 Jun 2025 09:53:35 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Amith",
"T. Mursheed",
""
],
[
"Castro",
"Alicia",
""
]
] |
2502.14722 | Qihong Lorena Li Hu | Qihong Lorena Li-Hu, Patricia Garc\'ia-Caspue\~nas, Andrea Ianiro, Stefano Discetti | Model-based time super-sampling of turbulent flow field sequences | null | null | 10.1103/2lqd-g9mt | null | physics.flu-dyn | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We propose a novel method for model-based time super-sampling of turbulent flow fields. The key enabler is the identification of an empirical Galerkin model from the projection of the Navier-Stokes equations on a data-tailored basis. The basis is obtained from a Proper Orthogonal Decomposition (POD) of the measured fields. Time super-sampling is thus achieved by a time-marching integration of the identified dynamical system, taking the original snapshots as initial conditions. Temporal continuity of the reconstructed velocity fields is achieved through a forward-backwards integration between consecutive measured Particle Image Velocimetry measurements of a turbulent jet flow. The results are compared with the interpolation of the POD temporal coefficients and the low-order reconstruction of data measured at a higher sampling rate. In both cases, the results obtained show the ability of the method to reconstruct the dynamics of the flow with small errors during several flow characteristic times. | [
{
"version": "v1",
"created": "Thu, 20 Feb 2025 16:48:25 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Li-Hu",
"Qihong Lorena",
""
],
[
"García-Caspueñas",
"Patricia",
""
],
[
"Ianiro",
"Andrea",
""
],
[
"Discetti",
"Stefano",
""
]
] |
2502.14948 | Zi Lin | Zi Lin, Sheng Shen, Jingbo Shang, Jason Weston, Yixin Nie | Learning to Solve and Verify: A Self-Play Framework for Code and Test Generation | 14 pages, 5 figures | null | null | null | cs.SE | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Recent advances in large language models (LLMs) have improved their performance on coding benchmarks. However, improvement is plateauing due to the exhaustion of readily available high-quality data. Prior work has shown the potential of synthetic self-instruct data, but naively training on a model's own outputs can cause error accumulation, especially in coding tasks, where generalization may collapse due to overly simple or erroneous training data, highlighting the need for rigorous quality checks on synthetic data. In this work, we explore an effective approach whereby the model itself verifies the correctness of its own data. We thus propose Sol-Ver, a self-play solver-verifier framework that jointly improves a single model's code and test generation capacity. By iteratively refining code (LLM-as-a-solver) and tests (LLM-as-a-verifier) together, we boost both capabilities without relying on human annotations or larger teacher models. Experiments with the Llama 3.1 8B model demonstrate substantial performance enhancements, achieving average relative improvements of 19.63% in code generation and 17.49% in test generation on MBPP and LiveCodeBench. | [
{
"version": "v1",
"created": "Thu, 20 Feb 2025 18:32:19 GMT"
},
{
"version": "v2",
"created": "Mon, 24 Feb 2025 21:24:46 GMT"
},
{
"version": "v3",
"created": "Wed, 04 Jun 2025 20:52:16 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Lin",
"Zi",
""
],
[
"Shen",
"Sheng",
""
],
[
"Shang",
"Jingbo",
""
],
[
"Weston",
"Jason",
""
],
[
"Nie",
"Yixin",
""
]
] |
2502.15652 | Haoxuan Li | Fengxiang Cheng, Haoxuan Li, Fenrong Liu, Robert van Rooij, Kun Zhang, Zhouchen Lin | Empowering LLMs with Logical Reasoning: A Comprehensive Survey | Accepted by IJCAI 2025 (Survey Track) | null | null | null | cs.AI cs.CL | http://creativecommons.org/licenses/by/4.0/ | Large language models (LLMs) have achieved remarkable successes on various tasks. However, recent studies have found that there are still significant challenges to the logical reasoning abilities of LLMs, which can be categorized into the following two aspects: (1) Logical question answering: LLMs often fail to generate the correct answer within a complex logical problem which requires sophisticated deductive, inductive or abductive reasoning given a collection of premises and constrains. (2) Logical consistency: LLMs are prone to producing responses contradicting themselves across different questions. For example, a state-of-the-art question-answering LLM Macaw, answers Yes to both questions Is a magpie a bird? and Does a bird have wings? but answers No to Does a magpie have wings?. To facilitate this research direction, we comprehensively investigate the most cutting-edge methods and propose a detailed taxonomy. Specifically, to accurately answer complex logic questions, previous methods can be categorized based on reliance on external solvers, prompts, and fine-tuning. To avoid logical contradictions, we discuss concepts and solutions of various logical consistencies, including implication, negation, transitivity, factuality consistencies, and their composites. In addition, we review commonly used benchmark datasets and evaluation metrics, and discuss promising research directions, such as extending to modal logic to account for uncertainty and developing efficient algorithms that simultaneously satisfy multiple logical consistencies. | [
{
"version": "v1",
"created": "Fri, 21 Feb 2025 18:20:35 GMT"
},
{
"version": "v2",
"created": "Mon, 24 Feb 2025 19:01:38 GMT"
},
{
"version": "v3",
"created": "Wed, 04 Jun 2025 19:38:43 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Cheng",
"Fengxiang",
""
],
[
"Li",
"Haoxuan",
""
],
[
"Liu",
"Fenrong",
""
],
[
"van Rooij",
"Robert",
""
],
[
"Zhang",
"Kun",
""
],
[
"Lin",
"Zhouchen",
""
]
] |
2502.15821 | Keane Wei Yang Ong | Keane Ong, Rui Mao, Deeksha Varshney, Erik Cambria and Gianmarco Mengaldo | Towards Robust ESG Analysis Against Greenwashing Risks: Aspect-Action Analysis with Cross-Category Generalization | Proceedings of the Association for Computational Linguistics Main Conference (ACL 2025) | null | null | null | cs.CL cs.AI | http://creativecommons.org/licenses/by/4.0/ | Sustainability reports are key for evaluating companies' environmental, social and governance, ESG performance, but their content is increasingly obscured by greenwashing - sustainability claims that are misleading, exaggerated, and fabricated. Yet, existing NLP approaches for ESG analysis lack robustness against greenwashing risks, often extracting insights that reflect misleading or exaggerated sustainability claims rather than objective ESG performance. To bridge this gap, we introduce A3CG - Aspect-Action Analysis with Cross-Category Generalization, as a novel dataset to improve the robustness of ESG analysis amid the prevalence of greenwashing. By explicitly linking sustainability aspects with their associated actions, A3CG facilitates a more fine-grained and transparent evaluation of sustainability claims, ensuring that insights are grounded in verifiable actions rather than vague or misleading rhetoric. Additionally, A3CG emphasizes cross-category generalization. This ensures robust model performance in aspect-action analysis even when companies change their reports to selectively favor certain sustainability areas. Through experiments on A3CG, we analyze state-of-the-art supervised models and LLMs, uncovering their limitations and outlining key directions for future research. | [
{
"version": "v1",
"created": "Thu, 20 Feb 2025 03:01:08 GMT"
},
{
"version": "v2",
"created": "Thu, 05 Jun 2025 12:05:19 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Ong",
"Keane",
""
],
[
"Mao",
"Rui",
""
],
[
"Varshney",
"Deeksha",
""
],
[
"Cambria",
"Erik",
""
],
[
"Mengaldo",
"Gianmarco",
""
]
] |
2502.17327 | Inbar Gat | Inbar Gat, Sigal Raab, Guy Tevet, Yuval Reshef, Amit H. Bermano, Daniel Cohen-Or | AnyTop: Character Animation Diffusion with Any Topology | SIGGRAPH 2025. Video: https://www.youtube.com/watch?v=NWOdkM5hAbE, Project page: https://anytop2025.github.io/Anytop-page, Code: https://github.com/Anytop2025/Anytop | null | null | null | cs.GR cs.AI cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Generating motion for arbitrary skeletons is a longstanding challenge in computer graphics, remaining largely unexplored due to the scarcity of diverse datasets and the irregular nature of the data. In this work, we introduce AnyTop, a diffusion model that generates motions for diverse characters with distinct motion dynamics, using only their skeletal structure as input. Our work features a transformer-based denoising network, tailored for arbitrary skeleton learning, integrating topology information into the traditional attention mechanism. Additionally, by incorporating textual joint descriptions into the latent feature representation, AnyTop learns semantic correspondences between joints across diverse skeletons. Our evaluation demonstrates that AnyTop generalizes well, even with as few as three training examples per topology, and can produce motions for unseen skeletons as well. Furthermore, our model's latent space is highly informative, enabling downstream tasks such as joint correspondence, temporal segmentation and motion editing. Our webpage, https://anytop2025.github.io/Anytop-page, includes links to videos and code. | [
{
"version": "v1",
"created": "Mon, 24 Feb 2025 17:00:36 GMT"
},
{
"version": "v2",
"created": "Thu, 05 Jun 2025 15:23:33 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Gat",
"Inbar",
""
],
[
"Raab",
"Sigal",
""
],
[
"Tevet",
"Guy",
""
],
[
"Reshef",
"Yuval",
""
],
[
"Bermano",
"Amit H.",
""
],
[
"Cohen-Or",
"Daniel",
""
]
] |
2502.17394 | Tanmay Parekh | Tanmay Parekh, Yuxuan Dong, Lucas Bandarkar, Artin Kim, I-Hung Hsu, Kai-Wei Chang, Nanyun Peng | SNaRe: Domain-aware Data Generation for Low-Resource Event Detection | Under review at ACL ARR May 2025 | null | null | null | cs.CL cs.AI | http://creativecommons.org/licenses/by/4.0/ | Event Detection (ED) -- the task of identifying event mentions from natural language text -- is critical for enabling reasoning in highly specialized domains such as biomedicine, law, and epidemiology. Data generation has proven to be effective in broadening its utility to wider applications without requiring expensive expert annotations. However, when existing generation approaches are applied to specialized domains, they struggle with label noise, where annotations are incorrect, and domain drift, characterized by a distributional mismatch between generated sentences and the target domain. To address these issues, we introduce SNaRe, a domain-aware synthetic data generation framework composed of three components: Scout, Narrator, and Refiner. Scout extracts triggers from unlabeled target domain data and curates a high-quality domain-specific trigger list using corpus-level statistics to mitigate domain drift. Narrator, conditioned on these triggers, generates high-quality domain-aligned sentences, and Refiner identifies additional event mentions, ensuring high annotation quality. Experimentation on three diverse domain ED datasets reveals how SNaRe outperforms the best baseline, achieving average F1 gains of 3-7% in the zero-shot/few-shot settings and 4-20% F1 improvement for multilingual generation. Analyzing the generated trigger hit rate and human evaluation substantiates SNaRe's stronger annotation quality and reduced domain drift. | [
{
"version": "v1",
"created": "Mon, 24 Feb 2025 18:20:42 GMT"
},
{
"version": "v2",
"created": "Thu, 05 Jun 2025 15:45:00 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Parekh",
"Tanmay",
""
],
[
"Dong",
"Yuxuan",
""
],
[
"Bandarkar",
"Lucas",
""
],
[
"Kim",
"Artin",
""
],
[
"Hsu",
"I-Hung",
""
],
[
"Chang",
"Kai-Wei",
""
],
[
"Peng",
"Nanyun",
""
]
] |
2502.17419 | Zhong-Zhi Li | Zhong-Zhi Li, Duzhen Zhang, Ming-Liang Zhang, Jiaxin Zhang, Zengyan Liu, Yuxuan Yao, Haotian Xu, Junhao Zheng, Pei-Jie Wang, Xiuyi Chen, Yingying Zhang, Fei Yin, Jiahua Dong, Zhiwei Li, Bao-Long Bi, Ling-Rui Mei, Junfeng Fang, Xiao Liang, Zhijiang Guo, Le Song, Cheng-Lin Liu | From System 1 to System 2: A Survey of Reasoning Large Language Models | Slow-thinking, Large Language Models, Human-like Reasoning, Decision Making in AI, AGI | null | null | null | cs.AI | http://creativecommons.org/licenses/by/4.0/ | Achieving human-level intelligence requires refining the transition from the fast, intuitive System 1 to the slower, more deliberate System 2 reasoning. While System 1 excels in quick, heuristic decisions, System 2 relies on logical reasoning for more accurate judgments and reduced biases. Foundational Large Language Models (LLMs) excel at fast decision-making but lack the depth for complex reasoning, as they have not yet fully embraced the step-by-step analysis characteristic of true System 2 thinking. Recently, reasoning LLMs like OpenAI's o1/o3 and DeepSeek's R1 have demonstrated expert-level performance in fields such as mathematics and coding, closely mimicking the deliberate reasoning of System 2 and showcasing human-like cognitive abilities. This survey begins with a brief overview of the progress in foundational LLMs and the early development of System 2 technologies, exploring how their combination has paved the way for reasoning LLMs. Next, we discuss how to construct reasoning LLMs, analyzing their features, the core methods enabling advanced reasoning, and the evolution of various reasoning LLMs. Additionally, we provide an overview of reasoning benchmarks, offering an in-depth comparison of the performance of representative reasoning LLMs. Finally, we explore promising directions for advancing reasoning LLMs and maintain a real-time \href{https://github.com/zzli2022/Awesome-Slow-Reason-System}{GitHub Repository} to track the latest developments. We hope this survey will serve as a valuable resource to inspire innovation and drive progress in this rapidly evolving field. | [
{
"version": "v1",
"created": "Mon, 24 Feb 2025 18:50:52 GMT"
},
{
"version": "v2",
"created": "Tue, 25 Feb 2025 17:15:00 GMT"
},
{
"version": "v3",
"created": "Fri, 25 Apr 2025 08:15:51 GMT"
},
{
"version": "v4",
"created": "Thu, 05 Jun 2025 03:18:45 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Li",
"Zhong-Zhi",
""
],
[
"Zhang",
"Duzhen",
""
],
[
"Zhang",
"Ming-Liang",
""
],
[
"Zhang",
"Jiaxin",
""
],
[
"Liu",
"Zengyan",
""
],
[
"Yao",
"Yuxuan",
""
],
[
"Xu",
"Haotian",
""
],
[
"Zheng",
"Junhao",
""
],
[
"Wang",
"Pei-Jie",
""
],
[
"Chen",
"Xiuyi",
""
],
[
"Zhang",
"Yingying",
""
],
[
"Yin",
"Fei",
""
],
[
"Dong",
"Jiahua",
""
],
[
"Li",
"Zhiwei",
""
],
[
"Bi",
"Bao-Long",
""
],
[
"Mei",
"Ling-Rui",
""
],
[
"Fang",
"Junfeng",
""
],
[
"Liang",
"Xiao",
""
],
[
"Guo",
"Zhijiang",
""
],
[
"Song",
"Le",
""
],
[
"Liu",
"Cheng-Lin",
""
]
] |
2502.17709 | Yu Zhou | Yu Zhou, Bingxuan Li, Mohan Tang, Xiaomeng Jin, Te-Lin Wu, Kuan-Hao Huang, Heng Ji, Kai-Wei Chang, Nanyun Peng | Contrastive Visual Data Augmentation | null | ICML 2025 | null | null | cs.CV cs.AI cs.CL cs.LG cs.MM | http://creativecommons.org/licenses/by/4.0/ | Large multimodal models (LMMs) often struggle to recognize novel concepts, as they rely on pre-trained knowledge and have limited ability to capture subtle visual details. Domain-specific knowledge gaps in training also make them prone to confusing visually similar, commonly misrepresented, or low-resource concepts. To help LMMs better align nuanced visual features with language, improving their ability to recognize and reason about novel or rare concepts, we propose a Contrastive visual Data Augmentation (CoDA) strategy. CoDA extracts key contrastive textual and visual features of target concepts against the known concepts they are misrecognized as, and then uses multimodal generative models to produce targeted synthetic data. Automatic filtering of extracted features and augmented images is implemented to guarantee their quality, as verified by human annotators. We show the effectiveness and efficiency of CoDA on low-resource concept and diverse scene recognition datasets including INaturalist and SUN. We additionally collect NovelSpecies, a benchmark dataset consisting of newly discovered animal species that are guaranteed to be unseen by LMMs. LLaVA-1.6 1-shot updating results on these three datasets show CoDA significantly improves SOTA visual data augmentation strategies by 12.3% (NovelSpecies), 5.1% (SUN), and 6.0% (iNat) absolute gains in accuracy. | [
{
"version": "v1",
"created": "Mon, 24 Feb 2025 23:05:31 GMT"
},
{
"version": "v2",
"created": "Thu, 05 Jun 2025 02:50:14 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Zhou",
"Yu",
""
],
[
"Li",
"Bingxuan",
""
],
[
"Tang",
"Mohan",
""
],
[
"Jin",
"Xiaomeng",
""
],
[
"Wu",
"Te-Lin",
""
],
[
"Huang",
"Kuan-Hao",
""
],
[
"Ji",
"Heng",
""
],
[
"Chang",
"Kai-Wei",
""
],
[
"Peng",
"Nanyun",
""
]
] |
2502.18137 | Jintao Zhang | Jintao Zhang, Chendong Xiang, Haofeng Huang, Jia Wei, Haocheng Xi, Jun Zhu, Jianfei Chen | SpargeAttention: Accurate and Training-free Sparse Attention Accelerating Any Model Inference | @inproceedings{zhang2025spargeattn, title={Spargeattn: Accurate sparse attention accelerating any model inference}, author={Zhang, Jintao and Xiang, Chendong and Huang, Haofeng and Wei, Jia and Xi, Haocheng and Zhu, Jun and Chen, Jianfei}, booktitle={International Conference on Machine Learning (ICML)}, year={2025} } | Proceedings of the 42 nd International Conference on Machine Learning, PMLR 267, 2025 (ICML 2025) | null | null | cs.LG cs.AI cs.CV cs.PF | http://creativecommons.org/licenses/by/4.0/ | An efficient attention implementation is essential for large models due to its quadratic time complexity. Fortunately, attention commonly exhibits sparsity, i.e., many values in the attention map are near zero, allowing for the omission of corresponding computations. Many studies have utilized the sparse pattern to accelerate attention. However, most existing works focus on optimizing attention within specific models by exploiting certain sparse patterns of the attention map. A universal sparse attention that guarantees both the speedup and end-to-end performance of diverse models remains elusive. In this paper, we propose SpargeAttn, a universal sparse and quantized attention for any model. Our method uses a two-stage online filter: in the first stage, we rapidly and accurately predict the attention map, enabling the skip of some matrix multiplications in attention. In the second stage, we design an online softmax-aware filter that incurs no extra overhead and further skips some matrix multiplications. Experiments show that our method significantly accelerates diverse models, including language, image, and video generation, without sacrificing end-to-end metrics. The codes are available at https://github.com/thu-ml/SpargeAttn. | [
{
"version": "v1",
"created": "Tue, 25 Feb 2025 12:02:17 GMT"
},
{
"version": "v2",
"created": "Thu, 01 May 2025 05:38:46 GMT"
},
{
"version": "v3",
"created": "Sat, 31 May 2025 16:22:36 GMT"
},
{
"version": "v4",
"created": "Thu, 05 Jun 2025 03:03:14 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Zhang",
"Jintao",
""
],
[
"Xiang",
"Chendong",
""
],
[
"Huang",
"Haofeng",
""
],
[
"Wei",
"Jia",
""
],
[
"Xi",
"Haocheng",
""
],
[
"Zhu",
"Jun",
""
],
[
"Chen",
"Jianfei",
""
]
] |
2502.18334 | Hans Hao-Hsun Hsu | Hans Hao-Hsun Hsu, Shikun Liu, Han Zhao, Pan Li | Structural Alignment Improves Graph Test-Time Adaptation | null | null | null | null | cs.LG | http://creativecommons.org/licenses/by/4.0/ | Graph-based learning excels at capturing interaction patterns in diverse domains like recommendation, fraud detection, and particle physics. However, its performance often degrades under distribution shifts, especially those altering network connectivity. Current methods to address these shifts typically require retraining with the source dataset, which is often infeasible due to computational or privacy limitations. We introduce Test-Time Structural Alignment (TSA), a novel algorithm for Graph Test-Time Adaptation (GTTA) that aligns graph structures during inference without accessing the source data. Grounded in a theoretical understanding of graph data distribution shifts, TSA employs three synergistic strategies: uncertainty-aware neighborhood weighting to accommodate neighbor label distribution shifts, adaptive balancing of self-node and aggregated neighborhood representations based on their signal-to-noise ratio, and decision boundary refinement to correct residual label and feature shifts. Extensive experiments on synthetic and real-world datasets demonstrate TSA's consistent outperformance of both non-graph TTA methods and state-of-the-art GTTA baselines. | [
{
"version": "v1",
"created": "Tue, 25 Feb 2025 16:26:25 GMT"
},
{
"version": "v2",
"created": "Sun, 25 May 2025 02:10:13 GMT"
},
{
"version": "v3",
"created": "Thu, 05 Jun 2025 14:44:56 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Hsu",
"Hans Hao-Hsun",
""
],
[
"Liu",
"Shikun",
""
],
[
"Zhao",
"Han",
""
],
[
"Li",
"Pan",
""
]
] |
2502.18470 | Dazhou Yu | Dazhou Yu, Riyang Bao, Ruiyu Ning, Jinghong Peng, Gengchen Mai, Liang Zhao | Spatial-RAG: Spatial Retrieval Augmented Generation for Real-World Spatial Reasoning Questions | null | null | null | null | cs.IR cs.ET cs.LG | http://creativecommons.org/licenses/by-nc-nd/4.0/ | Spatial reasoning remains a challenge for Large Language Models (LLMs), which struggle with spatial data retrieval and reasoning. We propose Spatial Retrieval-Augmented Generation (Spatial-RAG), a framework that extends RAG to spatial tasks by integrating sparse spatial retrieval (spatial databases) and dense semantic retrieval (LLM-based similarity). A multi-objective ranking strategy balances spatial constraints and semantic relevance, while an LLM-guided generator ensures coherent responses. Experiments on a real-world tourism dataset show that Spatial-RAG significantly improves spatial question answering, bridging the gap between LLMs and spatial intelligence. | [
{
"version": "v1",
"created": "Tue, 04 Feb 2025 01:30:06 GMT"
},
{
"version": "v2",
"created": "Thu, 27 Feb 2025 05:17:57 GMT"
},
{
"version": "v3",
"created": "Fri, 14 Mar 2025 02:48:55 GMT"
},
{
"version": "v4",
"created": "Thu, 05 Jun 2025 05:50:15 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Yu",
"Dazhou",
""
],
[
"Bao",
"Riyang",
""
],
[
"Ning",
"Ruiyu",
""
],
[
"Peng",
"Jinghong",
""
],
[
"Mai",
"Gengchen",
""
],
[
"Zhao",
"Liang",
""
]
] |
2502.18504 | Aman Goel | Aman Goel, Xian Carrie Wu, Zhe Wang, Dmitriy Bespalov, Yanjun Qi | TurboFuzzLLM: Turbocharging Mutation-based Fuzzing for Effectively Jailbreaking Large Language Models in Practice | Oral presentation at NAACL 2025 industry track | null | null | null | cs.CR cs.AI cs.CL cs.LG | http://creativecommons.org/licenses/by/4.0/ | Jailbreaking large-language models (LLMs) involves testing their robustness against adversarial prompts and evaluating their ability to withstand prompt attacks that could elicit unauthorized or malicious responses. In this paper, we present TurboFuzzLLM, a mutation-based fuzzing technique for efficiently finding a collection of effective jailbreaking templates that, when combined with harmful questions, can lead a target LLM to produce harmful responses through black-box access via user prompts. We describe the limitations of directly applying existing template-based attacking techniques in practice, and present functional and efficiency-focused upgrades we added to mutation-based fuzzing to generate effective jailbreaking templates automatically. TurboFuzzLLM achieves $\geq$ 95\% attack success rates (ASR) on public datasets for leading LLMs (including GPT-4o \& GPT-4 Turbo), shows impressive generalizability to unseen harmful questions, and helps in improving model defenses to prompt attacks. TurboFuzzLLM is available open source at https://github.com/amazon-science/TurboFuzzLLM. | [
{
"version": "v1",
"created": "Fri, 21 Feb 2025 21:10:12 GMT"
},
{
"version": "v2",
"created": "Wed, 04 Jun 2025 23:08:28 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Goel",
"Aman",
""
],
[
"Wu",
"Xian Carrie",
""
],
[
"Wang",
"Zhe",
""
],
[
"Bespalov",
"Dmitriy",
""
],
[
"Qi",
"Yanjun",
""
]
] |
2502.18826 | Yuxiao Wen | Yuxiao Wen | Adversarial Combinatorial Semi-bandits with Graph Feedback | To appear in ICML 2025 | null | null | null | cs.LG cs.IT math.IT stat.ML | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | In combinatorial semi-bandits, a learner repeatedly selects from a combinatorial decision set of arms, receives the realized sum of rewards, and observes the rewards of the individual selected arms as feedback. In this paper, we extend this framework to include \emph{graph feedback}, where the learner observes the rewards of all neighboring arms of the selected arms in a feedback graph $G$. We establish that the optimal regret over a time horizon $T$ scales as $\widetilde{\Theta}(S\sqrt{T}+\sqrt{\alpha ST})$, where $S$ is the size of the combinatorial decisions and $\alpha$ is the independence number of $G$. This result interpolates between the known regrets $\widetilde\Theta(S\sqrt{T})$ under full information (i.e., $G$ is complete) and $\widetilde\Theta(\sqrt{KST})$ under the semi-bandit feedback (i.e., $G$ has only self-loops), where $K$ is the total number of arms. A key technical ingredient is to realize a convexified action using a random decision vector with negative correlations. We also show that online stochastic mirror descent (OSMD) that only realizes convexified actions in expectation is suboptimal. | [
{
"version": "v1",
"created": "Wed, 26 Feb 2025 05:01:11 GMT"
},
{
"version": "v2",
"created": "Fri, 28 Feb 2025 00:20:14 GMT"
},
{
"version": "v3",
"created": "Fri, 02 May 2025 03:02:10 GMT"
},
{
"version": "v4",
"created": "Thu, 05 Jun 2025 05:09:33 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Wen",
"Yuxiao",
""
]
] |
2502.19145 | Pierre Peign\'e-Lefebvre | Pierre Peigne-Lefebvre, Mikolaj Kniejski, Filip Sondej, Matthieu David, Jason Hoelscher-Obermaier, Christian Schroeder de Witt, Esben Kran | Multi-Agent Security Tax: Trading Off Security and Collaboration Capabilities in Multi-Agent Systems | Accepted to AAAI 2025 Conference | null | null | null | cs.AI cs.MA | http://creativecommons.org/licenses/by-nc-nd/4.0/ | As AI agents are increasingly adopted to collaborate on complex objectives, ensuring the security of autonomous multi-agent systems becomes crucial. We develop simulations of agents collaborating on shared objectives to study these security risks and security trade-offs. We focus on scenarios where an attacker compromises one agent, using it to steer the entire system toward misaligned outcomes by corrupting other agents. In this context, we observe infectious malicious prompts - the multi-hop spreading of malicious instructions. To mitigate this risk, we evaluated several strategies: two "vaccination" approaches that insert false memories of safely handling malicious input into the agents' memory stream, and two versions of a generic safety instruction strategy. While these defenses reduce the spread and fulfillment of malicious instructions in our experiments, they tend to decrease collaboration capability in the agent network. Our findings illustrate potential trade-off between security and collaborative efficiency in multi-agent systems, providing insights for designing more secure yet effective AI collaborations. | [
{
"version": "v1",
"created": "Wed, 26 Feb 2025 14:00:35 GMT"
},
{
"version": "v2",
"created": "Wed, 04 Jun 2025 18:25:06 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Peigne-Lefebvre",
"Pierre",
""
],
[
"Kniejski",
"Mikolaj",
""
],
[
"Sondej",
"Filip",
""
],
[
"David",
"Matthieu",
""
],
[
"Hoelscher-Obermaier",
"Jason",
""
],
[
"de Witt",
"Christian Schroeder",
""
],
[
"Kran",
"Esben",
""
]
] |
2502.19402 | Seungwook Han | Seungwook Han, Jyothish Pari, Samuel J. Gershman, Pulkit Agrawal | General Intelligence Requires Reward-based Pretraining | https://improbableai.notion.site/General-Intelligence-Requires-Reward-Based-Pretraining-2023b66e4cf580d3ab44c7860b75d25f?pvs=73 | null | null | null | cs.LG | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Large Language Models (LLMs) have demonstrated impressive real-world utility, exemplifying artificial useful intelligence (AUI). However, their ability to reason adaptively and robustly -- the hallmarks of artificial general intelligence (AGI) -- remains fragile. While LLMs seemingly succeed in commonsense reasoning, programming, and mathematics, they struggle to generalize algorithmic understanding across novel contexts. Our experiments with algorithmic tasks in esoteric programming languages reveal that LLM's reasoning overfits to the training data and is limited in its transferability. We hypothesize that the core issue underlying such limited transferability is the coupling of reasoning and knowledge in LLMs.
To transition from AUI to AGI, we propose disentangling knowledge and reasoning through three key directions: (1) pretaining to reason using RL from scratch as an alternative to the widely used next-token prediction pretraining, (2) using a curriculum of synthetic tasks to ease the learning of a reasoning prior for RL that can then be transferred to natural language tasks, and (3) learning more generalizable reasoning functions using a small context window to reduce exploiting spurious correlations between tokens. Such a reasoning system coupled with a trained retrieval system and a large external memory bank as a knowledge store can overcome several limitations of existing architectures at learning to reason in novel scenarios. | [
{
"version": "v1",
"created": "Wed, 26 Feb 2025 18:51:12 GMT"
},
{
"version": "v2",
"created": "Thu, 05 Jun 2025 01:41:47 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Han",
"Seungwook",
""
],
[
"Pari",
"Jyothish",
""
],
[
"Gershman",
"Samuel J.",
""
],
[
"Agrawal",
"Pulkit",
""
]
] |
2502.20044 | Jiwan Poudel | Jiwan Poudel, Alessandro Bacchetta, Jian-Ping Chen, Dustin Keller, Ishara Fernando, Elena Long, David Ruth, Nathaly Santiesteban, Kalf Slifer | Spin 1 Transverse Momentum Dependent Tensor Structure Functions in CLAS12 | null | null | null | null | hep-ph hep-ex | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We propose to analyze CLAS12 RG-C data to study the tensor transverse-momentum-dependent parton distribution functions (TMDs) on deuteron data. The deuteron is the lightest nucleus with spin-1, in essence a weakly bound system of two spin-1/2 nucleons. However, one of the most intriguing characteristics of the deuteron is that the tensor polarized structure provides direct access to the quark and gluon distribution of light nuclear system, which cannot be naively constructed from the proton and neutron. We will study the tensor polarized structure functions with the Semi-inclusive Deep Inelastic Scattering (SIDIS) $eD \arrow eP_{h}X$ and Inclusive processes in the available data on deuterated ammonia (ND3) target. We will perform the first ever SIDIS analysis extraction of the tensor structure functions, which can be interpreted in term of completely unexplored tensor polarized TMDs. Our analysis will focus on the extraction of the tensor structure functions b1 from inclusive process, and $F_{U(LL),T}$ and $F^{cos 2\phi_{h}}_{U(LL)}$ from SIDIS. These last two structure functions carry information related to two tensor-polarized TMDs, $f_{1LL}$ and $h^{\perp}_{1LL}$. These initial exploratory measurements of tensor-polarized structure functions will enable the first extraction of spin-1 TMDs and motivate more precise future measurements. | [
{
"version": "v1",
"created": "Thu, 27 Feb 2025 12:35:45 GMT"
},
{
"version": "v2",
"created": "Wed, 04 Jun 2025 22:33:08 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Poudel",
"Jiwan",
""
],
[
"Bacchetta",
"Alessandro",
""
],
[
"Chen",
"Jian-Ping",
""
],
[
"Keller",
"Dustin",
""
],
[
"Fernando",
"Ishara",
""
],
[
"Long",
"Elena",
""
],
[
"Ruth",
"David",
""
],
[
"Santiesteban",
"Nathaly",
""
],
[
"Slifer",
"Kalf",
""
]
] |
2502.20916 | Stefano Roberto Soleti | LiquidO Collaboration: S. R. Soleti, J. J. G\'omez-Cadenas, J. Apilluelo, L. Asquith, E. F. Bannister, N. P. Barradas, C. L. Baylis, J. L. Beney, M. Berberan e Santos, X. de la Bernardie, T. J. C. Bezerra, M. Bongrand, C. Bourgeois, D. Breton, J. Busto, K. Burns, A. Cabrera, A. Cadiou, E. Calvo, M. de Carlos Generowicz, E. Chauveau, B. J. Cattermole, M. Chen, P. Chimenti, D. F. Cowen, S. Kr. Das, S. Dusini, A. Earle, M. Felizardo, C. Frigerio Martins, J. Gal\'an, J. A. Garc\'ia, R. Gazzini, A. Gibson-Foster, C. Girard-Carillo, W. C. Griffith, M. Guiti\`ere, F. Haddad, J. Hartnell, A. Holin, I. G. Irastorza, I. Jovanovic, A. Kling, L. Koch, P. Lasorak, J. F. Le Du, F. Lefevre, P. Loaiza, J. A. Lock, G. Luz\'on, J. Maalmi, J. P. Malhado, F. Mantovani, J. G. Marques, C. Marquet, M. Mart\'inez, D. Navas-Nicol\'as, H. Nunokawa, J. P. Ochoa-Ricoux, T. Palmeira, C. Palomares, D. Petyt, P. Pillot, A. Pin, J. C. C. Porter, M. S. Pravikoff, S. Richards, N. Rodrigues, M. Roche, R. Rosero, B. Roskovec, N. Roy, M. L. Sarsa, A. Serafini, C. Shepherd-Themistocleous, W. Shorrock, M. Silva, L. Simard, D. Stocco, V. Strati, J. S. Stutzmann, F. Suekane, N. Tuccori, A. Verdugo, B. Viaud, S. M. Wakely, G. Wendel, A. S. Wilhelm, A. W. R. Wong, M. Yeh, F. Yermia | COCOA: a compact Compton camera for astrophysical observation of MeV-scale gamma rays | 14 pages, 17 figures | Astropart. Phys. 172 (2025) 103135 | 10.1016/j.astropartphys.2025.103135 | null | astro-ph.IM physics.ins-det | http://creativecommons.org/licenses/by/4.0/ | COCOA (COmpact COmpton cAmera) is a next-generation gamma-ray telescope designed for astrophysical observations in the MeV energy range. The detector comprises a scatterer volume employing the LiquidO detection technology and an array of scintillating crystals acting as absorber. Surrounding plastic scintillator panels serve as a veto system for charged particles. The detector's compact, scalable design enables flexible deployment on microsatellites or high-altitude balloons. Gamma rays at MeV energies have not been well explored historically (the so-called "MeV gap") and COCOA has the potential to improve the sensitivity in this energy band. | [
{
"version": "v1",
"created": "Fri, 28 Feb 2025 10:17:07 GMT"
},
{
"version": "v2",
"created": "Tue, 25 Mar 2025 04:12:27 GMT"
},
{
"version": "v3",
"created": "Mon, 12 May 2025 17:52:35 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"LiquidO Collaboration",
"",
""
],
[
"Soleti",
"S. R.",
""
],
[
"Gómez-Cadenas",
"J. J.",
""
],
[
"Apilluelo",
"J.",
""
],
[
"Asquith",
"L.",
""
],
[
"Bannister",
"E. F.",
""
],
[
"Barradas",
"N. P.",
""
],
[
"Baylis",
"C. L.",
""
],
[
"Beney",
"J. L.",
""
],
[
"Santos",
"M. Berberan e",
""
],
[
"de la Bernardie",
"X.",
""
],
[
"Bezerra",
"T. J. C.",
""
],
[
"Bongrand",
"M.",
""
],
[
"Bourgeois",
"C.",
""
],
[
"Breton",
"D.",
""
],
[
"Busto",
"J.",
""
],
[
"Burns",
"K.",
""
],
[
"Cabrera",
"A.",
""
],
[
"Cadiou",
"A.",
""
],
[
"Calvo",
"E.",
""
],
[
"Generowicz",
"M. de Carlos",
""
],
[
"Chauveau",
"E.",
""
],
[
"Cattermole",
"B. J.",
""
],
[
"Chen",
"M.",
""
],
[
"Chimenti",
"P.",
""
],
[
"Cowen",
"D. F.",
""
],
[
"Das",
"S. Kr.",
""
],
[
"Dusini",
"S.",
""
],
[
"Earle",
"A.",
""
],
[
"Felizardo",
"M.",
""
],
[
"Martins",
"C. Frigerio",
""
],
[
"Galán",
"J.",
""
],
[
"García",
"J. A.",
""
],
[
"Gazzini",
"R.",
""
],
[
"Gibson-Foster",
"A.",
""
],
[
"Girard-Carillo",
"C.",
""
],
[
"Griffith",
"W. C.",
""
],
[
"Guitière",
"M.",
""
],
[
"Haddad",
"F.",
""
],
[
"Hartnell",
"J.",
""
],
[
"Holin",
"A.",
""
],
[
"Irastorza",
"I. G.",
""
],
[
"Jovanovic",
"I.",
""
],
[
"Kling",
"A.",
""
],
[
"Koch",
"L.",
""
],
[
"Lasorak",
"P.",
""
],
[
"Du",
"J. F. Le",
""
],
[
"Lefevre",
"F.",
""
],
[
"Loaiza",
"P.",
""
],
[
"Lock",
"J. A.",
""
],
[
"Luzón",
"G.",
""
],
[
"Maalmi",
"J.",
""
],
[
"Malhado",
"J. P.",
""
],
[
"Mantovani",
"F.",
""
],
[
"Marques",
"J. G.",
""
],
[
"Marquet",
"C.",
""
],
[
"Martínez",
"M.",
""
],
[
"Navas-Nicolás",
"D.",
""
],
[
"Nunokawa",
"H.",
""
],
[
"Ochoa-Ricoux",
"J. P.",
""
],
[
"Palmeira",
"T.",
""
],
[
"Palomares",
"C.",
""
],
[
"Petyt",
"D.",
""
],
[
"Pillot",
"P.",
""
],
[
"Pin",
"A.",
""
],
[
"Porter",
"J. C. C.",
""
],
[
"Pravikoff",
"M. S.",
""
],
[
"Richards",
"S.",
""
],
[
"Rodrigues",
"N.",
""
],
[
"Roche",
"M.",
""
],
[
"Rosero",
"R.",
""
],
[
"Roskovec",
"B.",
""
],
[
"Roy",
"N.",
""
],
[
"Sarsa",
"M. L.",
""
],
[
"Serafini",
"A.",
""
],
[
"Shepherd-Themistocleous",
"C.",
""
],
[
"Shorrock",
"W.",
""
],
[
"Silva",
"M.",
""
],
[
"Simard",
"L.",
""
],
[
"Stocco",
"D.",
""
],
[
"Strati",
"V.",
""
],
[
"Stutzmann",
"J. S.",
""
],
[
"Suekane",
"F.",
""
],
[
"Tuccori",
"N.",
""
],
[
"Verdugo",
"A.",
""
],
[
"Viaud",
"B.",
""
],
[
"Wakely",
"S. M.",
""
],
[
"Wendel",
"G.",
""
],
[
"Wilhelm",
"A. S.",
""
],
[
"Wong",
"A. W. R.",
""
],
[
"Yeh",
"M.",
""
],
[
"Yermia",
"F.",
""
]
] |
2502.21015 | Jaydeb Sarkar | Nilanjan Das, Soma Das, and Jaydeb Sarkar | Invariant subspaces and the $C_{00}$-property of Brownian Shifts | 26 pages. Thoroughly revised, and the results are now developed in the vector-valued setting | null | null | null | math.PR math.CV math.FA math.OA | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We introduce Brownian shifts on vector-valued Hardy spaces and describe their invariant subspaces. We then consider the restriction of Brownian shifts to their invariant subspaces and classify when they are unitarily equivalent. Additionally, we prove an asymptotic property stating that normalized Brownian shifts belong to the classical $C_{00}$-class. | [
{
"version": "v1",
"created": "Fri, 28 Feb 2025 13:01:53 GMT"
},
{
"version": "v2",
"created": "Thu, 05 Jun 2025 07:03:29 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Das",
"Nilanjan",
""
],
[
"Das",
"Soma",
""
],
[
"Sarkar",
"Jaydeb",
""
]
] |
2503.00038 | Yu Yan | Yu Yan, Sheng Sun, Zenghao Duan, Teli Liu, Min Liu, Zhiyi Yin, Jiangyu Lei, Qi Li | From Benign import Toxic: Jailbreaking the Language Model via Adversarial Metaphors | arXiv admin note: substantial text overlap with arXiv:2412.12145 | null | null | null | cs.CL cs.AI cs.CR | http://creativecommons.org/licenses/by-nc-sa/4.0/ | Current studies have exposed the risk of Large Language Models (LLMs) generating harmful content by jailbreak attacks. However, they overlook that the direct generation of harmful content from scratch is more difficult than inducing LLM to calibrate benign content into harmful forms. In our study, we introduce a novel attack framework that exploits AdVersArial meTAphoR (AVATAR) to induce the LLM to calibrate malicious metaphors for jailbreaking. Specifically, to answer harmful queries, AVATAR adaptively identifies a set of benign but logically related metaphors as the initial seed. Then, driven by these metaphors, the target LLM is induced to reason and calibrate about the metaphorical content, thus jailbroken by either directly outputting harmful responses or calibrating residuals between metaphorical and professional harmful content. Experimental results demonstrate that AVATAR can effectively and transferable jailbreak LLMs and achieve a state-of-the-art attack success rate across multiple advanced LLMs. | [
{
"version": "v1",
"created": "Tue, 25 Feb 2025 08:41:25 GMT"
},
{
"version": "v2",
"created": "Thu, 05 Jun 2025 17:10:34 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Yan",
"Yu",
""
],
[
"Sun",
"Sheng",
""
],
[
"Duan",
"Zenghao",
""
],
[
"Liu",
"Teli",
""
],
[
"Liu",
"Min",
""
],
[
"Yin",
"Zhiyi",
""
],
[
"Lei",
"Jiangyu",
""
],
[
"Li",
"Qi",
""
]
] |
2503.00271 | Kelechi Gabriel Kalu | Kelechi G. Kalu, Sofia Okorafor, Tanmay Singla, Santiago Torres-Arias, James C. Davis | Why Johnny Signs with Next-Generation Tools: A Usability Case Study of Sigstore | 21 Pages | null | null | null | cs.SE cs.CR | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Software signing is the most robust method for ensuring the integrity and authenticity of components in a software supply chain. However, traditional signing tools place key management and signer identification burdens on practitioners, leading to both security vulnerabilities and usability challenges. Next-generation signing tools such as Sigstore have automated some of these concerns, but little is known about their usability and adoption dynamics. This knowledge gap hampers the integration of signing into the software engineering process.
To fill this gap, we conducted a usability study of Sigstore, a pioneering and widely adopted exemplar in this space. Through 18 interviews, we explored (1) the factors practitioners consider when selecting a signing tool, (2) the problems and advantages associated with the tooling choices of practitioners, and (3) practitioners' signing-tool usage has evolved over time. Our findings illuminate the usability factors of next-generation signing tools and yield recommendations for toolmakers, including: (1) enhance integration flexibility through officially supported plugins and APIs, and (2) balance transparency with privacy by offering configurable logging options for enterprise use. | [
{
"version": "v1",
"created": "Sat, 01 Mar 2025 00:59:18 GMT"
},
{
"version": "v2",
"created": "Tue, 04 Mar 2025 19:55:28 GMT"
},
{
"version": "v3",
"created": "Fri, 14 Mar 2025 00:30:15 GMT"
},
{
"version": "v4",
"created": "Wed, 04 Jun 2025 22:10:22 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Kalu",
"Kelechi G.",
""
],
[
"Okorafor",
"Sofia",
""
],
[
"Singla",
"Tanmay",
""
],
[
"Torres-Arias",
"Santiago",
""
],
[
"Davis",
"James C.",
""
]
] |
2503.00555 | Tiansheng Huang | Tiansheng Huang, Sihao Hu, Fatih Ilhan, Selim Furkan Tekin, Zachary Yahn, Yichang Xu, Ling Liu | Safety Tax: Safety Alignment Makes Your Large Reasoning Models Less Reasonable | null | null | null | null | cs.CR cs.AI cs.LG | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Safety alignment is an important procedure before the official deployment of a Large Language Model (LLM). While safety alignment has been extensively studied for LLM, there is still a large research gap for Large Reasoning Models (LRMs) that equip with improved reasoning capability. We in this paper systematically examine a simplified pipeline for producing safety aligned LRMs. With our evaluation of various LRMs, we deliver two main findings: i) Safety alignment can be done upon the LRM to restore its safety capability. ii) Safety alignment leads to a degradation of the reasoning capability of LRMs. The two findings show that there exists a trade-off between reasoning and safety capability with the sequential LRM production pipeline. The discovered trade-off, which we name Safety Tax, should shed light on future endeavors of safety research on LRMs. As a by-product, we curate a dataset called DirectRefusal, which might serve as an alternative dataset for safety alignment. Our source code is available at https://github.com/git-disl/Safety-Tax. | [
{
"version": "v1",
"created": "Sat, 01 Mar 2025 16:42:01 GMT"
},
{
"version": "v2",
"created": "Thu, 05 Jun 2025 03:20:54 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Huang",
"Tiansheng",
""
],
[
"Hu",
"Sihao",
""
],
[
"Ilhan",
"Fatih",
""
],
[
"Tekin",
"Selim Furkan",
""
],
[
"Yahn",
"Zachary",
""
],
[
"Xu",
"Yichang",
""
],
[
"Liu",
"Ling",
""
]
] |
2503.00847 | Yanran Chen | Moritz Altemeyer, Steffen Eger, Johannes Daxenberger, Yanran Chen, Tim Altendorf, Philipp Cimiano, Benjamin Schiller | Argument Summarization and its Evaluation in the Era of Large Language Models | null | null | null | null | cs.CL | http://creativecommons.org/licenses/by-sa/4.0/ | Large Language Models (LLMs) have revolutionized various Natural Language Generation (NLG) tasks, including Argument Summarization (ArgSum), a key subfield of Argument Mining (AM). This paper investigates the integration of state-of-the-art LLMs into ArgSum, including for its evaluation. In particular, we propose a novel prompt-based evaluation scheme, and validate it through a novel human benchmark dataset. Our work makes three main contributions: (i) the integration of LLMs into existing ArgSum frameworks, (ii) the development of a new LLM-based ArgSum system, benchmarked against prior methods, and (iii) the introduction of an advanced LLM-based evaluation scheme. We demonstrate that the use of LLMs substantially improves both the generation and evaluation of argument summaries, achieving state-of-the-art results and advancing the field of ArgSum. We also show that among the four LLMs integrated in (i) and (ii), Qwen-3-32B, despite having the fewest parameters, performs best, even surpassing GPT-4o, while LLaMA-3.3-70B consistently underperforms. | [
{
"version": "v1",
"created": "Sun, 02 Mar 2025 10:49:10 GMT"
},
{
"version": "v2",
"created": "Mon, 17 Mar 2025 20:25:48 GMT"
},
{
"version": "v3",
"created": "Thu, 05 Jun 2025 14:03:53 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Altemeyer",
"Moritz",
""
],
[
"Eger",
"Steffen",
""
],
[
"Daxenberger",
"Johannes",
""
],
[
"Chen",
"Yanran",
""
],
[
"Altendorf",
"Tim",
""
],
[
"Cimiano",
"Philipp",
""
],
[
"Schiller",
"Benjamin",
""
]
] |
2503.01678 | Sushanth Varada | Sushanth Varada, Christian Sp\r{a}nsl\"att, Matteo Acciai | Exchange-phase erasure in anyonic Hong-Ou-Mandel interferometry | 6 + 4 pages, 4 figures | Phys. Rev. B 111, L201407 (2025) | 10.1103/PhysRevB.111.L201407 | null | cond-mat.mes-hall cond-mat.str-el | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Two-particle interferometry is an important tool for extracting the exchange statistics of quantum particles. We theoretically investigate the prospects of such interferometry to probe the statistics of point-like anyonic excitations injected in a Hong-Ou-Mandel (HOM) setup based on a quantum point contact device in the fractional quantum Hall regime. We compute the standard HOM ratio, i.e., the ratio of tunneling noises for two- and one-particle injections, and find that for point-like anyons, it only depends on the temperature and the anyon scaling dimension. Importantly, the latter is not necessarily related to the exchange phase. In fact, we establish that the HOM ratio does not reveal the exchange phase of the injected anyons: For injection-time delays that are small compared to the thermal time scale, we find that the exchange phase accumulated due to time-domain braiding between injected and thermally activated anyons is erased due to two mutually canceling sub-processes. In contrast, for time delays large compared to the thermal time, only a single sub-process contributes to the braiding, but the accumulated phase is canceled in the HOM ratio. These findings suggest caution when interpreting HOM interferometry experiments with anyons and approaches beyond the standard HOM ratio are thus necessary to extract anyonic statistics with two-particle interferometry experiments. | [
{
"version": "v1",
"created": "Mon, 03 Mar 2025 15:50:50 GMT"
},
{
"version": "v2",
"created": "Thu, 05 Jun 2025 12:04:00 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Varada",
"Sushanth",
""
],
[
"Spånslätt",
"Christian",
""
],
[
"Acciai",
"Matteo",
""
]
] |
2503.01779 | Luca Fabrizio Di Cerbo | Luca F. Di Cerbo, Alexander Dranishnikov, Ekansh Jauhari | Curvature, macroscopic dimensions, and symmetric products of surfaces | We add Remark 5.3 concerning the holomorphic bisectional curvature. We add more details for the proof of Proposition 5.12 (Proposition 5.11 in version 3), and we highlight the fact that the case g=1 is special. Finally, we add Corollary 7.9. 41 pages, no figures | null | null | MPIM-Bonn-2024 | math.GT math.AG math.DG | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We present a detailed study of the curvature and symplectic asphericity properties of symmetric products of surfaces. We show that these spaces can be used to answer nuanced questions arising in the study of closed Riemannian manifolds with positive scalar curvature. For example, we prove that symmetric products of surfaces sharply distinguish between two distinct notions of macroscopic dimension introduced by Gromov and the second-named author. As a natural generalization of this circle of ideas, we address the Gromov--Lawson and Gromov conjectures in the Kaehler projective setting and draw new connections between the theories of the minimal model, positivity in algebraic geometry, and macroscopic dimensions. | [
{
"version": "v1",
"created": "Mon, 03 Mar 2025 18:10:08 GMT"
},
{
"version": "v2",
"created": "Wed, 12 Mar 2025 18:02:57 GMT"
},
{
"version": "v3",
"created": "Tue, 25 Mar 2025 15:39:17 GMT"
},
{
"version": "v4",
"created": "Thu, 05 Jun 2025 16:57:54 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Di Cerbo",
"Luca F.",
""
],
[
"Dranishnikov",
"Alexander",
""
],
[
"Jauhari",
"Ekansh",
""
]
] |
2503.01821 | Xingyu Zhu | Xingyu Zhu, Abhishek Panigrahi, Sanjeev Arora | On the Power of Context-Enhanced Learning in LLMs | 77 pages, 18 figures; ICML 2025 Main Conference | null | null | null | cs.LG | http://creativecommons.org/licenses/by-sa/4.0/ | We formalize a new concept for LLMs, context-enhanced learning. It involves standard gradient-based learning on text except that the context is enhanced with additional data on which no auto-regressive gradients are computed. This setting is a gradient-based analog of usual in-context learning (ICL) and appears in some recent works. Using a multi-step reasoning task, we prove in a simplified setting that context-enhanced learning can be exponentially more sample-efficient than standard learning when the model is capable of ICL. At a mechanistic level, we find that the benefit of context-enhancement arises from a more accurate gradient learning signal. We also experimentally demonstrate that it appears hard to detect or recover learning materials that were used in the context during training. This may have implications for data security as well as copyright. | [
{
"version": "v1",
"created": "Mon, 03 Mar 2025 18:46:45 GMT"
},
{
"version": "v2",
"created": "Wed, 04 Jun 2025 22:53:41 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Zhu",
"Xingyu",
""
],
[
"Panigrahi",
"Abhishek",
""
],
[
"Arora",
"Sanjeev",
""
]
] |
2503.02197 | Ming Li | Zhixun Chen, Ming Li, Yuxuan Huang, Yali Du, Meng Fang, Tianyi Zhou | ATLaS: Agent Tuning via Learning Critical Steps | ACL2025, Camera-ready | null | null | null | cs.CL cs.AI cs.LG | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Large Language Model (LLM) agents have demonstrated remarkable generalization capabilities across multi-domain tasks. Existing agent tuning approaches typically employ supervised finetuning on entire expert trajectories. However, behavior-cloning of full trajectories can introduce expert bias and weaken generalization to states not covered by the expert data. Additionally, critical steps, such as planning, complex reasoning for intermediate subtasks, and strategic decision-making, are essential to success in agent tasks, so learning these steps is the key to improving LLM agents. For more effective and efficient agent tuning, we propose ATLaS that identifies the critical steps in expert trajectories and finetunes LLMs solely on these steps with reduced costs. By steering the training's focus to a few critical steps, our method mitigates the risk of overfitting entire trajectories and promotes generalization across different environments and tasks. In extensive experiments, an LLM finetuned on only 30% critical steps selected by ATLaS outperforms the LLM finetuned on all steps and recent open-source LLM agents. ATLaS maintains and improves base LLM skills as generalist agents interacting with diverse environments. | [
{
"version": "v1",
"created": "Tue, 04 Mar 2025 02:14:55 GMT"
},
{
"version": "v2",
"created": "Thu, 05 Jun 2025 01:42:08 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Chen",
"Zhixun",
""
],
[
"Li",
"Ming",
""
],
[
"Huang",
"Yuxuan",
""
],
[
"Du",
"Yali",
""
],
[
"Fang",
"Meng",
""
],
[
"Zhou",
"Tianyi",
""
]
] |
2503.02407 | Nikita Kazeev | Nikita Kazeev, Wei Nong, Ignat Romanov, Ruiming Zhu, Andrey Ustyuzhanin, Shuya Yamazaki, Kedar Hippalgaonkar | Wyckoff Transformer: Generation of Symmetric Crystals | https://github.com/SymmetryAdvantage/WyckoffTransformer | Proceedings of the 42nd International Conference on Machine Learning, Vancouver, Canada. PMLR 267, 2025 | null | null | cond-mat.mtrl-sci cs.LG physics.comp-ph | http://creativecommons.org/licenses/by/4.0/ | Crystal symmetry plays a fundamental role in determining its physical, chemical, and electronic properties such as electrical and thermal conductivity, optical and polarization behavior, and mechanical strength. Almost all known crystalline materials have internal symmetry. However, this is often inadequately addressed by existing generative models, making the consistent generation of stable and symmetrically valid crystal structures a significant challenge. We introduce WyFormer, a generative model that directly tackles this by formally conditioning on space group symmetry. It achieves this by using Wyckoff positions as the basis for an elegant, compressed, and discrete structure representation. To model the distribution, we develop a permutation-invariant autoregressive model based on the Transformer encoder and an absence of positional encoding. Extensive experimentation demonstrates WyFormer's compelling combination of attributes: it achieves best-in-class symmetry-conditioned generation, incorporates a physics-motivated inductive bias, produces structures with competitive stability, predicts material properties with competitive accuracy even without atomic coordinates, and exhibits unparalleled inference speed. | [
{
"version": "v1",
"created": "Tue, 04 Mar 2025 08:50:10 GMT"
},
{
"version": "v2",
"created": "Fri, 07 Mar 2025 07:46:34 GMT"
},
{
"version": "v3",
"created": "Wed, 04 Jun 2025 08:26:03 GMT"
},
{
"version": "v4",
"created": "Thu, 05 Jun 2025 07:50:47 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Kazeev",
"Nikita",
""
],
[
"Nong",
"Wei",
""
],
[
"Romanov",
"Ignat",
""
],
[
"Zhu",
"Ruiming",
""
],
[
"Ustyuzhanin",
"Andrey",
""
],
[
"Yamazaki",
"Shuya",
""
],
[
"Hippalgaonkar",
"Kedar",
""
]
] |
2503.03076 | Dr. Goutam Manna | Samit Ganguly, Goutam Manna, Debashis Gangopadhyay, Eduardo Guendelman and Abhijit Bhattacharyya | Non-Affine Extensions of the Raychaudhuri Equation in the K-essence Framework | 15 pages, 6 figures, 4 tables, Accepted from Physics Letters B | Physics Letters B (2025) | 10.1016/j.physletb.2025.139622 | null | gr-qc hep-th | http://creativecommons.org/licenses/by-sa/4.0/ | We present a new avenue of the Raychaudhuri Equation (RE) by introducing a non-affine parametrization within the k-essence framework. This modification accounts for non-geodesic flow curves, leading to emergent repulsive effects in cosmic evolution. Using a DBI-type k-essence Lagrangian, we derive a modified RE and demonstrate its ability to address the Hubble tension while predicting a natural emergence of a dynamical dark energy equation of state. Our Bayesian analysis, constrained by cosmological data, supports the theoretical scaling relation of the k-essence field ($\dot{\phi}$) and the cosmic scale factor ($a$). Furthermore, we reinterpret the modified RE as an anti-damped harmonic oscillator, we found a caustic avoidance signature, it may reveal classical or quantum-like effects in cosmic expansion. These results suggest a deep connection between scalar field dynamics and modified gravity, offering new perspectives on the nature of the expansion history of the universe. | [
{
"version": "v1",
"created": "Wed, 05 Mar 2025 00:31:42 GMT"
},
{
"version": "v2",
"created": "Thu, 05 Jun 2025 03:26:48 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Ganguly",
"Samit",
""
],
[
"Manna",
"Goutam",
""
],
[
"Gangopadhyay",
"Debashis",
""
],
[
"Guendelman",
"Eduardo",
""
],
[
"Bhattacharyya",
"Abhijit",
""
]
] |
2503.03417 | Jabez Magomere Jm | Jabez Magomere, Emanuele La Malfa, Manuel Tonneau, Ashkan Kazemi, Scott Hale | When Claims Evolve: Evaluating and Enhancing the Robustness of Embedding Models Against Misinformation Edits | Accepted to ACL 2025 Findings | null | null | null | cs.CL cs.AI | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Online misinformation remains a critical challenge, and fact-checkers increasingly rely on claim matching systems that use sentence embedding models to retrieve relevant fact-checks. However, as users interact with claims online, they often introduce edits, and it remains unclear whether current embedding models used in retrieval are robust to such edits. To investigate this, we introduce a perturbation framework that generates valid and natural claim variations, enabling us to assess the robustness of a wide-range of sentence embedding models in a multi-stage retrieval pipeline and evaluate the effectiveness of various mitigation approaches. Our evaluation reveals that standard embedding models exhibit notable performance drops on edited claims, while LLM-distilled embedding models offer improved robustness at a higher computational cost. Although a strong reranker helps to reduce the performance drop, it cannot fully compensate for first-stage retrieval gaps. To address these retrieval gaps, we evaluate train- and inference-time mitigation approaches, demonstrating that they can improve in-domain robustness by up to 17 percentage points and boost out-of-domain generalization by 10 percentage points. Overall, our findings provide practical improvements to claim-matching systems, enabling more reliable fact-checking of evolving misinformation. Code and data are available at https://github.com/JabezNzomo99/claim-matching-robustness. | [
{
"version": "v1",
"created": "Wed, 05 Mar 2025 11:47:32 GMT"
},
{
"version": "v2",
"created": "Thu, 06 Mar 2025 11:00:35 GMT"
},
{
"version": "v3",
"created": "Thu, 05 Jun 2025 10:17:20 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Magomere",
"Jabez",
""
],
[
"La Malfa",
"Emanuele",
""
],
[
"Tonneau",
"Manuel",
""
],
[
"Kazemi",
"Ashkan",
""
],
[
"Hale",
"Scott",
""
]
] |
2503.03758 | Ken-ichiro Arita | Kenichiro Arita | Classical periodic orbits in extended phase space for spherical harmonic oscillator with spin-orbit coupling | 10 pages, 9 figures, published version | Phys. Rev. E 111 (2025) 064204 | 10.1103/PhysRevE.111.064204 | null | nlin.CD nucl-th | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | A complete analysis of classical periodic orbits (POs) and their bifurcations was conducted in spherical harmonic oscillator system with spin-orbit coupling. The motion of the spin is explicitly considered using the spin canonical variables derived by semiclassical approximation to the spin coherent state path integral representation. In addition to the diametric and two circular PO families with frozen spin, solutions that bridge two circular POs are found in which orbital motion is coupled to spin precession. In addition, each bridge encounters a secondary bifurcation on the way from one circular PO to the other and generates a new PO, that survives at higher energies while maintaining a constant period. The generic expressions for those POs are obtained explicitly, and all the above peculiar bifurcation scenarios are described fully analytically. | [
{
"version": "v1",
"created": "Tue, 18 Feb 2025 07:41:52 GMT"
},
{
"version": "v2",
"created": "Wed, 04 Jun 2025 22:53:14 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Arita",
"Kenichiro",
""
]
] |
2503.04256 | Haotian Fu | Yixiang Sun, Haotian Fu, Michael Littman, George Konidaris | Knowledge Retention for Continual Model-Based Reinforcement Learning | null | null | null | null | cs.LG cs.AI | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We propose DRAGO, a novel approach for continual model-based reinforcement learning aimed at improving the incremental development of world models across a sequence of tasks that differ in their reward functions but not the state space or dynamics. DRAGO comprises two key components: Synthetic Experience Rehearsal, which leverages generative models to create synthetic experiences from past tasks, allowing the agent to reinforce previously learned dynamics without storing data, and Regaining Memories Through Exploration, which introduces an intrinsic reward mechanism to guide the agent toward revisiting relevant states from prior tasks. Together, these components enable the agent to maintain a comprehensive and continually developing world model, facilitating more effective learning and adaptation across diverse environments. Empirical evaluations demonstrate that DRAGO is able to preserve knowledge across tasks, achieving superior performance in various continual learning scenarios. | [
{
"version": "v1",
"created": "Thu, 06 Mar 2025 09:38:14 GMT"
},
{
"version": "v2",
"created": "Tue, 03 Jun 2025 20:10:44 GMT"
},
{
"version": "v3",
"created": "Thu, 05 Jun 2025 02:54:12 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Sun",
"Yixiang",
""
],
[
"Fu",
"Haotian",
""
],
[
"Littman",
"Michael",
""
],
[
"Konidaris",
"George",
""
]
] |
2503.04647 | Wen Yang | Wen Yang, Junhong Wu, Chen Wang, Chengqing Zong, Jiajun Zhang | Implicit Cross-Lingual Rewarding for Efficient Multilingual Preference Alignment | Camera ready version for ACL 2025 Findings | null | null | null | cs.CL cs.AI | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Direct Preference Optimization (DPO) has become a prominent method for aligning Large Language Models (LLMs) with human preferences. While DPO has enabled significant progress in aligning English LLMs, multilingual preference alignment is hampered by data scarcity. To address this, we propose a novel approach that $\textit{captures}$ learned preferences from well-aligned English models by implicit rewards and $\textit{transfers}$ them to other languages through iterative training. Specifically, we derive an implicit reward model from the logits of an English DPO-aligned model and its corresponding reference model. This reward model is then leveraged to annotate preference relations in cross-lingual instruction-following pairs, using English instructions to evaluate multilingual responses. The annotated data is subsequently used for multilingual DPO fine-tuning, facilitating preference knowledge transfer from English to other languages. Fine-tuning Llama3 for two iterations resulted in a 12.72% average improvement in Win Rate and a 5.97% increase in Length Control Win Rate across all training languages on the X-AlpacaEval leaderboard. Our findings demonstrate that leveraging existing English-aligned models can enable efficient and effective multilingual preference alignment, significantly reducing the need for extensive multilingual preference data. The code is available at https://github.com/ZNLP/Implicit-Cross-Lingual-Rewarding | [
{
"version": "v1",
"created": "Thu, 06 Mar 2025 17:33:01 GMT"
},
{
"version": "v2",
"created": "Thu, 05 Jun 2025 02:40:44 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Yang",
"Wen",
""
],
[
"Wu",
"Junhong",
""
],
[
"Wang",
"Chen",
""
],
[
"Zong",
"Chengqing",
""
],
[
"Zhang",
"Jiajun",
""
]
] |
2503.05408 | P\'al Andr\'as Papp | Toni B\"ohnlein, P\'al Andr\'as Papp, Raphael S. Steiner, Christos K. Matzoros, A. N. Yzelman | Efficient Parallel Scheduling for Sparse Triangular Solvers | null | null | null | null | cs.DC | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We develop and analyze new scheduling algorithms for solving sparse triangular linear systems (SpTRSV) in parallel. Our approach produces highly efficient synchronous schedules for the forward- and backward-substitution algorithm. Compared to state-of-the-art baselines HDagg and SpMP, we achieve a $3.32 \times$ and $1.42 \times$ geometric-mean speed-up, respectively. We achieve this by obtaining an up to $12.07 \times$ geometric-mean reduction in the number of synchronization barriers over HDagg, whilst maintaining a balanced workload, and by applying a matrix reordering step for locality. We show that our improvements are consistent across a variety of input matrices and hardware architectures. | [
{
"version": "v1",
"created": "Fri, 07 Mar 2025 13:29:12 GMT"
},
{
"version": "v2",
"created": "Thu, 05 Jun 2025 12:31:43 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Böhnlein",
"Toni",
""
],
[
"Papp",
"Pál András",
""
],
[
"Steiner",
"Raphael S.",
""
],
[
"Matzoros",
"Christos K.",
""
],
[
"Yzelman",
"A. N.",
""
]
] |
2503.05714 | Rozin Hasin | Rozin Hasin | Prosthetics of the Indian State: The e-Shram Portal for Unorganized Workers in India | null | null | null | null | cs.CY cs.ET cs.IR cs.SY eess.SY | http://creativecommons.org/licenses/by-nc-nd/4.0/ | This research paper examines the digital portal/database for unorganized workers in the informal sector economy of India today: e-Shram. Using affordance theory, I criticize the operationalization of this database for the labourers, alongside problems of accessibility and perception. | [
{
"version": "v1",
"created": "Sun, 16 Feb 2025 10:27:29 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Hasin",
"Rozin",
""
]
] |
2503.05740 | Zhengbang Yang | Zhengbang Yang, Junyuan Hong, Yijiang Pang, Jiayu Zhou, Zhuangdi Zhu | ChatWise: A Strategy-Guided Chatbot for Enhancing Cognitive Support in Older Adults | null | null | null | null | cs.CY cs.AI cs.CL | http://creativecommons.org/licenses/by/4.0/ | Cognitive health in older adults presents a growing challenge. Although conversational interventions show feasibility in improving cognitive wellness, human caregiver resources remain overloaded. AI-based chatbots have shown promise, yet existing work is often limited to implicit strategies or heavily depends on training and label resources. In response, we propose a strategy-guided AI chatbot named ChatWise that follows a dual-level conversation reasoning framework. It integrates macro-level strategy planning and micro-level utterance generation to enable engaging, multi-turn dialogue tailored to older adults. Empirical results show that ChatWise closely aligns with professional human caregiver behaviors in offline evaluation using real clinic data, and achieves positive user cognitive and emotional responses in interactive simulations with digital twins, which significantly outperforms AI baselines that follow implicit conversation generation. | [
{
"version": "v1",
"created": "Wed, 19 Feb 2025 21:32:09 GMT"
},
{
"version": "v2",
"created": "Thu, 05 Jun 2025 01:16:33 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Yang",
"Zhengbang",
""
],
[
"Hong",
"Junyuan",
""
],
[
"Pang",
"Yijiang",
""
],
[
"Zhou",
"Jiayu",
""
],
[
"Zhu",
"Zhuangdi",
""
]
] |
2503.05927 | Dasha Zaremba | Daria Zaremba, Kim Venn, Christian R. Hayes, Rapha\"el Errani, Triana Cornejo, Jennifer Glover, Jaclyn Jensen, Alan W. McConnachie, Julio F. Navarro, John Pazder, Federico Sestito, Andr\'e Anthony, Dave Andersen, Gabriella Baker, Timothy Chin, Vladimir Churilov, Ruben Diaz, Tony Farrell, Veronica Firpo, Manuel Gomez-Jimenez, David Henderson, Venu M. Kalari, Jon Lawrence, Steve Margheim, Bryan Miller, J. Gordon Robertson, Roque Ruiz-Carmona, Katherine Silversides, Karleyne Silva, Peter J. Young, and Ross Zhelem | GHOST commissioning science results -- IV: Chemodynamical analyses of Milky Way satellites Sagittarius II and Aquarius II | 32 pages, 25 figures, 6 tables | null | null | null | astro-ph.GA astro-ph.CO astro-ph.IM | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We present Gemini/GHOST high-resolution spectra of five stars observed in two low surface brightness Milky Way satellites, Sagittarius II (Sgr2) and Aquarius II (Aqu2). For Aqu2, the velocities and metallicities of the two stars are consistent with membership in a dark matter-dominated ultra faint dwarf galaxy (UFD). The chemical abundance ratios suggest inefficient star formation from only one or a few supernovae (e.g., low Na, Sr, Ba), and enriched potassium (K) from super-AGB stars. For Sgr2, the velocity and metallicity dispersions of its members are not clearly resolved and our detailed chemical abundances show typical ratios for metal-poor stars, with low dispersions. There is only one exception - we report the discovery of an r-process enhanced star (Sgr2584, [Eu/Fe]= $+0.7 \pm 0.2 $; thus, an r-I star). As r-I stars are found in both UFDs (Tuc III, Tuc IV, Grus II) and globular clusters (M15 and M92), then this does not help to further classify the nature of Sgr2. Our exploration of Sgr2 demonstrates the difficulty in classifying some of the faintest (ambiguous) satellites. We advocate for additional diagnostics in analysing the ambiguous systems, such as exploring radial segregation (by mass and/or chemistry), N-body simulations, and the need for dark matter to survive Galactic tidal effects. The spectra analysed in this paper were taken as part of the GHOST commissioning observations, testing faint observation limits (G < 18.8) and the single and double IFU observing modes. | [
{
"version": "v1",
"created": "Fri, 07 Mar 2025 20:48:22 GMT"
},
{
"version": "v2",
"created": "Fri, 09 May 2025 16:11:31 GMT"
},
{
"version": "v3",
"created": "Wed, 04 Jun 2025 20:42:46 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Zaremba",
"Daria",
""
],
[
"Venn",
"Kim",
""
],
[
"Hayes",
"Christian R.",
""
],
[
"Errani",
"Raphaël",
""
],
[
"Cornejo",
"Triana",
""
],
[
"Glover",
"Jennifer",
""
],
[
"Jensen",
"Jaclyn",
""
],
[
"McConnachie",
"Alan W.",
""
],
[
"Navarro",
"Julio F.",
""
],
[
"Pazder",
"John",
""
],
[
"Sestito",
"Federico",
""
],
[
"Anthony",
"André",
""
],
[
"Andersen",
"Dave",
""
],
[
"Baker",
"Gabriella",
""
],
[
"Chin",
"Timothy",
""
],
[
"Churilov",
"Vladimir",
""
],
[
"Diaz",
"Ruben",
""
],
[
"Farrell",
"Tony",
""
],
[
"Firpo",
"Veronica",
""
],
[
"Gomez-Jimenez",
"Manuel",
""
],
[
"Henderson",
"David",
""
],
[
"Kalari",
"Venu M.",
""
],
[
"Lawrence",
"Jon",
""
],
[
"Margheim",
"Steve",
""
],
[
"Miller",
"Bryan",
""
],
[
"Robertson",
"J. Gordon",
""
],
[
"Ruiz-Carmona",
"Roque",
""
],
[
"Silversides",
"Katherine",
""
],
[
"Silva",
"Karleyne",
""
],
[
"Young",
"Peter J.",
""
],
[
"Zhelem",
"Ross",
""
]
] |
2503.07323 | Yubo Zhao | Yubo Zhao, Qi Wu, Yifan Wang, Yu-Wing Tai, Chi-Keung Tang | Navigating Motion Agents in Dynamic and Cluttered Environments through LLM Reasoning | null | null | null | null | cs.AI cs.CV cs.RO | http://creativecommons.org/licenses/by/4.0/ | This paper advances motion agents empowered by large language models (LLMs) toward autonomous navigation in dynamic and cluttered environments, significantly surpassing first and recent seminal but limited studies on LLM's spatial reasoning, where movements are restricted in four directions in simple, static environments in the presence of only single agents much less multiple agents. Specifically, we investigate LLMs as spatial reasoners to overcome these limitations by uniformly encoding environments (e.g., real indoor floorplans), agents which can be dynamic obstacles and their paths as discrete tokens akin to language tokens. Our training-free framework supports multi-agent coordination, closed-loop replanning, and dynamic obstacle avoidance without retraining or fine-tuning. We show that LLMs can generalize across agents, tasks, and environments using only text-based interactions, opening new possibilities for semantically grounded, interactive navigation in both simulation and embodied systems. | [
{
"version": "v1",
"created": "Mon, 10 Mar 2025 13:39:09 GMT"
},
{
"version": "v2",
"created": "Thu, 05 Jun 2025 12:17:03 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Zhao",
"Yubo",
""
],
[
"Wu",
"Qi",
""
],
[
"Wang",
"Yifan",
""
],
[
"Tai",
"Yu-Wing",
""
],
[
"Tang",
"Chi-Keung",
""
]
] |
2503.07697 | Michael-Andrei Panaitescu-Liess | Michael-Andrei Panaitescu-Liess, Pankayaraj Pathmanathan, Yigitcan Kaya, Zora Che, Bang An, Sicheng Zhu, Aakriti Agrawal, Furong Huang | PoisonedParrot: Subtle Data Poisoning Attacks to Elicit Copyright-Infringing Content from Large Language Models | 18 pages, 18 figures. Accepted at NAACL 2025 | null | null | null | cs.LG cs.CR | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | As the capabilities of large language models (LLMs) continue to expand, their usage has become increasingly prevalent. However, as reflected in numerous ongoing lawsuits regarding LLM-generated content, addressing copyright infringement remains a significant challenge. In this paper, we introduce PoisonedParrot: the first stealthy data poisoning attack that induces an LLM to generate copyrighted content even when the model has not been directly trained on the specific copyrighted material. PoisonedParrot integrates small fragments of copyrighted text into the poison samples using an off-the-shelf LLM. Despite its simplicity, evaluated in a wide range of experiments, PoisonedParrot is surprisingly effective at priming the model to generate copyrighted content with no discernible side effects. Moreover, we discover that existing defenses are largely ineffective against our attack. Finally, we make the first attempt at mitigating copyright-infringement poisoning attacks by proposing a defense: ParrotTrap. We encourage the community to explore this emerging threat model further. | [
{
"version": "v1",
"created": "Mon, 10 Mar 2025 17:13:30 GMT"
},
{
"version": "v2",
"created": "Thu, 05 Jun 2025 04:56:24 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Panaitescu-Liess",
"Michael-Andrei",
""
],
[
"Pathmanathan",
"Pankayaraj",
""
],
[
"Kaya",
"Yigitcan",
""
],
[
"Che",
"Zora",
""
],
[
"An",
"Bang",
""
],
[
"Zhu",
"Sicheng",
""
],
[
"Agrawal",
"Aakriti",
""
],
[
"Huang",
"Furong",
""
]
] |
2503.08109 | Pramod Padmanabhan Mr. | Somnath Maity, Pramod Padmanabhan, Vladimir Korepin | Non-hermitian integrable systems from constant non-invertible solutions of the Yang-Baxter equation | 36 pages + references | JHEP05(2025)206 | 10.1007/JHEP05(2025)206 | null | hep-th cond-mat.stat-mech math-ph math.MP nlin.SI | http://creativecommons.org/licenses/by-nc-nd/4.0/ | We construct invertible spectral parameter dependent Yang-Baxter solutions ($R$-matrices) by Baxterizing constant non-invertible Yang-Baxter solutions. The solutions are algebraic (representation independent). They are constructed using supersymmetry (SUSY) algebras. The resulting $R$-matrices are regular leading to local non-hermitian Hamiltonians written in terms of the SUSY generators. As particular examples we Baxterize the $4\times 4$ constant non-invertible solutions of Hietarinta leading to nearest-neighbor Hamiltonians. On comparing with the literature we find two of the models are new. Apart from being non-hermitian, many of them are also non-diagonalizable with interesting spectrums. With appropriate representations of the SUSY generators we obtain spin chains in all local Hilbert space dimensions. | [
{
"version": "v1",
"created": "Tue, 11 Mar 2025 07:19:55 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Maity",
"Somnath",
""
],
[
"Padmanabhan",
"Pramod",
""
],
[
"Korepin",
"Vladimir",
""
]
] |
2503.09110 | Tariq Aziz | Tariq Aziz, Meng-Long Song, Liu Ye, and Dong Wang | Entropic Diagram Characterization of Quantum Coherence: Degenerate Distillation and the Maximum Eigenvalue Uncertainty Bound | 27 pages, 5 figures. Comments are welcome. Accepted by Physical Review Research | Phys. Rev. Research 7, 023221, 2025 | 10.1103/PhysRevResearch.7.023221 | null | quant-ph | http://creativecommons.org/licenses/by-nc-nd/4.0/ | We develop a rigorous framework for quantifying quantum coherence in finite-dimensional systems by applying the Schur-Horn majorization theorem to relate eigenvalue distributions and diagonal entries of density matrices. Building on this foundation, we introduce a versatile suite of coherence measures, including the relative cross-entropy of coherence and its partial variants, that satisfy all resource theoretic axioms under incoherent operations. This unifying approach clarifies the geometric boundaries of physically realizable states in von Neumann-Tsallis entropy space and uncovers the phenomenon of degenerate coherence distillation where symmetry in the eigenvalue spectrum enables enhanced coherence extraction in higher-dimensional systems. In addition, we strengthen the entropy-based uncertainty relation by refining the Maassen-Uffink bound to account for the largest eigenvalues across distinct measurement bases. This refinement forges a deeper connection between entropy and uncertainty, which yields operationally meaningful constraints for quantum information tasks. Altogether, our findings illustrate the power of majorization in resource-theoretic analyses of quantum coherence, which offer valuable tools for both fundamental research and real-world applications in quantum information processing. | [
{
"version": "v1",
"created": "Wed, 12 Mar 2025 06:51:11 GMT"
},
{
"version": "v2",
"created": "Tue, 20 May 2025 03:45:08 GMT"
},
{
"version": "v3",
"created": "Wed, 21 May 2025 09:47:31 GMT"
},
{
"version": "v4",
"created": "Wed, 28 May 2025 12:35:47 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Aziz",
"Tariq",
""
],
[
"Song",
"Meng-Long",
""
],
[
"Ye",
"Liu",
""
],
[
"Wang",
"Dong",
""
]
] |
2503.09117 | Qizhou Wang | Yue Wang and Qizhou Wang and Feng Liu and Wei Huang and Yali Du and Xiaojiang Du and Bo Han | GRU: Mitigating the Trade-off between Unlearning and Retention for LLMs | null | null | null | null | cs.LG cs.CL | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Large language model (LLM) unlearning has demonstrated its essential role in removing privacy and copyright-related responses, crucial for their legal and safe applications. However, the pursuit of complete unlearning often comes with substantial costs due to its compromises in their general functionality, leading to a notorious trade-off between unlearning and retention. It motivates this paper to explore enhanced unlearning schemes that can mitigate this trade-off. Specifically, we propose Gradient Rectified Unlearning (GRU), an improved framework that regulates the directions of gradient updates during the unlearning procedure such that their side impacts on other, unrelated responses can be minimized. GRU is easy and general to implement, demonstrating practical effectiveness across a variety of well-established unlearning benchmarks. | [
{
"version": "v1",
"created": "Wed, 12 Mar 2025 07:08:54 GMT"
},
{
"version": "v2",
"created": "Wed, 04 Jun 2025 02:58:11 GMT"
},
{
"version": "v3",
"created": "Thu, 05 Jun 2025 13:34:42 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Wang",
"Yue",
""
],
[
"Wang",
"Qizhou",
""
],
[
"Liu",
"Feng",
""
],
[
"Huang",
"Wei",
""
],
[
"Du",
"Yali",
""
],
[
"Du",
"Xiaojiang",
""
],
[
"Han",
"Bo",
""
]
] |
2503.09845 | Evgeny Mukhin | Keshav Dahiya and Evgeny Mukhin | Intertwiners of representations of untwisted quantum affine algebras and Yangians revisited | Latex, 43 pages, 10 figures | Journal of Mathematical Physics, 1 May 2025; 66 (5): 051701 | 10.1063/5.0274972 | null | math.QA math-ph math.MP | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We discuss applications of the $q$-characters to the computation of the $R$-matrices. In particular, we describe the $R$-matrix acting in the tensor square of the first fundamental representation of E$_8$ and in a number of other cases, where the decomposition of the tensor squares with respect to non-affine quantum algebra has non-trivial multiplicities. As an illustration, we also recover $R$-matrices acting in the multiplicity free-case on the tensor squares of the first fundamental representations of all other types of untwisted quantum affine algebras. The answer is written in terms of projectors related to the decomposition of the tensor squares with respect to non-affine quantum algebras. Then we give explicit expressions for the $R$-matrices in terms of matrix units with respect to a natural basis (except for the case of E$_8$). We give similar formulas for the Yangian $R$-matrices. | [
{
"version": "v1",
"created": "Wed, 12 Mar 2025 21:07:38 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Dahiya",
"Keshav",
""
],
[
"Mukhin",
"Evgeny",
""
]
] |
2503.09872 | Kaveh Ahadi | Antonio Gonzalez, Samuel J. Poage, Bernardo Langa, Jr., Deepak Sapkota, Salva Salmani-Rezaie, Shalinee Chikara, Michael D. Williams, David A. Muller, Kasra Sardashti, and Kaveh Ahadi | Superconductivity in tin telluride films grown by molecular beam epitaxy | null | null | null | null | cond-mat.supr-con cond-mat.mtrl-sci quant-ph | http://creativecommons.org/licenses/by/4.0/ | The intersection of superconductivity and ferroelectricity hosts a wide range of exotic quantum phenomena. Here, we report on the observation of superconductivity in high-quality tin telluride films grown by molecular beam epitaxy. Unintentionally doped tin telluride undergoes a ferroelectric transition at ~100 K. The critical temperature of superconductivity increases monotonically with indium concentration. The critical field of superconductivity, however, does not follow the same behavior as critical temperature with indium concentration and exhibits a carrier-density-dependent violation of the Pauli limit. The electron-phonon coupling, from the McMillan formula, exhibits a systematic enhancement with indium concentration, suggesting a potential violation of BCS weak coupling at high indium concentrations. | [
{
"version": "v1",
"created": "Wed, 12 Mar 2025 22:08:47 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Gonzalez",
"Antonio",
""
],
[
"Poage",
"Samuel J.",
""
],
[
"Langa,",
"Bernardo",
"Jr."
],
[
"Sapkota",
"Deepak",
""
],
[
"Salmani-Rezaie",
"Salva",
""
],
[
"Chikara",
"Shalinee",
""
],
[
"Williams",
"Michael D.",
""
],
[
"Muller",
"David A.",
""
],
[
"Sardashti",
"Kasra",
""
],
[
"Ahadi",
"Kaveh",
""
]
] |
2503.09920 | Marcos Vinicius De Sousa Silva | Carlos F. S. Pereira, Marcos V. de S. Silva, H. Belich, Denis C. Rodrigues, J\'ulio C. Fabris, and Manuel E. Rodrigues | Black-bounce solutions in a k-essence theory under the effects of bumblebee gravity | 15 pages, 7 figures. Comments are welcome. V2: Minor modifications to match published version | Physical Review D 111, 124005 (2025) | 10.1103/m3gq-1t5l | null | gr-qc | http://creativecommons.org/licenses/by/4.0/ | In the present study, we analyze the effects of violation of Lorentz symmetry for black-bounce solutions in a $k$-essence theory that has the form of a power law for the configuration $n=1/3$. We perform such analysis for a known model explored in previous work $\Sigma^2=x^2+a^2$ and complement the proposal with a new black-bounce model for the area functions $\Sigma^2_1=\sqrt{x^4+d^4}$. This model has the Schwarzschild-de Sitter asymptomatic behavior for $x\to{-\infty}$, and we investigate the scalar field, potential, and energy conditions for both models. We have shown that the violation of Lorentz symmetry can be generated through $k$-essence without the need for an additional field. These results corroborate the validation of other previously investigated wormhole solutions. | [
{
"version": "v1",
"created": "Thu, 13 Mar 2025 00:23:13 GMT"
},
{
"version": "v2",
"created": "Thu, 05 Jun 2025 17:54:42 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Pereira",
"Carlos F. S.",
""
],
[
"Silva",
"Marcos V. de S.",
""
],
[
"Belich",
"H.",
""
],
[
"Rodrigues",
"Denis C.",
""
],
[
"Fabris",
"Júlio C.",
""
],
[
"Rodrigues",
"Manuel E.",
""
]
] |
2503.11315 | JeongHun Yeo | Jeong Hun Yeo, Hyeongseop Rha, Se Jin Park, Yong Man Ro | MMS-LLaMA: Efficient LLM-based Audio-Visual Speech Recognition with Minimal Multimodal Speech Tokens | Accepted at Findings of ACL 2025. The code and models are available https://github.com/JeongHun0716/MMS-LLaMA | null | null | null | cs.CV cs.MM cs.SD eess.AS | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Audio-Visual Speech Recognition (AVSR) achieves robust speech recognition in noisy environments by combining auditory and visual information. However, recent Large Language Model (LLM) based AVSR systems incur high computational costs due to the high temporal resolution of audio-visual speech processed by LLMs. In this work, we introduce an efficient multimodal speech LLM framework that minimizes token length while preserving essential linguistic content. Our approach employs an early AV-fusion module for streamlined feature integration, an audio-visual speech Q-Former that dynamically allocates tokens based on input duration, and a refined query allocation strategy with a speech rate predictor to adjust token allocation according to speaking speed of each audio sample. Extensive experiments on the LRS3 dataset show that our method achieves state-of-the-art performance with a WER of 0.72% while using only 3.5 tokens per second. Moreover, our approach not only reduces token usage by 86% compared to the previous multimodal speech LLM framework, but also improves computational efficiency by reducing FLOPs by 35.7%. | [
{
"version": "v1",
"created": "Fri, 14 Mar 2025 11:31:30 GMT"
},
{
"version": "v2",
"created": "Thu, 05 Jun 2025 05:58:37 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Yeo",
"Jeong Hun",
""
],
[
"Rha",
"Hyeongseop",
""
],
[
"Park",
"Se Jin",
""
],
[
"Ro",
"Yong Man",
""
]
] |
2503.12716 | Keshav Dahiya | Keshav Dahiya and Evgeny Mukhin | Intertwiners of representations of twisted quantum affine algebras | 24 pages. arXiv admin note: substantial text overlap with arXiv:2503.09845 | Journal of Mathematical Physics, 1 May 2025; 66 (5): 051703 | 10.1063/5.0275522 | null | math.QA math-ph math.MP | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We use the $q$-characters to compute explicit expressions of the $R$-matrices for first fundamental representations of all types of twisted quantum affine algebras. | [
{
"version": "v1",
"created": "Mon, 17 Mar 2025 00:58:07 GMT"
},
{
"version": "v2",
"created": "Wed, 09 Apr 2025 21:59:20 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Dahiya",
"Keshav",
""
],
[
"Mukhin",
"Evgeny",
""
]
] |
2503.12730 | Abir Harrasse | Abir Harrasse, Philip Quirke, Clement Neo, Dhruv Nathawani and Amir Abdullah | TinySQL: A Progressive Text-to-SQL Dataset for Mechanistic Interpretability Research | 9 pages, 19 figures, 7 tables, 18 trained models | null | null | null | cs.LG cs.AI cs.DB | http://creativecommons.org/licenses/by-sa/4.0/ | Mechanistic interpretability research faces a gap between analyzing simple circuits in toy tasks and discovering features in large models. To bridge this gap, we propose text-to-SQL generation as an ideal task to study, as it combines the formal structure of toy tasks with real-world complexity. We introduce TinySQL, a synthetic dataset, progressing from basic to advanced SQL operations, and train models ranging from 33M to 1B parameters to establish a comprehensive testbed for interpretability. We apply multiple complementary interpretability techniques, including Edge Attribution Patching and Sparse Autoencoders, to identify minimal circuits and components supporting SQL generation. We compare circuits for different SQL subskills, evaluating their minimality, reliability, and identifiability. Finally, we conduct a layerwise logit lens analysis to reveal how models compose SQL queries across layers: from intent recognition to schema resolution to structured generation. Our work provides a robust framework for probing and comparing interpretability methods in a structured, progressively complex setting. | [
{
"version": "v1",
"created": "Mon, 17 Mar 2025 01:47:50 GMT"
},
{
"version": "v2",
"created": "Thu, 05 Jun 2025 11:07:16 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Harrasse",
"Abir",
""
],
[
"Quirke",
"Philip",
""
],
[
"Neo",
"Clement",
""
],
[
"Nathawani",
"Dhruv",
""
],
[
"Abdullah",
"Amir",
""
]
] |
2503.12930 | Anthony Frion | Anthony Frion (Lab-STICC\_OSE, IMT Atlantique - MEE), Lucas Drumetz (IMT Atlantique - MEE, Lab-STICC\_OSE), Mauro Dalla Mura (GIPSA-SIGMAPHY), Guillaume Tochon (GIPSA-SIGMAPHY), Abdeldjalil A\"issa-El-Bey (IMT Atlantique - MEE, Lab-STICC\_COSYDE) | Augmented Invertible Koopman Autoencoder for long-term time series forecasting | null | Transactions on Machine Learning Research Journal, 2025 | null | null | cs.LG stat.ML | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Following the introduction of Dynamic Mode Decomposition and its numerous extensions, many neural autoencoder-based implementations of the Koopman operator have recently been proposed. This class of methods appears to be of interest for modeling dynamical systems, either through direct long-term prediction of the evolution of the state or as a powerful embedding for downstream methods. In particular, a recent line of work has developed invertible Koopman autoencoders (IKAEs), which provide an exact reconstruction of the input state thanks to their analytically invertible encoder, based on coupling layer normalizing flow models. We identify that the conservation of the dimension imposed by the normalizing flows is a limitation for the IKAE models, and thus we propose to augment the latent state with a second, non-invertible encoder network. This results in our new model: the Augmented Invertible Koopman AutoEncoder (AIKAE). We demonstrate the relevance of the AIKAE through a series of long-term time series forecasting experiments, on satellite image time series as well as on a benchmark involving predictions based on a large lookback window of observations. | [
{
"version": "v1",
"created": "Mon, 17 Mar 2025 08:40:50 GMT"
},
{
"version": "v2",
"created": "Thu, 05 Jun 2025 09:23:32 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Frion",
"Anthony",
"",
"Lab-STICC\\_OSE, IMT Atlantique - MEE"
],
[
"Drumetz",
"Lucas",
"",
"IMT Atlantique - MEE, Lab-STICC\\_OSE"
],
[
"Mura",
"Mauro Dalla",
"",
"GIPSA-SIGMAPHY"
],
[
"Tochon",
"Guillaume",
"",
"GIPSA-SIGMAPHY"
],
[
"Aïssa-El-Bey",
"Abdeldjalil",
"",
"IMT Atlantique - MEE, Lab-STICC\\_COSYDE"
]
] |
2503.15398 | Fedor Levkovich-Maslyuk | Fedor Levkovich-Maslyuk | Separation of variables for higher rank integrable models: review of recent progress | 35 pages; pedagogical review for J Phys A; v2: published version | null | 10.1088/1751-8121/ade103 | null | hep-th cond-mat.stat-mech math-ph math.MP nlin.SI | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Separation of variables (SoV) is a powerful method expected to be applicable for a wide range of quantum integrable systems, from models in condensed matter physics to gauge and string theories. Yet its full implementation for many higher rank examples, such as SU(N) spin chains with N>2, has remained elusive for a long time. In this pedagogical review we discuss the major progress achieved recently in understanding SoV for models of this type. In particular, for rational SU(N) spin chains we describe different constructions of the SoV basis, novel compact forms for spin chain eigenstates, the functional SoV approach, and explicit computation of the SoV measure. We also discuss key first applications of these results, namely the new compact determinant representations for many observables such as scalar products and correlators. | [
{
"version": "v1",
"created": "Wed, 19 Mar 2025 16:37:21 GMT"
},
{
"version": "v2",
"created": "Thu, 05 Jun 2025 16:28:33 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Levkovich-Maslyuk",
"Fedor",
""
]
] |
2503.15490 | Anil Doshi | Anil R. Doshi and Alastair Moore | Toward a Human-AI Task Tensor: A Taxonomy for Organizing Work in the Age of Generative AI | null | Forthcoming in the Handbook of AI and Strategy (Csaszar & Jia, eds.; Edward Elgar Publishing) 2026 | null | null | cs.HC | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We introduce a framework for understanding the impact of generative AI on human work, which we call the human-AI task tensor. A tensor is a structured framework that organizes tasks along multiple interdependent dimensions. Our human-AI task tensor introduces a systematic approach to studying how humans and AI interact to perform tasks, and has eight dimensions: task definition, AI contribution, interaction modality, audit requirement, output definition, decision-making authority, AI structure, and human persona. After describing the eight dimensions of the tensor, we provide illustrative frameworks (derived from projections of the tensor) and a human-AI task canvas that provide analytical tractability and practical insight for organizational decision-making. We demonstrate how the human-AI task tensor can be used to organize emerging and future research on generative AI. We propose that the human-AI task tensor offers a starting point for understanding how work will be performed with the emergence of generative AI. | [
{
"version": "v1",
"created": "Mon, 06 Jan 2025 13:41:08 GMT"
},
{
"version": "v2",
"created": "Wed, 23 Apr 2025 10:58:46 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Doshi",
"Anil R.",
""
],
[
"Moore",
"Alastair",
""
]
] |
2503.15819 | Junyi Shen | Junyi Shen, Tetsuro Miyazaki, Kenji Kawashima | Control Pneumatic Soft Bending Actuator with Online Learning Pneumatic Physical Reservoir Computing | 8 pages, 13 figures, IEEE-RAS International Conference on Soft Robotics (RoboSoft 2025) | 2025 IEEE 8th International Conference on Soft Robotics (RoboSoft) | 10.1109/RoboSoft63089.2025.11020910 | null | cs.RO cs.LG cs.SY eess.SY | http://creativecommons.org/licenses/by/4.0/ | The intrinsic nonlinearities of soft robots present significant control but simultaneously provide them with rich computational potential. Reservoir computing (RC) has shown effectiveness in online learning systems for controlling nonlinear systems such as soft actuators. Conventional RC can be extended into physical reservoir computing (PRC) by leveraging the nonlinear dynamics of soft actuators for computation. This paper introduces a PRC-based online learning framework to control the motion of a pneumatic soft bending actuator, utilizing another pneumatic soft actuator as the PRC model. Unlike conventional designs requiring two RC models, the proposed control system employs a more compact architecture with a single RC model. Additionally, the framework enables zero-shot online learning, addressing limitations of previous PRC-based control systems reliant on offline training. Simulations and experiments validated the performance of the proposed system. Experimental results indicate that the PRC model achieved superior control performance compared to a linear model, reducing the root-mean-square error (RMSE) by an average of over 37% in bending motion control tasks. The proposed PRC-based online learning control framework provides a novel approach for harnessing physical systems' inherent nonlinearities to enhance the control of soft actuators. | [
{
"version": "v1",
"created": "Thu, 20 Mar 2025 03:09:46 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Shen",
"Junyi",
""
],
[
"Miyazaki",
"Tetsuro",
""
],
[
"Kawashima",
"Kenji",
""
]
] |
2503.16108 | Patricia Gimeno-Estivill | Patricia Gimeno-Estivill, Tuomas Lappi, Heikki M\"antysaari | Inclusive $\mathrm{D}^0$ photoproduction in ultraperipheral collisions | Public code in https://github.com/patgies/inclusive-D0-UPC | null | null | null | hep-ph nucl-th | http://creativecommons.org/licenses/by/4.0/ | We compute the differential cross section for inclusive $\mathrm{D}^0$ production in ultraperipheral collisions within the Color Glass Condensate framework. Our predictions are found to be in relatively good agreement with the CMS data at small transverse momentum, which is the region of validity of our approach. Furthermore, we quantify saturation effects by a nuclear modification ratio $R_{pA}$ for $\mathrm{D}^0$ photoproduction and examine both analytically and numerically the collinear factorization limit of the $\mathrm{D}^0$ differential photoproduction cross section. | [
{
"version": "v1",
"created": "Thu, 20 Mar 2025 12:54:44 GMT"
},
{
"version": "v2",
"created": "Wed, 04 Jun 2025 19:21:08 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Gimeno-Estivill",
"Patricia",
""
],
[
"Lappi",
"Tuomas",
""
],
[
"Mäntysaari",
"Heikki",
""
]
] |
2503.16691 | Andrew Finley Dr. | Paul B. May and Andrew O. Finley | Spatial-temporal prediction of forest attributes using latent Gaussian models and inventory data | null | null | null | null | stat.AP stat.ME | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | The USDA Forest Inventory and Analysis (FIA) program conducts a national forest inventory for the United States through a network of permanent field plots. FIA produces estimates of area averages and totals for plot-measured forest variables through design-based inference, assuming a fixed population and a probability sample of field plot locations. The fixed-population assumption and characteristics of the FIA sampling scheme make it difficult to estimate change in forest variables over time using design-based inference. We propose spatial-temporal models based on Gaussian processes as a flexible tool for forest inventory data, capable of inferring forest variables and change thereof over arbitrary spatial and temporal domains. It is shown to be beneficial for the covariance function governing the latent Gaussian process to account for variation at multiple scales, separating spatially local variation from ecosystem-scale variation. We demonstrate a model for forest biomass density, inferring 20 years of biomass change within two US National Forests. | [
{
"version": "v1",
"created": "Thu, 20 Mar 2025 20:19:19 GMT"
},
{
"version": "v2",
"created": "Thu, 05 Jun 2025 13:52:38 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"May",
"Paul B.",
""
],
[
"Finley",
"Andrew O.",
""
]
] |
2503.16920 | Mohammad Golam Kibria | Mohammad Golam Kibria, Lauren Kucirka, Javed Mostafa | Assessing AI Explainability: A Usability Study Using a Novel Framework Involving Clinicians | 10 pages, 4 figures, Accepted to the 13th IEEE International Conference on Healthcare Informatics (IEEE ICHI 2025) | null | null | null | cs.HC | http://creativecommons.org/licenses/by-nc-nd/4.0/ | An AI design framework was developed based on three core principles, namely understandability, trust, and usability. The framework was conceptualized by synthesizing evidence from the literature and by consulting with experts. The initial version of the AI Explainability Framework was validated based on an in-depth expert engagement and review process. For evaluation purposes, an AI-anchored prototype, incorporating novel explainability features, was built and deployed online. The primary function of the prototype was to predict the postpartum depression risk using analytics models. The development of the prototype was carried out in an iterative fashion, based on a pilot-level formative evaluation, followed by refinements and summative evaluation. The System Explainability Scale (SES) metric was developed to measure the influence of the three dimensions of the AI Explainability Framework. For the summative stage, a comprehensive usability test was conducted involving 20 clinicians, and the SES metric was used to assess clinicians` satisfaction with the tool. On a 5-point rating system, the tool received high scores for the usability dimension, followed by trust and understandability. The average explainability score was 4.56. In terms of understandability, trust, and usability, the average score was 4.51, 4.53 and 4.71 respectively. Overall, the 13-item SES metric showed strong internal consistency with Cronbach`s alpha of 0.84 and a positive correlation coefficient (Spearman`s rho = 0.81, p<0.001) between the composite SES score and explainability. A major finding was that the framework, combined with the SES usability metric, provides a straightforward approach for developing AI-based healthcare tools that lower the challenges associated with explainability. | [
{
"version": "v1",
"created": "Fri, 21 Mar 2025 07:32:41 GMT"
},
{
"version": "v2",
"created": "Thu, 05 Jun 2025 05:07:18 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Kibria",
"Mohammad Golam",
""
],
[
"Kucirka",
"Lauren",
""
],
[
"Mostafa",
"Javed",
""
]
] |
2503.17248 | Val\'ery Rousseau | Val G. Rousseau | Einstein's Cat -- A Thought Experiment Against Anti-Relativist Claims | 4 pages, 4 figures | null | null | null | physics.pop-ph physics.class-ph quant-ph | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | When faced with overwhelming evidence supporting the reality of time dilation, confirmed in particular by the Hafele-Keating experiment, some anti-relativists reluctantly concede that time dilation applies to light clocks. However, they argue that the Theory of Relativity remains flawed, claiming that time dilation applies to light clocks only, not to massive objects. They assert that atomic clocks, which operate based on microwave radiation, merely create the illusion that the Hafele-Keating experiment confirms the theory. To refute this misconception, we introduce a thought experiment inspired by Schrodinger's cat, in which the fate of Einstein's cat depends on a "Sync-or-Die clock", an imaginary device that tests the synchronization between a light clock and a mechanical clock, potentially triggering the release of poison. By analyzing this scenario from both the inertial frame where the device is at rest and another in which it moves at constant velocity, we demonstrate that time dilation must apply to the mechanical clock in exactly the same way as it does to the light clock, highlighting the universality of relativistic time dilation. | [
{
"version": "v1",
"created": "Fri, 21 Mar 2025 15:53:39 GMT"
},
{
"version": "v2",
"created": "Thu, 05 Jun 2025 00:11:05 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Rousseau",
"Val G.",
""
]
] |
2503.17464 | Nicol\'as Allo-G\'omez | Nicol\'as Allo-G\'omez | Other Examples of Principal Ideal Domains that are not Euclidean Domains | 9 pages. Submitted | null | null | null | math.AC math.RA | http://creativecommons.org/licenses/by-nc-nd/4.0/ | It is a well-known and easily established fact that every Euclidean domain is also a principal ideal domain. However, the converse statement is not true, and this is usually shown by exhibiting as a counterexample the ring of algebraic integers in a certain, very specific quadratic field, and the proof that this works is quite unnatural and technical. In this article, we will present a family of counterexamples constructed using real closed fields. | [
{
"version": "v1",
"created": "Fri, 21 Mar 2025 18:20:56 GMT"
},
{
"version": "v2",
"created": "Wed, 04 Jun 2025 23:32:12 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Allo-Gómez",
"Nicolás",
""
]
] |
2503.17644 | Mudit Gaur Mr. | Mudit Gaur, Utsav Singh, Amrit Singh Bedi, Raghu Pasupathu, Vaneet Aggarwal | On The Sample Complexity Bounds In Bilevel Reinforcement Learning | This is updated version of the paper 2410.15610 | null | null | null | cs.LG cs.AI | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Bilevel reinforcement learning (BRL) has emerged as a powerful framework for aligning generative models, yet its theoretical foundations, especially sample complexity bounds, remain underexplored. In this work, we present the first sample complexity bound for BRL, establishing a rate of $\mathcal{O}(\epsilon^{-3})$ in continuous state-action spaces. Traditional MDP analysis techniques do not extend to BRL due to its nested structure and non-convex lower-level problems. We overcome these challenges by leveraging the Polyak-{\L}ojasiewicz (PL) condition and the MDP structure to obtain closed-form gradients, enabling tight sample complexity analysis. Our analysis also extends to general bi-level optimization settings with non-convex lower levels, where we achieve state-of-the-art sample complexity results of $\mathcal{O}(\epsilon^{-3})$ improving upon existing bounds of $\mathcal{O}(\epsilon^{-6})$. Additionally, we address the computational bottleneck of hypergradient estimation by proposing a fully first-order, Hessian-free algorithm suitable for large-scale problems. | [
{
"version": "v1",
"created": "Sat, 22 Mar 2025 04:22:04 GMT"
},
{
"version": "v2",
"created": "Fri, 23 May 2025 19:57:37 GMT"
},
{
"version": "v3",
"created": "Thu, 05 Jun 2025 04:48:59 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Gaur",
"Mudit",
""
],
[
"Singh",
"Utsav",
""
],
[
"Bedi",
"Amrit Singh",
""
],
[
"Pasupathu",
"Raghu",
""
],
[
"Aggarwal",
"Vaneet",
""
]
] |
2503.17838 | Haozhe Shu | Haozhe Shu (1 and 2), Mingpei Lin (2) ((1) Mathematical Institute, Tohoku University, Sendai, Japan, (2) Advanced Institute for Material Research, Tohoku University, Sendai, Japan) | Analysis of pitchfork bifurcations and symmetry breaking in the elliptic restricted three-body problem | null | null | null | null | math.DS math-ph math.MP | http://creativecommons.org/licenses/by/4.0/ | A unified framework is proposed to quantitatively characterize pitchfork bifurcations and associated symmetry breaking in the elliptic restricted three-body problem (ERTBP). It is known that planar/vertical Lyapunov orbits and Lissajous orbits near the collinear libration points undergo pitchfork bifurcations with varying orbital energy. These bifurcations induce symmetry breaking, generating bifurcated families including halo/quasi-halo orbits, axial/quasi-axial orbits, and their corresponding invariant manifolds. Traditional semi-analytical methods for constructing halo orbits, based on resonant bifurcation mechanisms, have obstacles in fully exploiting the intrinsic symmetry breaking characteristics in pitchfork bifurcations. In this paper, a unified trigonometric series-based framework is proposed to analyze these bifurcated families from the perspective of coupling-induced bifurcation mechanisms. By introducing a coupling coefficient and various bifurcation equations into the ERTBP, different symmetry breaking is achieved when the coupling coefficient is non-zero. This unified semi-analytical framework captures bifurcations of both periodic/quasi-periodic and transit/non-transit orbits. Furthermore, it reveals that pitchfork bifurcation solutions in the ERTBP fundamentally depend solely on the orbital eccentricity and three amplitude parameters of the system's degrees of freedom, governing both the elliptic direction and the hyperbolic one. | [
{
"version": "v1",
"created": "Sat, 22 Mar 2025 18:46:05 GMT"
},
{
"version": "v2",
"created": "Thu, 05 Jun 2025 06:24:37 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Shu",
"Haozhe",
"",
"1 and 2"
],
[
"Lin",
"Mingpei",
"",
"Advanced Institute for Material Research, Tohoku University, Sendai, Japan"
]
] |
2503.18033 | Dvir Samuel | Dvir Samuel, Matan Levy, Nir Darshan, Gal Chechik, Rami Ben-Ari | OmnimatteZero: Fast Training-free Omnimatte with Pre-trained Video Diffusion Models | Project Page: https://dvirsamuel.github.io/omnimattezero.github.io/ | null | null | null | cs.CV | http://creativecommons.org/licenses/by/4.0/ | In Omnimatte, one aims to decompose a given video into semantically meaningful layers, including the background and individual objects along with their associated effects, such as shadows and reflections. Existing methods often require extensive training or costly self-supervised optimization. In this paper, we present OmnimatteZero, a training-free approach that leverages off-the-shelf pre-trained video diffusion models for omnimatte. It can remove objects from videos, extract individual object layers along with their effects, and composite those objects onto new videos. These are accomplished by adapting zero-shot image inpainting techniques for video object removal, a task they fail to handle effectively out-of-the-box. To overcome this, we introduce temporal and spatial attention guidance modules that steer the diffusion process for accurate object removal and temporally consistent background reconstruction. We further show that self-attention maps capture information about the object and its footprints and use them to inpaint the object's effects, leaving a clean background. Additionally, through simple latent arithmetic, object layers can be isolated and recombined seamlessly with new video layers to produce new videos. Evaluations show that OmnimatteZero not only achieves superior performance in terms of background reconstruction but also sets a new record for the fastest Omnimatte approach, achieving real-time performance with minimal frame runtime. | [
{
"version": "v1",
"created": "Sun, 23 Mar 2025 11:26:48 GMT"
},
{
"version": "v2",
"created": "Thu, 05 Jun 2025 07:10:09 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Samuel",
"Dvir",
""
],
[
"Levy",
"Matan",
""
],
[
"Darshan",
"Nir",
""
],
[
"Chechik",
"Gal",
""
],
[
"Ben-Ari",
"Rami",
""
]
] |
2503.18617 | Tomasz R\'o\.za\'nski | Tomasz R\'o\.za\'nski, Yuan-Sen Ting | Scaling Laws for Emulation of Stellar Spectra | 25 pages, 11 figures, accepted for publication in the Open Journal of Astrophysics | null | null | null | astro-ph.IM astro-ph.SR cs.LG | http://creativecommons.org/licenses/by/4.0/ | Neural network-based emulators for the inference of stellar parameters and elemental abundances represent an increasingly popular methodology in modern spectroscopic surveys. However, these approaches are often constrained by their emulation precision and domain transfer capabilities. Greater generalizability has previously been achieved only with significantly larger model architectures, as demonstrated by Transformer-based models in natural language processing. This observation aligns with neural scaling laws, where model performance predictably improves with increased model size, computational resources allocated to model training, and training data volume. In this study, we demonstrate that these scaling laws also apply to Transformer-based spectral emulators in astronomy. Building upon our previous work with TransformerPayne and incorporating Maximum Update Parametrization techniques from natural language models, we provide training guidelines for scaling models to achieve optimal performance. Our results show that within the explored parameter space, clear scaling relationships emerge. These findings suggest that optimal computational resource allocation requires balanced scaling. Specifically, given a tenfold increase in training compute, achieving an optimal seven-fold reduction in mean squared error necessitates an approximately 2.5-fold increase in dataset size and a 3.8-fold increase in model size. This study establishes a foundation for developing spectral foundational models with enhanced domain transfer capabilities. | [
{
"version": "v1",
"created": "Mon, 24 Mar 2025 12:20:24 GMT"
},
{
"version": "v2",
"created": "Wed, 04 Jun 2025 20:55:17 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Różański",
"Tomasz",
""
],
[
"Ting",
"Yuan-Sen",
""
]
] |
2503.18789 | Boris Rubinstein | Boris Y. Rubinstein | On the Sylvester program and Cayley algorithm for vector partition reduction | 23 pages, submitted to The Ramanujan Journal | null | null | null | math.NT math.CO | http://creativecommons.org/licenses/by/4.0/ | A vector partition problem asks for a number of nonnegative integer solutions to a system of several linear Diophantine equations with integer nonnegative coefficients. J.J. Sylvester put forward an idea of reduction of vector partition to a sum of scalar partitions. In the simplest case of two equations with positive coefficients A. Cayley performed a reduction of the corresponding double partition to a sum of scalar partitions using an algorithm subject to a set of conditions on the coefficients. We suggested a modification of the original Cayley algorithm for the cases when these conditions are not satisfied. This result is extended to arbitrary number of the Diophantine equations to accomplish the Sylvester program of the vector partition reduction to a combination of scalar partitions. | [
{
"version": "v1",
"created": "Mon, 24 Mar 2025 15:36:27 GMT"
},
{
"version": "v2",
"created": "Sun, 06 Apr 2025 20:11:37 GMT"
},
{
"version": "v3",
"created": "Wed, 04 Jun 2025 20:58:12 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Rubinstein",
"Boris Y.",
""
]
] |
2503.19136 | Alexander Terenin | Sidhanth Holalkere and David S. Bindel and Silvia Sell\'an and Alexander Terenin | Stochastic Poisson Surface Reconstruction with One Solve using Geometric Gaussian Processes | null | International Conference on Machine Learning, 2025 | null | null | cs.GR cs.CV cs.LG stat.ML | http://creativecommons.org/licenses/by/4.0/ | Poisson Surface Reconstruction is a widely-used algorithm for reconstructing a surface from an oriented point cloud. To facilitate applications where only partial surface information is available, or scanning is performed sequentially, a recent line of work proposes to incorporate uncertainty into the reconstructed surface via Gaussian process models. The resulting algorithms first perform Gaussian process interpolation, then solve a set of volumetric partial differential equations globally in space, resulting in a computationally expensive two-stage procedure. In this work, we apply recently-developed techniques from geometric Gaussian processes to combine interpolation and surface reconstruction into a single stage, requiring only one linear solve per sample. The resulting reconstructed surface samples can be queried locally in space, without the use of problem-dependent volumetric meshes or grids. These capabilities enable one to (a) perform probabilistic collision detection locally around the region of interest, (b) perform ray casting without evaluating points not on the ray's trajectory, and (c) perform next-view planning on a per-ray basis. They also do not requiring one to approximate kernel matrix inverses with diagonal matrices as part of intermediate computations, unlike prior methods. Results show that our approach provides a cleaner, more-principled, and more-flexible stochastic surface reconstruction pipeline. | [
{
"version": "v1",
"created": "Mon, 24 Mar 2025 20:47:51 GMT"
},
{
"version": "v2",
"created": "Thu, 05 Jun 2025 16:54:25 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Holalkere",
"Sidhanth",
""
],
[
"Bindel",
"David S.",
""
],
[
"Sellán",
"Silvia",
""
],
[
"Terenin",
"Alexander",
""
]
] |
2503.19600 | Jan Lang | David E Edmunds and Jan Lang | Notes on Non-Compact Maps and the Importance of Bernstein Numbers | 24 pages | null | null | null | math.FA | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | In this review paper we study non-compact operators and embeddings between function spaces, highlighting interesting phenomena and the significance of Bernstein numbers. In particular, we demonstrate that for non-compact maps the usual $s$-numbers (e.g., approximation, Kolmogorov, and entropy numbers) fail to reveal finer structural properties, and one must instead consider concepts such as strict singularity and Bernstein numbers. | [
{
"version": "v1",
"created": "Tue, 25 Mar 2025 12:30:56 GMT"
},
{
"version": "v2",
"created": "Wed, 04 Jun 2025 20:21:55 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Edmunds",
"David E",
""
],
[
"Lang",
"Jan",
""
]
] |
2503.20431 | Rohit Goswami MInstP MBCS MRSC | Rohit Goswami (1,2 and 3) and Ashwini Kumar Rawat (1) and Sonaly Goswami (1) and Debabrata Goswami (1) ((1) Department of Chemistry, Indian Institute of Kanpur, India (2) Science Institute, Iceland (3) Quansight Labs, Austin, TX, USA) | Compositional Analysis of Fragrance Accords Using Femtosecond Thermal Lens Spectroscopy | 26 pages, 6 figures, 14 supplementary figures | null | null | null | physics.chem-ph | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Femtosecond thermal lens spectroscopy (FTLS) is a powerful analytical tool, yet its application to complex, multi-component mixtures like fragrance accords remains limited. Here, we introduce and validate a unified metric, the Femtosecond Thermal Lens Integrated Magnitude (FTL-IM), to characterize such mixtures. The FTL-IM, derived from the integrated signal area, provides a direct, model-free measure of the total thermo-optical response, including critical convective effects. Applying the FTL-IM to complex six-component accords, we demonstrate its utility in predicting a mixture's thermal response from its composition through linear additivity with respect to component mole fractions. Our method quantifies the accords' behavior, revealing both the baseline contributions of components and the dominant, non-linear effects of highly-active species like Methyl Anthranilate. This consistency is validated across single-beam Z-scan, dual-beam Z-scan, and time-resolved FTLS measurements. The metric also demonstrates the necessity of single-beam measurements for interpreting dual-beam data. This work establishes a rapid, quantitative method for fragrance analysis, offering advantages for quality control by directly linking a mixture's bulk thermo-optical properties to its composition. | [
{
"version": "v1",
"created": "Wed, 26 Mar 2025 11:06:13 GMT"
},
{
"version": "v2",
"created": "Thu, 05 Jun 2025 15:19:39 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Goswami",
"Rohit",
"",
"1,2 and 3"
],
[
"Rawat",
"Ashwini Kumar",
"",
"Department of Chemistry, Indian Institute of Kanpur, India"
],
[
"Goswami",
"Sonaly",
"",
"Department of Chemistry, Indian Institute of Kanpur, India"
],
[
"Goswami",
"Debabrata",
"",
"Department of Chemistry, Indian Institute of Kanpur, India"
]
] |
2503.21292 | Leonidas Mouchliadis Dr | Maria Kefalogianni, Leonidas Mouchliadis, Emmanuel Stratakis, Sotiris Psilodimitrakopoulos | Shell-Core Structural Anisotropy in Starch Granules Revealed by Polarization-Dependent Third-Harmonic Generation Imaging | 15 pages, 7 figures | null | null | null | physics.optics | http://creativecommons.org/licenses/by-nc-sa/4.0/ | We investigate the polarization-dependent third-harmonic generation (P-THG) response of starch granules, revealing distinct structural organization between their inner and outer regions. By rotating the linear polarization of the excitation beam, we uncover two characteristic P-THG modulation patterns within single granules, corresponding to an outer shell and a structurally distinct inner core. These patterns are analyzed using a theoretical model based on orthorhombic symmetry, allowing us to extract key ratios of third-order nonlinear susceptibility tensor components and determine the average molecular orientation within each region. In particular, we define an anisotropy ratio, AR=\c{hi}_xxxx^((3) )/\c{hi}_yyyy^((3) ) as a quantitative descriptor of local molecular alignment. Our results show that the shell and core exhibit significantly different AR values, underscoring a transition in molecular organization across the granule. This study establishes P-THG as a powerful contrast mechanism for probing sub-micron structural heterogeneity in biological materials beyond conventional THG intensity imaging. | [
{
"version": "v1",
"created": "Thu, 27 Mar 2025 09:22:08 GMT"
},
{
"version": "v2",
"created": "Thu, 05 Jun 2025 08:12:28 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Kefalogianni",
"Maria",
""
],
[
"Mouchliadis",
"Leonidas",
""
],
[
"Stratakis",
"Emmanuel",
""
],
[
"Psilodimitrakopoulos",
"Sotiris",
""
]
] |
2503.21670 | Rajvee Sheth | Rajvee Sheth, Himanshu Beniwal, Mayank Singh | COMI-LINGUA: Expert Annotated Large-Scale Dataset for Multitask NLP in Hindi-English Code-Mixing | null | null | null | null | cs.CL cs.AI | http://creativecommons.org/licenses/by/4.0/ | We introduce COMI-LINGUA, the largest manually annotated Hindi-English code-mixed dataset, comprising 125K+ high-quality instances across five core NLP tasks: Matrix Language Identification, Token-level Language Identification, POS Tagging, Named Entity Recognition (NER), and Machine Translation. Each instance is annotated by three bilingual annotators, yielding over 376K expert annotations with strong inter-annotator agreement (Fleiss' Kappa $\geq$ 0.81). The rigorously preprocessed and filtered dataset covers both Devanagari and Roman scripts and spans diverse domains, ensuring real-world linguistic coverage. Evaluation reveals that closed-source LLMs significantly outperform traditional tools and open-source models. Notably, one-shot prompting consistently boosts performance across tasks, especially in structure-sensitive predictions like POS and NER, highlighting the effectiveness of prompt-based adaptation in code-mixed, low-resource settings. COMI-LINGUA is publicly available at: https://github.com/lingo-iitgn/CodeMixing_Project. | [
{
"version": "v1",
"created": "Thu, 27 Mar 2025 16:36:39 GMT"
},
{
"version": "v2",
"created": "Thu, 05 Jun 2025 04:46:46 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Sheth",
"Rajvee",
""
],
[
"Beniwal",
"Himanshu",
""
],
[
"Singh",
"Mayank",
""
]
] |
2503.22124 | Peng Lin | Peng Lin, Haopeng Yang, Gui Gui, Mengxiang Zeng and Weihua Gui | Scheduling problems of aircraft on a same runway and dual runways | null | null | null | null | math.OC | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | In this paper, the scheduling problems of landing and takeoff aircraft on a same runway and on dual runways are addressed. In contrast to the approaches based on mixed-integer optimization models in existing works, our approach focuses on the minimum separation times between aircraft by introducing some necessary assumptions and new concepts including relevance, breakpoint aircraft, path and class-monotonically-decreasing sequence. Four scheduling problems are discussed including landing scheduling problem, takeoff scheduling problem, and mixed landing and takeoff scheduling problems on a same runway and on dual runways with the consideration of conversions between different aircraft sequences in typical scenarios. Two real-time optimal algorithms are proposed for the four scheduling problems by fully exploiting the combinations of different classes of aircraft, and necessary definitions, lemmas and theorems are presented for the optimal convergence of the algorithms. Numerical examples are presented to show the effectiveness of the proposed algorithms. In particular, when 100 aircraft are considered, by using the algorithm in this paper, the optimal solution can be obtained in less than 5 seconds, while by using the CPLEX software to solve the mix-integer optimization model, the optimal solution cannot be obtained within 1 hour. | [
{
"version": "v1",
"created": "Fri, 28 Mar 2025 03:56:12 GMT"
},
{
"version": "v2",
"created": "Sun, 13 Apr 2025 08:19:48 GMT"
},
{
"version": "v3",
"created": "Wed, 16 Apr 2025 07:28:35 GMT"
},
{
"version": "v4",
"created": "Sun, 20 Apr 2025 11:27:48 GMT"
},
{
"version": "v5",
"created": "Thu, 24 Apr 2025 02:59:30 GMT"
},
{
"version": "v6",
"created": "Fri, 02 May 2025 03:04:02 GMT"
},
{
"version": "v7",
"created": "Tue, 27 May 2025 11:04:23 GMT"
},
{
"version": "v8",
"created": "Thu, 05 Jun 2025 03:02:31 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Lin",
"Peng",
""
],
[
"Yang",
"Haopeng",
""
],
[
"Gui",
"Gui",
""
],
[
"Zeng",
"Mengxiang",
""
],
[
"Gui",
"Weihua",
""
]
] |
2503.22353 | Yubo Li | Yubo Li, Yidi Miao, Xueying Ding, Ramayya Krishnan, Rema Padman | Firm or Fickle? Evaluating Large Language Models Consistency in Sequential Interactions | 8 pages, 5 figures | null | null | null | cs.CL cs.AI | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Large Language Models (LLMs) have shown remarkable capabilities across various tasks, but their deployment in high-stake domains requires consistent and coherent behavior across multiple rounds of user interaction. This paper introduces a comprehensive framework for evaluating and improving LLM response consistency, making three key contributions. Code and data are available at: https://github.com/yubol-bobo/MT-Consistency. First, we introduce Position-Weighted Consistency (PWC), a metric designed to capture both the importance of early-stage stability and recovery patterns in multi-turn interactions. Second, we present MT-Consistency, a carefully curated benchmark dataset spanning diverse domains and difficulty levels, specifically designed to evaluate LLM consistency under various challenging follow-up scenarios. Third, we introduce Confidence-Aware Response Generation (CARG), a framework that significantly improves response stability by explicitly integrating internal model confidence scores during the generation process. Experimental results demonstrate that CARG significantly improves response stability without sacrificing accuracy, offering a practical path toward more dependable LLM behavior in critical, real-world deployments. | [
{
"version": "v1",
"created": "Fri, 28 Mar 2025 11:49:56 GMT"
},
{
"version": "v2",
"created": "Thu, 29 May 2025 14:35:00 GMT"
},
{
"version": "v3",
"created": "Fri, 30 May 2025 13:25:15 GMT"
},
{
"version": "v4",
"created": "Thu, 05 Jun 2025 08:39:20 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Li",
"Yubo",
""
],
[
"Miao",
"Yidi",
""
],
[
"Ding",
"Xueying",
""
],
[
"Krishnan",
"Ramayya",
""
],
[
"Padman",
"Rema",
""
]
] |
2503.22992 | Ole Hans | Ole Hans, Benedikt Walter and J\"urgen Adamy | Evaluation of Remote Driver Performance in Urban Environment Operational Design Domains | This work has been submitted to the IEEE for possible publication | null | 10.1109/OJITS.2025.3574692 | null | eess.SY cs.RO cs.SY | http://creativecommons.org/licenses/by/4.0/ | Remote driving has emerged as a solution for enabling human intervention in scenarios where Automated Driving Systems (ADS) face challenges, particularly in urban Operational Design Domains (ODDs). This study evaluates the performance of Remote Drivers (RDs) of passenger cars in a representative urban ODD in Las Vegas, focusing on the influence of cumulative driving experience and targeted training approaches. Using performance metrics such as efficiency, braking, acceleration, and steering, the study shows that driving experience can lead to noticeable improvements of RDs and demonstrates how experience up to 600 km correlates with improved vehicle control. In addition, driving efficiency exhibited a positive trend with increasing kilometers, particularly during the first 300 km of experience, which reaches a plateau from 400 km within a range of 0.35 to 0.42 km/min in the defined ODD. The research further compares ODD-specific training methods, where the detailed ODD training approaches attains notable advantages over other training approaches. The findings underscore the importance of tailored ODD training in enhancing RD performance, safety, and scalability for Remote Driving System (RDS) in real-world applications, while identifying opportunities for optimizing training protocols to address both routine and extreme scenarios. The study provides a robust foundation for advancing RDS deployment within urban environments, contributing to the development of scalable and safety-critical remote operation standards. | [
{
"version": "v1",
"created": "Sat, 29 Mar 2025 06:49:00 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Hans",
"Ole",
""
],
[
"Walter",
"Benedikt",
""
],
[
"Adamy",
"Jürgen",
""
]
] |
2503.24102 | Yewei Song | Yewei Song, Lujun Li, Cedric Lothritz, Saad Ezzini, Lama Sleem, Niccolo Gentile, Radu State, Tegawend\'e F. Bissyand\'e, Jacques Klein | Is LLM the Silver Bullet to Low-Resource Languages Machine Translation? | null | null | null | null | cs.CL | http://creativecommons.org/licenses/by-nc-sa/4.0/ | Low-Resource Languages (LRLs) present significant challenges in natural language processing due to their limited linguistic resources and underrepresentation in standard datasets. While recent advances in Large Language Models (LLMs) and Neural Machine Translation have substantially improved translation capabilities for high-resource languages, performance disparities persist for LRLs, particularly impacting privacy-sensitive and resource-constrained scenarios. This paper systematically evaluates current LLMs in 200 languages using the FLORES-200 benchmark and demonstrates their limitations in LRL translation capability. We also explore alternative data sources, including news articles and bilingual dictionaries, and demonstrate how knowledge distillation from large pre-trained teacher models can significantly improve the performance of small LLMs on LRL translation tasks. For example, this approach increases EN->LB with the LLM-as-a-Judge score on the validation set from 0.36 to 0.89 for Llama-3.2-3B. Furthermore, we examine different fine-tuning configurations, providing practical insights on optimal data scale, training efficiency, and the preservation of generalization capabilities of models under study. | [
{
"version": "v1",
"created": "Mon, 31 Mar 2025 13:56:03 GMT"
},
{
"version": "v2",
"created": "Thu, 05 Jun 2025 17:55:07 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Song",
"Yewei",
""
],
[
"Li",
"Lujun",
""
],
[
"Lothritz",
"Cedric",
""
],
[
"Ezzini",
"Saad",
""
],
[
"Sleem",
"Lama",
""
],
[
"Gentile",
"Niccolo",
""
],
[
"State",
"Radu",
""
],
[
"Bissyandé",
"Tegawendé F.",
""
],
[
"Klein",
"Jacques",
""
]
] |
2504.00155 | Karol Kolodziej | Karol Kolodziej | Monte Carlo phase space integration of multiparticle cross sections with carlomat_4.5 | 12 pages, 3 figures, minor changes matching the version published in Comput. Phys. Commun | null | null | null | hep-ph | http://creativecommons.org/licenses/by/4.0/ | Multidimensional phase space integrals must be calculated in order to obtain predictions for total or differential cross sections, or to simulate unweighted events of multiparticle reactions. The corresponding matrix elements, already in the leading order, receive contributions typically from dozens of thousands of the Feynman diagrams, many of which often involve strong peaks due to denominators of some Feynman propagators approaching their minima. As the number of peaks exceeds by far the number of integration variables, such integrals can practically be performed within the multichannel Monte Carlo approach, with different phase space parameterizations, each designed to smooth possibly a few peaks at a time. This obviously requires a lot different phase space parameterizations which, if possible, should be generated and combined in a single multichannel Monte Carlo procedure in a fully automatic way.A few different approaches to the calculation of the multidimensional phase space integrals have been incorporated in version 4.5 of the multipurpose Monte Carlo program carlomat. The present work illustrates how carlomat_4.5 can facilitate the challenging task of calculating the multidimensional phase space integrals. | [
{
"version": "v1",
"created": "Mon, 31 Mar 2025 19:09:03 GMT"
},
{
"version": "v2",
"created": "Wed, 04 Jun 2025 18:52:46 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Kolodziej",
"Karol",
""
]
] |
2504.01359 | Zhenghua Xu | Qinghai Huo, Guangbin Ren, and Zhenghua Xu | Monogenic functions over real alternative *-algebras: fundamental results and applications | 23 pages. In this version, Subsection 2.3 was added | null | null | null | math.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | The concept of monogenic functions over real alternative $\ast$-algebras has recently been introduced to unify several classical monogenic (or regular) functions theories in hypercomplex analysis, including quaternionic, octonionic, and Clifford analysis. This paper explores the fundamental properties of these monogenic functions, focusing on the Cauchy-Pompeiu integral formula and Taylor series expansion in hypercomplex subspaces, among which the non-commutativity and especially non-associativity of multiplications demand full considerations. The theory presented herein provides a robust framework for understanding monogenic functions in the context of real alternative $\ast$-algebras, shedding light on the interplay between algebraic structures and hypercomplex analysis. | [
{
"version": "v1",
"created": "Wed, 02 Apr 2025 05:04:50 GMT"
},
{
"version": "v2",
"created": "Thu, 05 Jun 2025 00:47:04 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Huo",
"Qinghai",
""
],
[
"Ren",
"Guangbin",
""
],
[
"Xu",
"Zhenghua",
""
]
] |
2504.01394 | Pankaj Mishra | Sanu Kumar Gangwar, Rajamanickam Ravisankar, Henrique Fabrelli, Paulsamy Muruganandam, and Pankaj Kumar Mishra | Double unstable avoided crossings and complex domain patterns formation in spin-orbit coupled spin-1 condensates | 17 pages, 15 figures | Phys. Rev. A 111, 063303 (2025) | 10.1103/PhysRevA.111.063303 | null | cond-mat.quant-gas | http://creativecommons.org/licenses/by/4.0/ | We analyze the impact of spin-orbit and Rabi couplings on the dynamical stability of spin-orbit-coupled spin-1 Bose-Einstein condensates for ferromagnetic (FM) and antiferromagnetic (AFM) interactions. Determining the collective excitation spectrum through Bogoliubov-de-Gennes theory, we characterize the dynamical stability regime via modulational instability. For AFM interactions, the eigenspectrum reveals the presence of both stable and unstable avoided crossings (UAC), with the first-excited branch undergoing a double unstable avoided crossing. In contrast, with ferromagnetic interactions, only a single UAC, which occurs between the low-lying and first-excited branches, is observed. Furthermore, the eigenvectors demonstrate the transition from density-like to spin-like behaviour, as the collective excitation shows the transition from stable to unstable mode for both the FM and AFM interactions. In the multi-band instability state, eigenvectors display spin-density mixed mode, while they show spin-flip nature in the avoided crossing regime. Our analysis suggests that spin-orbit coupling enhances the instability gain, while Rabi coupling plays the opposite role. Finally, we corroborate our analytical findings of stable and unstable regimes through numerical simulations of the dynamical evolution of the condensates by introducing the perturbations upon quenching the trap strength. The dynamical phases show the formation of complex domains with AFM interaction, which may be attributed to the double unstable avoided crossings in such a system. | [
{
"version": "v1",
"created": "Wed, 02 Apr 2025 06:19:52 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Gangwar",
"Sanu Kumar",
""
],
[
"Ravisankar",
"Rajamanickam",
""
],
[
"Fabrelli",
"Henrique",
""
],
[
"Muruganandam",
"Paulsamy",
""
],
[
"Mishra",
"Pankaj Kumar",
""
]
] |
2504.01483 | Ruiyang Liu | Siran Li, Ruiyang Liu, Chen Liu, Zhendong Wang, Gaofeng He, Yong-Lu Li, Xiaogang Jin, Huamin Wang | GarmageNet: A Multimodal Generative Framework for Sewing Pattern Design and Generic Garment Modeling | null | null | null | null | cs.GR cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Realistic digital garment modeling remains a labor-intensive task due to the intricate process of translating 2D sewing patterns into high-fidelity, simulation-ready 3D garments. We introduce GarmageNet, a unified generative framework that automates the creation of 2D sewing patterns, the construction of sewing relationships, and the synthesis of 3D garment initializations compatible with physics-based simulation. Central to our approach is Garmage, a novel garment representation that encodes each panel as a structured geometry image, effectively bridging the semantic and geometric gap between 2D structural patterns and 3D garment shapes. GarmageNet employs a latent diffusion transformer to synthesize panel-wise geometry images and integrates GarmageJigsaw, a neural module for predicting point-to-point sewing connections along panel contours. To support training and evaluation, we build GarmageSet, a large-scale dataset comprising over 10,000 professionally designed garments with detailed structural and style annotations. Our method demonstrates versatility and efficacy across multiple application scenarios, including scalable garment generation from multi-modal design concepts (text prompts, sketches, photographs), automatic modeling from raw flat sewing patterns, pattern recovery from unstructured point clouds, and progressive garment editing using conventional instructions-laying the foundation for fully automated, production-ready pipelines in digital fashion. Project page: https://style3d.github.io/garmagenet. | [
{
"version": "v1",
"created": "Wed, 02 Apr 2025 08:37:32 GMT"
},
{
"version": "v2",
"created": "Thu, 05 Jun 2025 08:21:51 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Li",
"Siran",
""
],
[
"Liu",
"Ruiyang",
""
],
[
"Liu",
"Chen",
""
],
[
"Wang",
"Zhendong",
""
],
[
"He",
"Gaofeng",
""
],
[
"Li",
"Yong-Lu",
""
],
[
"Jin",
"Xiaogang",
""
],
[
"Wang",
"Huamin",
""
]
] |
2504.01759 | Dongyan Sui | Dongyan Sui, Haotian Pu, Siyang Leng, Stefan Vlaski | Alpha-Beta HMM: Hidden Markov Model Filtering with Equal Exit Probabilities and a Step-Size Parameter | v2: Journal extension, submitted to IEEE TAC. Conference version remains available as v1 | null | null | null | eess.SY cs.SY stat.AP | http://creativecommons.org/licenses/by-nc-nd/4.0/ | The hidden Markov model (HMM) provides a powerful framework for inference in time-varying environments, where the underlying state evolves according to a Markov chain. To address the optimal filtering problem in general dynamic settings, we propose the $\alpha\beta$-HMM algorithm, which simplifies the state transition model to a Markov chain with equal exit probabilities and introduces a step-size parameter to balance the influence of observational data and the model. By analyzing the algorithm's dynamics in stationary environments, we uncover a fundamental trade-off between inference accuracy and adaptation capability, highlighting how key parameters and observation quality impact performance. A comprehensive theoretical analysis of the nonlinear dynamical system governing the evolution of the log-belief ratio, along with supporting numerical experiments, demonstrates that the proposed approach effectively balances adaptability and inference performance in dynamic environments. | [
{
"version": "v1",
"created": "Wed, 02 Apr 2025 14:16:13 GMT"
},
{
"version": "v2",
"created": "Wed, 04 Jun 2025 20:18:49 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Sui",
"Dongyan",
""
],
[
"Pu",
"Haotian",
""
],
[
"Leng",
"Siyang",
""
],
[
"Vlaski",
"Stefan",
""
]
] |
2504.01814 | Pok Man Lo | Udita Shukla, Pok Man Lo | Energy Density Functional of Confined Quarks: an Improved Ansatz | typos corrected, version appears in JSPC | null | null | null | nucl-th astro-ph.HE hep-ph hep-th | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Density Functional Theory (DFT) is a robust framework for modeling interacting many-body systems, including the equation of state (EoS) of dense matter. Many models, however, rely on energy functionals based on assumptions that have not been rigorously validated. We critically analyze a commonly used ansatz for confinement, where the energy functional scales with density as $U \propto n^{\frac{2}{3}}$ . Our findings, derived from a systematic non-local energy functional, reveal that this scaling does not capture the dynamics of confinement. Instead, the energy functional evolves from $n^2$ at low densities to $n$ at high densities, governed by an infrared cutoff. These results suggest that models relying on such assumptions should be revisited to ensure more reliable EoS construction. | [
{
"version": "v1",
"created": "Wed, 02 Apr 2025 15:20:03 GMT"
},
{
"version": "v2",
"created": "Thu, 05 Jun 2025 15:38:16 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Shukla",
"Udita",
""
],
[
"Lo",
"Pok Man",
""
]
] |
2504.02234 | Jacy Reese Anthis | Jacy Reese Anthis, Ryan Liu, Sean M. Richardson, Austin C. Kozlowski, Bernard Koch, James Evans, Erik Brynjolfsson, Michael Bernstein | LLM Social Simulations Are a Promising Research Method | Published at ICML 2025 | null | null | null | cs.HC cs.AI cs.CL cs.CY | http://creativecommons.org/licenses/by/4.0/ | Accurate and verifiable large language model (LLM) simulations of human research subjects promise an accessible data source for understanding human behavior and training new AI systems. However, results to date have been limited, and few social scientists have adopted this method. In this position paper, we argue that the promise of LLM social simulations can be achieved by addressing five tractable challenges. We ground our argument in a review of empirical comparisons between LLMs and human research subjects, commentaries on the topic, and related work. We identify promising directions, including context-rich prompting and fine-tuning with social science datasets. We believe that LLM social simulations can already be used for pilot and exploratory studies, and more widespread use may soon be possible with rapidly advancing LLM capabilities. Researchers should prioritize developing conceptual models and iterative evaluations to make the best use of new AI systems. | [
{
"version": "v1",
"created": "Thu, 03 Apr 2025 03:01:26 GMT"
},
{
"version": "v2",
"created": "Thu, 05 Jun 2025 15:12:55 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Anthis",
"Jacy Reese",
""
],
[
"Liu",
"Ryan",
""
],
[
"Richardson",
"Sean M.",
""
],
[
"Kozlowski",
"Austin C.",
""
],
[
"Koch",
"Bernard",
""
],
[
"Evans",
"James",
""
],
[
"Brynjolfsson",
"Erik",
""
],
[
"Bernstein",
"Michael",
""
]
] |
2504.03370 | Adeel A. Khan | Adeel A. Khan | Equivariant homology of stacks | 19 pages | null | null | null | math.AT math.AG | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We use sheaf theory and the six operations to define and study the (equivariant) homology of stacks. The construction makes sense in the algebraic, complex-analytic, or even topological categories. | [
{
"version": "v1",
"created": "Fri, 04 Apr 2025 11:41:31 GMT"
},
{
"version": "v2",
"created": "Wed, 04 Jun 2025 21:53:49 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Khan",
"Adeel A.",
""
]
] |
2504.04209 | Topi Sirki\"a | Topi Sirki\"a, Matti Heikinheimo, Kimmo Tuominen | Axion Superradiance in Dipole Magnetic Fields of Pulsars | 22 pages, 2 figures. Added discussion on plasma effects. Matches published version | null | null | HIP-2025-11/TH | hep-ph astro-ph.HE | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We consider constraints on the axion-photon coupling by superradiance due to a plasma instability in the magnetospheres of millisecond pulsars. We compute the growth rate of a superradiant axion cloud in a dipole magnetic field, and give a semi-analytical formula for the superradiance rate for the lowest state. By requiring the associated instability time to be longer than the characteristic age of the supermassive black-widow millisecond pulsar PSR J0952-0607, we examine the pulsar-timing array constraints on axions of mass $\sim 10^{-12}\text{ eV}$. We find that competitive axion bounds from plasma instabilities are unlikely unless a new high spin pulsar is discovered. | [
{
"version": "v1",
"created": "Sat, 05 Apr 2025 15:40:21 GMT"
},
{
"version": "v2",
"created": "Thu, 05 Jun 2025 09:49:54 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Sirkiä",
"Topi",
""
],
[
"Heikinheimo",
"Matti",
""
],
[
"Tuominen",
"Kimmo",
""
]
] |
2504.05250 | Mustafa Burak Gurbuz | Mustafa Burak Gurbuz, Xingyu Zheng, Constantine Dovrolis | PEAKS: Selecting Key Training Examples Incrementally via Prediction Error Anchored by Kernel Similarity | null | null | null | null | cs.LG stat.ML | http://creativecommons.org/licenses/by/4.0/ | As deep learning continues to be driven by ever-larger datasets, understanding which examples are most important for generalization has become a critical question. While progress in data selection continues, emerging applications require studying this problem in dynamic contexts. To bridge this gap, we pose the Incremental Data Selection (IDS) problem, where examples arrive as a continuous stream, and need to be selected without access to the full data source. In this setting, the learner must incrementally build a training dataset of predefined size while simultaneously learning the underlying task. We find that in IDS, the impact of a new sample on the model state depends fundamentally on both its geometric relationship in the feature space and its prediction error. Leveraging this insight, we propose PEAKS (Prediction Error Anchored by Kernel Similarity), an efficient data selection method tailored for IDS. Our comprehensive evaluations demonstrate that PEAKS consistently outperforms existing selection strategies. Furthermore, PEAKS yields increasingly better performance returns than random selection as training data size grows on real-world datasets. The code is available at https://github.com/BurakGurbuz97/PEAKS. | [
{
"version": "v1",
"created": "Mon, 07 Apr 2025 16:42:09 GMT"
},
{
"version": "v2",
"created": "Tue, 08 Apr 2025 02:48:22 GMT"
},
{
"version": "v3",
"created": "Wed, 04 Jun 2025 21:14:36 GMT"
}
] | 2025-06-06T00:00:00 | [
[
"Gurbuz",
"Mustafa Burak",
""
],
[
"Zheng",
"Xingyu",
""
],
[
"Dovrolis",
"Constantine",
""
]
] |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.