id
string | submitter
string | authors
string | title
string | comments
string | journal-ref
string | doi
string | report-no
string | categories
string | license
string | abstract
string | versions
list | update_date
timestamp[s] | authors_parsed
sequence |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
2410.16270 | Lingyu Li | Lingyu Li, Yixu Wang, Haiquan Zhao, Shuqi Kong, Yan Teng, Chunbo Li, Yingchun Wang | Reflection-Bench: Evaluating Epistemic Agency in Large Language Models | 29 pages, 19 figures, 9 tables | null | null | null | cs.AI | http://creativecommons.org/licenses/by/4.0/ | With large language models (LLMs) increasingly deployed as cognitive engines for AI agents, the reliability and effectiveness critically hinge on their intrinsic epistemic agency, which remains understudied. Epistemic agency, the ability to flexibly construct, adapt, and monitor beliefs about dynamic environments, represents a base-model-level capacity independent of specific tools, modules, or applications. We characterize the holistic process underlying epistemic agency, which unfolds in seven interrelated dimensions: prediction, decision-making, perception, memory, counterfactual thinking, belief updating, and meta-reflection. Correspondingly, we propose Reflection-Bench, a cognitive-psychology-inspired benchmark consisting of seven tasks with long-term relevance and minimization of data leakage. Through a comprehensive evaluation of 16 models using three prompting strategies, we identify a clear three-tier performance hierarchy and significant limitations of current LLMs, particularly in meta-reflection capabilities. While state-of-the-art LLMs demonstrate rudimentary signs of epistemic agency, our findings suggest several promising research directions, including enhancing core cognitive functions, improving cross-functional coordination, and developing adaptive processing mechanisms. Our code and data are available at https://github.com/AI45Lab/ReflectionBench. | [
{
"version": "v1",
"created": "Mon, 21 Oct 2024 17:59:50 GMT"
},
{
"version": "v2",
"created": "Fri, 30 May 2025 05:15:52 GMT"
},
{
"version": "v3",
"created": "Wed, 04 Jun 2025 11:45:19 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Li",
"Lingyu",
""
],
[
"Wang",
"Yixu",
""
],
[
"Zhao",
"Haiquan",
""
],
[
"Kong",
"Shuqi",
""
],
[
"Teng",
"Yan",
""
],
[
"Li",
"Chunbo",
""
],
[
"Wang",
"Yingchun",
""
]
] |
2410.16428 | Lantian Li Mr. | Wan Lin and Junhui Chen and Tianhao Wang and Zhenyu Zhou and Lantian Li and Dong Wang | Neural Scoring: A Refreshed End-to-End Approach for Speaker Recognition in Complex Conditions | null | null | null | null | cs.SD eess.AS | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Modern speaker verification systems primarily rely on speaker embeddings and cosine similarity. While effective, these methods struggle with multi-talker speech due to the unidentifiability of embedding vectors. We propose Neural Scoring (NS), a novel end-to-end framework that directly estimates verification posterior probabilities without relying on test-side embeddings, making it more powerful and robust to complex conditions, e.g., with multiple talkers. To address the challenge of training such end-to-end models, we introduce a multi-enrollment training strategy, which pairs each test utterance with multiple enrolled speakers and proves essential to the model's success. Experiments on the VoxCeleb dataset demonstrate that NS consistently outperforms both the baseline and several competitive methods, achieving an overall 70.36% reduction in Equal Error Rate (EER) compared to the baseline. | [
{
"version": "v1",
"created": "Mon, 21 Oct 2024 18:48:54 GMT"
},
{
"version": "v2",
"created": "Tue, 03 Jun 2025 23:35:07 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Lin",
"Wan",
""
],
[
"Chen",
"Junhui",
""
],
[
"Wang",
"Tianhao",
""
],
[
"Zhou",
"Zhenyu",
""
],
[
"Li",
"Lantian",
""
],
[
"Wang",
"Dong",
""
]
] |
2410.16502 | Jason Chan | Jason Chan, Robert Gaizauskas, Zhixue Zhao | RULEBREAKERS: Challenging LLMs at the Crossroads between Formal Logic and Human-like Reasoning | Preprint. Accepted by ICML 2025 | null | null | null | cs.CL | http://creativecommons.org/licenses/by/4.0/ | Formal logic enables computers to reason in natural language by representing sentences in symbolic forms and applying rules to derive conclusions. However, in what our study characterizes as "rulebreaker" scenarios, this method can lead to conclusions that are typically not inferred or accepted by humans given their common sense and factual knowledge. Inspired by works in cognitive science, we create RULEBREAKERS, the first dataset for rigorously evaluating the ability of large language models (LLMs) to recognize and respond to rulebreakers (versus non-rulebreakers) in a human-like manner. Evaluating seven LLMs, we find that most models, including GPT-4o, achieve mediocre accuracy on RULEBREAKERS and exhibit some tendency to over-rigidly apply logical rules unlike what is expected from typical human reasoners. Further analysis suggests that this apparent failure is potentially associated with the models' poor utilization of their world knowledge and their attention distribution patterns. Whilst revealing a limitation of current LLMs, our study also provides a timely counterbalance to a growing body of recent works that propose methods relying on formal logic to improve LLMs' general reasoning capabilities, highlighting their risk of further increasing divergence between LLMs and human-like reasoning. | [
{
"version": "v1",
"created": "Mon, 21 Oct 2024 20:48:16 GMT"
},
{
"version": "v2",
"created": "Thu, 29 May 2025 16:31:57 GMT"
},
{
"version": "v3",
"created": "Wed, 04 Jun 2025 07:55:15 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Chan",
"Jason",
""
],
[
"Gaizauskas",
"Robert",
""
],
[
"Zhao",
"Zhixue",
""
]
] |
2410.17296 | Mattis Van 't Schip | Mattis van 't Schip | The Internet of Forgotten Things: European Cybersecurity Regulation and IoT Manufacturer Cessation | null | null | 10.1016/j.clsr.2025.106152 | null | cs.CY | http://creativecommons.org/licenses/by-nc-nd/4.0/ | Many modern consumer devices rely on network connections and cloud services to perform their core functions. This dependency is especially present in Internet of Things (IoT) devices, which combine hardware and software with network connections (e.g., a 'smart' doorbell with a camera). This paper argues that current European product legislation, which aims to protect consumers of, inter alia, IoT devices, has a blind spot for an increasing problem in the competitive IoT market: manufacturer cessation. Without the manufacturer's cloud servers, many IoT devices cannot perform core functions such as data analysis. If an IoT manufacturer ceases their operations, consumers of the manufacturer's devices are thus often left with a dysfunctional device and, as the paper shows, hardly any legal remedies. This paper therefore investigates three properties that could support legislators in finding a solution for IoT manufacturer cessation: i) pre-emptive measures, aimed at ii) manufacturer-independent iii) collective control. The paper finally shows how these three properties already align with current legislative processes surrounding 'interoperability' and open-source software development. | [
{
"version": "v1",
"created": "Tue, 22 Oct 2024 06:57:49 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Schip",
"Mattis van 't",
""
]
] |
2410.17341 | Theo O'Neill | Theo J. O'Neill, Alyssa A. Goodman, Juan D. Soler, Catherine Zucker, Jiwon Jesse Han | A 3D Model of the Local Bubble's Magnetic Field: Insights from Dust and Starlight Polarization | 30 pages, 17 figures. Accepted by ApJ | null | null | null | astro-ph.GA | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Clustered stellar feedback creates expanding voids in the magnetized interstellar medium known as superbubbles. Although theory suggests that superbubble expansion is influenced by interstellar magnetic fields, direct observational data on 3D superbubble magnetic field geometry is limited. The Sun's location inside the Local Bubble provides a unique opportunity to infer a superbubble's 3D magnetic field orientation, under the assumptions that: $\mathrm{I}$) the Local Bubble's surface is the primary contributor to plane-of-the-sky polarization observations across much of the sky, and $\mathrm{II}$) the Local Bubble's magnetic field is tangent to its dust-traced shell. In this work, we validate these assumptions and construct a model of the Local Bubble's 3D B-field orientation from $\textit{Planck}$ 353 GHz polarization observations and a 3D-dust-derived model of the Local Bubble's shell. We test Assumption $\mathrm{I}$ by examining correlations between the Local Bubble's 3D geometry, dust polarization, and starlight polarization. We find that the Local Bubble likely dominates the polarized signal in the majority of lines of sight. We jointly test Assumptions $\mathrm{I}$ and $\mathrm{II}$ by applying our reconstruction method to a simulated superbubble, where we successfully reconstruct the 3D magnetic field orientation over the bulk of its surface. Finally, we use our 3D B-field model to infer the initial magnetic field orientation in the solar neighborhood prior to the Local Bubble's formation, and derive an orientation parallel to the present-day Local Arm of the galaxy. These findings provide new insights into the co-evolution of superbubbles and the magnetized interstellar medium. | [
{
"version": "v1",
"created": "Tue, 22 Oct 2024 18:18:03 GMT"
},
{
"version": "v2",
"created": "Wed, 04 Jun 2025 14:38:07 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"O'Neill",
"Theo J.",
""
],
[
"Goodman",
"Alyssa A.",
""
],
[
"Soler",
"Juan D.",
""
],
[
"Zucker",
"Catherine",
""
],
[
"Han",
"Jiwon Jesse",
""
]
] |
2410.17407 | Joeri De Ro | Joeri De Ro | Morita theory for dynamical von Neumann algebras | 24 pages + references. Author accepted version, for publication in International Mathematics Research Notices | null | null | null | math.OA math.QA | http://creativecommons.org/licenses/by/4.0/ | Given a locally compact quantum group $\mathbb{G}$ and two $\mathbb{G}$-$W^*$-algebras $\alpha: A\curvearrowleft \mathbb{G}$ and $\beta: B\curvearrowleft \mathbb{G}$, we study the notion of equivariant $W^*$-Morita equivalence $(A, \alpha)\sim_{\mathbb{G}} (B, \beta)$, which is an equivariant version of Rieffel's notion of $W^*$-Morita equivalence. We prove that important dynamical properties of $\mathbb{G}$-$W^*$-algebras, such as (inner) amenability, are preserved under equivariant Morita equivalence. For a coideal von Neumann algebra $L^\infty(\mathbb{K}\backslash \mathbb{G})\subseteq L^\infty(\mathbb{G})$ with dual coideal von Neumann algebra $L^\infty(\check{\mathbb{K}})\subseteq L^\infty(\check{\mathbb{G}})$, we use a natural $\check{\mathbb{G}}$-$W^*$-Morita equivalence $L^\infty(\mathbb{K}\backslash \mathbb{G})\rtimes_\Delta \mathbb{G} \sim_{\check{\mathbb{G}}} L^\infty(\check{\mathbb{K}})$ to relate dynamical properties of $L^\infty(\mathbb{K}\backslash \mathbb{G})$ with dynamical properties of $L^\infty(\check{\mathbb{K}})$. We use this to refine some recent results established by Anderson-Sackaney and Khosravi. This refinement allows us to answer a question of Kalantar, Kasprzak, Skalski and Vergnioux, namely that for $\mathbb{H}$ a closed quantum subgroup of the compact quantum group $\mathbb{G}$, coamenability of $\mathbb{H}\backslash \mathbb{G}$ and relative amenability of $\ell^\infty(\check{\mathbb{H}})$ in $\ell^\infty(\check{\mathbb{G}})$ are equivalent. Moreover, if $\mathbb{G}$ is compact, we study the relation between $\mathbb{G}$-$W^*$-Morita equivalence of $(A, \alpha)$ and $(B, \beta)$ and $\mathbb{G}$-$C^*$-Morita equivalence of the associated $\mathbb{G}$-$C^*$-algebras $(\mathcal{R}(A), \alpha)$ and $(\mathcal{R}(B), \beta)$ of regular elements. | [
{
"version": "v1",
"created": "Tue, 22 Oct 2024 20:28:58 GMT"
},
{
"version": "v2",
"created": "Tue, 18 Feb 2025 15:25:02 GMT"
},
{
"version": "v3",
"created": "Tue, 03 Jun 2025 18:00:08 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"De Ro",
"Joeri",
""
]
] |
2410.19052 | Jiabao Wang | Jia-Bao Wang, Zi-Hao Dong, and Yi Zhang | 1D Spontaneous Symmetry Breaking in thermal equilibrium via Non-Hermitian Construction | 7 pages, 5 figures | null | null | null | quant-ph cond-mat.mes-hall cond-mat.str-el physics.comp-ph | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Spontaneous symmetry breaking generally circumvents one-dimensional systems with local interactions in thermal equilibrium. Here, we analyze a category of one-dimensional Hermitian models via local non-Hermitian constructions. Notably, spontaneous symmetry breaking and long-range order may emerge at finite temperatures in such systems under periodic boundary conditions, in sharp contrast to Hermitian constructions. We demonstrate clear numerical evidence, such as order parameters and specific heat, supporting phase diagrams with robust ordered phases. Non-Hermitian physics plays a vital role in prohibiting domain-wall proliferation and promoting spontaneous symmetry breaking. The fermions exhibit an exotic topological nature in their path-integral windings, which uphold nonzero integers -- commonly a non-Hermitian signature -- in the ordered phases, thus offering a novel and spontaneous origin for both symmetry breaking and non-Hermiticity. | [
{
"version": "v1",
"created": "Thu, 24 Oct 2024 18:00:41 GMT"
},
{
"version": "v2",
"created": "Wed, 04 Jun 2025 12:43:59 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Wang",
"Jia-Bao",
""
],
[
"Dong",
"Zi-Hao",
""
],
[
"Zhang",
"Yi",
""
]
] |
2410.19585 | Michael Hanke | Michael Hanke, Roswitha M\"arz | On the computation of accurate initial conditions for linear higher-index differential-algebraic equations and its application in initial value solvers | 51 pages | null | null | null | math.NA cs.NA | http://creativecommons.org/licenses/by/4.0/ | In contrast to regular ordinary differential equations, the problem of accurately setting initial conditions just emerges in the context of differential-algebraic equations where the dynamic degree of freedom of the system is smaller than the absolute dimension of the described process, and the actual lower-dimensional configuration space of the system is deeply implicit. For linear higher-index differential-algebraic equations, we develop an appropriate numerical method based on properties of canonical subspaces and on the so-called geometric reduction. Taking into account the fact that higher-index differential-algebraic equations lead to ill-posed problems in naturally given norms, we modify this approach to serve as transfer conditions from one time-window to the next in a time stepping procedure and combine it with window-wise overdetermined least-squares collocation to construct the first fully numerical solvers for higher-index initial-value problems. | [
{
"version": "v1",
"created": "Fri, 25 Oct 2024 14:27:58 GMT"
},
{
"version": "v2",
"created": "Wed, 04 Jun 2025 16:31:56 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Hanke",
"Michael",
""
],
[
"März",
"Roswitha",
""
]
] |
2410.20426 | Yongkang Li | Yongkang Li, Huisheng Shu, Litan Yan | Exact Temporal Variation for Fractional Stochastic Heat Equation Driven by Space-Time White Noise | null | null | null | null | math.PR | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | In this paper, we consider the exact fractional variation for the temporal process of the solution to the fractional stochastic heat equation on $\mathbb{R}$ driven by a space-time white noise, and as an application we give the estimate of drift parameter. | [
{
"version": "v1",
"created": "Sun, 27 Oct 2024 12:49:26 GMT"
},
{
"version": "v2",
"created": "Thu, 21 Nov 2024 13:24:13 GMT"
},
{
"version": "v3",
"created": "Wed, 19 Feb 2025 09:08:52 GMT"
},
{
"version": "v4",
"created": "Wed, 04 Jun 2025 12:33:14 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Li",
"Yongkang",
""
],
[
"Shu",
"Huisheng",
""
],
[
"Yan",
"Litan",
""
]
] |
2410.20529 | Asier Alonso-Bardaji | Asier Alonso-Bardaji | Formation of nonsingular spherical black holes with holonomy corrections | Minor changes to match the published version | Phys.Rev.D 111 (2025) 8, 084023 | 10.1103/PhysRevD.111.084023 | null | gr-qc | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We present a canonical model of spherical gravity with covariant corrections motivated by loop quantum gravity. The effective Hamiltonian defines univocally a family of geometries that generalizes the Lema\^itre-Tolman-Bondi spacetimes, and they can be matched to the vacuum of the theory across a timelike hypersurface comoving with the flow of matter. Such is precisely the complete spacetime picture of a spherical star subject to its own gravitational pull. The singularity gets replaced with a spacelike boundary in the trapped region of spacetime, where the curvature remains finite, and the area of the orbits of the spherical symmetry group attains its infimum. Observers falling into the black hole are doomed to travel forever towards this boundary without ever reaching it. The theory also predicts the formation of stable black-hole remnants of Planckian mass. | [
{
"version": "v1",
"created": "Sun, 27 Oct 2024 17:47:20 GMT"
},
{
"version": "v2",
"created": "Wed, 04 Jun 2025 07:59:02 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Alonso-Bardaji",
"Asier",
""
]
] |
2410.22647 | Yuetian Luo | Yuetian Luo and Chao Gao | Adaptive Robust Confidence Intervals | null | null | null | null | math.ST stat.ME stat.TH | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | This paper studies the construction of adaptive confidence intervals under Huber's contamination model when the contamination proportion is unknown. For the robust confidence interval of a Gaussian mean, we show that the optimal length of an adaptive interval must be exponentially wider than that of a non-adaptive one. An optimal construction is achieved through simultaneous uncertainty quantification of quantiles at all levels. The results are further extended beyond the Gaussian location model by addressing a general family of robust hypothesis testing. In contrast to adaptive robust estimation, our findings reveal that the optimal length of an adaptive robust confidence interval critically depends on the distribution's shape. | [
{
"version": "v1",
"created": "Wed, 30 Oct 2024 02:27:02 GMT"
},
{
"version": "v2",
"created": "Tue, 03 Jun 2025 23:25:06 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Luo",
"Yuetian",
""
],
[
"Gao",
"Chao",
""
]
] |
2410.23705 | Aiko Schmei{\ss}er | Laura Barros, Aiko Schmei{\ss}er | Job Loss and Political Entry | null | null | null | null | econ.GN q-fin.EC | http://creativecommons.org/licenses/by/4.0/ | The supply of politicians affects the quality of democratic institutions. Yet little is known about the economic motivations that drive individuals into politics. This paper examines how experiencing a job loss affects individuals' decisions to enter political life and explores its implications for political selection. Using highly granular administrative data linking individual records of political participation with comprehensive employer-employee data for all formal workers in Brazil, and leveraging mass layoffs for causal identification, we find that job loss significantly increases the likelihood of joining a political party and running for local office. Layoff-induced candidates are positively selected on various competence measures, suggesting that economic shocks improve the quality of political entrants. Further, we observe that the increase in candidacies is more pronounced among laid-off individuals with greater financial incentives from holding office and higher predicted income losses. Finally, using a regression discontinuity design, we find that eligibility for unemployment benefits increases the likelihood of becoming a party member and running for local office. These results are consistent with the reduction in private-sector opportunity costs and the increased time resources explaining the rise in political entry. | [
{
"version": "v1",
"created": "Thu, 31 Oct 2024 07:51:04 GMT"
},
{
"version": "v2",
"created": "Mon, 11 Nov 2024 13:49:06 GMT"
},
{
"version": "v3",
"created": "Wed, 04 Jun 2025 08:29:47 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Barros",
"Laura",
""
],
[
"Schmeißer",
"Aiko",
""
]
] |
2411.00296 | Ilya (Eli) Shnitman | Ilya Shnitman | Applying numerosity to surreal integration | 11 pages | null | null | null | math.GM | http://creativecommons.org/licenses/by-sa/4.0/ | We introduce the concept of numerosity, a measure of a set's size, more precise than cardinality and satisfying the Euclid's principle "the whole is greater than a part" as a surreal number and provide an explicit formula for expressing numerosities of some sequences as surreal expressions, with examples. Then (mapping surreal numbers to divergent integrals and interpreting them as Laplace transforms) we introduce a surreal-valued function that shares many properties with the Dirac Delta distribution and employ it to derive some integration properties of surreal values. Finally, we provide a formula for integrating a surreal-valued function over surreal domain, employing numerosity in a way, similar to Lebesgue measure and derive some novel formulas, connecting surreal numbers via integration. The suggested here method of surreal integration is different from methods suggested by other authors in not satisfying the property of linearity regarding infinite factors. In our opinion, the struggle to keep this rule intact is the reason for many failed attempts to define the surreal integration. For the suggested formulas, code in Wolfram Mathematica language is provided. | [
{
"version": "v1",
"created": "Fri, 01 Nov 2024 01:34:20 GMT"
},
{
"version": "v2",
"created": "Sun, 10 Nov 2024 13:12:20 GMT"
},
{
"version": "v3",
"created": "Sat, 08 Feb 2025 11:32:41 GMT"
},
{
"version": "v4",
"created": "Tue, 08 Apr 2025 07:49:30 GMT"
},
{
"version": "v5",
"created": "Tue, 03 Jun 2025 20:10:48 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Shnitman",
"Ilya",
""
]
] |
2411.00441 | Roman Bause | Roman Bause, Nithesh Balasubramanian, Ties Fikkers, Eifion H. Prinsen, Kees Steinebach, Arian Jadbabaie, Nicholas R. Hutzler, I. Agust\'in Aucar, Luk\'a\v{s} F. Pa\v{s}teka, Anastasia Borschevsky and Steven Hoekstra | Prospects for measuring the electron's electric dipole moment with polyatomic molecules in an optical lattice | null | null | 10.1103/8ltl-7wsb | null | physics.atom-ph | http://creativecommons.org/licenses/by/4.0/ | We present the conceptual design of an experiment to measure the electron's electric dipole moment (eEDM) using $^{138}$BaOH molecules in an optical lattice. The BaOH molecule is laser-coolable and highly sensitive to the eEDM, making it an attractive candidate for such a precision measurement, and capturing it in an optical lattice offers potentially very long coherence times. We study possibilities and limitations of this approach, identify the most crucial limiting factors and ways to overcome them. The proposed apparatus can reach a statistical error of $10^{-30}\,e\,$cm by measuring spin precession on a total number of $5 \times 10^9$ molecules over a span of 120 days. | [
{
"version": "v1",
"created": "Fri, 01 Nov 2024 08:24:10 GMT"
},
{
"version": "v2",
"created": "Wed, 04 Jun 2025 09:47:43 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Bause",
"Roman",
""
],
[
"Balasubramanian",
"Nithesh",
""
],
[
"Fikkers",
"Ties",
""
],
[
"Prinsen",
"Eifion H.",
""
],
[
"Steinebach",
"Kees",
""
],
[
"Jadbabaie",
"Arian",
""
],
[
"Hutzler",
"Nicholas R.",
""
],
[
"Aucar",
"I. Agustín",
""
],
[
"Pašteka",
"Lukáš F.",
""
],
[
"Borschevsky",
"Anastasia",
""
],
[
"Hoekstra",
"Steven",
""
]
] |
2411.01747 | Dang Nguyen | Dang Nguyen, Viet Dac Lai, Seunghyun Yoon, Ryan A. Rossi, Handong Zhao, Ruiyi Zhang, Puneet Mathur, Nedim Lipka, Yu Wang, Trung Bui, Franck Dernoncourt, Tianyi Zhou | DynaSaur: Large Language Agents Beyond Predefined Actions | 19 pages, 10 figures | null | null | null | cs.CL | http://creativecommons.org/licenses/by-nc-sa/4.0/ | Existing LLM agent systems typically select actions from a fixed and predefined set at every step. While this approach is effective in closed, narrowly scoped environments, it presents two major challenges for real-world, open-ended scenarios: (1) it significantly restricts the planning and acting capabilities of LLM agents, and (2) it requires substantial human effort to enumerate and implement all possible actions, which is impractical in complex environments with a vast number of potential actions. To address these limitations, we propose an LLM agent framework that can dynamically create and compose actions as needed. In this framework, the agent interacts with its environment by generating and executing programs written in a general-purpose programming language. Moreover, generated actions are accumulated over time for future reuse. Our extensive experiments across multiple benchmarks show that this framework significantly improves flexibility and outperforms prior methods that rely on a fixed action set. Notably, it enables LLM agents to adapt and recover in scenarios where predefined actions are insufficient or fail due to unforeseen edge cases. Our code can be found in https://github.com/adobe-research/dynasaur. | [
{
"version": "v1",
"created": "Mon, 04 Nov 2024 02:08:59 GMT"
},
{
"version": "v2",
"created": "Wed, 04 Jun 2025 16:26:58 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Nguyen",
"Dang",
""
],
[
"Lai",
"Viet Dac",
""
],
[
"Yoon",
"Seunghyun",
""
],
[
"Rossi",
"Ryan A.",
""
],
[
"Zhao",
"Handong",
""
],
[
"Zhang",
"Ruiyi",
""
],
[
"Mathur",
"Puneet",
""
],
[
"Lipka",
"Nedim",
""
],
[
"Wang",
"Yu",
""
],
[
"Bui",
"Trung",
""
],
[
"Dernoncourt",
"Franck",
""
],
[
"Zhou",
"Tianyi",
""
]
] |
2411.02360 | Saubhik Sarkar | Saubhik Sarkar and Abolfazl Bayat | Noisy Stark probes as quantum-enhanced sensors | Close to published version | Phys. Rev. A 111, 062602 (2025) | 10.1103/PhysRevA.111.062602 | null | quant-ph | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Wannier-Stark localization has been proven to be a resource for quantum-enhanced sensitivity for precise estimation of a gradient field. An extremely promising feature of such probes is their ability to showcase such enhanced scaling even dynamically with system size, on top of the quadratic scaling in time. In this paper, we address the issue of decoherence that occurs during time evolution and characterize how that affects the sensing performance. We determine the parameter domains in which the enhancement is sustained under dephasing dynamics. In addition, we consider an effective non-Hermitian description of the open quantum system dynamics for describing the effect of decoherence on the sensing performance of the probe. By investigating the static and dynamic properties of the non-Hermitian Hamiltonians, we show that quantum-enhanced sensitivity can indeed be sustained over certain range of decoherence strength for Wannier-Stark probes. This is demonstrated with two examples of non-Hermitian systems with non-reciprocal couplings. | [
{
"version": "v1",
"created": "Mon, 04 Nov 2024 18:26:21 GMT"
},
{
"version": "v2",
"created": "Tue, 24 Dec 2024 14:25:10 GMT"
},
{
"version": "v3",
"created": "Wed, 04 Jun 2025 03:24:58 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Sarkar",
"Saubhik",
""
],
[
"Bayat",
"Abolfazl",
""
]
] |
2411.02430 | Lin Wang | Lin Wang, Xiaocui Yang, Shi Feng, Daling Wang, Yifei Zhang, Zhitao Zhang | Generative Emotion Cause Explanation in Multimodal Conversations | null | null | 10.1145/3731715.3733348 | null | cs.CL cs.AI cs.CV cs.LG | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Multimodal conversation, a crucial form of human communication, carries rich emotional content, making the exploration of the causes of emotions within it a research endeavor of significant importance. However, existing research on the causes of emotions typically employs an utterance selection method within a single textual modality to locate causal utterances. This approach remains limited to coarse-grained assessments, lacks nuanced explanations of emotional causation, and demonstrates inadequate capability in identifying multimodal emotional triggers. Therefore, we introduce a task-\textbf{Multimodal Emotion Cause Explanation in Conversation (MECEC)}. This task aims to generate a summary based on the multimodal context of conversations, clearly and intuitively describing the reasons that trigger a given emotion. To adapt to this task, we develop a new dataset (ECEM) based on the MELD dataset. ECEM combines video clips with detailed explanations of character emotions, helping to explore the causal factors behind emotional expression in multimodal conversations. A novel approach, FAME-Net, is further proposed, that harnesses the power of Large Language Models (LLMs) to analyze visual data and accurately interpret the emotions conveyed through facial expressions in videos. By exploiting the contagion effect of facial emotions, FAME-Net effectively captures the emotional causes of individuals engaged in conversations. Our experimental results on the newly constructed dataset show that FAME-Net outperforms several excellent baselines. Code and dataset are available at https://github.com/3222345200/FAME-Net. | [
{
"version": "v1",
"created": "Fri, 01 Nov 2024 09:16:30 GMT"
},
{
"version": "v2",
"created": "Tue, 03 Jun 2025 07:51:21 GMT"
},
{
"version": "v3",
"created": "Wed, 04 Jun 2025 02:46:12 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Wang",
"Lin",
""
],
[
"Yang",
"Xiaocui",
""
],
[
"Feng",
"Shi",
""
],
[
"Wang",
"Daling",
""
],
[
"Zhang",
"Yifei",
""
],
[
"Zhang",
"Zhitao",
""
]
] |
2411.03230 | Chaithanya Rayudu | Chaithanya Rayudu | Fermionic Independent Set and Laplacian of an independence complex are QMA-hard | 14 pages | null | null | null | quant-ph cs.CC | http://creativecommons.org/licenses/by/4.0/ | The Independent Set is a well known NP-hard optimization problem. In this work, we define a fermionic generalization of the Independent Set problem and prove that the optimization problem is QMA-hard in a $k$-particle subspace using perturbative gadgets. We discuss how the Fermionic Independent Set is related to the problem of computing the minimum eigenvalue of the $k^{\text{th}}$-Laplacian of an independence complex of a vertex weighted graph. Consequently, we use the same perturbative gadget to prove QMA-hardness of the later problem resolving an open conjecture from arXiv:2311.17234 and give the first example of a natural topological data analysis problem that is QMA-hard. | [
{
"version": "v1",
"created": "Tue, 05 Nov 2024 16:21:48 GMT"
},
{
"version": "v2",
"created": "Tue, 03 Jun 2025 21:57:45 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Rayudu",
"Chaithanya",
""
]
] |
2411.03347 | Qiqi Jiang | Qiqi Jiang, Guang-Xing Li, Chandra B. Singh | Dispersion relation for the linear theory of relativistic Rayleigh Taylor instability in magnetized medium revisited | null | null | null | null | astro-ph.HE physics.flu-dyn | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | The Rayleigh-Taylor instability (RTI) arises at the interface between two fluids of different densities, notably when a heavier fluid lies above a lighter one in an effective gravitational field. In astrophysical systems with high velocities, relativistic corrections are necessary. We investigate the linear theory of relativistic Rayleigh-Taylor instability (R-RTI) in a magnetized medium, where fluids can move parallel to the interface at relativistic velocities. We chose an "intermediate frame" where fluids on each side of the interface move in opposite directions with identical Lorentz factors gamma_*. This symmetry facilitates analytical derivations and the study of relativistic effects on the instability's dynamics. We derive the correct version of the R-RTI. We find that the instability is activated when the Atwood number A = (rho1 h1 - rho2 h2) / (rho1 h1 + rho2 h2) > 0, where rho1 and rho2 are densities measured in the rest frame of the fluids, and that this criterion does not contain relativistic corrections. The relativistic effect is mostly contained in the Lorentz transformation of the gravitational acceleration g' = g / (gamma_*)^2, which, combined with time dilation, leads to a much slower growth of instability (omega' = omega0 / gamma_*), and a slightly elongated length of the unstable patch, due to weaker g in the intermediate frame. Taking time dilation into account, when viewed in the rest frame of the medium, we expect the instability to grow at a much reduced rate. The analytical results should guide further explorations of instability in systems such as microquasars (muQSOs), active galactic nuclei (AGNs), gamma-ray bursts (GRBs), and radio pulsars (PSRs), where the apparent stability of the jet can be attributed to either the intrinsic stability (e.g. the Atwood number) or the much prolonged duration through which R-RTI can grow. | [
{
"version": "v1",
"created": "Sun, 03 Nov 2024 18:38:48 GMT"
},
{
"version": "v2",
"created": "Wed, 04 Jun 2025 13:00:27 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Jiang",
"Qiqi",
""
],
[
"Li",
"Guang-Xing",
""
],
[
"Singh",
"Chandra B.",
""
]
] |
2411.03962 | Zhangcheng Qiang | Zhangcheng Qiang, Kerry Taylor, Weiqing Wang | How Does A Text Preprocessing Pipeline Affect Ontology Syntactic Matching? | 13 pages, 14 figures, 4 tables | null | null | null | cs.CL | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | The classic text preprocessing pipeline, comprising Tokenisation, Normalisation, Stop Words Removal, and Stemming/Lemmatisation, has been implemented in many systems for syntactic ontology matching (OM). However, the lack of standardisation in text preprocessing creates diversity in mapping results. In this paper, we investigate the effect of the text preprocessing pipeline on syntactic OM in 8 Ontology Alignment Evaluation Initiative (OAEI) tracks with 49 distinct alignments. We find that Phase 1 text preprocessing (Tokenisation and Normalisation) is more effective than Phase 2 text preprocessing (Stop Words Removal and Stemming/Lemmatisation). We propose two novel approaches to repair unwanted false mappings caused by Phase 2 text preprocessing. One is an ad hoc logic-based repair approach that employs an ontology-specific check to find common words that cause false mappings. These words are stored in a reserved word set and applied before the text preprocessing. By leveraging the power of large language models (LLMs), we also propose a post hoc LLM-based repair approach. This approach utilises the strong background knowledge provided by LLMs to repair non-existent and counter-intuitive false mappings after the text preprocessing. It also overcomes the tendency towards unstable true mappings by injecting the classic text preprocessing pipeline via function calling. The experimental results show that these two approaches can improve the matching correctness and the overall matching performance. | [
{
"version": "v1",
"created": "Wed, 06 Nov 2024 14:51:02 GMT"
},
{
"version": "v2",
"created": "Thu, 21 Nov 2024 12:00:23 GMT"
},
{
"version": "v3",
"created": "Sat, 23 Nov 2024 10:18:33 GMT"
},
{
"version": "v4",
"created": "Sun, 02 Mar 2025 13:52:25 GMT"
},
{
"version": "v5",
"created": "Mon, 31 Mar 2025 16:35:00 GMT"
},
{
"version": "v6",
"created": "Sun, 25 May 2025 13:34:15 GMT"
},
{
"version": "v7",
"created": "Wed, 04 Jun 2025 15:11:00 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Qiang",
"Zhangcheng",
""
],
[
"Taylor",
"Kerry",
""
],
[
"Wang",
"Weiqing",
""
]
] |
2411.04281 | Xingran Chen | Xingran Chen, Zhenke Wu, Xu Shi, Hyunghoon Cho, Bhramar Mukherjee | Generating Synthetic Electronic Health Record Data: a Methodological Scoping Review with Benchmarking on Phenotype Data and Open-Source Software | null | null | 10.1093/jamia/ocaf082 | null | cs.LG cs.AI stat.ML | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We conduct a scoping review of existing approaches for synthetic EHR data generation, and benchmark major methods with proposed open-source software to offer recommendations for practitioners. We search three academic databases for our scoping review. Methods are benchmarked on open-source EHR datasets, MIMIC-III/IV. Seven existing methods covering major categories and two baseline methods are implemented and compared. Evaluation metrics concern data fidelity, downstream utility, privacy protection, and computational cost. 42 studies are identified and classified into five categories. Seven open-source methods covering all categories are selected, trained on MIMIC-III, and evaluated on MIMIC-III or MIMIC-IV for transportability considerations. Among them, GAN-based methods demonstrate competitive performance in fidelity and utility on MIMIC-III; rule-based methods excel in privacy protection. Similar findings are observed on MIMIC-IV, except that GAN-based methods further outperform the baseline methods in preserving fidelity. A Python package, "SynthEHRella", is provided to integrate various choices of approaches and evaluation metrics, enabling more streamlined exploration and evaluation of multiple methods. We found that method choice is governed by the relative importance of the evaluation metrics in downstream use cases. We provide a decision tree to guide the choice among the benchmarked methods. Based on the decision tree, GAN-based methods excel when distributional shifts exist between the training and testing populations. Otherwise, CorGAN and MedGAN are most suitable for association modeling and predictive modeling, respectively. Future research should prioritize enhancing fidelity of the synthetic data while controlling privacy exposure, and comprehensive benchmarking of longitudinal or conditional generation methods. | [
{
"version": "v1",
"created": "Wed, 06 Nov 2024 21:59:19 GMT"
},
{
"version": "v2",
"created": "Tue, 03 Jun 2025 21:38:16 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Chen",
"Xingran",
""
],
[
"Wu",
"Zhenke",
""
],
[
"Shi",
"Xu",
""
],
[
"Cho",
"Hyunghoon",
""
],
[
"Mukherjee",
"Bhramar",
""
]
] |
2411.04312 | Ying-Ying Lee | Ying-Ying Lee and Chu-An Liu | Lee Bounds with a Continuous Treatment in Sample Selection | null | null | null | null | econ.EM | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We study causal inference in sample selection models where a continuous or multivalued treatment affects both outcomes and their observability (e.g., employment or survey responses). We generalized the widely used Lee (2009)'s bounds for binary treatment effects. Our key innovation is a sufficient treatment values assumption that imposes weak restrictions on selection heterogeneity and is implicit in separable threshold-crossing models, including monotone effects on selection. Our double debiased machine learning estimator enables nonparametric and high-dimensional methods, using covariates to tighten the bounds and capture heterogeneity. Applications to Job Corps and CCC program evaluations reinforce prior findings under weaker assumptions. | [
{
"version": "v1",
"created": "Wed, 06 Nov 2024 23:30:27 GMT"
},
{
"version": "v2",
"created": "Sun, 12 Jan 2025 22:44:54 GMT"
},
{
"version": "v3",
"created": "Sun, 09 Feb 2025 23:25:43 GMT"
},
{
"version": "v4",
"created": "Tue, 03 Jun 2025 23:06:53 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Lee",
"Ying-Ying",
""
],
[
"Liu",
"Chu-An",
""
]
] |
2411.04900 | Xin Wang | Stephen F. King, Xin Wang, Ye-Ling Zhou | Modular domain walls and gravitational waves | 24 pages, 6 figures, v2 includes minor changes, accepted for publication in JCAP | null | null | null | hep-ph astro-ph.CO hep-th | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We discuss modular domain walls and gravitational waves in a class of supersymmetric models where quark and lepton flavour symmetry emerges from modular symmetry. In such models a single modulus field $\tau$ is often assumed to be stabilised at or near certain fixed point values such as $\tau = {\rm i}$ and $\tau = \omega$ (the cube root of unity), in its fundamental domain. We show that, in the global supersymmetry limit of certain classes of potentials, the vacua at these fixed points may be degenerate, leading to the formation of modular domain walls in the early Universe. Taking supergravity effects into account, in the background of a fixed dilaton field $S$, the degeneracy may be lifted, leading to a bias term in the potential allowing the domain walls to collapse. We study the resulting gravitational wave spectra arising from the dynamics of such modular domain walls, and assess their observability by current and future experiments, as a window into modular flavour symmetry. | [
{
"version": "v1",
"created": "Thu, 07 Nov 2024 17:42:32 GMT"
},
{
"version": "v2",
"created": "Wed, 04 Jun 2025 14:56:28 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"King",
"Stephen F.",
""
],
[
"Wang",
"Xin",
""
],
[
"Zhou",
"Ye-Ling",
""
]
] |
2411.04920 | Simon Razniewski | Yujia Hu, Tuan-Phong Nguyen, Shrestha Ghosh, Simon Razniewski | Enabling LLM Knowledge Analysis via Extensive Materialization | 14 pages, 4 tables, 12 figures | ACL 2025 | null | null | cs.CL cs.AI cs.DB | http://creativecommons.org/licenses/by/4.0/ | Large language models (LLMs) have majorly advanced NLP and AI, and next to their ability to perform a wide range of procedural tasks, a major success factor is their internalized factual knowledge. Since Petroni et al. (2019), analyzing this knowledge has gained attention. However, most approaches investigate one question at a time via modest-sized pre-defined samples, introducing an ``availability bias'' (Tversky&Kahnemann, 1973) that prevents the analysis of knowledge (or beliefs) of LLMs beyond the experimenter's predisposition.
To address this challenge, we propose a novel methodology to comprehensively materialize an LLM's factual knowledge through recursive querying and result consolidation. Our approach is a milestone for LLM research, for the first time providing constructive insights into the scope and structure of LLM knowledge (or beliefs).
As a prototype, we build GPTKB, a knowledge base (KB) comprising 101 million relational triples for over 2.9 million entities from GPT-4o-mini. We use GPTKB to exemplarily analyze GPT-4o-mini's factual knowledge in terms of scale, accuracy, bias, cutoff and consistency, at the same time. GPTKB is accessible at https://gptkb.org | [
{
"version": "v1",
"created": "Thu, 07 Nov 2024 17:57:03 GMT"
},
{
"version": "v2",
"created": "Fri, 08 Nov 2024 16:06:09 GMT"
},
{
"version": "v3",
"created": "Mon, 16 Dec 2024 14:05:03 GMT"
},
{
"version": "v4",
"created": "Wed, 04 Jun 2025 10:58:03 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Hu",
"Yujia",
""
],
[
"Nguyen",
"Tuan-Phong",
""
],
[
"Ghosh",
"Shrestha",
""
],
[
"Razniewski",
"Simon",
""
]
] |
2411.05042 | Iryna Hartsock | Iryna Hartsock and Cyrillo Araujo and Les Folio and Ghulam Rasool | Improving Radiology Report Conciseness and Structure via Local Large Language Models | published version | null | 10.1007/s10278-025-01510-w | null | cs.CL cs.AI | http://creativecommons.org/licenses/by-nc-nd/4.0/ | Radiology reports are often lengthy and unstructured, posing challenges for referring physicians to quickly identify critical imaging findings while increasing the risk of missed information. This retrospective study aimed to enhance radiology reports by making them concise and well-structured, with findings organized by relevant organs. To achieve this, we utilized private large language models (LLMs) deployed locally within our institution's firewall, ensuring data security and minimizing computational costs. Using a dataset of 814 radiology reports from seven board-certified body radiologists at Moffitt Cancer Center, we tested five prompting strategies within the LangChain framework. After evaluating several models, the Mixtral LLM demonstrated superior adherence to formatting requirements compared to alternatives like Llama. The optimal strategy involved condensing reports first and then applying structured formatting based on specific instructions, reducing verbosity while improving clarity. Across all radiologists and reports, the Mixtral LLM reduced redundant word counts by more than 53%. These findings highlight the potential of locally deployed, open-source LLMs to streamline radiology reporting. By generating concise, well-structured reports, these models enhance information retrieval and better meet the needs of referring physicians, ultimately improving clinical workflows. | [
{
"version": "v1",
"created": "Wed, 06 Nov 2024 19:00:57 GMT"
},
{
"version": "v2",
"created": "Tue, 03 Jun 2025 22:53:07 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Hartsock",
"Iryna",
""
],
[
"Araujo",
"Cyrillo",
""
],
[
"Folio",
"Les",
""
],
[
"Rasool",
"Ghulam",
""
]
] |
2411.05239 | Leshem Choshen | Moshik Hershcovitch, Andrew Wood, Leshem Choshen, Guy Girmonsky, Roy Leibovitz, Ilias Ennmouri, Michal Malka, Peter Chin, Swaminathan Sundararaman, Danny Harnik | ZipNN: Lossless Compression for AI Models | IEEE Cloud. arXiv admin note: text overlap with arXiv:2404.15198 | null | null | null | cs.LG cs.IT math.IT | http://creativecommons.org/licenses/by/4.0/ | With the growth of model sizes and the scale of their deployment, their sheer size burdens the infrastructure requiring more network and more storage to accommodate these. While there is a vast model compression literature deleting parts of the model weights for faster inference, we investigate a more traditional type of compression - one that represents the model in a compact form and is coupled with a decompression algorithm that returns it to its original form and size - namely lossless compression.
We present ZipNN a lossless compression tailored to neural networks. Somewhat surprisingly, we show that specific lossless compression can gain significant network and storage reduction on popular models, often saving 33% and at times reducing over 50% of the model size. We investigate the source of model compressibility and introduce specialized compression variants tailored for models that further increase the effectiveness of compression. On popular models (e.g. Llama 3) ZipNN shows space savings that are over 17% better than vanilla compression while also improving compression and decompression speeds by 62%. We estimate that these methods could save over an ExaByte per month of network traffic downloaded from a large model hub like Hugging Face. | [
{
"version": "v1",
"created": "Thu, 07 Nov 2024 23:28:23 GMT"
},
{
"version": "v2",
"created": "Wed, 04 Jun 2025 15:59:06 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Hershcovitch",
"Moshik",
""
],
[
"Wood",
"Andrew",
""
],
[
"Choshen",
"Leshem",
""
],
[
"Girmonsky",
"Guy",
""
],
[
"Leibovitz",
"Roy",
""
],
[
"Ennmouri",
"Ilias",
""
],
[
"Malka",
"Michal",
""
],
[
"Chin",
"Peter",
""
],
[
"Sundararaman",
"Swaminathan",
""
],
[
"Harnik",
"Danny",
""
]
] |
2411.05441 | Jules Berry | Jules Berry (IRMAR, INSA Rennes, UR), Fausto Colantoni (UNIROMA) | Sticky diffusions on star graphs : characterization and It{\^o} formula | null | null | null | null | math.PR | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We investigate continuous diffusions on star graphs with sticky behavior at the vertex. These are Markov processes with continuous paths having a positive occupation time at the vertex. We characterize sticky diffusions as time-changed nonsticky diffusions by adapting the classical technique of It{\^o} and McKean. We prove a form of It{\^o} formula, also known as Freidlin-Sheu formula, for this type of process. As an intermediate step, we also obtain a stochastic differential equation satisfied by the radial component of the process. These results generalize those already known for sticky diffusions on a half-line and skew sticky diffusions on the real line. | [
{
"version": "v1",
"created": "Fri, 08 Nov 2024 09:37:58 GMT"
},
{
"version": "v2",
"created": "Wed, 04 Jun 2025 12:06:09 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Berry",
"Jules",
"",
"IRMAR, INSA Rennes, UR"
],
[
"Colantoni",
"Fausto",
"",
"UNIROMA"
]
] |
2411.05561 | Laure Ciernik | Laure Ciernik and Lorenz Linhardt and Marco Morik and Jonas Dippel and Simon Kornblith and Lukas Muttenthaler | Objective drives the consistency of representational similarity across datasets | 26 pages | null | null | null | cs.CV cs.AI cs.LG | http://creativecommons.org/licenses/by/4.0/ | The Platonic Representation Hypothesis claims that recent foundation models are converging to a shared representation space as a function of their downstream task performance, irrespective of the objectives and data modalities used to train these models (Huh et al., 2024). Representational similarity is generally measured for individual datasets and is not necessarily consistent across datasets. Thus, one may wonder whether this convergence of model representations is confounded by the datasets commonly used in machine learning. Here, we propose a systematic way to measure how representational similarity between models varies with the set of stimuli used to construct the representations. We find that the objective function is a crucial factor in determining the consistency of representational similarities across datasets. Specifically, self-supervised vision models learn representations whose relative pairwise similarities generalize better from one dataset to another compared to those of image classification or image-text models. Moreover, the correspondence between representational similarities and the models' task behavior is dataset-dependent, being most strongly pronounced for single-domain datasets. Our work provides a framework for analyzing similarities of model representations across datasets and linking those similarities to differences in task behavior. | [
{
"version": "v1",
"created": "Fri, 08 Nov 2024 13:35:45 GMT"
},
{
"version": "v2",
"created": "Wed, 04 Jun 2025 13:59:06 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Ciernik",
"Laure",
""
],
[
"Linhardt",
"Lorenz",
""
],
[
"Morik",
"Marco",
""
],
[
"Dippel",
"Jonas",
""
],
[
"Kornblith",
"Simon",
""
],
[
"Muttenthaler",
"Lukas",
""
]
] |
2411.07035 | Teagan Clarke | Teagan A. Clarke, Paul D. Lasky, Eric Thrane | Inferring jet physics from neutron star - black hole mergers with gravitational waves | 10 pages, 5 figures, 2 tables, published in ApJ | null | 10.3847/1538-4357/adc804 | null | astro-ph.HE gr-qc | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Neutron star - black hole (NSBH) mergers that undergo tidal disruption may launch jets that could power a gamma-ray burst. We use a population of simulated NSBH systems to measure jet parameters from the gravitational waves emitted by these systems. The conditions during the tidal disruption and merger phase required to power a gamma-ray burst are uncertain. It is likely that the system must achieve some minimum remnant baryonic mass after the merger before a jet can be launched to power a gamma-ray burst. Assuming a fiducial neutron star equation of state, we show how Bayesian hierarchical inference can be used to infer the minimum remnant mass required to launch a gamma-ray burst jet as well as the maximum gamma-ray burst viewing angle to detect a gamma-ray burst. We find that with 200 NSBH observations, we can measure the minimum disk mass to within 0.01 solar masses at 90% credibility. We simultaneously infer the maximum gamma-ray burst viewing angle to within 13 degrees at 90% credibility. We conclude that upcoming upgrades to the LIGO observatories may provide important new insights into the physics of NSBH jets. | [
{
"version": "v1",
"created": "Mon, 11 Nov 2024 14:43:25 GMT"
},
{
"version": "v2",
"created": "Wed, 04 Jun 2025 07:54:44 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Clarke",
"Teagan A.",
""
],
[
"Lasky",
"Paul D.",
""
],
[
"Thrane",
"Eric",
""
]
] |
2411.07161 | Young Min Cho | Young-Min Cho, Raphael Shu, Nilaksh Das, Tamer Alkhouli, Yi-An Lai, Jason Cai, Monica Sunkara, Yi Zhang, Dan Roth | RoundTable: Investigating Group Decision-Making Mechanism in Multi-Agent Collaboration | preprint | null | null | null | cs.MA cs.AI | http://creativecommons.org/licenses/by/4.0/ | Effective group decision-making is critical in Multi-Agent Systems (MAS). Yet, how different mechanisms for reaching consensus impact collaboration quality and efficiency remains understudied. We conduct a systematic study on group decision-making mechanisms in a decentralized setting. Through controlled experiments, we analyze how different voting rules affect decision quality and efficiency in a multi-round collaboration. Results reveal that majority voting often cause inefficient collaboration due to its strict acceptance criteria. At the extreme, unanimous voting gives 87% lower initial performance than the best-performing method. Our qualitative analysis of cross-agent communication shows that messages become longer and more repetitive over time: while message length increases by 84%, similarity to the previous round increases to 90%. Based on these insights, language-based early stopping methods make the performance 13% closer to oracle while reducing rounds by 50%. Our findings highlight the crucial role of group decision-making in optimizing MAS collaboration. | [
{
"version": "v1",
"created": "Mon, 11 Nov 2024 17:37:47 GMT"
},
{
"version": "v2",
"created": "Tue, 03 Jun 2025 22:35:00 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Cho",
"Young-Min",
""
],
[
"Shu",
"Raphael",
""
],
[
"Das",
"Nilaksh",
""
],
[
"Alkhouli",
"Tamer",
""
],
[
"Lai",
"Yi-An",
""
],
[
"Cai",
"Jason",
""
],
[
"Sunkara",
"Monica",
""
],
[
"Zhang",
"Yi",
""
],
[
"Roth",
"Dan",
""
]
] |
2411.07293 | Dario Antolini | Dario Antolini, Nick Early | The Chirotropical Grassmannian | General improvements to the exposition. Added details in Section 6 on CEGM integrands and canonical forms. 13 pages, 1 figure | Special volume on Positive Geometry, Le Matematiche 80 (1) (2025), 123-142 | null | null | math.CO hep-th math.AG | http://creativecommons.org/licenses/by/4.0/ | Recent developments in particle physics have revealed deep connections between scattering amplitudes and tropical geometry. From the heart of this relationship emerged the chirotropical Grassmannian $\text{Trop}^\chi \text{G}(k,n)$ and the chirotropical Dressian $\text{Dr}^\chi(k,n)$, polyhedral fans built from uniform realizable chirotopes that encode the combinatorial structure of Generalized Feynman Diagrams. We prove that $\text{Trop}^\chi \text{G}(3,n) = \text{Dr}^\chi(3,n)$ for $n = 6,7,8$, and develop algorithms to compute these objects from their rays modulo lineality. Using these algorithms, we compute all chirotropical Grassmannians $\text{Trop}^\chi \text{G}(3,n)$ for $n = 6,7,8$ across all isomorphism classes of chirotopes. We prove that each chirotopal configuration space $X^\chi(3,6)$ is diffeomorphic to a polytope and propose an associated canonical logarithmic differential form. Finally, we show that the equality between chirotropical Grassmannian and Dressian fails for $(k,n) = (4,8)$. | [
{
"version": "v1",
"created": "Mon, 11 Nov 2024 19:00:04 GMT"
},
{
"version": "v2",
"created": "Sat, 14 Dec 2024 19:17:11 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Antolini",
"Dario",
""
],
[
"Early",
"Nick",
""
]
] |
2411.07502 | Grigor Sargsyan | G. H. Sargsyan, Kazuki Yoshida, Kazuyuki Ogata, K. D. Launey, J. E. Escher, D. Langr, and T. Dytrych | Ab initio informed 20Ne(p, p$\alpha$)16O reaction elucidates the emergence of alpha clustering from chiral potentials | 7 pages, 3 figures | Phys. Lett. B 866 (2025) 139563 | 10.1016/j.physletb.2025.139563 | null | nucl-th | http://creativecommons.org/licenses/by/4.0/ | We report on the first \textit{ab initio} informed $\alpha$ knock-out reaction in the intermediate-mass region, with the aim to probe the underlying chiral potential and its impact on the emergence of alpha clustering in this mass region. The theoretical predictions of the $\alpha+^{16}$O clustering in the $^{20}$Ne ground state, based on the \textit{ab initio} symmetry-adapted no-core shell model, yield a triple differential cross section for $^{20}$Ne(p, p$\alpha$)$^{16}$O that is in a remarkable agreement with the data. This allows us to examine predictions of surface and in-medium $\alpha$-cluster features from a chiral potential and to compare these to the successful antisymmetrized molecular dynamics approach. | [
{
"version": "v1",
"created": "Tue, 12 Nov 2024 03:00:01 GMT"
},
{
"version": "v2",
"created": "Tue, 03 Jun 2025 21:14:13 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Sargsyan",
"G. H.",
""
],
[
"Yoshida",
"Kazuki",
""
],
[
"Ogata",
"Kazuyuki",
""
],
[
"Launey",
"K. D.",
""
],
[
"Escher",
"J. E.",
""
],
[
"Langr",
"D.",
""
],
[
"Dytrych",
"T.",
""
]
] |
2411.08243 | Khaoula Chehbouni | Khaoula Chehbouni, Jonathan Cola\c{c}o Carr, Yash More, Jackie CK Cheung, Golnoosh Farnadi | Beyond the Safety Bundle: Auditing the Helpful and Harmless Dataset | NAACL Main Conference 2025 - Accepted as an Oral | null | null | null | cs.CL cs.CY | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | In an effort to mitigate the harms of large language models (LLMs), learning from human feedback (LHF) has been used to steer LLMs towards outputs that are intended to be both less harmful and more helpful. Despite the widespread adoption of LHF in practice, the quality of this feedback and its effectiveness as a safety mitigation technique remain unclear. This study addresses these issues by auditing the widely-used Helpful and Harmless (HH) dataset by Anthropic. Our work includes: (1) a thorough investigation of the dataset's content through both manual and automated evaluation; (2) experiments demonstrating the dataset's impact on models' safety; and (3) an analysis of the 100 most influential papers citing this dataset. Through our audit, we showcase how conceptualization failures and quality issues identified in the HH dataset can create additional harms by leading to disparate safety behaviors across demographic groups. Our findings highlight the need for more nuanced, context-sensitive approaches to safety mitigation in LLMs. | [
{
"version": "v1",
"created": "Tue, 12 Nov 2024 23:43:20 GMT"
},
{
"version": "v2",
"created": "Tue, 29 Apr 2025 13:58:44 GMT"
},
{
"version": "v3",
"created": "Tue, 03 Jun 2025 19:37:50 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Chehbouni",
"Khaoula",
""
],
[
"Carr",
"Jonathan Colaço",
""
],
[
"More",
"Yash",
""
],
[
"Cheung",
"Jackie CK",
""
],
[
"Farnadi",
"Golnoosh",
""
]
] |
2411.08536 | Hongyu Xiang | Li Guo, Wenchuan Hu, Hongyu Xiang, Bin Zhang | Extended Shuffle Product for Multiple Zeta Values | 24pages, 0 figures, | null | null | null | math.NT math.RA | http://creativecommons.org/publicdomain/zero/1.0/ | The shuffle algebra on positive integers encodes the usual multiple zeta values (MZVs) (with positive arguments) thanks to the representations of MZVs by iterated Chen integrals of Kontsevich. Together with the quasi-shuffle (stuffle) algebra, it provides the algebraic framework to study relations among MZVs.
This paper enlarges the shuffle algebra uniquely to what we call the extended shuffle algebra that encodes convergent multiple zeta series with arbitrary integer arguments, not just the positive ones in the usual case. To achieved this goal, we first replace the Rota-Baxter operator of weight zero (the integral operator) that characterizes the shuffle product by the differential operator which extends the shuffle product to the larger space. We then show that the subspace corresponding to the convergent MZVs with integer arguments becomes a subalgebra under this extended shuffle product. Furthermore, by lifting the extended shuffle algebra to the locality algebra of Chen symbols, we prove that taking summations of fractions from Chen symbols defines an algebra homomorphism from the above subalgebra to the subalgebra of real numbers spanned by convergent multiple zeta series. | [
{
"version": "v1",
"created": "Wed, 13 Nov 2024 11:34:07 GMT"
},
{
"version": "v2",
"created": "Mon, 24 Mar 2025 07:26:19 GMT"
},
{
"version": "v3",
"created": "Wed, 04 Jun 2025 03:01:22 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Guo",
"Li",
""
],
[
"Hu",
"Wenchuan",
""
],
[
"Xiang",
"Hongyu",
""
],
[
"Zhang",
"Bin",
""
]
] |
2411.09620 | Edoardo Bellincioni | Edoardo Bellincioni, Detlef Lohse, Sander G. Huisman | Melting of floating ice cylinders in fresh and saline environments | null | null | null | null | physics.flu-dyn | http://creativecommons.org/licenses/by-sa/4.0/ | We experimentally investigated the melting of floating ice cylinders. Experiments were carried out in a tank, with ice cylinders with radii between 5 cm and 12 cm, floating horizontally with their axis perpendicular to gravity. The water in the tank was at room temperature, with salinities ranging from 0 g/L to 35 g/L. These conditions correspond to Rayleigh numbers in the range $10^5 \lesssim \mathrm{Ra} \lesssim 10^9$. The relative density and thus the floating behaviour was varied by employing ice made of H$_2$O-D$_2$O mixtures. In addition, we explored a two-layer stable stratification. We studied the morphological evolution of the cross-section of the cylinders and interpreted our observations in the context of their interaction with the convective flow. The cylinders only capsize in fresh water but not when the ambient is saline. This behaviour can be explained by the balance between the torques exerted by buoyancy and drag, which change as the cylinder melts and rotates. We modelled the oscillatory motion of the cylinders after a capsize as a damped non-linear oscillator. The downward plume of the ice cylinders follows the expected scalings for a line-source plume. The plume's Reynolds number scales with Rayleigh number in two regimes, namely $\mathrm{Re} \propto \mathrm{Ra}^{1/2}$ for $\mathrm{Ra} < O(10^7)$ and $\mathrm{Re} \propto \mathrm{Ra}^{1/3}$ for $\mathrm{Ra} > O(10^7)$, and the heat transfer (nondimensional as Nusselt number) scales as $\mathrm{Nu} \propto \mathrm{Ra}^{1/3}$. Although the addition of salt substantially alters the solutal, thermal and momentum boundary layers, these scaling relations hold irrespectively of the initial size or the water salinity. While important differences exist between our experiments and real icebergs, our results can qualitatively be connected to natural phenomena occurring in fjords and around isolated icebergs. | [
{
"version": "v1",
"created": "Thu, 14 Nov 2024 17:40:20 GMT"
},
{
"version": "v2",
"created": "Wed, 04 Jun 2025 17:00:41 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Bellincioni",
"Edoardo",
""
],
[
"Lohse",
"Detlef",
""
],
[
"Huisman",
"Sander G.",
""
]
] |
2411.10371 | Zefan Zeng | Qing Cheng, Zefan Zeng, Xingchen Hu, Yuehang Si, Zhong Liu | A Survey of Event Causality Identification: Taxonomy, Challenges, Assessment, and Prospects | null | null | null | null | cs.CL cs.AI | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Event Causality Identification (ECI) has emerged as a pivotal task in natural language processing (NLP), aimed at automatically detecting causal relationships between events in text. In this comprehensive survey, we systematically elucidate the foundational principles and technical frameworks of ECI, proposing a novel classification framework to categorize and clarify existing methods. {We discuss associated challenges, provide quantitative evaluations, and outline future directions for this dynamic and rapidly evolving field. We first delineate key definitions, problem formalization, and evaluation protocols of ECI. Our classification framework organizes ECI methods based on two primary tasks: Sentence-level Event Causality Identification (SECI) and Document-level Event Causality Identification (DECI). For SECI, we review methods including feature pattern-based matching, machine learning-based classification, deep semantic encoding, prompt-based fine-tuning, and causal knowledge pre-training, alongside common data augmentation strategies. For DECI, we focus on techniques such as deep semantic encoding, event graph reasoning, and prompt-based fine-tuning. We dedicate specific discussions to advancements in multi-lingual and cross-lingual ECI as well as zero-shot ECI leveraging Large Language Models (LLMs). Furthermore, we analyze the strengths, limitations, and unresolved challenges of each method. Extensive quantitative evaluations are conducted on four benchmark datasets to assess various ECI methods. Finally, we explore future research directions. | [
{
"version": "v1",
"created": "Fri, 15 Nov 2024 17:19:42 GMT"
},
{
"version": "v2",
"created": "Mon, 25 Nov 2024 16:55:09 GMT"
},
{
"version": "v3",
"created": "Wed, 04 Jun 2025 07:35:00 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Cheng",
"Qing",
""
],
[
"Zeng",
"Zefan",
""
],
[
"Hu",
"Xingchen",
""
],
[
"Si",
"Yuehang",
""
],
[
"Liu",
"Zhong",
""
]
] |
2411.10490 | Kris Luyten | Gilles Eerlings, Sebe Vanbrabant, Jori Liesenborgs, Gustavo Rovelo Ruiz, Davy Vanacken, Kris Luyten | AI-Spectra: A Visual Dashboard for Model Multiplicity to Enhance Informed and Transparent Decision-Making | Accepted for publication in an LNCS Volume "Engineering Interactive Computer Systems - EICS 2024 - International Workshops and Doctoral Consortium, Selected Papers" | Engineering Interactive Computer Systems. EICS 2024 International Workshops. EICS 2024. Lecture Notes in Computer Science, vol 15518. Springer, Cham | 10.1007/978-3-031-91760-8_5 | null | cs.HC cs.AI | http://creativecommons.org/licenses/by-nc-sa/4.0/ | We present an approach, AI-Spectra, to leverage model multiplicity for interactive systems. Model multiplicity means using slightly different AI models yielding equally valid outcomes or predictions for the same task, thus relying on many simultaneous "expert advisors" that can have different opinions. Dealing with multiple AI models that generate potentially divergent results for the same task is challenging for users to deal with. It helps users understand and identify AI models are not always correct and might differ, but it can also result in an information overload when being confronted with multiple results instead of one. AI-Spectra leverages model multiplicity by using a visual dashboard designed for conveying what AI models generate which results while minimizing the cognitive effort to detect consensus among models and what type of models might have different opinions. We use a custom adaptation of Chernoff faces for AI-Spectra; Chernoff Bots. This visualization technique lets users quickly interpret complex, multivariate model configurations and compare predictions across multiple models. Our design is informed by building on established Human-AI Interaction guidelines and well know practices in information visualization. We validated our approach through a series of experiments training a wide variation of models with the MNIST dataset to perform number recognition. Our work contributes to the growing discourse on making AI systems more transparent, trustworthy, and effective through the strategic use of multiple models. | [
{
"version": "v1",
"created": "Thu, 14 Nov 2024 18:50:41 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Eerlings",
"Gilles",
""
],
[
"Vanbrabant",
"Sebe",
""
],
[
"Liesenborgs",
"Jori",
""
],
[
"Ruiz",
"Gustavo Rovelo",
""
],
[
"Vanacken",
"Davy",
""
],
[
"Luyten",
"Kris",
""
]
] |
2411.10992 | Tomoshi Takeda | Tomoshi Takeda, Toru Tamagawa, Teruaki Enoto, Takao Kitaguchi, Yo Kato, Tatehiro Mihara, Wataru Iwakiri, Masaki Numazawa, Naoyuki Ota, Sota Watanabe, Arata Jujo, Amira Aoyama, Satoko Iwata, Takuya Takahashi, Kaede Yamasaki, Chin-Ping Hu, Hiromitsu Takahashi, Akira Dohi, Nobuya Nishimura, Ryosuke Hirai, Yuto Yoshida, Hiroki Sato, Syoki Hayashi, Yuanhui Zhou, Keisuke Uchiyama, Hirokazu Odaka, Tsubasa Tamba and Kentaro Taniguchi | NinjaSat monitoring of Type-I X-ray bursts from the clocked burster SRGA J144459.2$-$604207 | 7 pages, 5 figures, accepted for publication in PASJ Letter | Publications of the Astronomical Society of Japan 77 (2025) | 10.1093/pasj/psaf003 | RIKEN-iTHEMS-Report-24 | astro-ph.HE | http://creativecommons.org/licenses/by/4.0/ | The CubeSat X-ray observatory NinjaSat was launched on 2023 November 11 and has provided opportunities for agile and flexible monitoring of bright X-ray sources. On 2024 February 23, the NinjaSat team started long-term observation of the new X-ray source SRGA J144459.2$-$604207 as the first scientific target, which was discovered on 2024 February 21 and recognized as the sixth clocked X-ray burster. Our 25-day observation covered almost the entire decay of this outburst from two days after the peak at $\sim$100 mCrab on February 23 until March 18 at a few mCrab level. The Gas Multiplier Counter onboard NinjaSat successfully detected 12 Type-I X-ray bursts with a typical burst duration of $\sim$20 s, shorter than other clocked burster systems. As the persistent X-ray emission declined by a factor of five, X-ray bursts showed a notable change in its morphology: the rise time became shorter from 4.4(7) s to 0.3(3) s (1$\sigma$ errors), and the peak amplitude increased by 44%. The burst recurrence time $\Delta t_{\rm rec}$ also became longer from 2 hr to 10 hr, following the relation of $\Delta t_{\rm rec} \propto F_{\rm per}^{-0.84}$, where $F_{\rm per}$ is the persistent X-ray flux, by applying a Markov chain Monte Carlo method. The short duration of bursts is explained by the He-enhanced composition of accretion matter and the relation between $\Delta t_{\textrm{rec}}$ and $F_{\rm per}$ by a massive neutron star. This study demonstrated that CubeSat pointing observations can provide valuable astronomical X-ray data. | [
{
"version": "v1",
"created": "Sun, 17 Nov 2024 07:40:09 GMT"
},
{
"version": "v2",
"created": "Tue, 21 Jan 2025 08:06:42 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Takeda",
"Tomoshi",
""
],
[
"Tamagawa",
"Toru",
""
],
[
"Enoto",
"Teruaki",
""
],
[
"Kitaguchi",
"Takao",
""
],
[
"Kato",
"Yo",
""
],
[
"Mihara",
"Tatehiro",
""
],
[
"Iwakiri",
"Wataru",
""
],
[
"Numazawa",
"Masaki",
""
],
[
"Ota",
"Naoyuki",
""
],
[
"Watanabe",
"Sota",
""
],
[
"Jujo",
"Arata",
""
],
[
"Aoyama",
"Amira",
""
],
[
"Iwata",
"Satoko",
""
],
[
"Takahashi",
"Takuya",
""
],
[
"Yamasaki",
"Kaede",
""
],
[
"Hu",
"Chin-Ping",
""
],
[
"Takahashi",
"Hiromitsu",
""
],
[
"Dohi",
"Akira",
""
],
[
"Nishimura",
"Nobuya",
""
],
[
"Hirai",
"Ryosuke",
""
],
[
"Yoshida",
"Yuto",
""
],
[
"Sato",
"Hiroki",
""
],
[
"Hayashi",
"Syoki",
""
],
[
"Zhou",
"Yuanhui",
""
],
[
"Uchiyama",
"Keisuke",
""
],
[
"Odaka",
"Hirokazu",
""
],
[
"Tamba",
"Tsubasa",
""
],
[
"Taniguchi",
"Kentaro",
""
]
] |
2411.11832 | Gabrielle Adams | Gabrielle Adams, Katharina Lodders | Effects of Metallicity on Graphite, TiC, and SiC Condensation in Carbon Stars | 33 pages, 9 figures, 2 tables. Accepted to The Astrophysical Journal, March 2025 | ApJ 985 35 (2025) | 10.3847/1538-4357/adc40f | null | astro-ph.GA | http://creativecommons.org/licenses/by-nc-nd/4.0/ | From transmission electron microscopy and other laboratory studies of presolar grains, the implicit condensation sequence of carbon-bearing condensates in circumstellar envelopes of carbon stars is (from first to last) TiC-graphite-SiC. We use thermochemical equilibrium condensation calculations and show that the condensation sequence of TiC, graphite, and SiC depends on metallicity in addition to C/O ratio and total pressure. Calculations were performed for a characteristic carbon star ratio of C/O = 1.2 from 1E-10 to 1E-4 bars total pressure and for uniform metallicity variations ranging from 0.01 to 100 times solar elemental abundances. TiC always condenses before SiC, and the carbide condensation temperatures increase with increasing metallicity and total pressure. Graphite, however, can condense in a cooling circumstellar envelope before TiC, between TiC and SiC, or after SiC, depending on the carbon-bearing gas chemistry, which is dependent on metallicity and total pressure. Analytical expressions for the graphite, TiC, and SiC condensation temperatures as functions of metallicity and total pressure are presented. The inferred sequence from laboratory presolar grain studies, TiC-graphite-SiC, is favored under equilibrium conditions at solar and subsolar metallicities between ~1E-5 to 1E-8 bar total pressure within circumstellar envelopes of carbon stars with nominal C/O = 1.2. We also explored the dependence of the sequence at C/O ratios of 1.1 and 3.0 and found that as the C/O ratio increases, the TiC-graphite-SiC region shifts towards higher total pressures and lower metallicities. | [
{
"version": "v1",
"created": "Mon, 18 Nov 2024 18:51:46 GMT"
},
{
"version": "v2",
"created": "Tue, 03 Jun 2025 19:17:08 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Adams",
"Gabrielle",
""
],
[
"Lodders",
"Katharina",
""
]
] |
2411.12288 | Stefan Graser | Stefan Graser and Martin Schrepp and Stephan B\"ohm | Construction of the UXAR-CT -- a User eXperience Questionnaire for Augmented Reality in Corporate Training | 9 pages, 9 figures, CENTRIC 2024 : The Seventeenth International Conference on Advances in Human-oriented and Personalized Mechanisms, Technologies, and Services | Proceedings of the Seventeenth International Conference on Advances in Human-oriented and Personalized Mechanisms, Technologies, and Services CENTRIC 2024, Venice, pp. 14 - 22, ISSN 2308-3492 | null | 978-1-68558-193-0 | cs.HC | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Measuring User Experience (UX) with questionnaires is essential for developing and improving products. However, no domain-specific standardized UX questionnaire exists for Augmented Reality (AR) in Corporate Training (CT). Thus, this study introduces the UXAR-CT questionnaire - an AR-specific UX questionnaire for CT environments. We describe the construction procedure and the evaluation process of the questionnaire. A set of candidate items was constructed, and a larger sample of participants evaluated several AR-based learning scenarios with these items. Based on the results, we performed a Principal Component Analysis (PCA) to identify relevant measurement items for each scale. The three best-fitting items were selected based on the results to form the final questionnaire. The first results regarding scale quality indicate a high level of internal consistency. The final version of the UXAR-CT questionnaire is provided and will be evaluated in further research. | [
{
"version": "v1",
"created": "Tue, 19 Nov 2024 07:17:37 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Graser",
"Stefan",
""
],
[
"Schrepp",
"Martin",
""
],
[
"Böhm",
"Stephan",
""
]
] |
2411.12289 | Stefan Graser | Stefan Graser and Anastasia Snimshchikova and Martin Schrepp and Stephan B\"ohm | Enhancing UX Research Activities Using GenAI -- Potential Applications and Challenges | 13 pages, 4 figures, CENTRIC 2024 : The Seventeenth International Conference on Advances in Human-oriented and Personalized Mechanisms, Technologies, and Services | Proceedings of the Seventeenth International Conference on Advances in Human-oriented and Personalized Mechanisms, Technologies, and Services CENTRIC 2024, Venice, pp. 1 -13, ISSN 2308-3492 | null | null | cs.HC | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | User Experience (UX) Research covers various methods for gathering the users' subjective impressions of a product. For this, practitioners face different activities and tasks related to the research process. This includes processing a large amount of data based on qualitative and quantitative data. However, this can be very laborious in practice. Thus, the application of GenAI can support UX research activities. This paper provides a practical perspective on this topic. Based on previous studies, we present different use cases indicating the potential of GenAI in UX research. Moreover, we provide insights into an exploratory study using GenAI along an entire UX research process. Results show that Large Language Models (LLMs) are useful for various tasks. Thus, the research activities can be carried out more efficiently. However, the researcher should always review results to ensure quality. In summary, we want to express the potential of GenAI enhancing UX research | [
{
"version": "v1",
"created": "Tue, 19 Nov 2024 07:18:51 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Graser",
"Stefan",
""
],
[
"Snimshchikova",
"Anastasia",
""
],
[
"Schrepp",
"Martin",
""
],
[
"Böhm",
"Stephan",
""
]
] |
2411.12777 | Stefan Graser | Stefan Graser and Felix Kirschenlohr and Stephan B\"ohm | User Experience Evaluation of Augmented Reality: A Systematic Literature Review | 18 pages, 11 figures, CENTRIC 2024 : The Seventeenth International Conference on Advances in Human-oriented and Personalized Mechanisms, Technologies, and Services | Proceedings of the Seventeenth International Conference on Advances in Human-oriented and Personalized Mechanisms, Technologies, and Services CENTRIC 2024, Venice, pp. 23 - 40, ISSN 2308-3492 | null | null | cs.HC | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Due to technological development, Augmented Reality (AR) can be applied in different domains. However, innovative technologies refer to new interaction paradigms, thus creating a new experience for the user. This so-called User Experience (UX) is essential for developing and designing interactive products. Moreover, UX must be measured to get insights into the user's perception and, thus, to improve innovative technologies. We conducted a Systematic Literature Review (SLR) to provide an overview of the current research concerning UX evaluation of AR. In particular, we aim to identify (1) research referring to UX evaluation of AR and (2) articles containing AR-specific UX models or frameworks concerning the theoretical foundation. The SLR is a five-step approach including five scopes. From a total of 498 records based on eight search terms referring to two databases, 30 relevant articles were identified and further analyzed. Results show that most approaches concerning UX evaluation of AR are quantitative. In summary, five UX models/frameworks were identified. Concerning the UX evaluation results of AR in Training and Education, the UX was consistently positive. Negative aspects refer to errors and deficiencies concerning the AR system and its functionality. No specific metric for UX evaluation of AR in the field of Training and Education exists. Only three AR-specific standardized UX questionnaires could be found. However, the questionnaires do not refer to the field of Training and Education. Thus, there is a lack of research in the field of UX evaluation of AR in Training and Education. | [
{
"version": "v1",
"created": "Tue, 19 Nov 2024 07:19:28 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Graser",
"Stefan",
""
],
[
"Kirschenlohr",
"Felix",
""
],
[
"Böhm",
"Stephan",
""
]
] |
2411.13118 | Stefan Graser | Stefan Graser and Stephan B\"ohm and Martin Schrepp | Using ChatGPT-4 for the Identification of Common UX Factors within a Pool of Measurement Items from Established UX Questionnaires | 10 pages, 1 figure, The Sixteenth International Conference on Advances in Human-oriented and Personalized Mechanisms, Technologies, and Services CENTRIC 2023 | Proceedings of the Sixteenth International Conference on Advances in Human-oriented and Personalized Mechanisms, Technologies, and Services CENTRIC 2023, Valencia, pp 19 - 28, ISSN 2308-3492 | null | null | cs.HC | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Measuring User Experience (UX) with standardized questionnaires is a widely used method. A questionnaire is based on different scales that represent UX factors and items. However, the questionnaires have no common ground concerning naming different factors and the items used to measure them. This study aims to identify general UX factors based on the formulation of the measurement items. Items from a set of 40 established UX questionnaires were analyzed by Generative AI (GenAI) to identify semantically similar items and to cluster similar topics. We used the LLM ChatGPT-4 for this analysis. Results show that ChatGPT-4 can classify items into meaningful topics and thus help to create a deeper understanding of the structure of the UX research field. In addition, we show that ChatGPT-4 can filter items related to a predefined UX concept out of a pool of UX items. | [
{
"version": "v1",
"created": "Wed, 20 Nov 2024 08:26:40 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Graser",
"Stefan",
""
],
[
"Böhm",
"Stephan",
""
],
[
"Schrepp",
"Martin",
""
]
] |
2411.13143 | Yanze Chen | Yanze Chen | Eisenstein Series on Metaplectic Covers and Multiple Dirichlet Series | 43 pages. Fixed several typos | null | null | null | math.NT math.RT | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We computed the first Whittaker coefficient of an Eisenstein series on a global metaplectic group induced from the torus and related the result with a Weyl group multiple Dirichlet series attached to the (dual) root system of the group under a mild assumption on the root system and the degree of the metaplectic cover. This confirms a conjecture of Brubaker-Bump-Friedberg. | [
{
"version": "v1",
"created": "Wed, 20 Nov 2024 09:18:35 GMT"
},
{
"version": "v2",
"created": "Wed, 04 Jun 2025 03:31:09 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Chen",
"Yanze",
""
]
] |
2411.13187 | Erica Coppolillo | Erica Coppolillo, Federico Cinus, Marco Minici, Francesco Bonchi, Giuseppe Manco | Engagement-Driven Content Generation with Large Language Models | null | null | null | null | cs.LG cs.AI | http://creativecommons.org/licenses/by-nc-sa/4.0/ | Large Language Models (LLMs) demonstrate significant persuasive capabilities in one-on-one interactions, but their influence within social networks, where interconnected users and complex opinion dynamics pose unique challenges, remains underexplored. This paper addresses the research question: \emph{Can LLMs generate meaningful content that maximizes user engagement on social networks?}
To answer this, we propose a pipeline using reinforcement learning with simulated feedback, where the network's response to LLM-generated content (i.e., the reward) is simulated through a formal engagement model. This approach bypasses the temporal cost and complexity of live experiments, enabling an efficient feedback loop between the LLM and the network under study. It also allows to control over endogenous factors such as the LLM's position within the social network and the distribution of opinions on a given topic. Our approach is adaptive to the opinion distribution of the underlying network and agnostic to the specifics of the engagement model, which is embedded as a plug-and-play component. Such flexibility makes it suitable for more complex engagement tasks and interventions in computational social science.
Using our framework, we analyze the performance of LLMs in generating social engagement under different conditions, showcasing their full potential in this task. The experimental code is publicly available at https://github.com/mminici/Engagement-Driven-Content-Generation. | [
{
"version": "v1",
"created": "Wed, 20 Nov 2024 10:40:08 GMT"
},
{
"version": "v2",
"created": "Thu, 21 Nov 2024 08:56:57 GMT"
},
{
"version": "v3",
"created": "Fri, 22 Nov 2024 13:05:40 GMT"
},
{
"version": "v4",
"created": "Wed, 04 Jun 2025 16:02:28 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Coppolillo",
"Erica",
""
],
[
"Cinus",
"Federico",
""
],
[
"Minici",
"Marco",
""
],
[
"Bonchi",
"Francesco",
""
],
[
"Manco",
"Giuseppe",
""
]
] |
2411.13630 | Amit Kumar Das | Amit Kumar Das, Debabrata Biswas | Competitive binding of Activator-Repressor in Stochastic Gene Expression | 27 pages,47 figures | null | null | null | q-bio.MN physics.bio-ph | http://creativecommons.org/licenses/by/4.0/ | Regulation of gene expression is the consequence of interactions between the promoter of the gene and the transcription factors (TFs). In this paper, we explore the features of a genetic network where the TFs (activators and repressors) bind the promoter in a competitive way. We develop an analytical theory that offers detailed reaction kinetics of the competitive activator-repressor system which could be the powerful tools for extensive study and analysis of the genetic circuit in future research. Moreover, the theoretical approach helps us to find a most probable set of parameter values which was unavailable in experiments. We study the noisy behaviour of the circuit and compare the profile with the network where the activator and repressor bind the promoter non-competitively. We further notice that, due to the effect of transcriptional reinitiation in the presence of the activator and repressor molecules, there exits some anomalous characteristic features in the mean expressions and noise profiles. We find that, in presence of the reinitiation the noise in transcriptional level remains low while it is higher in translational level than the noise when the reinitiation is absent. In addition, it is possible to reduce the noise further below the Poissonian level in competitive circuit than the non-competitive one with the help of some noise reducing parameters. | [
{
"version": "v1",
"created": "Wed, 20 Nov 2024 18:30:39 GMT"
},
{
"version": "v2",
"created": "Wed, 04 Jun 2025 16:38:56 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Das",
"Amit Kumar",
""
],
[
"Biswas",
"Debabrata",
""
]
] |
2411.13818 | Xiangyu Ding | Xiangyu Ding, Lisa Hui Sun | Proof of Merca's stronger conjecture on truncated Jacobi triple product series | null | null | null | null | math.NT math.CO | http://creativecommons.org/licenses/by/4.0/ | In the study of theta series and partition functions, Andrews and Merca, Guo and Zeng independently conjectured that a truncated Jacobi triple product series has nonnegative coefficients. This conjecture was proved analytically by Mao and combinatorially by Yee. In 2021, Merca proposed a stronger version of the conjecture, that is, for positive integers $1\leq S<R$ with $k\geq 1$, the coefficient of $q^n$ in
the theta series
\[
\frac{(-1)^{k} \sum_{j=k}^{\infty}(-1)^j q^{R j(j+1) / 2}\left(q^{-Sj}-q^{( j+1) S}\right)}{\left(q^S, q^{R-S}; q^R\right)_{\infty}}
\]
is nonnegative. Recently, some very special cases of this conjecture have been proved and studied. For any given $R, S$ and $k$, we take $s=S/(S,R), r=R/(S,R)$ which are coprime, equivalently. In this paper, we confirm Merca's stronger conjecture for sufficiently large $n$. Furthermore, for given $r, s$ and $k$, we provide a systematic method to determine an integer $N(r, s, k)$ such that Merca's stronger conjecture holds for $ n\geq N(r,s,k) $. More precisely, we decompose the infinite product in the denominator of the above theta series into two parts, one of which can be interpreted as the generating function of partitions with certain restricted parts and the other is a nonmodular infinite product.
We derive the general upper and lower bounds for the coefficients of these two parts by using the partition theoretical method and the circle method, respectively. Further multiplying the partition part by the numerator of the theta series and considering the convolution with the nonmodular infinite product, we obtain the constant $N(r,s,k)$ and confirm Merca's stronger conjecture when $n\geq N(r,s,k)$. Moreover, we also show that when $k$ is sufficiently large, this conjecture holds directly for any $n\geq 0$. | [
{
"version": "v1",
"created": "Thu, 21 Nov 2024 03:48:27 GMT"
},
{
"version": "v2",
"created": "Mon, 31 Mar 2025 07:33:17 GMT"
},
{
"version": "v3",
"created": "Wed, 04 Jun 2025 02:48:08 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Ding",
"Xiangyu",
""
],
[
"Sun",
"Lisa Hui",
""
]
] |
2411.13946 | Stefan Graser | Stefan Graser and Stephan B\"ohm | A Systematic Literature Review on Technology Acceptance Research on Augmented Reality in the Field of Training and Education | 8 pages, 7 figures, The Fifteenth International Conference on Advances in Human-oriented and Personalized Mechanisms, Technologies, and Services CENTRIC 2022 | Proceedings of the Fifteenth International Conference on Advances in Human-oriented and Personalized Mechanisms, Technologies, and Services CENTRIC 2022, Lisbon, pp. 20 -28, ISSN 2308-3492 | null | null | cs.HC | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Augmented Reality (AR) is an emerging technology that ranks among the top innovations in interactive media. With the emergence of new technologies, the question about the factors influencing user acceptance arises. Many research models on the user acceptance of technologies were developed and extended to answer this question in the last decades. This research paper provides an overview of the current state in the scientific literature on user acceptance factors of AR in training and education. We conducted a systematic literature review, identifying 45 scientific papers on technology acceptance of augmented reality. Twenty-two papers refer more specifically to the field of training and education. Overall, 33 different technology acceptance models and 34 acceptance variables were identified. Based on the results, there is a great potential for further research. | [
{
"version": "v1",
"created": "Thu, 21 Nov 2024 08:58:34 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Graser",
"Stefan",
""
],
[
"Böhm",
"Stephan",
""
]
] |
2411.14862 | Cheng Shang | Bo-Wang Zhang, Cheng Shang, J. Y. Sun, Zhuo-Cheng Gu, and X. X. Yi | Manipulating spectral transitions and photonic transmission in a non-Hermitian optical system through nanoparticle perturbations | 9 pages, 3 figures | Phys. Rev. A 111, 063702 (2025) | 10.1103/PhysRevA.111.063702 | null | physics.optics quant-ph | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | In recent years, extensive research has been dedicated to the study of parity-time ($\mathcal{PT}$) symmetry, which involves the engineered balance of gain and loss in non-Hermitian optics. Complementary to $\mathcal{PT}$ symmetry, the concept of anti-$\mathcal{PT}$ symmetry has emerged as a natural framework for describing the dynamics of open systems with dissipations. In this work, we study spectral transitions and photon transmission in a linear spinning resonator perturbed by nanoparticles. First, we show that by precisely controlling the nanoparticle perturbations, the eigenvalues (or spectra) of a non-Hermitian system satisfying anti-$\mathcal{PT}$ symmetry can transit to that of a quasi-closed Hermitian system. Second, we outline the essential conditions for constructing a quasi-closed system and analyze its dynamic behavior with respect to photon transmission. By adjusting the rotational angular velocity of the spinning resonator and the strength of the nanoparticle perturbations, the quasi-closed system enables a variety of photon distribution behaviors, which may have significant applications in quantum devices. Our findings offer valuable insights for the design of dissipative quantum devices under realistic conditions and for understanding their responses to external perturbations. | [
{
"version": "v1",
"created": "Fri, 22 Nov 2024 11:22:34 GMT"
},
{
"version": "v2",
"created": "Thu, 09 Jan 2025 17:37:25 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Zhang",
"Bo-Wang",
""
],
[
"Shang",
"Cheng",
""
],
[
"Sun",
"J. Y.",
""
],
[
"Gu",
"Zhuo-Cheng",
""
],
[
"Yi",
"X. X.",
""
]
] |
2411.15466 | Chaehun Shin | Chaehun Shin, Jooyoung Choi, Heeseung Kim, Sungroh Yoon | Large-Scale Text-to-Image Model with Inpainting is a Zero-Shot Subject-Driven Image Generator | CVPR 2025 | null | null | null | cs.CV | http://creativecommons.org/licenses/by/4.0/ | Subject-driven text-to-image generation aims to produce images of a new subject within a desired context by accurately capturing both the visual characteristics of the subject and the semantic content of a text prompt. Traditional methods rely on time- and resource-intensive fine-tuning for subject alignment, while recent zero-shot approaches leverage on-the-fly image prompting, often sacrificing subject alignment. In this paper, we introduce Diptych Prompting, a novel zero-shot approach that reinterprets as an inpainting task with precise subject alignment by leveraging the emergent property of diptych generation in large-scale text-to-image models. Diptych Prompting arranges an incomplete diptych with the reference image in the left panel, and performs text-conditioned inpainting on the right panel. We further prevent unwanted content leakage by removing the background in the reference image and improve fine-grained details in the generated subject by enhancing attention weights between the panels during inpainting. Experimental results confirm that our approach significantly outperforms zero-shot image prompting methods, resulting in images that are visually preferred by users. Additionally, our method supports not only subject-driven generation but also stylized image generation and subject-driven image editing, demonstrating versatility across diverse image generation applications. Project page: https://diptychprompting.github.io/ | [
{
"version": "v1",
"created": "Sat, 23 Nov 2024 06:17:43 GMT"
},
{
"version": "v2",
"created": "Wed, 04 Jun 2025 01:35:55 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Shin",
"Chaehun",
""
],
[
"Choi",
"Jooyoung",
""
],
[
"Kim",
"Heeseung",
""
],
[
"Yoon",
"Sungroh",
""
]
] |
2411.15641 | Andrei Derevianko | H. W. T. Morgan, H. B. Tran Tan, R. Elwell, A. N.Alexandrova, Eric R. Hudson, Andrei Derevianko | Theory of internal conversion of the thorium-229 nuclear isomer in solid-state hosts | Discussions of NEET process and particle-hole interaction added | null | 10.1103/9s8f-hv1f | null | physics.atom-ph | http://creativecommons.org/licenses/by/4.0/ | Laser excitation of thorium-229 nuclei in doped wide bandgap crystals has been demonstrated recently, opening the possibility of developing ultrastable solid-state clocks and sensitive searches for new physics. We develop a quantitative theory of the internal conversion of isomeric thorium-229 in solid-state hosts. The internal conversion of the isomer proceeds by resonantly exciting a valence band electron to a defect state, accompanied by multi-phonon emission. We demonstrate that, if the process is energetically allowed, it generally quenches the isomer on timescales much faster than the isomer's radiative lifetime, despite thorium being in the +4 charge state in the valence band. | [
{
"version": "v1",
"created": "Sat, 23 Nov 2024 19:47:22 GMT"
},
{
"version": "v2",
"created": "Tue, 03 Jun 2025 21:19:58 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Morgan",
"H. W. T.",
""
],
[
"Tan",
"H. B. Tran",
""
],
[
"Elwell",
"R.",
""
],
[
"Alexandrova",
"A. N.",
""
],
[
"Hudson",
"Eric R.",
""
],
[
"Derevianko",
"Andrei",
""
]
] |
2411.15983 | Marco Salvalaglio | Caihao Qiu, David J. Srolovitz, Gregory S. Rohrer, Jian Han, Marco Salvalaglio | Why Grain Growth is Not Curvature Flow | 11 pages, 5 figures. Supplementary Information attached (additional 7 pages, 5 figures) | null | null | null | cond-mat.mtrl-sci | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Grain growth in polycrystals is traditionally considered a capillarity-driven process, where grain boundaries (GBs) migrate toward their centers of curvature (i.e., mean curvature flow) with a velocity proportional to the local curvature (including extensions to account for anisotropic GB energy and mobility). Experimental and simulation evidence shows that this simplistic view is untrue. We demonstrate that the failure of the classical mean curvature flow description of grain growth mainly originates from the shear deformation naturally coupled with GB motion (i.e., shear coupling). Our findings are built on large-scale microstructure evolution simulations incorporating the fundamental (crystallography-respecting) microscopic mechanism of GB migration. The nature of the deviations from curvature flow revealed in our simulations is consistent with observations in recent experimental studies on different materials. This work also demonstrates how to incorporate the mechanical effects that are essential to the accurate prediction of microstructure evolution. | [
{
"version": "v1",
"created": "Sun, 24 Nov 2024 21:06:45 GMT"
},
{
"version": "v2",
"created": "Tue, 06 May 2025 11:37:35 GMT"
},
{
"version": "v3",
"created": "Wed, 04 Jun 2025 15:20:49 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Qiu",
"Caihao",
""
],
[
"Srolovitz",
"David J.",
""
],
[
"Rohrer",
"Gregory S.",
""
],
[
"Han",
"Jian",
""
],
[
"Salvalaglio",
"Marco",
""
]
] |
2411.16258 | Nina Dimova | N. Dimova, J.S. Barnard, D. Bortoletto, G. Crevatin, M. Gallagher-Jones, R. Goldsbrough, D. Hynds, A. Kirkland, L. O'Ryan, R. Plackett, I. Shipsey, D. Weatherill, D. Wood | Measurement of the Resolution of the Timepix4 Detector for 100 keV and 200 keV Electrons for Transmission Electron Microscopy | 20 pages, 11 figures; submitted to NIMA; corresponding author: N. Dimova | null | 10.1016/j.nima.2025.170335 | null | physics.ins-det | http://creativecommons.org/licenses/by-sa/4.0/ | We have evaluated the imaging capabilities of the Timepix4 hybrid silicon pixel detector for 100 keV and 200 keV electrons in a Transmission Electron Microscope (TEM). Using the knife-edge method, we have measured the Modulation Transfer Function (MTF) at both energies. Our results show a decrease in MTF response at Nyquist (spatial) frequency, dropping from approximately 0.16 at 100 keV to 0.0046 at 200 keV. However, by using the temporal structure of the detected events, including the arrival time and amplitude provided by the Timepix4, we enhanced the spatial discrimination of electron arrival. This approach improved the MTF at Nyquist by factors of 2.12 for 100 keV and 3.16 for 200 keV. These findings demonstrate that the blurring effects caused by extended electron trajectories within the sensing layer can be partially corrected in the image data. | [
{
"version": "v1",
"created": "Mon, 25 Nov 2024 10:20:44 GMT"
},
{
"version": "v2",
"created": "Wed, 04 Jun 2025 10:57:43 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Dimova",
"N.",
""
],
[
"Barnard",
"J. S.",
""
],
[
"Bortoletto",
"D.",
""
],
[
"Crevatin",
"G.",
""
],
[
"Gallagher-Jones",
"M.",
""
],
[
"Goldsbrough",
"R.",
""
],
[
"Hynds",
"D.",
""
],
[
"Kirkland",
"A.",
""
],
[
"O'Ryan",
"L.",
""
],
[
"Plackett",
"R.",
""
],
[
"Shipsey",
"I.",
""
],
[
"Weatherill",
"D.",
""
],
[
"Wood",
"D.",
""
]
] |
2411.16288 | Ferdinand Ihringer | Ferdinand Ihringer | A Survey of Cameron-Liebler Sets and Low Degree Boolean Functions in Grassmann Graphs | 26 pages (20 text, 6 references); accepted by the conference volume of Sum(m)it280 | null | null | null | math.CO cs.DM | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We survey results for Cameron-Liebler sets and low degree Boolean functions for Hamming graphs, Johnson graphs and Grassmann graphs from the point of view of association schemes. This survey covers selected results in finite geometry, Boolean function analysis, design theory, coding theory, and cryptography. | [
{
"version": "v1",
"created": "Mon, 25 Nov 2024 11:12:35 GMT"
},
{
"version": "v2",
"created": "Wed, 04 Jun 2025 09:04:18 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Ihringer",
"Ferdinand",
""
]
] |
2411.16666 | Jiaan Han | Jiaan Han, Junxiao Chen, Yanzhe Fu | CatNet: Controlling the False Discovery Rate in LSTM with SHAP Feature Importance and Gaussian Mirrors | null | null | null | null | stat.ML cs.AI cs.LG q-fin.ST | http://creativecommons.org/licenses/by-nc-sa/4.0/ | We introduce CatNet, an algorithm that effectively controls False Discovery Rate (FDR) and selects significant features in LSTM. CatNet employs the derivative of SHAP values to quantify the feature importance, and constructs a vector-formed mirror statistic for FDR control with the Gaussian Mirror algorithm. To avoid instability due to nonlinear or temporal correlations among features, we also propose a new kernel-based independence measure. CatNet performs robustly on different model settings with both simulated and real-world data, which reduces overfitting and improves interpretability of the model. Our framework that introduces SHAP for feature importance in FDR control algorithms and improves Gaussian Mirror can be naturally extended to other time-series or sequential deep learning models. | [
{
"version": "v1",
"created": "Mon, 25 Nov 2024 18:53:37 GMT"
},
{
"version": "v2",
"created": "Tue, 26 Nov 2024 16:23:19 GMT"
},
{
"version": "v3",
"created": "Wed, 04 Jun 2025 17:23:25 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Han",
"Jiaan",
""
],
[
"Chen",
"Junxiao",
""
],
[
"Fu",
"Yanzhe",
""
]
] |
2411.16826 | Edoardo Di Martino | Edoardo Di Martino, Alessandro Galeazzi, Michele Starnini, Walter Quattrociocchi, Matteo Cinelli | Ideological Fragmentation of the Social Media Ecosystem: From echo chambers to echo platforms | null | null | null | null | cs.CY cs.SI physics.soc-ph | http://creativecommons.org/licenses/by/4.0/ | The entertainment-driven nature of social media encourages users to engage with like-minded individuals and consume content aligned with their beliefs, limiting exposure to diverse perspectives. Simultaneously, users migrate between platforms, either due to moderation policies like de-platforming or in search of environments better suited to their preferences. These dynamics drive the specialization of the social media ecosystem, shifting from internal echo chambers to "echo platforms"--entire platforms functioning as ideologically homogeneous niches. To systematically analyze this phenomenon in political discussions, we propose a quantitative approach based on three key dimensions: platform centrality, news consumption, and user base composition. We analyze 117 million posts related to the 2020 US Presidential elections from nine social media platforms--Facebook, Reddit, Twitter, YouTube, BitChute, Gab, Parler, Scored, and Voat. Our findings reveal significant differences among platforms in their centrality within the ecosystem, the reliability of circulated news, and the ideological diversity of their users, highlighting a clear divide between mainstream and alt-tech platforms. The latter occupy a peripheral role, feature a higher prevalence of unreliable content, and exhibit greater ideological uniformity. These results highlight the key dimensions shaping the fragmentation and polarization of the social media landscape. | [
{
"version": "v1",
"created": "Mon, 25 Nov 2024 18:45:03 GMT"
},
{
"version": "v2",
"created": "Wed, 04 Jun 2025 13:07:39 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Di Martino",
"Edoardo",
""
],
[
"Galeazzi",
"Alessandro",
""
],
[
"Starnini",
"Michele",
""
],
[
"Quattrociocchi",
"Walter",
""
],
[
"Cinelli",
"Matteo",
""
]
] |
2411.17089 | Lei Gao | Chaoyi Jiang, Lei Gao, Hossein Entezari Zarch, Murali Annavaram | KVPR: Efficient LLM Inference with I/O-Aware KV Cache Partial Recomputation | ACL Findings 2025 | null | null | null | cs.LG cs.DC cs.PF | http://creativecommons.org/licenses/by-nc-nd/4.0/ | Inference for Large Language Models (LLMs) is computationally demanding. To reduce the cost of auto-regressive decoding, Key-Value (KV) cache is used to store intermediate activations, which significantly lowers the computational overhead for token generation. However, the memory required for the KV cache grows rapidly, often exceeding the capacity of GPU memory. A cost-effective alternative is to offload KV cache to CPU memory, which alleviates GPU memory pressure, but shifts the bottleneck to the limited bandwidth of the PCIe connection between the CPU and GPU. Existing methods attempt to address these issues by overlapping GPU computation with I/O or employing CPU-GPU heterogeneous execution, but they are hindered by excessive data movement and dependence on CPU capabilities. Fully overlapping PCIe communication latency gets challenging as the size of the KV cache grows and/or the GPU compute capabilities increase. In this paper, we introduce KVPR, an efficient I/O-aware LLM inference method where the CPU first transfers a partial set of activations, from which the GPU can start recomputing the KV cache values. While the GPU recomputes the partial KV cache, the remaining portion of the KV cache is transferred concurrently from the CPU. This approach overlaps GPU recomputation with KV cache transfer to minimize idle GPU time and maximize inference performance. KVPR is fully automated by integrating a profiler module that utilizes input characteristics and system hardware information, a scheduler module to optimize the distribution of computation and communication workloads, and a runtime module to efficiently execute the derived execution plan. Experimental results show that KVPR achieves up to 35.8% lower latency and 46.2% higher throughput during decoding compared to state-of-the-art approaches. The code is available at https://github.com/chaoyij/KVPR. | [
{
"version": "v1",
"created": "Tue, 26 Nov 2024 04:03:14 GMT"
},
{
"version": "v2",
"created": "Wed, 04 Jun 2025 16:08:50 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Jiang",
"Chaoyi",
""
],
[
"Gao",
"Lei",
""
],
[
"Zarch",
"Hossein Entezari",
""
],
[
"Annavaram",
"Murali",
""
]
] |
2411.17098 | Wei-Hong Liang | Hai-Peng Li, Wei-Hong Liang, Chu-Wen Xiao, Ju-Jun Xie, Eulogio Oset | Determination of the binding and $DK$ probability of the $D^{*}_{s0}(2317)$ from the $(\bar{D}\bar K)^-$ mass distributions in $\Lambda_{b}\to \Lambda_{c} (\bar{D}\bar K)^-$ decays | 8 pages, 5 figures, 6 tables; v2: discussion added, references added, version to appear in Eur.Phys.J.C | null | null | null | hep-ph | http://creativecommons.org/licenses/by/4.0/ | We study the $\Lambda_{b}\to\Lambda_{c}\bar{D}^{0}K^{-}$ and $\Lambda_{b}\to \Lambda_{c}D^{-}\bar{K}^{0}$ decays which proceed via a Cabibbo and $N_c$ favored process of external emission, and we determine the $\bar{D}^{0}K^{-}$ and $D^{-}\bar{K}^{0}$ mass distributions close to the $\bar{D} \bar{K}$ threshold. For this, we use the tree level contribution plus the rescattering of the meson-meson components, using the extension of the local hidden gauge approach to the charm sector that produces the $D^*_{s0}(2317)$ resonance. We observe a large enhancement of the mass distributions close to threshold due to the presence of this resonance below threshold. Next we undertake the inverse problem of extracting the maximum information on the interaction of the $\bar{D} \bar{K}$ channels from these distributions, and using the resampling method we find that from these data one can obtain precise values of the scattering lengths and effective ranges, the existence of an $I=0$ bound state with a precision of about $4 \;\rm MeV$ in the mass, plus the $\bar{D} \bar{K}$ molecular probability of this state with reasonable precision. Given the fact that the $\Lambda_{b}\to\Lambda_{c}\bar{D}^{0}K^{-}$ decay is already measured by the LHCb collaboration, it is expected that in the next runs with more statistics of the decay, these mass distributions can be measured with precision and the method proposed here can be used to determine the nature of the $D^*_{s0}(2317)$, which is still an issue of debate. | [
{
"version": "v1",
"created": "Tue, 26 Nov 2024 04:32:00 GMT"
},
{
"version": "v2",
"created": "Wed, 04 Jun 2025 15:23:01 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Li",
"Hai-Peng",
""
],
[
"Liang",
"Wei-Hong",
""
],
[
"Xiao",
"Chu-Wen",
""
],
[
"Xie",
"Ju-Jun",
""
],
[
"Oset",
"Eulogio",
""
]
] |
2411.17467 | Xuweiyi Chen | Xuweiyi Chen, Zezhou Cheng | Learning 3D Representations from Procedural 3D Programs | SynData4CV @ CVPR2025 | Project Page: https://point-mae-zero.cs.virginia.edu/ | null | null | null | cs.CV | http://creativecommons.org/licenses/by/4.0/ | Self-supervised learning has emerged as a promising approach for acquiring transferable 3D representations from unlabeled 3D point clouds. Unlike 2D images, which are widely accessible, acquiring 3D assets requires specialized expertise or professional 3D scanning equipment, making it difficult to scale and raising copyright concerns. To address these challenges, we propose learning 3D representations from procedural 3D programs that automatically generate 3D shapes using simple primitives and augmentations. Remarkably, despite lacking semantic content, the 3D representations learned from the procedurally generated 3D shapes perform on par with state-of-the-art representations learned from semantically recognizable 3D models (e.g., airplanes) across various downstream 3D tasks, including shape classification, part segmentation, and masked point cloud completion. We provide a detailed analysis on factors that make a good 3D procedural program. Extensive experiments further suggest that current self-supervised learning methods on point clouds do not rely on the semantics of 3D shapes, shedding light on the nature of 3D representations learned. | [
{
"version": "v1",
"created": "Mon, 25 Nov 2024 18:59:57 GMT"
},
{
"version": "v2",
"created": "Wed, 04 Jun 2025 11:12:41 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Chen",
"Xuweiyi",
""
],
[
"Cheng",
"Zezhou",
""
]
] |
2411.18065 | Faraz Tahmasebi | Faraz Tahmasebi, Yian Wang, Benji Y.H. Huang, Hyoukjun Kwon | FlexiBit: Fully Flexible Precision Bit-parallel Accelerator Architecture for Arbitrary Mixed Precision AI | 11 pages, 19 figures, 5 tables, 4 pseudo-codes | null | null | null | cs.AR | http://creativecommons.org/licenses/by/4.0/ | Recent research has shown that large language models (LLMs) can utilize low-precision floating point (FP) quantization to deliver high efficiency while maintaining original model accuracy. In particular, recent works have shown the effectiveness of non-power-of-two precisions, such as FP6 and FP5, and diverse sensitivity to low-precision arithmetic of LLM layers, which motivates mixed precision arithmetic including non-power-of-two precisions in LLMs. Although low-precision algorithmically leads to low computational overheads, such benefits cannot be fully exploited due to hardware constraints that support a limited set of power-of-two precisions (e.g., FP8, 16, 32, and 64 in NVIDIA H100 Tensor Core). In addition, the hardware compute units are designed to support standard formats (e.g., E4M3 and E5M2 for FP8). Such practices require re-designing the hardware whenever new precision and format emerge, which leads to high hardware replacement costs to exploit the benefits of new precisions and formats. Therefore, in this paper, we propose a new accelerator architecture, FlexiBit, which efficiently supports FP and INT arithmetic in arbitrary precisions and formats. Unlike previous bit-serial designs, which also provide flexibility but at the cost of performance due to its bit-wise temporal processing nature, FlexiBit's architecture enables bit-parallel processing of any precision and format without compute unit underutilization. FlexiBit's new capability to exploit non-power of two precision and format led to 1.66x and 1.62x higher performance per area on GPT-3 in FP6 targeting a cloud-scale accelerator, compared to a Tensor Core-like architecture and a state-of-the-art bit-parallel flexible precision accelerator, BitFusion, respectively. Also, the bit-parallel nature of FlexiBit's architecture led to 3.9x higher performance/area compared to a state-of-the-art bit-serial architecture. | [
{
"version": "v1",
"created": "Wed, 27 Nov 2024 05:16:26 GMT"
},
{
"version": "v2",
"created": "Wed, 04 Jun 2025 01:25:53 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Tahmasebi",
"Faraz",
""
],
[
"Wang",
"Yian",
""
],
[
"Huang",
"Benji Y. H.",
""
],
[
"Kwon",
"Hyoukjun",
""
]
] |
2411.18445 | Rathan Samala | Lavanya V Salian, Samala Rathan, Rakesh Kumar | Compact finite-difference scheme for some Sobolev type equations with Dirichlet boundary conditions | 29 pages, 13 figues, 12 tables | null | null | null | math.NA cs.NA | http://creativecommons.org/licenses/by/4.0/ | This study aims to construct a stable, high-order compact finite difference method for solving Sobolev-type equations with Dirichlet boundary conditions in one-space dimension. Approximation of higher-order mixed derivatives in some specific Sobolev-type equations requires a bigger stencil information. One can approximate such derivatives on compact stencils, which are higher-order accurate and take less stencil information but are implicit and sparse. Spatial derivatives in this work are approximated using the sixth-order compact finite difference method (Compact6), while temporal derivatives are handled with the explicit forward Euler difference scheme. We examine the accuracy and convergence behavior of the proposed scheme. Using the von Neumann stability analysis, we establish $L_2-$stability theory for the linear case. We derive conditions under which fully discrete schemes are stable. Also, the amplification factor $\mathcal{C}(\theta)$ is analyzed to ensure the decay property over time. Real parts of $\mathcal{C}(\theta)$ lying on the negative real axis confirm the exponential decay of the solution. A series of numerical experiments were performed to verify the effectiveness of the proposed scheme. These tests include both one dimensional and two-dimensional cases of cases of advection-free and advection-diffusion flows. They also cover applications to the equal width equation, such as the propagation of a single solitary wave, interactions between two and three solitary waves, undular bore formation, and the Benjamin-Bona-Mahony-Burgers equation. | [
{
"version": "v1",
"created": "Wed, 27 Nov 2024 15:35:57 GMT"
},
{
"version": "v2",
"created": "Wed, 04 Jun 2025 02:30:36 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Salian",
"Lavanya V",
""
],
[
"Rathan",
"Samala",
""
],
[
"Kumar",
"Rakesh",
""
]
] |
2411.19647 | Shaowen Wang | Shaowen Wang, Anan Liu, Jian Xiao, Huan Liu, Yuekui Yang, Cong Xu, Qianqian Pu, Suncong Zheng, Wei Zhang, Di Wang, Jie Jiang, Jian Li | CAdam: Confidence-Based Optimization for Online Learning | null | null | null | null | cs.LG cs.AI stat.ML | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Modern recommendation systems frequently employ online learning to dynamically update their models with freshly collected data. The most commonly used optimizer for updating neural networks in these contexts is the Adam optimizer, which integrates momentum ($m_t$) and adaptive learning rate ($v_t$). However, the volatile nature of online learning data, characterized by its frequent distribution shifts and presence of noise, poses significant challenges to Adam's standard optimization process: (1) Adam may use outdated momentum and the average of squared gradients, resulting in slower adaptation to distribution changes, and (2) Adam's performance is adversely affected by data noise. To mitigate these issues, we introduce CAdam, a confidence-based optimization strategy that assesses the consistency between the momentum and the gradient for each parameter dimension before deciding on updates. If momentum and gradient are in sync, CAdam proceeds with parameter updates according to Adam's original formulation; if not, it temporarily withholds updates and monitors potential shifts in data distribution in subsequent iterations. This method allows CAdam to distinguish between the true distributional shifts and mere noise, and to adapt more quickly to new data distributions. In various settings with distribution shift or noise, our experiments demonstrate that CAdam surpasses other well-known optimizers, including the original Adam. Furthermore, in large-scale A/B testing within a live recommendation system, CAdam significantly enhances model performance compared to Adam, leading to substantial increases in the system's gross merchandise volume (GMV). | [
{
"version": "v1",
"created": "Fri, 29 Nov 2024 12:00:27 GMT"
},
{
"version": "v2",
"created": "Wed, 04 Jun 2025 14:58:21 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Wang",
"Shaowen",
""
],
[
"Liu",
"Anan",
""
],
[
"Xiao",
"Jian",
""
],
[
"Liu",
"Huan",
""
],
[
"Yang",
"Yuekui",
""
],
[
"Xu",
"Cong",
""
],
[
"Pu",
"Qianqian",
""
],
[
"Zheng",
"Suncong",
""
],
[
"Zhang",
"Wei",
""
],
[
"Wang",
"Di",
""
],
[
"Jiang",
"Jie",
""
],
[
"Li",
"Jian",
""
]
] |
2412.00194 | Pradip Kattel | Pradip Kattel, Parameshwar R. Pasnoori, J. H. Pixley, and Natan Andrei | Edge modes and boundary impurities in the anisotropic Heisenberg spin chain | 20+22 pages, 27 figures | Phys. Rev. B 111, 174430 ( 2025) | 10.1103/PhysRevB.111.174430 | null | cond-mat.str-el cond-mat.stat-mech hep-th math-ph math.MP quant-ph | http://creativecommons.org/licenses/by/4.0/ | We present a comprehensive analysis of boundary phenomena in a spin-$\frac{1}{2}$ anisotropic Heisenberg chain (XXZ-$\frac{1}{2}$) in the gapped antiferromagnetic phase, with a particular focus on the interplay between fractionalized spin-$\frac{1}{4} $ edge modes and a coupled spin-$\frac{1}{2}$ impurity at the edge. Employing a combination of Bethe Ansatz, exact diagonalization, and density matrix renormalization group (DMRG) methods, we explore the intricate phase diagram that emerges when the impurity is coupled either integrably or non-integrably to the chain. For integrable antiferromagnetic impurity couplings, we identify two distinct phases: the Kondo phase, where the impurity is screened by a multiparticle Kondo effect, and the antiferromagnetic bound mode phase, where an exponentially localized bound state screens the impurity in the ground state. When coupled ferromagnetically while maintaining integrability, the impurity behaves as a free spin-$\frac{1}{2}$, leading to either a ferromagnetic bound mode phase, where the impurity remains free in the ground state but may be screened at higher energy excitations or an unscreened (or local moment) phase where impurity remains unscreened in every eigenstate whereas for non-integrable ferromagnetic coupling, the impurity is not free. In the case of non-integrable antiferromagnetic coupling, a third phase emerges, characterized by mid-gap excitations with two degenerate states below the mass gap on top of the Kondo and antiferromagnetic bound mode phases, further enriching the phase diagram. Our findings highlight the nuanced behavior of boundary impurities in gapped antiferromagnetic systems, offering new insights into Kondo effects and impurity screening in the presence of fractionalized edge modes and bulk antiferromagnetic order. | [
{
"version": "v1",
"created": "Fri, 29 Nov 2024 19:00:06 GMT"
},
{
"version": "v2",
"created": "Tue, 13 May 2025 02:53:39 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Kattel",
"Pradip",
""
],
[
"Pasnoori",
"Parameshwar R.",
""
],
[
"Pixley",
"J. H.",
""
],
[
"Andrei",
"Natan",
""
]
] |
2412.00418 | Yu Shi | Yu Shi, Yiqi Wang, WeiXuan Lang, Jiaxin Zhang, Pan Dong, Aiping Li | Mixture of Experts for Node Classification | null | null | null | null | cs.SI cs.AI | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Nodes in the real-world graphs exhibit diverse patterns in numerous aspects, such as degree and homophily. However, most existent node predictors fail to capture a wide range of node patterns or to make predictions based on distinct node patterns, resulting in unsatisfactory classification performance. In this paper, we reveal that different node predictors are good at handling nodes with specific patterns and only apply one node predictor uniformly could lead to suboptimal result. To mitigate this gap, we propose a mixture of experts framework, MoE-NP, for node classification. Specifically, MoE-NP combines a mixture of node predictors and strategically selects models based on node patterns. Experimental results from a range of real-world datasets demonstrate significant performance improvements from MoE-NP. | [
{
"version": "v1",
"created": "Sat, 30 Nov 2024 10:05:03 GMT"
},
{
"version": "v2",
"created": "Wed, 12 Mar 2025 12:33:46 GMT"
},
{
"version": "v3",
"created": "Wed, 04 Jun 2025 03:57:43 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Shi",
"Yu",
""
],
[
"Wang",
"Yiqi",
""
],
[
"Lang",
"WeiXuan",
""
],
[
"Zhang",
"Jiaxin",
""
],
[
"Dong",
"Pan",
""
],
[
"Li",
"Aiping",
""
]
] |
2412.00779 | Junhee Ryu | Hongjie Dong, Junhee Ryu | Sobolev estimates for parabolic and elliptic equations in divergence form with degenerate coefficients | 42 pages, minor revision, to appear in Trans. Amer. Math. Soc | null | null | null | math.AP | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We study a class of degenerate parabolic and elliptic equations in divergence form in the upper half space $\{x_d>0\}$. The leading coefficients are of the form $x_d^2a_{ij}$, where $a_{ij}$ are bounded, uniformly elliptic, and measurable in $(t,x_d)$ except $a_{dd}$, which is measurable in $t$ or $x_d$. Additionally, they have small bounded mean oscillations in the other spatial variables. We obtain the well-posedness and regularity of solutions in weighted mixed-norm Sobolev spaces. | [
{
"version": "v1",
"created": "Sun, 01 Dec 2024 12:02:54 GMT"
},
{
"version": "v2",
"created": "Wed, 04 Jun 2025 04:40:42 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Dong",
"Hongjie",
""
],
[
"Ryu",
"Junhee",
""
]
] |
2412.01886 | Ryohei Kobayashi | Ryohei Kobayashi, Yuyang Li, Hanyu Xue, Po-Shen Hsin, Yu-An Chen | Generalized Statistics on Lattice | 40 pages, 11 figures. added comments on gapless systems in Sec.VI A. changed title | null | null | null | quant-ph cond-mat.str-el hep-th math.QA | http://creativecommons.org/licenses/by/4.0/ | Statistics of excitations play an essential role in understanding phases of matter. In this paper, we introduce a universal method for studying the generalized statistics of Abelian particles and extended excitations in lattices of any dimension. We compute the statistics using the Berry phase of a sequence of unitary operators that transports the excitations while canceling local ambiguities at each step. The sequence is derived from locality, using the Smith normal form. We show that the statistics are quantized invariants. Our method unifies the statistics for the braiding and fusion of particles and loops, and leads to the discovery of novel statistics for membrane excitations. The statistics can be interpreted as the quantum anomaly of a generalized global symmetry, which manifests as an obstruction to gauging the symmetry on lattices. Furthermore, we show that non-trivial statistics forbid short-range entangled states, establishing the dynamical consequence of anomalies in microscopic lattice models. | [
{
"version": "v1",
"created": "Mon, 02 Dec 2024 19:00:00 GMT"
},
{
"version": "v2",
"created": "Mon, 09 Dec 2024 00:13:19 GMT"
},
{
"version": "v3",
"created": "Wed, 04 Jun 2025 17:44:31 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Kobayashi",
"Ryohei",
""
],
[
"Li",
"Yuyang",
""
],
[
"Xue",
"Hanyu",
""
],
[
"Hsin",
"Po-Shen",
""
],
[
"Chen",
"Yu-An",
""
]
] |
2412.03293 | Yichen Zhu | Junjie Wen, Minjie Zhu, Yichen Zhu, Zhibin Tang, Jinming Li, Zhongyi Zhou, Chengmeng Li, Xiaoyu Liu, Yaxin Peng, Chaomin Shen, Feifei Feng | Diffusion-VLA: Generalizable and Interpretable Robot Foundation Model via Self-Generated Reasoning | Accepted by ICML 2025. The project page is available at: http://diffusion-vla.github.io | null | null | null | cs.RO cs.CV | http://creativecommons.org/licenses/by/4.0/ | In this paper, we present DiffusionVLA, a novel framework that seamlessly combines the autoregression model with the diffusion model for learning visuomotor policy. Central to our approach is a next-token prediction objective, enabling the model to reason effectively over the user's query in the context of current observations. Subsequently, a diffusion model is attached to generate robust action outputs. To enhance policy learning through self-reasoning, we introduce a novel reasoning injection module that integrates reasoning phrases directly into the policy learning process. The whole framework is simple and flexible, making it easy to deploy and upgrade. We conduct extensive experiments using multiple real robots to validate the effectiveness of DiffusionVLA. Our tests include a challenging factory sorting task, where DiffusionVLA successfully categorizes objects, including those not seen during training. We observe that the reasoning module makes the model interpretable. It allows observers to understand the model thought process and identify potential causes of policy failures. Additionally, we test DiffusionVLA on a zero-shot bin-picking task, achieving 63.7\% accuracy on 102 previously unseen objects. Our method demonstrates robustness to visual changes, such as distractors and new backgrounds, and easily adapts to new embodiments. Furthermore, DiffusionVLA can follow novel instructions and retain conversational ability. Notably, DiffusionVLA is data-efficient and fast at inference; our smallest DiffusionVLA-2B runs 82Hz on a single A6000 GPU and can train from scratch on less than 50 demonstrations for a complex task. Finally, we scale the model from 2B to 72B parameters, showcasing improved generalization capabilities with increased model size. | [
{
"version": "v1",
"created": "Wed, 04 Dec 2024 13:11:38 GMT"
},
{
"version": "v2",
"created": "Tue, 13 May 2025 09:37:27 GMT"
},
{
"version": "v3",
"created": "Wed, 04 Jun 2025 08:30:06 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Wen",
"Junjie",
""
],
[
"Zhu",
"Minjie",
""
],
[
"Zhu",
"Yichen",
""
],
[
"Tang",
"Zhibin",
""
],
[
"Li",
"Jinming",
""
],
[
"Zhou",
"Zhongyi",
""
],
[
"Li",
"Chengmeng",
""
],
[
"Liu",
"Xiaoyu",
""
],
[
"Peng",
"Yaxin",
""
],
[
"Shen",
"Chaomin",
""
],
[
"Feng",
"Feifei",
""
]
] |
2412.03477 | Wasilij Barsukow | Wasilij Barsukow, Janina Kern, Christian Klingenberg, Lisa Lechner | Analysis of the multi-dimensional semi-discrete Active Flux method using the Fourier transform | null | null | null | null | math.NA cs.NA | http://creativecommons.org/licenses/by-nc-sa/4.0/ | The degrees of freedom of Active Flux are cell averages and point values along the cell boundaries. These latter are shared between neighbouring cells, which gives rise to a globally continuous reconstruction. The semi-discrete Active Flux method uses its degrees of freedom to obtain Finite Difference approxi\-mations to the spatial derivatives which are used in the point value update. The averages are updated using a quadrature of the flux and making use of the point values as quadrature points. The integration in time employs standard Runge-Kutta methods. We show that this generalization of the Active Flux method in two and three spatial dimensions is stationarity preserving for linear acoustics on Cartesian grids, and present an analysis of numerical diffusion and stability. | [
{
"version": "v1",
"created": "Wed, 04 Dec 2024 17:07:59 GMT"
},
{
"version": "v2",
"created": "Wed, 04 Jun 2025 08:19:19 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Barsukow",
"Wasilij",
""
],
[
"Kern",
"Janina",
""
],
[
"Klingenberg",
"Christian",
""
],
[
"Lechner",
"Lisa",
""
]
] |
2412.03747 | The CMS Collaboration | CMS Collaboration | Model-agnostic search for dijet resonances with anomalous jet substructure in proton-proton collisions at $\sqrt{s}$ = 13 TeV | Replaced with the published version. Added the journal reference and the DOI. All the figures and tables can be found at http://cms-results.web.cern.ch/cms-results/public-results/publications/EXO-22-026 (CMS Public Pages) | Rep. Prog. Phys. 88 (2025) 067802 | 10.1088/1361-6633/add762 | CMS-EXO-22-026, CERN-EP-2024-291 | hep-ex | http://creativecommons.org/licenses/by/4.0/ | This paper presents a model-agnostic search for narrow resonances in the dijet final state in the mass range 1.8-6 TeV. The signal is assumed to produce jets with substructure atypical of jets initiated by light quarks or gluons, with minimal additional assumptions. Search regions are obtained by utilizing multivariate machine-learning methods to select jets with anomalous substructure. A collection of complementary anomaly detection methods - based on unsupervised, weakly supervised, and semisupervised algorithms - are used in order to maximize the sensitivity to unknown new physics signatures. These algorithms are applied to data corresponding to an integrated luminosity of 138 fb$^{-1}$, recorded by the CMS experiment at the LHC, at a center-of-mass energy of 13 TeV. No significant excesses above background expectations are seen. Exclusion limits are derived on the production cross section of benchmark signal models varying in resonance mass, jet mass, and jet substructure. Many of these signatures have not been previously sought, making several of the limits reported on the corresponding benchmark models the first ever. When compared to benchmark inclusive and substructure-based search strategies, the anomaly detection methods are found to significantly enhance the sensitivity to a variety of models. | [
{
"version": "v1",
"created": "Wed, 04 Dec 2024 22:34:03 GMT"
},
{
"version": "v2",
"created": "Tue, 03 Jun 2025 17:14:54 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"CMS Collaboration",
"",
""
]
] |
2412.04184 | Shailendra Bhandari | Shailendra Bhandari, Pedro Lencastre, Rujeena Mathema, Alexander Szorkovszky, Anis Yazidi, Pedro Lind | Modeling Eye Gaze Velocity Trajectories using GANs with Spectral Loss for Enhanced Fidelity | 16 | Nature Scientific reports 2025 | 10.1038/s41598-025-05286-5 | null | cs.NE cs.LG | http://creativecommons.org/licenses/by/4.0/ | Accurate modeling of eye gaze dynamics is essential for advancement in human-computer interaction, neurological diagnostics, and cognitive research. Traditional generative models like Markov models often fail to capture the complex temporal dependencies and distributional nuance inherent in eye gaze trajectories data. This study introduces a GAN framework employing LSTM and CNN generators and discriminators to generate high-fidelity synthetic eye gaze velocity trajectories. We conducted a comprehensive evaluation of four GAN architectures: CNN-CNN, LSTM-CNN, CNN-LSTM, and LSTM-LSTM trained under two conditions: using only adversarial loss and using a weighted combination of adversarial and spectral losses. Our findings reveal that the LSTM-CNN architecture trained with this new loss function exhibits the closest alignment to the real data distribution, effectively capturing both the distribution tails and the intricate temporal dependencies. The inclusion of spectral regularization significantly enhances the GANs ability to replicate the spectral characteristics of eye gaze movements, leading to a more stable learning process and improved data fidelity. Comparative analysis with an HMM optimized to four hidden states further highlights the advantages of the LSTM-CNN GAN. Statistical metrics show that the HMM-generated data significantly diverges from the real data in terms of mean, standard deviation, skewness, and kurtosis. In contrast, the LSTM-CNN model closely matches the real data across these statistics, affirming its capacity to model the complexity of eye gaze dynamics effectively. These results position the spectrally regularized LSTM-CNN GAN as a robust tool for generating synthetic eye gaze velocity data with high fidelity. | [
{
"version": "v1",
"created": "Thu, 05 Dec 2024 14:23:40 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Bhandari",
"Shailendra",
""
],
[
"Lencastre",
"Pedro",
""
],
[
"Mathema",
"Rujeena",
""
],
[
"Szorkovszky",
"Alexander",
""
],
[
"Yazidi",
"Anis",
""
],
[
"Lind",
"Pedro",
""
]
] |
2412.04249 | Guillaume Lhost | Guillaume Lhost and Geoffrey Comp\`ere | Approach to the separatrix with eccentric orbits | 38 pages, 11 figures | null | null | null | gr-qc hep-th | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Eccentric binary compact mergers are prime targets of current and future gravitational wave observatories. In the small mass ratio expansion, post-adiabatic inspirals have been modeled up to the separatrix, where first-principle modeling currently ends. In this paper, we derive the analytic late time solution to the adiabatic inspiral in terms of self-force coefficients at the separatrix. We identify the role of the Lambert $W_{-1}$ function as a key mathematical ingredient in the approach to the separatrix. | [
{
"version": "v1",
"created": "Thu, 05 Dec 2024 15:30:00 GMT"
},
{
"version": "v2",
"created": "Wed, 04 Jun 2025 14:08:19 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Lhost",
"Guillaume",
""
],
[
"Compère",
"Geoffrey",
""
]
] |
2412.04300 | Ziwei Huang | Ziwei Huang, Wanggui He, Quanyu Long, Yandi Wang, Haoyuan Li, Zhelun Yu, Fangxun Shu, Long Chan, Hao Jiang, Fei Wu, Leilei Gan | T2I-FactualBench: Benchmarking the Factuality of Text-to-Image Models with Knowledge-Intensive Concepts | null | null | null | null | cs.CV cs.AI | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Evaluating the quality of synthesized images remains a significant challenge in the development of text-to-image (T2I) generation. Most existing studies in this area primarily focus on evaluating text-image alignment, image quality, and object composition capabilities, with comparatively fewer studies addressing the evaluation of the factuality of T2I models, particularly when the concepts involved are knowledge-intensive. To mitigate this gap, we present T2I-FactualBench in this work - the largest benchmark to date in terms of the number of concepts and prompts specifically designed to evaluate the factuality of knowledge-intensive concept generation. T2I-FactualBench consists of a three-tiered knowledge-intensive text-to-image generation framework, ranging from the basic memorization of individual knowledge concepts to the more complex composition of multiple knowledge concepts. We further introduce a multi-round visual question answering (VQA) based evaluation framework to assess the factuality of three-tiered knowledge-intensive text-to-image generation tasks. Experiments on T2I-FactualBench indicate that current state-of-the-art (SOTA) T2I models still leave significant room for improvement. | [
{
"version": "v1",
"created": "Thu, 05 Dec 2024 16:21:01 GMT"
},
{
"version": "v2",
"created": "Sat, 07 Dec 2024 17:25:28 GMT"
},
{
"version": "v3",
"created": "Wed, 04 Jun 2025 03:29:18 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Huang",
"Ziwei",
""
],
[
"He",
"Wanggui",
""
],
[
"Long",
"Quanyu",
""
],
[
"Wang",
"Yandi",
""
],
[
"Li",
"Haoyuan",
""
],
[
"Yu",
"Zhelun",
""
],
[
"Shu",
"Fangxun",
""
],
[
"Chan",
"Long",
""
],
[
"Jiang",
"Hao",
""
],
[
"Wu",
"Fei",
""
],
[
"Gan",
"Leilei",
""
]
] |
2412.04381 | Emiel Claasen | Emiel Claasen, Mehregan Doroudiani | Type II superstring amplitude at one-loop and transcendentality | 6 pages; Submission includes ancillary data file; v3: Sections rearranged and minor changes - version appeared in PRL | null | 10.1103/PhysRevLett.134.201601 | null | hep-th math.NT | http://creativecommons.org/licenses/by/4.0/ | We calculate the four-graviton scattering amplitude in Type II superstring theory at one loop up to seventh order in the low-energy expansion through the recently developed iterated integral formalism of Modular Graph Functions (MGFs). The machinery of the novel method allows us to propose a general form of the amplitude, which suggests that the expansion is expressible in terms of single-valued multiple zeta values and logarithmic derivatives of the Riemann zeta function at positive and negative odd integers. Furthermore, we comment on the transcendental behavior of the amplitude. | [
{
"version": "v1",
"created": "Thu, 05 Dec 2024 17:57:46 GMT"
},
{
"version": "v2",
"created": "Thu, 23 Jan 2025 13:21:37 GMT"
},
{
"version": "v3",
"created": "Wed, 04 Jun 2025 13:21:47 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Claasen",
"Emiel",
""
],
[
"Doroudiani",
"Mehregan",
""
]
] |
2412.05237 | JiaWei Guo | Jarvis Guo, Tuney Zheng, Yuelin Bai, Bo Li, Yubo Wang, King Zhu, Yizhi Li, Graham Neubig, Wenhu Chen, Xiang Yue | MAmmoTH-VL: Eliciting Multimodal Reasoning with Instruction Tuning at Scale | ACL 2025 Main | null | null | null | cs.CL cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Open-source multimodal large language models (MLLMs) have shown significant potential in a broad range of multimodal tasks. However, their reasoning capabilities remain constrained by existing instruction-tuning datasets, which were predominately repurposed from academic datasets such as VQA, AI2D, and ChartQA. These datasets target simplistic tasks, and only provide phrase-level answers without any intermediate rationales. To address these challenges, we introduce a scalable and cost-effective method to construct a large-scale multimodal instruction-tuning dataset with rich intermediate rationales designed to elicit CoT reasoning. Using only open models, we create a dataset containing 12M instruction-response pairs to cover diverse, reasoning-intensive tasks with detailed and faithful rationales. Experiments demonstrate that training MLLMs on this dataset significantly improves reasoning capabilities, achieving state-of-the-art performance on benchmarks such as MathVerse (+8.1%), MMMU-Pro (+7%), and MuirBench (+13.3%). Additionally, the model demonstrates notable improvements of up to 4% on non-reasoning-based benchmarks. Ablation studies further highlight the importance of key components, such as rewriting and self-filtering, in the dataset construction process. | [
{
"version": "v1",
"created": "Fri, 06 Dec 2024 18:14:24 GMT"
},
{
"version": "v2",
"created": "Wed, 04 Jun 2025 10:07:57 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Guo",
"Jarvis",
""
],
[
"Zheng",
"Tuney",
""
],
[
"Bai",
"Yuelin",
""
],
[
"Li",
"Bo",
""
],
[
"Wang",
"Yubo",
""
],
[
"Zhu",
"King",
""
],
[
"Li",
"Yizhi",
""
],
[
"Neubig",
"Graham",
""
],
[
"Chen",
"Wenhu",
""
],
[
"Yue",
"Xiang",
""
]
] |
2412.06141 | Peng Xia | Kangyu Zhu, Peng Xia, Yun Li, Hongtu Zhu, Sheng Wang, Huaxiu Yao | MMedPO: Aligning Medical Vision-Language Models with Clinical-Aware Multimodal Preference Optimization | ICML 2025 | null | null | null | cs.CV cs.AI cs.CL cs.LG | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | The advancement of Large Vision-Language Models (LVLMs) has propelled their application in the medical field. However, Medical LVLMs (Med-LVLMs) encounter factuality challenges due to modality misalignment, where the models prioritize textual knowledge over visual input, leading to hallucinations that contradict information in medical images. Previous attempts to enhance modality alignment in Med-LVLMs through preference optimization have inadequately mitigated clinical relevance in preference data, making these samples easily distinguishable and reducing alignment effectiveness. To address this challenge, we propose MMedPO, a novel multimodal medical preference optimization approach that considers the clinical relevance of preference samples to enhance Med-LVLM alignment. MMedPO curates multimodal preference data by introducing two types of dispreference: (1) plausible hallucinations injected through target Med-LVLMs or GPT-4o to produce medically inaccurate responses, and (2) lesion region neglect achieved through local lesion-noising, disrupting visual understanding of critical areas. We then calculate clinical relevance for each sample based on scores from multiple Med-LLMs and visual tools, and integrate these scores into the preference optimization process as weights, enabling effective alignment. Our experiments demonstrate that MMedPO significantly enhances factual accuracy in Med-LVLMs, achieving substantial improvements over existing preference optimization methods by averaging 14.2% and 51.7% across the Med-VQA and report generation tasks. Our code are available in https://github.com/aiming-lab/MMedPO. | [
{
"version": "v1",
"created": "Mon, 09 Dec 2024 01:50:39 GMT"
},
{
"version": "v2",
"created": "Mon, 19 May 2025 03:03:31 GMT"
},
{
"version": "v3",
"created": "Wed, 21 May 2025 08:39:51 GMT"
},
{
"version": "v4",
"created": "Wed, 04 Jun 2025 07:13:53 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Zhu",
"Kangyu",
""
],
[
"Xia",
"Peng",
""
],
[
"Li",
"Yun",
""
],
[
"Zhu",
"Hongtu",
""
],
[
"Wang",
"Sheng",
""
],
[
"Yao",
"Huaxiu",
""
]
] |
2412.06551 | Haoran Guan Mr | Haoran Guan and Yuwei Fan | Deterministic and randomized LU-Householder CholeskyQR | null | null | null | null | math.NA cs.NA | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | In this work, we develop LU-Householder CholeskyQR (LHC) for QR factorization of tall-skinny matrices. Similar to LU-CholeskyQR \cite{LUChol}, LHC does not require a sufficient condition of $\kappa_{2}(X)$ for the input tall-skinny matrix $X \in \mathbb{R}^{m\times n}$. This characteristic ensures the algorithm's reliability in the real-world applications, which is different from other CholeskyQR-type algorithms. To address the issue of numerical breakdown in Cholesky factorization of LU-CholeskyQR when the $L$-factor is ill-conditioned, LHC employs HouseholderQR to generate the upper-triangular factor alternatively. In order to accelerate LHC for tall-skinny matrices, we incorporate the latest sketching techniques to develop randomized versions of LHC, SLHC with single-sketching and SSLHC with multi-sketching. To ensure numerical stability, LHC2, SLHC3, and SSLHC3 are constructed with CholeskyQR(2) after the preconditioning steps. We provide rounding error analysis of these new algorithms. Numerical experiments demonstrate the better applicability of our new algorithms compared to LU-CholeskyQR2 while maintaining numerical stability. With the sketching technique, our randomized algorithms, SLHC3 and SSLHC3, show significant acceleration over LHC2 in our tests. Among them, SLHC3 with the Gaussian sketch can deal with more ill-conditioned $X$ with different $m$ and $n$ for $X$. SSLHC3 with multi-sketching is applicable when $\frac{m}{n}$ is large enough and is more efficient than SLHC3. Both of them are robust enough after numerous experiments. | [
{
"version": "v1",
"created": "Mon, 09 Dec 2024 14:59:55 GMT"
},
{
"version": "v2",
"created": "Tue, 11 Feb 2025 10:53:42 GMT"
},
{
"version": "v3",
"created": "Wed, 04 Jun 2025 14:43:52 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Guan",
"Haoran",
""
],
[
"Fan",
"Yuwei",
""
]
] |
2412.06679 | Eva M. Gonz\'alez-Ruiz | Eva M. Gonz\'alez-Ruiz, Johannes Bjerlin, Oliver August Dall'Alba Sandberg, and Anders S. S{\o}rensen | Two-photon correlations and HOM visibility from an imperfect single-photon source | 14 pages, 9 figures + appendix | null | 10.1103/PhysRevApplied.23.054063 | null | quant-ph | http://creativecommons.org/licenses/by/4.0/ | We study the single photon purity of a resonantly driven single-photon source in the realistic scenario where the excitation laser can leak into the detection path. We find that the duration of the excitation pulse strongly influences the quality of the single-photon source. We calculate the influence of this on the effective parameters describing the most relevant properties, including the two-photon component and Hong-Ou-Mandel (HOM) visibility. Furthermore, we analyze how these properties can be strongly affected by frequency filtering of the outgoing field. Our results highlight that the relation between the two-photon component of the emission and the HOM visibility is more complicated than typically assumed in the literature, and depends on the specific details of the source. | [
{
"version": "v1",
"created": "Mon, 09 Dec 2024 17:19:10 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"González-Ruiz",
"Eva M.",
""
],
[
"Bjerlin",
"Johannes",
""
],
[
"Sandberg",
"Oliver August Dall'Alba",
""
],
[
"Sørensen",
"Anders S.",
""
]
] |
2412.07169 | Tal Zeevi | Tal Zeevi, Ravid Shwartz-Ziv, Yann LeCun, Lawrence H. Staib, John A. Onofrey | Rate-In: Information-Driven Adaptive Dropout Rates for Improved Inference-Time Uncertainty Estimation | Accepted to the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) 2025. Code available at: https://github.com/code-supplement-25/rate-in | null | null | null | cs.LG cs.CV stat.ML | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Accurate uncertainty estimation is crucial for deploying neural networks in risk-sensitive applications such as medical diagnosis. Monte Carlo Dropout is a widely used technique for approximating predictive uncertainty by performing stochastic forward passes with dropout during inference. However, using static dropout rates across all layers and inputs can lead to suboptimal uncertainty estimates, as it fails to adapt to the varying characteristics of individual inputs and network layers. Existing approaches optimize dropout rates during training using labeled data, resulting in fixed inference-time parameters that cannot adjust to new data distributions, compromising uncertainty estimates in Monte Carlo simulations.
In this paper, we propose Rate-In, an algorithm that dynamically adjusts dropout rates during inference by quantifying the information loss induced by dropout in each layer's feature maps. By treating dropout as controlled noise injection and leveraging information-theoretic principles, Rate-In adapts dropout rates per layer and per input instance without requiring ground truth labels. By quantifying the functional information loss in feature maps, we adaptively tune dropout rates to maintain perceptual quality across diverse medical imaging tasks and architectural configurations. Our extensive empirical study on synthetic data and real-world medical imaging tasks demonstrates that Rate-In improves calibration and sharpens uncertainty estimates compared to fixed or heuristic dropout rates without compromising predictive performance. Rate-In offers a practical, unsupervised, inference-time approach to optimizing dropout for more reliable predictive uncertainty estimation in critical applications. | [
{
"version": "v1",
"created": "Tue, 10 Dec 2024 04:03:46 GMT"
},
{
"version": "v2",
"created": "Wed, 11 Dec 2024 12:50:45 GMT"
},
{
"version": "v3",
"created": "Tue, 14 Jan 2025 18:51:43 GMT"
},
{
"version": "v4",
"created": "Wed, 04 Jun 2025 03:01:01 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Zeevi",
"Tal",
""
],
[
"Shwartz-Ziv",
"Ravid",
""
],
[
"LeCun",
"Yann",
""
],
[
"Staib",
"Lawrence H.",
""
],
[
"Onofrey",
"John A.",
""
]
] |
2412.07326 | Asaf Shabtai | Yael Itzhakev, Amit Giloni, Yuval Elovici, Asaf Shabtai | Addressing Key Challenges of Adversarial Attacks and Defenses in the Tabular Domain: A Methodological Framework for Coherence and Consistency | null | null | null | null | cs.LG | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Machine learning models trained on tabular data are vulnerable to adversarial attacks, even in realistic scenarios where attackers only have access to the model's outputs. Since tabular data contains complex interdependencies among features, it presents a unique challenge for adversarial samples which must maintain coherence and respect these interdependencies to remain indistinguishable from benign data. Moreover, existing attack evaluation metrics-such as the success rate, perturbation magnitude, and query count-fail to account for this challenge. To address those gaps, we propose a technique for perturbing dependent features while preserving sample coherence. In addition, we introduce Class-Specific Anomaly Detection (CSAD), an effective novel anomaly detection approach, along with concrete metrics for assessing the quality of tabular adversarial attacks. CSAD evaluates adversarial samples relative to their predicted class distribution, rather than a broad benign distribution. It ensures that subtle adversarial perturbations, which may appear coherent in other classes, are correctly identified as anomalies. We integrate SHAP explainability techniques to detect inconsistencies in model decision-making, extending CSAD for SHAP-based anomaly detection. Our evaluation incorporates both anomaly detection rates with SHAP-based assessments to provide a more comprehensive measure of adversarial sample quality. We evaluate various attack strategies, examining black-box query-based and transferability-based gradient attacks across four target models. Experiments on benchmark tabular datasets reveal key differences in the attacker's risk and effort and attack quality, offering insights into the strengths, limitations, and trade-offs faced by attackers and defenders. Our findings lay the groundwork for future research on adversarial attacks and defense development in the tabular domain. | [
{
"version": "v1",
"created": "Tue, 10 Dec 2024 09:17:09 GMT"
},
{
"version": "v2",
"created": "Tue, 03 Jun 2025 21:06:19 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Itzhakev",
"Yael",
""
],
[
"Giloni",
"Amit",
""
],
[
"Elovici",
"Yuval",
""
],
[
"Shabtai",
"Asaf",
""
]
] |
2412.08337 | Faramarz Rahmani | Mehdi Sadeghi, Faramarz Rahmani | Thermodynamic Behavior of a 4D Nonminimal Maxwell-AdS Black Hole | 29 pages, 25 figures, f1 was corrected, | null | null | null | hep-th | http://creativecommons.org/licenses/by-nc-sa/4.0/ | In this paper, we derive a black hole solution within the Einstein Maxwell framework incorporating a nonminimal coupling between the Ricci tensor and the Maxwell field strength tensor, using a perturbative approach. We subsequently explore the thermodynamic phase transitions of the black hole in an extended phase space, analyzing both canonical and grand canonical ensembles. Our findings reveal that the system exhibits Van der Waals like behavior in both ensembles. Moreover, for sufficiently small values of electric charge and Maxwell potential, the thermodynamics is dominated by a Hawking Page phase transition. | [
{
"version": "v1",
"created": "Wed, 11 Dec 2024 12:24:46 GMT"
},
{
"version": "v2",
"created": "Thu, 12 Dec 2024 21:24:51 GMT"
},
{
"version": "v3",
"created": "Tue, 03 Jun 2025 18:26:21 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Sadeghi",
"Mehdi",
""
],
[
"Rahmani",
"Faramarz",
""
]
] |
2412.08674 | Xiangqiong Liu | Xiang-Qiong Liu | Thermomagnetic Effects of Quark Matter in the NJL Model: Application of Regularization Schemes | null | null | null | null | nucl-th | http://creativecommons.org/licenses/by-sa/4.0/ | In the SU(3) Nambu-Jona-Lasinio (NJL) model of a thermally magnetized medium, the regularization methods adopted for the thermodynamic potential and the mass gap equation are utilized to calculate the relevant thermodynamic quantities in thermomagnetic quark matter. When dealing with the thermodynamic quantities and the gap equation, three schemes can be chosen, namely magnetic field-independent regularization, soft cut-off regularization, and Pauli-Villars regularization. These three regularization schemes have different influences on the properties of thermomagnetic quark matter, and different choices of regularization schemes will also lead to different effects in the calculation of the properties of thermomagnetic quark matter. | [
{
"version": "v1",
"created": "Wed, 11 Dec 2024 08:20:56 GMT"
},
{
"version": "v2",
"created": "Wed, 04 Jun 2025 03:22:50 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Liu",
"Xiang-Qiong",
""
]
] |
2412.09398 | Wenlong Zhang | W. Zhang, T. Grismayer, and L. O. Silva | Anomalous pinch in electron-electron beam collision | 10 pages, 5 figures, 1 table, with appendices | null | null | null | physics.plasm-ph hep-ex physics.acc-ph physics.comp-ph | http://creativecommons.org/licenses/by/4.0/ | We show that an anomalous pinch can occur in ultrarelativistic electron-electron or positron-positron beam interaction, caused by the combined interplay of collective beam motion (disruption) and strong-field quantum electrodynamics (SF-QED). The locally created electron-positron pairs, from SF-QED effects, screen the self-fields of the beams and can invert the polarity of the Lorentz force resulting in a pinch of the beams. A theoretical model predicts the pinch condition and is confirmed by first-principles 3-dimensional particle-in-cell simulations. This anomalous pinch enhances density compression, increases the collision luminosity, and amplifies the local magnetic fields and the quantum parameter of the beam particles by several orders of magnitude. | [
{
"version": "v1",
"created": "Thu, 12 Dec 2024 16:03:40 GMT"
},
{
"version": "v2",
"created": "Wed, 04 Jun 2025 06:19:44 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Zhang",
"W.",
""
],
[
"Grismayer",
"T.",
""
],
[
"Silva",
"L. O.",
""
]
] |
2412.09586 | Fiona Ryan | Fiona Ryan, Ajay Bati, Sangmin Lee, Daniel Bolya, Judy Hoffman, James M. Rehg | Gaze-LLE: Gaze Target Estimation via Large-Scale Learned Encoders | CVPR 2025 Highlight | null | null | null | cs.CV | http://creativecommons.org/licenses/by/4.0/ | We address the problem of gaze target estimation, which aims to predict where a person is looking in a scene. Predicting a person's gaze target requires reasoning both about the person's appearance and the contents of the scene. Prior works have developed increasingly complex, hand-crafted pipelines for gaze target estimation that carefully fuse features from separate scene encoders, head encoders, and auxiliary models for signals like depth and pose. Motivated by the success of general-purpose feature extractors on a variety of visual tasks, we propose Gaze-LLE, a novel transformer framework that streamlines gaze target estimation by leveraging features from a frozen DINOv2 encoder. We extract a single feature representation for the scene, and apply a person-specific positional prompt to decode gaze with a lightweight module. We demonstrate state-of-the-art performance across several gaze benchmarks and provide extensive analysis to validate our design choices. Our code is available at: http://github.com/fkryan/gazelle . | [
{
"version": "v1",
"created": "Thu, 12 Dec 2024 18:55:30 GMT"
},
{
"version": "v2",
"created": "Wed, 04 Jun 2025 03:52:12 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Ryan",
"Fiona",
""
],
[
"Bati",
"Ajay",
""
],
[
"Lee",
"Sangmin",
""
],
[
"Bolya",
"Daniel",
""
],
[
"Hoffman",
"Judy",
""
],
[
"Rehg",
"James M.",
""
]
] |
2412.09840 | Martin Maas | Jianheng Ling, Pratik Worah, Yawen Wang, Yunchuan Kong, Anshul Kapoor, Chunlei Wang, Clifford Stein, Diwakar Gupta, Jason Behmer, Logan A. Bush, Prakash Ramanan, Rajesh Kumar, Thomas Chestna, Yajing Liu, Ying Liu, Ye Zhao, Kathryn S. McKinley, Meeyoung Park, Martin Maas | LAVA: Lifetime-Aware VM Allocation with Learned Distributions and Adaptation to Mispredictions | null | null | null | null | cs.DC | http://creativecommons.org/licenses/by/4.0/ | Scheduling virtual machines (VMs) on hosts in cloud data centers dictates efficiency and is an NP-hard problem with incomplete information. Prior work improved VM scheduling with predicted VM lifetimes. Our work further improves lifetime-aware scheduling using repredictions with lifetime distributions versus one-shot prediction. Our approach repredicts and adjusts VM and host lifetimes when incorrect predictions emerge. We also present novel approaches for defragmentation and regular system maintenance, which are essential to our data center reliability and optimizations, and are not explored in prior work. We show repredictions deliver a fundamental advance in effectiveness over one-shot prediction.
We call our novel combination of distribution-based lifetime predictions and scheduling algorithms Lifetime Aware VM Allocation (LAVA). LAVA reduces resource stranding and increases the number of empty hosts, which are critical for large VM scheduling, cloud system updates, and reducing dynamic energy consumption. Our approach runs in production within Google's hyperscale cloud data centers, where it improves efficiency by decreasing stranded compute and memory resources by ~3% and ~2% respectively. It increases empty hosts by 2.3-9.2 pp in production, reducing dynamic energy consumption, and increasing availability for large VMs and cloud system updates. We also show a reduction in VM migrations for host defragmentation and maintenance. In addition to our fleet-wide production deployment, we perform simulation studies to characterize the design space and show that our algorithm significantly outperforms the prior state of the art lifetime-based scheduling approach. | [
{
"version": "v1",
"created": "Fri, 13 Dec 2024 04:19:28 GMT"
},
{
"version": "v2",
"created": "Wed, 04 Jun 2025 01:30:15 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Ling",
"Jianheng",
""
],
[
"Worah",
"Pratik",
""
],
[
"Wang",
"Yawen",
""
],
[
"Kong",
"Yunchuan",
""
],
[
"Kapoor",
"Anshul",
""
],
[
"Wang",
"Chunlei",
""
],
[
"Stein",
"Clifford",
""
],
[
"Gupta",
"Diwakar",
""
],
[
"Behmer",
"Jason",
""
],
[
"Bush",
"Logan A.",
""
],
[
"Ramanan",
"Prakash",
""
],
[
"Kumar",
"Rajesh",
""
],
[
"Chestna",
"Thomas",
""
],
[
"Liu",
"Yajing",
""
],
[
"Liu",
"Ying",
""
],
[
"Zhao",
"Ye",
""
],
[
"McKinley",
"Kathryn S.",
""
],
[
"Park",
"Meeyoung",
""
],
[
"Maas",
"Martin",
""
]
] |
2412.10032 | Niclas Popp | Niclas Popp, Dan Zhang, Jan Hendrik Metzen, Matthias Hein, Lukas Schott | Single-Pass Object-Focused Data Selection | null | null | null | null | cs.CV | http://creativecommons.org/publicdomain/zero/1.0/ | While unlabeled image data is often plentiful, the costs of high-quality labels pose an important practical challenge: Which images should one select for labeling to use the annotation budget for a particular target task most effectively? To address this problem, we focus on single-pass data selection, which refers to the process of selecting all data to be annotated at once before training a downstream model. Prior methods for single-pass data selection rely on image-level representations and fail to reliably outperform random selection for object detection and segmentation. We propose Object-Focused Data Selection (OFDS) which leverages object-level features from foundation models and ensures semantic coverage of all target classes. In extensive experiments across tasks and target domains, OFDS consistently outperforms random selection and all baselines. The best results for constrained annotation budgets are obtained by combining human labels from OFDS with autolabels from foundation models. Moreover, using OFDS to select the initial labeled set for active learning yields consistent improvements | [
{
"version": "v1",
"created": "Fri, 13 Dec 2024 10:47:05 GMT"
},
{
"version": "v2",
"created": "Wed, 04 Jun 2025 17:06:22 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Popp",
"Niclas",
""
],
[
"Zhang",
"Dan",
""
],
[
"Metzen",
"Jan Hendrik",
""
],
[
"Hein",
"Matthias",
""
],
[
"Schott",
"Lukas",
""
]
] |
2412.10573 | Yiwen Gu | Yiwen Gu, Mahir Patel, Margrit Betke | ExeChecker: Where Did I Go Wrong? | null | null | 10.1007/978-3-031-92591-7_21 | null | cs.CV cs.HC cs.LG | http://creativecommons.org/licenses/by-nc-sa/4.0/ | In this paper, we present a contrastive learning based framework, ExeChecker, for the interpretation of rehabilitation exercises. Our work builds upon state-of-the-art advances in the area of human pose estimation, graph-attention neural networks, and transformer interpretablity. The downstream task is to assist rehabilitation by providing informative feedback to users while they are performing prescribed exercises. We utilize a contrastive learning strategy during training. Given a tuple of correctly and incorrectly executed exercises, our model is able to identify and highlight those joints that are involved in an incorrect movement and thus require the user's attention. We collected an in-house dataset, ExeCheck, with paired recordings of both correct and incorrect execution of exercises. In our experiments, we tested our method on this dataset as well as the UI-PRMD dataset and found ExeCheck outperformed the baseline method using pairwise sequence alignment in identifying joints of physical relevance in rehabilitation exercises. | [
{
"version": "v1",
"created": "Fri, 13 Dec 2024 21:34:54 GMT"
},
{
"version": "v2",
"created": "Wed, 04 Jun 2025 03:36:07 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Gu",
"Yiwen",
""
],
[
"Patel",
"Mahir",
""
],
[
"Betke",
"Margrit",
""
]
] |
2412.11050 | Yujin Wang Mr | Yujin Wang, Quanfeng Liu, Jiaqi Fan, Jinlong Hong, Hongqing Chu, Mengjian Tian, Bingzhao Gao, Hong Chen | RAC3: Retrieval-Augmented Corner Case Comprehension for Autonomous Driving with Vision-Language Models | 14 pages, 7 figures | null | null | null | cs.CV cs.AI | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Understanding and addressing corner cases is essential for ensuring the safety and reliability of autonomous driving systems. Vision-language models (VLMs) play a crucial role in enhancing scenario comprehension, yet they face significant challenges, such as hallucination and insufficient real-world grounding, which compromise their performance in critical driving scenarios. In this work, RAC3, a novel framework designed to enhance the performance of VLMs in corner case comprehension, is proposed. RAC3 integrates a frequency-spatial fusion (FSF) image encoder, a cross-modal alignment training method for embedding models with hard and semi-hard negative mining, and a fast querying and retrieval pipeline based on K-Means clustering and hierarchical navigable small world (HNSW) indexing. A multimodal chain-of-thought (CoT) prompting strategy to guide analogical reasoning and reduce hallucinations during inference is introduced. Moreover, an update mechanism is integrated into RAC3 to ensure continual learning within the framework. Extensive experiments on the CODA and nuScenes datasets demonstrate that RAC3 significantly improves corner case comprehension across multiple downstream tasks. Compared to prior state-of-the-art methods, RAC3 achieves the highest final score of 74.46 on the CODA-LM benchmark and shows consistent performance gains when integrated with end-to-end frameworks like DriveLM. These results demonstrate the effectiveness of retrieval-augmented strategies and cross-modal alignment for safer and more interpretable autonomous driving. | [
{
"version": "v1",
"created": "Sun, 15 Dec 2024 04:51:30 GMT"
},
{
"version": "v2",
"created": "Sun, 13 Apr 2025 05:30:02 GMT"
},
{
"version": "v3",
"created": "Wed, 04 Jun 2025 13:20:02 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Wang",
"Yujin",
""
],
[
"Liu",
"Quanfeng",
""
],
[
"Fan",
"Jiaqi",
""
],
[
"Hong",
"Jinlong",
""
],
[
"Chu",
"Hongqing",
""
],
[
"Tian",
"Mengjian",
""
],
[
"Gao",
"Bingzhao",
""
],
[
"Chen",
"Hong",
""
]
] |
2412.11097 | Hisanori Oshima | Hisanori Oshima, Ken Mochizuki, Ryusuke Hamazaki, Yohei Fuji | Topology and Spectrum in Measurement-Induced Phase Transitions | 8 + 9 pages, 3 + 4 figures, v4: published version | null | null | RIKEN-iTHEMS-Report-24 | quant-ph cond-mat.dis-nn cond-mat.stat-mech | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Competition among repetitive measurements of noncommuting observables and unitary dynamics can give rise to a wide variety of entanglement phases. Here, we propose a general framework based on Lyapunov analysis to characterize topological properties in monitored quantum systems through their spectrum and many-body topological invariants. We illustrate this framework by analyzing (1+1)-dimensional monitored circuits for Majorana fermions, which are known to exhibit topological and trivial area-law entangled phases as well as a critical phase with sub-volume-law entanglement. Through the Lyapunov analysis, we identify the presence (absence) of edge-localized zero modes inside the bulk gap in the topological (trivial) phase and a bulk gapless spectrum in the critical phase. Furthermore, by suitably exploiting the fermion parity with twisted measurement outcomes at the boundary, we construct a topological invariant that distinguishes the two area-law phases and dynamically characterizes the critical phase. Our framework thus provides a general route to extend the notion of bulk-edge correspondence to monitored quantum dynamics. | [
{
"version": "v1",
"created": "Sun, 15 Dec 2024 07:32:16 GMT"
},
{
"version": "v2",
"created": "Fri, 20 Dec 2024 15:21:18 GMT"
},
{
"version": "v3",
"created": "Mon, 23 Dec 2024 14:19:06 GMT"
},
{
"version": "v4",
"created": "Wed, 04 Jun 2025 08:38:13 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Oshima",
"Hisanori",
""
],
[
"Mochizuki",
"Ken",
""
],
[
"Hamazaki",
"Ryusuke",
""
],
[
"Fuji",
"Yohei",
""
]
] |
2412.11201 | Shu-Ao Liao | Shu-Ao Liao, Jin Zhang, and Li-Ping Yang | Phase diagram of Rydberg atoms in a two-leg rectangular ladder | 15 pages, 13 figures | null | 10.1103/PhysRevB.111.165154 | null | cond-mat.quant-gas hep-lat physics.atom-ph quant-ph | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Using the density matrix renormalization group algorithm, we map the ground-state phase diagram of a two-leg Rydberg ladder array with lattice spacings $a_x=2a_y$. We identify various density wave phases that spontaneously break the translational symmetry or the top-bottom reflection symmetry within the ladder. By increasing the laser detuning from zero, where the system is in a disordered phase that preserves all symmetries, we observe density wave orders with spontaneous breaking of the translational $\mathbb{Z}_p$ symmetries at intermediate detuning values, while the reflection symmetry is preserved. These orders exhibit nonzero bond orders with positive expectation values on every $p$th rung, thus labeled as $\mathbb{Z}_p^+$ phases. At larger detuning values, another spontaneous breaking of the reflection symmetry, which disrupted the bond orders on the rungs, occurs via an Ising phase transition. In these phases, either the top or the bottom site is occupied in a staggered way on every $p$th rung, breaking the translational $\mathbb{Z}_{2p}$ symmetry, thus labeled by $\mathbb{Z}_{2p}$ phases. We locate and characterize the 3-state Potts point and Ashkin-Teller point along the commensurate lines, as well as the direct chiral phase transitions between the disordered phase and the $\mathbb{Z}_p^+$ ($p = 3, 4$) phases. Critical exponents $\nu$ and $z$ are calculated for both conformal and chiral phase transition points. We finally identify two types of floating phases in the phase diagram: one characterized by a quasi-long-range incommensurate bond-order wave, and the other by a quasi-long-range incommensurate wave of density differences in the rungs. Our work motivates further applications of Rydberg atom arrays in quantum simulation. | [
{
"version": "v1",
"created": "Sun, 15 Dec 2024 14:28:59 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Liao",
"Shu-Ao",
""
],
[
"Zhang",
"Jin",
""
],
[
"Yang",
"Li-Ping",
""
]
] |
2412.11989 | ATLAS Collaboration | ATLAS Collaboration | Test of lepton flavour universality in $W$-boson decays into electrons and $\tau$-leptons using $pp$ collisions at $\sqrt{s}=13$ TeV with the ATLAS detector | 50 pages in total, author list starting page 33, 8 figures, 4 tables, published on JHEP. All figures including auxiliary figures are available at http://atlas.web.cern.ch/Atlas/GROUPS/PHYSICS/PAPERS/TOPQ-2023-10 | JHEP 05 (2025) 038 | 10.1007/JHEP05(2025)038 | CERN-EP-2024-315 | hep-ex | http://creativecommons.org/licenses/by/4.0/ | A measurement of the ratio of the branching fractions, $R_{\tau/e} = B(W \to \tau \nu)/ B(W \to e \nu)$, is performed using a sample of $W$ bosons originating from top-quark decays to final states containing $\tau$-leptons or electrons. This measurement uses $pp$ collisions at $\sqrt{s}=13$ TeV, collected by the ATLAS experiment at the Large Hadron Collider during Run 2, corresponding to an integrated luminosity of 140 fb$^{-1}$. The $W \to \tau \nu_\tau$ (with $\tau \to e \nu_e \nu_\tau$) and $W \to e \nu_e$ decays are distinguished using the differences in the impact parameter distributions and transverse momentum spectra of the electrons. The measured ratio of branching fractions $R_{\tau/e} = 0.975 \pm 0.012 \textrm{(stat.)} \pm 0.020 \textrm{(syst.)}$, is consistent with the Standard Model assumption of lepton flavour universality in $W$-boson decays. | [
{
"version": "v1",
"created": "Mon, 16 Dec 2024 17:14:14 GMT"
},
{
"version": "v2",
"created": "Wed, 04 Jun 2025 09:54:43 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"ATLAS Collaboration",
"",
""
]
] |
2412.12288 | Samuel Fedida | Samuel Fedida and Adrian Kent | Mixture equivalence principles and post-quantum theories of gravity | 17 pages, 6 figures | null | 10.1103/pttr-6kj7 | null | gr-qc quant-ph | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We examine the mixture equivalence principle (MEP), which states that proper and improper mixed states with the same density matrix are always experimentally indistinguishable, and a weaker version, which states that this is sometimes true in gravity theories. We point out that Moller-Rosenfeld semiclassical gravity violates the weak MEP and that nonlinear extensions of quantum mechanics violate the MEP. We further demonstrate that modifications of the Born rule in quantum theory also typically violate the MEP. We analyse such violations in the context of thermal baths, where proper and improper thermal states induce different physical situations. This has significant implications in the context of black hole physics. We argue that Moller-Rosenfeld semiclassical gravity is not the semiclassical limit of quantum gravity in the context of black hole spacetimes, even in the presence of $N\gg1$ matter fields. | [
{
"version": "v1",
"created": "Mon, 16 Dec 2024 19:02:22 GMT"
},
{
"version": "v2",
"created": "Mon, 02 Jun 2025 15:49:26 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Fedida",
"Samuel",
""
],
[
"Kent",
"Adrian",
""
]
] |
2412.12533 | Liuke Lyu | Liuke Lyu, Menghan Song, Ting-Tung Wang, Zi Yang Meng, William Witczak-Krempa | Multiparty Entanglement Microscopy of Quantum Ising models in 1d, 2d and 3d | 12+8 pages, 7+7 figures | Phys. Rev. B 111, 245108 (2025) | 10.1103/PhysRevB.111.245108 | null | cond-mat.str-el | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Entanglement microscopy reveals the true quantum correlations among the microscopic building blocks of many-body systems [Nat. Commun. 16, 96 (2025)]. Using this approach, we study the multipartite entanglement of the quantum Ising model in 1d, 2d, and 3d. We first obtain the full reduced density matrix (tomography) of subregions that have at most 4 sites via quantum Monte Carlo, exact diagonalization, and the exact solution in 1d. We then analyze both bipartite and genuine multipartite entanglement (GME) among the sites in the subregion. To do so, we use a variety of measures including the negativity, as well as a true measure of GME: the genuinely multipartite concurrence (or GME concurrence), and its computationally cheaper lower bound, $I_2$. We provide a complete proof that $I_2$ bounds the GME concurrence, and show how the symmetries of the state simplify its evaluation. For adjacent sites, we find 3- and 4-spin GME present across large portions of the phase diagram, reaching maximum near the quantum critical point. In 1d, we identify the singular scaling of the derivative $dI_2/dh$ approaching the critical point. We observe a sharp decrease of GME with increasing dimensionality, coherent with the monogamous nature of entanglement. Furthermore, we find that GME disappears for subregions consisting of non-adjacent sites in both 2d and 3d, offering a stark illustration of the short-ranged nature of entanglement in equilibrium quantum matter arXiv:2402.06677. Finally, we analyze the most collective form of entanglement by evaluating the GME concurrence among all spins in the lattice, which can be obtained from a simple observable: the single-site transverse magnetization. This global concurrence is larger in 1d compared to 2d/3d, but it is relatively less robust against perturbations such as local measurements. | [
{
"version": "v1",
"created": "Tue, 17 Dec 2024 04:45:19 GMT"
},
{
"version": "v2",
"created": "Wed, 04 Jun 2025 14:25:58 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Lyu",
"Liuke",
""
],
[
"Song",
"Menghan",
""
],
[
"Wang",
"Ting-Tung",
""
],
[
"Meng",
"Zi Yang",
""
],
[
"Witczak-Krempa",
"William",
""
]
] |
2412.13058 | Romain Br\'egier | Br\'egier Romain, Baradel Fabien, Lucas Thomas, Galaaoui Salma, Armando Matthieu, Weinzaepfel Philippe, Rogez Gr\'egory | CondiMen: Conditional Multi-Person Mesh Recovery | accepted to the RHOBIN workshop at CVPR 2025 | null | null | null | cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Multi-person human mesh recovery (HMR) consists in detecting all individuals in a given input image, and predicting the body shape, pose, and 3D location for each detected person. The dominant approaches to this task rely on neural networks trained to output a single prediction for each detected individual. In contrast, we propose CondiMen, a method that outputs a joint parametric distribution over likely poses, body shapes, intrinsics and distances to the camera, using a Bayesian network. This approach offers several advantages. First, a probability distribution can handle some inherent ambiguities of this task -- such as the uncertainty between a person's size and their distance to the camera, or simply the loss of information when projecting 3D data onto the 2D image plane. Second, the output distribution can be combined with additional information to produce better predictions, by using e.g. known camera or body shape parameters, or by exploiting multi-view observations. Third, one can efficiently extract the most likely predictions from the output distribution, making our proposed approach suitable for real-time applications. Empirically we find that our model i) achieves performance on par with or better than the state-of-the-art, ii) captures uncertainties and correlations inherent in pose estimation and iii) can exploit additional information at test time, such as multi-view consistency or body shape priors. CondiMen spices up the modeling of ambiguity, using just the right ingredients on hand. | [
{
"version": "v1",
"created": "Tue, 17 Dec 2024 16:22:56 GMT"
},
{
"version": "v2",
"created": "Wed, 04 Jun 2025 10:45:52 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Romain",
"Brégier",
""
],
[
"Fabien",
"Baradel",
""
],
[
"Thomas",
"Lucas",
""
],
[
"Salma",
"Galaaoui",
""
],
[
"Matthieu",
"Armando",
""
],
[
"Philippe",
"Weinzaepfel",
""
],
[
"Grégory",
"Rogez",
""
]
] |
2412.13191 | Alexander Taveira Blomenhofer | Alexander Taveira Blomenhofer and Monique Laurent | Moment-sos and spectral hierarchies for polynomial optimization on the sphere and quantum de Finetti theorems | 31 pages | null | null | null | math.OC | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We revisit the convergence analysis of two approximation hierarchies for polynomial optimization on the unit sphere. The first one is based on the moment-sos approach and gives semidefinite bounds for which Fang and Fawzi (2021) showed an analysis in $O(1/r^2)$ for the r-th level bound, using the polynomial kernel method. The second hierarchy was recently proposed by Lovitz and Johnston (2023) and gives spectral bounds for which they show a convergence rate in $O(1/r)$, using a quantum de Finetti theorem of Christandl et al. (2007) that applies to complex Hermitian matrices with a "double" symmetry. We investigate links between these approaches, in particular, via duality of moments and sums of squares.
Our main results include showing that the spectral bounds cannot have a convergence rate better than $O(1/r^2)$ and that they do not enjoy generic finite convergence. In addition, we propose alternative performance analyses that involve explicit constants depending on intrinsic parameters of the optimization problem. For this we develop a novel "banded" real de Finetti theorem that applies to real matrices with "double" symmetry. We also show how to use the polynomial kernel method to obtain a de Finetti type result in $O(1/r^2)$ for real maximally symmetric matrices, improving an earlier result in $O(1/r)$ of Doherty and Wehner (2012). | [
{
"version": "v1",
"created": "Tue, 17 Dec 2024 18:59:36 GMT"
},
{
"version": "v2",
"created": "Wed, 04 Jun 2025 14:39:43 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Blomenhofer",
"Alexander Taveira",
""
],
[
"Laurent",
"Monique",
""
]
] |
2412.14468 | Aditya Desai | Aditya Desai, Shuo Yang, Alejandro Cuadron, Matei Zaharia, Joseph E. Gonzalez, Ion Stoica | HashAttention: Semantic Sparsity for Faster Inference | Accepted at ICML'2025 | null | null | null | cs.LG cs.AI | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Leveraging long contexts is crucial for advanced AI systems, but attention computation poses a scalability challenge. While scaled dot-product attention (SDPA) exhibits token sparsity, i.e. only a few pivotal tokens significantly contribute to output, exploiting this sparsity remains challenging. Existing methods either suffer from quality degradation or require substantial additional resources. We show that identifying pivotal tokens is a Maximum Inner Product Search (MIPS) problem. However, existing MIPS solutions are not well-suited for SDPA, as they are not GPU-friendly and often underperform due to the separated query and key distributions. This paper introduces HashAttention, framing pivotal token identification as a recommendation problem. Given a query, HashAttention encodes keys and queries in Hamming space, capturing the required semantic similarity, using learned mapping functions. HashAttention efficiently identifies pivotal tokens for a given query using bitwise operations and computes attention using only these tokens, improving the overall attention efficiency. Trained on generic data, HashAttention reduces tokens used by up to $16\times$ with minimal quality loss, requiring only 32 bits of auxiliary memory per token. Sparsity can be further improved to $32\times$ through task-specific fine-tuning. On A100 GPU, at $32\times$ sparsity, incorporating HashAttention reduces attention latency by up to $4.3\times$ in GPT-FAST and $2.54\times$ in FlashDecode, and achieves up to $3.12\times$ higher throughput for GPT-FAST. | [
{
"version": "v1",
"created": "Thu, 19 Dec 2024 02:34:15 GMT"
},
{
"version": "v2",
"created": "Tue, 03 Jun 2025 21:55:57 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Desai",
"Aditya",
""
],
[
"Yang",
"Shuo",
""
],
[
"Cuadron",
"Alejandro",
""
],
[
"Zaharia",
"Matei",
""
],
[
"Gonzalez",
"Joseph E.",
""
],
[
"Stoica",
"Ion",
""
]
] |
2412.14706 | Jianrong Zhang | Jianrong Zhang and Hehe Fan and Yi Yang | EnergyMoGen: Compositional Human Motion Generation with Energy-Based Diffusion Model in Latent Space | Accepted to CVPR 2025. Project page: https://jiro-zhang.github.io/EnergyMoGen/ | null | null | null | cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Diffusion models, particularly latent diffusion models, have demonstrated remarkable success in text-driven human motion generation. However, it remains challenging for latent diffusion models to effectively compose multiple semantic concepts into a single, coherent motion sequence. To address this issue, we propose EnergyMoGen, which includes two spectrums of Energy-Based Models: (1) We interpret the diffusion model as a latent-aware energy-based model that generates motions by composing a set of diffusion models in latent space; (2) We introduce a semantic-aware energy model based on cross-attention, which enables semantic composition and adaptive gradient descent for text embeddings. To overcome the challenges of semantic inconsistency and motion distortion across these two spectrums, we introduce Synergistic Energy Fusion. This design allows the motion latent diffusion model to synthesize high-quality, complex motions by combining multiple energy terms corresponding to textual descriptions. Experiments show that our approach outperforms existing state-of-the-art models on various motion generation tasks, including text-to-motion generation, compositional motion generation, and multi-concept motion generation. Additionally, we demonstrate that our method can be used to extend motion datasets and improve the text-to-motion task. | [
{
"version": "v1",
"created": "Thu, 19 Dec 2024 10:19:43 GMT"
},
{
"version": "v2",
"created": "Wed, 04 Jun 2025 16:54:19 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Zhang",
"Jianrong",
""
],
[
"Fan",
"Hehe",
""
],
[
"Yang",
"Yi",
""
]
] |
2412.14874 | Soumya Adhikari | Soumya Adhikari, Aravind Aikot, Madhu Mishra and Bindusar Sahoo | Dilaton Weyl multiplets for $N = 3$ conformal supergravity in four dimensions | 25 pages, no figures; New references added by citing some new papers; Minor typos fixed; Final published version uploaded | JHEP 04 (2025) 062 | 10.1007/JHEP04(2025)062 | null | hep-th | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We construct a dilaton Weyl multiplet for $N = 3$ conformal supergravity in four dimensions. We couple an on-shell vector multiplet to the standard Weyl multiplet and use the field equations of the vector multiplet to replace some of the components of the auxiliary fields of the standard Weyl multiplet with the fields of the vector multiplet and some dual gauge fields. The R-symmetry of the multiplet is $SU(2) \times U(1) \times U(1)$. Furthermore, we gauge fix one of the two $U(1)$ symmetries and rewrite the result for the dilaton Weyl multiplet with $SU(2) \times U(1)$ R-symmetry. | [
{
"version": "v1",
"created": "Thu, 19 Dec 2024 14:13:32 GMT"
},
{
"version": "v2",
"created": "Tue, 07 Jan 2025 14:41:20 GMT"
},
{
"version": "v3",
"created": "Tue, 22 Apr 2025 13:41:10 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Adhikari",
"Soumya",
""
],
[
"Aikot",
"Aravind",
""
],
[
"Mishra",
"Madhu",
""
],
[
"Sahoo",
"Bindusar",
""
]
] |
2412.15168 | Sibin Yang | Sibin Yang, Gabe Schumm, and Anders W. Sandvik | Dynamic structure factor of a spin-1/2 Heisenberg chain with long-range interactions | 23 pages, 29 figures | Phys. Rev. B 111, 224404 (2025) | 10.1103/PhysRevB.111.224404 | null | cond-mat.str-el | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We study the dynamic structure factor $S(k,\omega)$ of the spin-1/2 chain with long-range, power-law decaying unfrustrated (sign
alternating) Heisenberg interactions $J_r \sim (-1)^{r-1} r^{-\alpha}$ by means of stochastic analytic continuation (SAC) of imaginary-time
correlations computed by quantum Monte Carlo calculations. We do so in both the long-range antiferromagnetic (AFM, for $\alpha \lesssim 2.23$) and
quasi-long-range-ordered (QLRO, for $\alpha \gtrsim 2.23$) ground-state phases, employing different SAC parametrizations of $S(k,\omega)$ to resolve sharp
edges characteristic of fractional quasi-particles and sharp peaks expected with conventional quasi-particles. In order to identify the most
statistically accurate parametrization, we apply a newly developed cross-validation method as a ``model selection'' tool. We confirm that the
spectral function contains a power-law divergent edge in the QLRO phase and a very sharp (likely $\delta$-function) magnon peak in the AFM phase.
From our SAC results, we extract the dispersion relation in the different regimes of the model, and in the AFM phase we extract the weight of
the magnon pole. In the limit where the model reduces to the conventional Heisenberg chain with nearest-neighbor interactions, our $S(k,\omega)$
agrees well with known Bethe ansatz results. In the AFM phase the low-energy dispersion relation is known to be nonlinear, $\omega_k \sim k^z$,
and we extract the corresponding dynamic exponent $z(\alpha)$, which in general is somewhat above the form obtained in linear spin-wave
theory. We also find a significant continuum above the magnon peak. This study serves as a benchmark for SAC/QMC studies of systems with a transition
from conventional to fractionalized quasi-particles. | [
{
"version": "v1",
"created": "Thu, 19 Dec 2024 18:43:01 GMT"
},
{
"version": "v2",
"created": "Wed, 04 Jun 2025 00:07:05 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Yang",
"Sibin",
""
],
[
"Schumm",
"Gabe",
""
],
[
"Sandvik",
"Anders W.",
""
]
] |
2412.15255 | Jonibek Mansurov | Jonibek Mansurov, Akhmed Sakip, Alham Fikri Aji | Data Laundering: Artificially Boosting Benchmark Results through Knowledge Distillation | 14 pages | null | null | null | cs.CL cs.AI | http://creativecommons.org/licenses/by-sa/4.0/ | In this paper, we show that knowledge distillation can be subverted to manipulate language model benchmark scores, revealing a critical vulnerability in current evaluation practices. We introduce "Data Laundering," a process that enables the covert transfer of benchmark-specific knowledge through seemingly legitimate intermediate training steps. Through extensive experiments with a 2-layer BERT student model, we show how this approach can achieve substantial improvements in benchmark accuracy (up to 75\% on GPQA) without developing genuine reasoning capabilities. Notably, this method can be exploited intentionally or even unintentionally, as researchers may inadvertently adopt this method and inflate scores without realising the implications. While our findings demonstrate the effectiveness of this technique, we present them as a cautionary tale highlighting the urgent need for more robust evaluation methods in AI. This work aims to contribute to the ongoing discussion about evaluation integrity in AI development and the need for benchmarks that more accurately reflect true model capabilities. The code is available at https://github.com/mbzuai-nlp/data_laundering. | [
{
"version": "v1",
"created": "Sun, 15 Dec 2024 19:38:48 GMT"
},
{
"version": "v2",
"created": "Wed, 04 Jun 2025 12:38:38 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Mansurov",
"Jonibek",
""
],
[
"Sakip",
"Akhmed",
""
],
[
"Aji",
"Alham Fikri",
""
]
] |
2412.15323 | Marcus St{\aa}lhammar Dr. | Marcus St{\aa}lhammar, and Lukas R{\o}dland | Abelian Spectral Topology of Multifold Exceptional Points | Updated version. 30 pages, 8 figures, 4 tables | null | null | null | cond-mat.mes-hall math-ph math.MP physics.optics quant-ph | http://creativecommons.org/licenses/by/4.0/ | The advent of non-Hermitian physics has enriched the plethora of topological phases to include phenomena without Hermitian counterparts. Despite being among the most well-studied uniquely non-Hermitian features, the topological properties of multifold exceptional points, $n$-fold spectral degeneracies (EP$n$s) at which also the corresponding eigenvectors coalesce, were only recently revealed in terms of topological resultant winding numbers and concomitant Abelian doubling theorems. Nevertheless, a more mathematically fundamental description of EP$n$s and their topological nature has remained an open question. To fill this void, in this article, we revisit the topological classification of EP$n$s in generic systems and systems with local symmetries, generalize it in terms of more mathematically tractable (local) similarity relations, and extend it to include all such similarities as well as non-local symmetries. Through the resultant vector, whose components are given in terms of the resultants between the corresponding characteristic polynomial and its derivatives, the topological nature of the resultant winding number is understood in several ways: in terms of i) the tenfold classification of (Hermitian) topological matter, ii) the framework of Mayer--Vietoris sequence, and iii) the classification of vector bundles. The classification scheme further predicts the existence of topological bulk Fermi arcs protected by a $\mathbb{Z}_2$-invariant, induced by non-local symmetries, dubbed $\mathbb{Z}_2$-protected Fermi arcs. Our work reveals the mathematical foundations on which the topological nature of EP$n$s resides, enriches the theoretical understanding of non-Hermitian spectral features, and will therefore find great use in modern experiments within both classical and quantum physics. | [
{
"version": "v1",
"created": "Thu, 19 Dec 2024 19:00:00 GMT"
},
{
"version": "v2",
"created": "Wed, 04 Jun 2025 12:30:50 GMT"
}
] | 2025-06-05T00:00:00 | [
[
"Stålhammar",
"Marcus",
""
],
[
"Rødland",
"Lukas",
""
]
] |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.