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 |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
2503.08585 | Shehreen Azad | Shehreen Azad, Vibhav Vineet, Yogesh Singh Rawat | HierarQ: Task-Aware Hierarchical Q-Former for Enhanced Video
Understanding | Accepted in CVPR 2025 | null | null | null | cs.CV | http://creativecommons.org/licenses/by/4.0/ | Despite advancements in multimodal large language models (MLLMs), current
approaches struggle in medium-to-long video understanding due to frame and
context length limitations. As a result, these models often depend on frame
sampling, which risks missing key information over time and lacks task-specific
relevance. To address these challenges, we introduce HierarQ, a task-aware
hierarchical Q-Former based framework that sequentially processes frames to
bypass the need for frame sampling, while avoiding LLM's context length
limitations. We introduce a lightweight two-stream language-guided feature
modulator to incorporate task awareness in video understanding, with the entity
stream capturing frame-level object information within a short context and the
scene stream identifying their broader interactions over longer period of time.
Each stream is supported by dedicated memory banks which enables our proposed
Hierachical Querying transformer (HierarQ) to effectively capture short and
long-term context. Extensive evaluations on 10 video benchmarks across video
understanding, question answering, and captioning tasks demonstrate HierarQ's
state-of-the-art performance across most datasets, proving its robustness and
efficiency for comprehensive video analysis.
| [
{
"version": "v1",
"created": "Tue, 11 Mar 2025 16:21:23 GMT"
},
{
"version": "v2",
"created": "Thu, 24 Apr 2025 17:42:16 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"Azad",
"Shehreen",
""
],
[
"Vineet",
"Vibhav",
""
],
[
"Rawat",
"Yogesh Singh",
""
]
] |
2503.09829 | Joohwan Seo | Joohwan Seo, Soochul Yoo, Junwoo Chang, Hyunseok An, Hyunwoo Ryu,
Soomi Lee, Arvind Kruthiventy, Jongeun Choi, and Roberto Horowitz | SE(3)-Equivariant Robot Learning and Control: A Tutorial Survey | Accepted to International Journcal of Control, Automation and Systems
(IJCAS) | null | null | null | cs.RO cs.LG cs.SY eess.SY | http://creativecommons.org/licenses/by-nc-nd/4.0/ | Recent advances in deep learning and Transformers have driven major
breakthroughs in robotics by employing techniques such as imitation learning,
reinforcement learning, and LLM-based multimodal perception and
decision-making. However, conventional deep learning and Transformer models
often struggle to process data with inherent symmetries and invariances,
typically relying on large datasets or extensive data augmentation. Equivariant
neural networks overcome these limitations by explicitly integrating symmetry
and invariance into their architectures, leading to improved efficiency and
generalization. This tutorial survey reviews a wide range of equivariant deep
learning and control methods for robotics, from classic to state-of-the-art,
with a focus on SE(3)-equivariant models that leverage the natural 3D
rotational and translational symmetries in visual robotic manipulation and
control design. Using unified mathematical notation, we begin by reviewing key
concepts from group theory, along with matrix Lie groups and Lie algebras. We
then introduce foundational group-equivariant neural network design and show
how the group-equivariance can be obtained through their structure. Next, we
discuss the applications of SE(3)-equivariant neural networks in robotics in
terms of imitation learning and reinforcement learning. The SE(3)-equivariant
control design is also reviewed from the perspective of geometric control.
Finally, we highlight the challenges and future directions of equivariant
methods in developing more robust, sample-efficient, and multi-modal real-world
robotic systems.
| [
{
"version": "v1",
"created": "Wed, 12 Mar 2025 20:47:40 GMT"
},
{
"version": "v2",
"created": "Tue, 18 Mar 2025 06:26:34 GMT"
},
{
"version": "v3",
"created": "Wed, 23 Apr 2025 20:00:44 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"Seo",
"Joohwan",
""
],
[
"Yoo",
"Soochul",
""
],
[
"Chang",
"Junwoo",
""
],
[
"An",
"Hyunseok",
""
],
[
"Ryu",
"Hyunwoo",
""
],
[
"Lee",
"Soomi",
""
],
[
"Kruthiventy",
"Arvind",
""
],
[
"Choi",
"Jongeun",
""
],
[
"Horowitz",
"Roberto",
""
]
] |
2503.10742 | Yudong Liu | Yudong Liu, Jingwei Sun, Yueqian Lin, Jingyang Zhang, Ming Yin, Qinsi
Wang, Jianyi Zhang, Hai Li, Yiran Chen | Keyframe-oriented Vision Token Pruning: Enhancing Efficiency of Large
Vision Language Models on Long-Form Video Processing | null | null | null | null | cs.LG cs.CL | http://creativecommons.org/licenses/by-nc-sa/4.0/ | Vision language models (VLMs) demonstrate strong capabilities in jointly
processing visual and textual data. However, they often incur substantial
computational overhead due to redundant visual information, particularly in
long-form video scenarios. Existing approaches predominantly focus on either
vision token pruning, which may overlook spatio-temporal dependencies, or
keyframe selection, which identifies informative frames but discards others,
thus disrupting contextual continuity. In this work, we propose KVTP
(Keyframe-oriented Vision Token Pruning), a novel framework that overcomes the
drawbacks of token pruning and keyframe selection. By adaptively assigning
pruning rates based on frame relevance to the query, KVTP effectively retains
essential contextual information while significantly reducing redundant
computation. To thoroughly evaluate the long-form video understanding
capacities of VLMs, we curated and reorganized subsets from VideoMME,
EgoSchema, and NextQA into a unified benchmark named SparseKV-QA that
highlights real-world scenarios with sparse but crucial events. Our experiments
with VLMs of various scales show that KVTP can reduce token usage by 80%
without compromising spatiotemporal and contextual consistency, significantly
cutting computation while maintaining the performance. These results
demonstrate our approach's effectiveness in efficient long-video processing,
facilitating more scalable VLM deployment.
| [
{
"version": "v1",
"created": "Thu, 13 Mar 2025 17:47:52 GMT"
},
{
"version": "v2",
"created": "Thu, 24 Apr 2025 14:53:57 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"Liu",
"Yudong",
""
],
[
"Sun",
"Jingwei",
""
],
[
"Lin",
"Yueqian",
""
],
[
"Zhang",
"Jingyang",
""
],
[
"Yin",
"Ming",
""
],
[
"Wang",
"Qinsi",
""
],
[
"Zhang",
"Jianyi",
""
],
[
"Li",
"Hai",
""
],
[
"Chen",
"Yiran",
""
]
] |
2503.13875 | Art Waeterschoot | Art Waeterschoot | Harmonic covers of skeleta | 44 pages, comments welcome | null | null | null | math.AG | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | The geometry of a toroidal scheme over a DVR is encoded in a $\mathbb{Z}$-PL
space known as the dual polyhedral complex. Any such dual complex is a
skeleton, i.e. a nonarchimedean analytic retract, and admits a combinatorial
divisor theory via specialisation. These structures on the dual complex
interact via a Poincar\'e-Lelong slope formula, which interprets
specialisations of divisors as Laplacians of PL functions. The main result
presented here shows that finite covers of toroidal schemes give harmonic
morphisms of dual complexes, i.e. morphisms that preserve the tropical Laplace
equation. A crucial ingredient is a balancing condition which is a variant of
the tropical multiplicity formula for dual complexes. We apply these results to
obtain a Riemann-Hurwitz formula for covers of skeleta in any dimension: the
Laplacian of the different function is the tropical relative canonical divisor.
| [
{
"version": "v1",
"created": "Tue, 18 Mar 2025 04:01:13 GMT"
},
{
"version": "v2",
"created": "Wed, 23 Apr 2025 22:26:07 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"Waeterschoot",
"Art",
""
]
] |
2503.13911 | Yacong Li | Ruobing Jiang, Yacong Li, Haobing Liu, Yanwei Yu | Incorporating Attributes and Multi-Scale Structures for Heterogeneous
Graph Contrastive Learning | null | Information Fusion 2025 | 10.1016/j.inffus.2025.103220 | null | cs.LG cs.SI | http://creativecommons.org/licenses/by/4.0/ | Heterogeneous graphs (HGs) are composed of multiple types of nodes and edges,
making it more effective in capturing the complex relational structures
inherent in the real world. However, in real-world scenarios, labeled data is
often difficult to obtain, which limits the applicability of semi-supervised
approaches. Self-supervised learning aims to enable models to automatically
learn useful features from data, effectively addressing the challenge of
limited labeling data. In this paper, we propose a novel contrastive learning
framework for heterogeneous graphs (ASHGCL), which incorporates three distinct
views, each focusing on node attributes, high-order and low-order structural
information, respectively, to effectively capture attribute information,
high-order structures, and low-order structures for node representation
learning. Furthermore, we introduce an attribute-enhanced positive sample
selection strategy that combines both structural information and attribute
information, effectively addressing the issue of sampling bias. Extensive
experiments on four real-world datasets show that ASHGCL outperforms
state-of-the-art unsupervised baselines and even surpasses some supervised
benchmarks.
| [
{
"version": "v1",
"created": "Tue, 18 Mar 2025 05:15:21 GMT"
},
{
"version": "v2",
"created": "Thu, 10 Apr 2025 09:07:02 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"Jiang",
"Ruobing",
""
],
[
"Li",
"Yacong",
""
],
[
"Liu",
"Haobing",
""
],
[
"Yu",
"Yanwei",
""
]
] |
2503.15840 | Junle Li | Junle Li, Meiqi Tian, and Bingzhuo Zhong | Automatic Generation of Safety-compliant Linear Temporal Logic via Large
Language Model: A Self-supervised Framework | null | null | null | null | cs.LO cs.FL | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Converting high-level tasks described by natural language into formal
specifications like Linear Temporal Logic (LTL) is a key step towards providing
formal safety guarantees over cyber-physical systems (CPS). While the
compliance of the formal specifications themselves against the safety
restrictions imposed on CPS is crucial for ensuring safety, most existing works
only focus on translation consistency between natural languages and formal
specifications. In this paper, we introduce AutoSafeLTL, a self-supervised
framework that utilizes large language models (LLMs) to automate the generation
of LTL specifications complying with a set of safety restrictions while
preserving their logical consistency and semantic accuracy. As a key insight,
our framework integrates Language Inclusion check with an automated
counterexample-guided modification mechanism to ensure the safety-compliance of
the resulting LTL specifications. In particular, we develop 1) an
LLM-as-an-Aligner, which performs atomic proposition matching between generated
LTL specifications and safety restrictions to enforce semantic alignment; and
2) an LLM-as-a-Critic, which automates LTL specification refinement by
interpreting counterexamples derived from Language Inclusion checks.
Experimental results demonstrate that our architecture effectively guarantees
safety-compliance for the generated LTL specifications, achieving a 0%
violation rate against imposed safety restrictions. This shows the potential of
our work in synergizing AI and formal verification techniques, enhancing
safety-aware specification generation and automatic verification for both AI
and critical CPS applications.
| [
{
"version": "v1",
"created": "Thu, 20 Mar 2025 04:40:29 GMT"
},
{
"version": "v2",
"created": "Thu, 24 Apr 2025 06:30:19 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"Li",
"Junle",
""
],
[
"Tian",
"Meiqi",
""
],
[
"Zhong",
"Bingzhuo",
""
]
] |
2503.16618 | Ioannis Apollon Paraskevas Mr | Joseph A. Dessi, Evgenios T.A. Kakariadis, and Ioannis Apollon
Paraskevas | A note on Arveson's hyperrigidity and non-degenerate C*-correspondences | 32 pages, references added, minor editorial changes | null | null | null | math.OA math.FA | http://creativecommons.org/licenses/by/4.0/ | We revisit the results of Kim, and of Katsoulis and Ramsey concerning
hyperrigidity for non-degenerate C*-correspondences. We show that the tensor
algebra is hyperrigid, if and only if Katsura's ideal acts non-degenerately, if
and only if Katsura's ideal acts non-degenerately under any representation.
This gives a positive answer to the question of Katsoulis and Ramsey, showing
that their necessary condition and their sufficient condition for hyperrigidity
of the tensor algebra are equivalent. Non-degeneracy of the left action of
Katsura's ideal was also shown by Kim to be equivalent to hyperrigidity for the
selfadjoint operator space associated with the C*-correspondence, and our
approach provides a simplified proof of this result as well. In the process we
revisit Arveson's criterion connecting maximality with the unique extension
property and hyperrigidity, in conjunction with the work of Salomon on
generating sets.
| [
{
"version": "v1",
"created": "Thu, 20 Mar 2025 18:16:32 GMT"
},
{
"version": "v2",
"created": "Thu, 24 Apr 2025 13:58:10 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"Dessi",
"Joseph A.",
""
],
[
"Kakariadis",
"Evgenios T. A.",
""
],
[
"Paraskevas",
"Ioannis Apollon",
""
]
] |
2503.17430 | Shun-Cai Zhao | Yi-Meng Huang, Zi-Ran Zhao, and Shun-Cai Zhao | Long-term excitation energy transfer predicted by a modified
convolutional neural networks in the FMO complexes | 11 pages, 10figures | null | null | null | physics.chem-ph cs.LG quant-ph | http://creativecommons.org/licenses/by-nc-nd/4.0/ | In machine learning (ML), the risk of recursive strategies overfitting
historical data has driven the development of convolutional neural networks
(CNNs) in simulating quantum dissipative dynamics. In this work, we propose an
efficient CNNs scheme incorporating novel redundant time-functions to predict
100 picosecond (ps) excitation energy transfer (EET) in Fenna-Matthews-Olson
(FMO) complexes, in which the original time $t$ is normalized by mapping it to
the [0, 1] range, allowing different functions focus on distinct time
intervals, thereby effectively capturing the multi-timescale characteristics of
EET dynamics. This method simplifies optimization and enhances learning
efficiency, and demonstrate the accuracy, robustness, and efficiency of our
approach in predicting quantum dissipative dynamics.
| [
{
"version": "v1",
"created": "Fri, 21 Mar 2025 12:40:39 GMT"
},
{
"version": "v2",
"created": "Tue, 25 Mar 2025 01:51:14 GMT"
},
{
"version": "v3",
"created": "Thu, 24 Apr 2025 04:26:48 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"Huang",
"Yi-Meng",
""
],
[
"Zhao",
"Zi-Ran",
""
],
[
"Zhao",
"Shun-Cai",
""
]
] |
2503.17653 | Manoj Mandal | Manoj Mandal, Sabyasachi Pal, G. K. Jaisawal, Anne Lohfink, Sachindra
Naik, Jaiverdhan Chauhan | Probing thermonuclear bursts and X-ray reflection features in Aql X-1
during 2024 outburst | Accepted in Journal of High Energy Astrophysics (JHEAP) | Journal of High Energy Astrophysics, 47 (2025), 100387 | 10.1016/j.jheap.2025.100387 | null | astro-ph.HE | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We report the broadband timing and spectral properties of the neutron star
low-mass X-ray binary Aql X-1 during the 2024 outburst with NICER, NuSTAR, and
Swift observatories. We detected six thermonuclear X-ray bursts during the
NICER and NuSTAR observations, with the observed X-ray burst profiles
exhibiting a strong energy dependence. The time-resolved burst spectra indicate
the presence of soft excess during the burst, which can be modeled by using a
variable persistent emission method ($f_a$ method), or the relxillNS reflection
model. We found that the reflection model can contribute $\sim$20% of total
emission as observed during the NICER burst. The reflection and blackbody
component fluxes are strongly correlated as observed during a burst. The excess
emission is possible due to the enhanced mass accretion rate to the neutron
star due to the Poynting-Rodertson drag and a fraction of burst emission may be
reflected from the disk. The bursts did not show photospheric radius expansion
during the peak. Moreover, we examined the burst-free accretion emission in the
broadband range with NuSTAR, NICER, and Swift at two epochs of the outburst.
The persistent emission showed X-ray reflection feature, which can be well
modeled with the relativistic reflection model relxillCp. The inner disk radius
(R$_{in}$) is found to be nearly 22 and 10 times $\rm R_{g}$ for two
observations, respectively. Assuming that the inner disk is truncated at the
magnetospheric radius, the magnetic field strength at the poles of the neutron
star is estimated to be $(0.6-1.9) \times 10^9$ G.
| [
{
"version": "v1",
"created": "Sat, 22 Mar 2025 05:06:14 GMT"
},
{
"version": "v2",
"created": "Thu, 24 Apr 2025 15:59:51 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"Mandal",
"Manoj",
""
],
[
"Pal",
"Sabyasachi",
""
],
[
"Jaisawal",
"G. K.",
""
],
[
"Lohfink",
"Anne",
""
],
[
"Naik",
"Sachindra",
""
],
[
"Chauhan",
"Jaiverdhan",
""
]
] |
2503.17766 | Jin-Jun Geng | Jin-Jun Geng, Ding-Fang Hu, Hao-Xuan Gao, Yi-Fang Liang, Yan-Long Hua,
Guo-Rui Zhang, Tian-Rui Sun, Bing Li, Yuan-Qi Liu, Fan Xu, Chen Deng,
Chen-Ran Hu, Ming Xu, Yong-Feng Huang, Miao-Miao Zhang, Min Fang, Jing-Zhi
Yan, Tao An, Xue-Feng Wu | GRB Timing: Decoding the Hidden Slow Jets in GRB 060729 | 15 pages, 6 figures, 2 tables, ApJL accepted | null | null | null | astro-ph.HE | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Gamma-ray bursts (GRBs) are luminous stellar explosions characterized by the
ejection of relativistic jets. This work proposes a novel paradigm to study
these GRB jets. By analyzing the timing information of prompt pulses and X-ray
flares, in conjunction with the multi-wavelength afterglow observations, we
identify three distinct jets in the extraordinary GRB 060729, with initial bulk
Lorentz factors ranging from approximately 20 to 80, smaller than typical
values of $> 100$. These three jets undergo two successive collisions,
producing the observed pair of X-ray flares. Following these interactions, the
system evolves into a fast, narrow jet and a slower, hollow jet that continues
to propagate in the circumburst medium, evidenced by the notable twin bumps
observed in the X-ray and optical afterglow of GRB 060729. Our findings
demonstrate that the timing of the early emission enables us to measure the
velocities of the GRB jets. The proposed paradigm enhances our understanding of
jet dynamics and shock interactions and serves as a powerful tool for probing
the physics of the central engine with the expanded sample in the current
golden era of GRB research.
| [
{
"version": "v1",
"created": "Sat, 22 Mar 2025 13:14:41 GMT"
},
{
"version": "v2",
"created": "Thu, 24 Apr 2025 02:04:08 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"Geng",
"Jin-Jun",
""
],
[
"Hu",
"Ding-Fang",
""
],
[
"Gao",
"Hao-Xuan",
""
],
[
"Liang",
"Yi-Fang",
""
],
[
"Hua",
"Yan-Long",
""
],
[
"Zhang",
"Guo-Rui",
""
],
[
"Sun",
"Tian-Rui",
""
],
[
"Li",
"Bing",
""
],
[
"Liu",
"Yuan-Qi",
""
],
[
"Xu",
"Fan",
""
],
[
"Deng",
"Chen",
""
],
[
"Hu",
"Chen-Ran",
""
],
[
"Xu",
"Ming",
""
],
[
"Huang",
"Yong-Feng",
""
],
[
"Zhang",
"Miao-Miao",
""
],
[
"Fang",
"Min",
""
],
[
"Yan",
"Jing-Zhi",
""
],
[
"An",
"Tao",
""
],
[
"Wu",
"Xue-Feng",
""
]
] |
2503.19089 | Po-Hsuan Lin | Po-Hsuan Lin, Yen Ling Tan | Cursed Job Market Signaling | 24 pages, 4 figures | null | null | null | econ.GN q-fin.EC | http://creativecommons.org/licenses/by-nc-sa/4.0/ | We study how cursedness, the tendency to neglect how other people's
strategies depend on their private information, affects information
transmission in Spence's job market signaling game. We characterize the Cursed
Sequential Equilibrium and show that as players become more cursed, the worker
obtains less education -- a costly signal that does not enhance productivity --
suggesting that cursedness improves the efficiency of information transmission.
However, this efficiency improvement depends on the richness of the message
space. Revisiting the job market signaling experiment by K\"ubler, M\"uller,
and Normann (2008), we find supportive evidence for our theory.
| [
{
"version": "v1",
"created": "Mon, 24 Mar 2025 19:21:29 GMT"
},
{
"version": "v2",
"created": "Mon, 7 Apr 2025 20:53:25 GMT"
},
{
"version": "v3",
"created": "Wed, 23 Apr 2025 20:11:40 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"Lin",
"Po-Hsuan",
""
],
[
"Tan",
"Yen Ling",
""
]
] |
2503.19175 | K. Singh Vikrant | K. S. Vikrant, Prosanto Biswas, and S. O. Reza Moheimani | A three-axis Nanopositioner based on Near-Field Acoustic Levitation and
Electromagnetic Actuation | null | null | null | null | physics.app-ph cs.SY eess.SY | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Near-field acoustic levitation (NFAL) enables nanometer-scale positioning
resolution and bandwidth exceeding several hundred hertz specifically along the
vertical (Z) direction, owing to its high acoustic stiffness and squeeze film
damping. However, its application to horizontal (XY) positioning is limited by
significantly lower acoustic stiffness and insufficient damping in horizontal
directions, resulting in reduced resolution and bandwidth. Moreover, NFAL-based
positioning systems typically lack multi-axis actuation capabilities due to
challenges in generating multi-directional acoustic forces. This work presents
a hybrid positioning approach that overcomes the mentioned limitations by
integrating NFAL with electromagnetic actuation. A planar magnetic platform is
acoustically levitated, while a coplanar current-carrying coil provides
horizontal trapping stiffness more than three orders of magnitude higher than
that achievable with acoustic forces alone. Additionally, the coil generates
three-dimensional electromagnetic forces, enabling multi-axis positioning
capability. Eddy currents induced in a thin copper sheet integrated with the
coil enhance horizontal damping by 52 times. We experimentally demonstrate
precise 3-axis linear motion with a root mean square (RMS) positioning
resolution better than 20 nm along all axes. The system achieves an in-plane
motion range of 1.42 mm with a bandwidth of 16 Hz and a Z-axis motion range of
40 micrometers with a positioning bandwidth of 171 Hz.
| [
{
"version": "v1",
"created": "Mon, 24 Mar 2025 21:54:53 GMT"
},
{
"version": "v2",
"created": "Wed, 23 Apr 2025 17:29:49 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"Vikrant",
"K. S.",
""
],
[
"Biswas",
"Prosanto",
""
],
[
"Moheimani",
"S. O. Reza",
""
]
] |
2503.19268 | Ephraim Linder | Ephraim Linder, Sofya Raskhodnikova, Adam Smith, Thomas Steinke | Privately Evaluating Untrusted Black-Box Functions | null | null | null | null | cs.DS | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We provide tools for sharing sensitive data when the data curator does not
know in advance what questions an (untrusted) analyst might ask about the data.
The analyst can specify a program that they want the curator to run on the
dataset. We model the program as a black-box function $f$. We study
differentially private algorithms, called privacy wrappers, that, given
black-box access to a real-valued function $f$ and a sensitive dataset $x$,
output an accurate approximation to $f(x)$. The dataset $x$ is modeled as a
finite subset of a possibly infinite set $U$, in which each entry represents
data of one individual. A privacy wrapper calls $f$ on the dataset $x$ and on
some subsets of $x$ and returns either an approximation to $f(x)$ or a
nonresponse symbol $\perp$. The wrapper may also use additional information
(that is, parameters) provided by the analyst, but differential privacy is
required for all values of these parameters. Correct setting of these
parameters will ensure better accuracy of the wrapper. The bottleneck in the
running time of our wrappers is the number of calls to $f$, which we refer to
as queries. Our goal is to design wrappers with high accuracy and low query
complexity. We introduce a novel setting, the automated sensitivity detection
setting, where the analyst supplies the black-box function $f$ and the intended
(finite) range of $f$. In the previously considered setting, the claimed
sensitivity bound setting, the analyst supplies additional parameters that
describe the sensitivity of $f$. We design privacy wrappers for both settings
and show that our wrappers are nearly optimal in terms of accuracy, locality
(i.e., the depth of the local neighborhood of the dataset $x$ they explore),
and query complexity. In the claimed sensitivity bound setting, we provide the
first accuracy guarantees that have no dependence on the size of the universe
$U$.
| [
{
"version": "v1",
"created": "Tue, 25 Mar 2025 02:04:13 GMT"
},
{
"version": "v2",
"created": "Thu, 24 Apr 2025 14:01:48 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"Linder",
"Ephraim",
""
],
[
"Raskhodnikova",
"Sofya",
""
],
[
"Smith",
"Adam",
""
],
[
"Steinke",
"Thomas",
""
]
] |
2503.20322 | Hao Ai | Hao Ai, Kunyi Wang, Zezhou Wang, Hao Lu, Jin Tian, Yaxin Luo, Peng
Xing, Jen-Yuan Huang, Huaxia Li, Gen luo | Dynamic Pyramid Network for Efficient Multimodal Large Language Model | null | null | null | null | cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Multimodal large language models (MLLMs) have demonstrated impressive
performance in various vision-language (VL) tasks, but their expensive
computations still limit the real-world application. To address this issue,
recent efforts aim to compress the visual features to save the computational
costs of MLLMs. However, direct visual compression methods, e.g. efficient
projectors, inevitably destroy the visual semantics in MLLM, especially in
difficult samples. To overcome this shortcoming, we propose a novel dynamic
pyramid network (DPN) for efficient MLLMs. Specifically, DPN formulates MLLM as
a hierarchical structure where visual features are gradually compressed with
increasing depth. In this case, even with a high compression ratio,
fine-grained visual information can still be perceived in shallow layers. To
maximize the benefit of DPN, we further propose an innovative Dynamic Pooling
Experts (DPE) that can dynamically choose the optimal visual compression rate
according to input features. With this design, harder samples will be assigned
larger computations, thus preserving the model performance. To validate our
approach, we conduct extensive experiments on two popular MLLMs and ten
benchmarks. Experimental results show that DPN can save up to 56% average FLOPs
on LLaVA while further achieving +0.74% performance gains. Besides, the
generalization ability of DPN is also validated on the existing high-resolution
MLLM called LLaVA-HR. The source code will be released at
https://github.com/aihao2000/DPN-LLaVA.
| [
{
"version": "v1",
"created": "Wed, 26 Mar 2025 08:44:11 GMT"
},
{
"version": "v2",
"created": "Thu, 24 Apr 2025 08:32:38 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"Ai",
"Hao",
""
],
[
"Wang",
"Kunyi",
""
],
[
"Wang",
"Zezhou",
""
],
[
"Lu",
"Hao",
""
],
[
"Tian",
"Jin",
""
],
[
"Luo",
"Yaxin",
""
],
[
"Xing",
"Peng",
""
],
[
"Huang",
"Jen-Yuan",
""
],
[
"Li",
"Huaxia",
""
],
[
"luo",
"Gen",
""
]
] |
2503.20457 | Christoph Widder | Christoph Widder, Johannes Zimmer, Tanja Schilling | On the generalized Langevin equation and the Mori projection operator
technique | null | null | null | null | math-ph cond-mat.stat-mech math.MP | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | In statistical physics, the Mori-Zwanzig projection operator formalism (also
called Nakajima-Zwanzig projection operator formalism) is used to derive a
linear integro-differential equation for observables in Hilbert space, the
generalized Langevin equation (GLE). This technique relies on the splitting of
the dynamics into a projected and an orthogonal part. We prove that the GLE
together with the second fluctuation dissipation theorem (2FDT) uniquely define
the fluctuating forces as well as the memory kernel. The GLE and 2FDT are an
immediate consequence of the existence and uniqueness of solutions of linear
Volterra equations. They neither rely on the Dyson identity nor on the concept
of orthogonal dynamics. This holds true for autonomous as well as
non-autonomous systems. Further results are obtained for the Mori projection
for autonomous systems, for which the fluctuating forces are orthogonal to the
observable of interest. In particular, we prove that the orthogonal dynamics is
a strongly continuous semigroup generated by $\overline{\mathcal{QL}}Q$, where
$\mathcal{L}$ is the generator of the time evolution operator, and
$\mathcal{P}=1-\mathcal{Q}$ is the Mori projection operator. As a consequence,
the corresponding orbit maps (e.g. the fluctuating forces) are the unique mild
solutions of the associated abstract Cauchy problem. Furthermore, we show that
the orthogonal dynamics is a unitary group, if $\mathcal{L}$ is skew-adjoint.
In this case, the fluctuating forces are stationary. In addition, we present a
proof of the GLE by means of semigroup theory, and we retrieve the commonly
used definitions for the fluctuating forces, memory kernel, and orthogonal
dynamics. Our results apply to general autonomous dynamical systems, whose time
evolution is given by a strongly continuous semigroup. This includes large
classes of systems in classical statistical mechanics.
| [
{
"version": "v1",
"created": "Wed, 26 Mar 2025 11:38:22 GMT"
},
{
"version": "v2",
"created": "Mon, 14 Apr 2025 15:45:06 GMT"
},
{
"version": "v3",
"created": "Thu, 24 Apr 2025 14:36:51 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"Widder",
"Christoph",
""
],
[
"Zimmer",
"Johannes",
""
],
[
"Schilling",
"Tanja",
""
]
] |
2503.20872 | Daniele De Bernardis | Daniele De Bernardis, Hugo Levy-Falk, Elena Fanella, Rocco Duquennoy,
Valerio Digiorgio, Giacomo Scalari, Maja Colautti, and Costanza Toninelli | Hybrid interfaces at the single quantum level in fluorescent molecules | null | null | null | null | quant-ph cond-mat.mes-hall | http://creativecommons.org/licenses/by/4.0/ | We theoretically investigate a single fluorescent molecule as a hybrid
quantum optical device, in which multiple external laser sources exert control
of the vibronic states. In the high-saturation regime, a coherent interaction
is established between the vibrational and electronic degrees of freedom, and
molecules can simulate several cavity QED models, whereby a specific
vibrational mode plays the role of the cavity mode. Focusing on the specific
example where the system is turned into an analogue simulator of the quantum
Rabi model, the steady state exhibits vibrational bi-modality resulting in a
statistical mixture of highly non-classical vibronic cat states. Applying our
paradigm to molecules with prominent spatial asymmetry and combining an optical
excitation with a THz(IR) driving, the system can be turned into a single
photon transducer. Two possible implementations are discussed based on the
coupling to a subwavelength THz patch antenna or a resonant metamaterial. In a
nutshell, this work assesses the role of molecules as an optomechanical quantum
toolbox for creating hybrid entangled states of electrons, photons, and
vibrations, hence enabling frequency conversion over very different energy
scales.
| [
{
"version": "v1",
"created": "Wed, 26 Mar 2025 18:00:03 GMT"
},
{
"version": "v2",
"created": "Tue, 1 Apr 2025 13:56:04 GMT"
},
{
"version": "v3",
"created": "Wed, 23 Apr 2025 18:00:04 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"De Bernardis",
"Daniele",
""
],
[
"Levy-Falk",
"Hugo",
""
],
[
"Fanella",
"Elena",
""
],
[
"Duquennoy",
"Rocco",
""
],
[
"Digiorgio",
"Valerio",
""
],
[
"Scalari",
"Giacomo",
""
],
[
"Colautti",
"Maja",
""
],
[
"Toninelli",
"Costanza",
""
]
] |
2503.21073 | Andrew Lee | Andrew Lee, Melanie Weber, Fernanda Vi\'egas, Martin Wattenberg | Shared Global and Local Geometry of Language Model Embeddings | null | null | null | null | cs.CL cs.LG | http://creativecommons.org/licenses/by/4.0/ | Researchers have recently suggested that models share common representations.
In our work, we find that token embeddings of language models exhibit common
geometric structure. First, we find ``global'' similarities: token embeddings
often share similar relative orientations. Next, we characterize local geometry
in two ways: (1) by using Locally Linear Embeddings, and (2) by defining a
simple measure for the intrinsic dimension of each token embedding. Our
intrinsic dimension demonstrates that token embeddings lie on a lower
dimensional manifold. We qualitatively show that tokens with lower intrinsic
dimensions often have semantically coherent clusters, while those with higher
intrinsic dimensions do not. Both characterizations allow us to find
similarities in the local geometry of token embeddings. Perhaps most
surprisingly, we find that alignment in token embeddings persists through the
hidden states of language models, allowing us to develop an application for
interpretability. Namely, we introduce Emb2Emb, a simple method to transfer
steering vectors from one language model to another, despite the two models
having different dimensions.
| [
{
"version": "v1",
"created": "Thu, 27 Mar 2025 01:17:06 GMT"
},
{
"version": "v2",
"created": "Thu, 24 Apr 2025 02:54:29 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"Lee",
"Andrew",
""
],
[
"Weber",
"Melanie",
""
],
[
"Viégas",
"Fernanda",
""
],
[
"Wattenberg",
"Martin",
""
]
] |
2503.21184 | Muthu Priyal Venkatachalam | Jagdev Singh, R. Ramesh, B.Raghavendra Prasad, V. Muthu Priyal, K.
Sasikumar Raja, S.N. Venkata, P.U.Kamath, V. Natarajan, S.Pawankumar, V.U.
Sanal Krishnan, P.Savarimuthu, Shalabh Mishra, Varun Kumar, Chavali Sumana,
S. Bhavana Hegde, D. Utkarsha, Amit Kumar, S. Nagabhushana, S.Kathiravan, P.
Vemareddy, C.Kathiravan, K. Nagaraju, Belur Ravindra, Wageesh Mishra | Visible Emission Line Coronagraph (VELC) onboard ADITYA-L1 | To appear in solar physics | null | null | null | astro-ph.SR astro-ph.IM | http://creativecommons.org/publicdomain/zero/1.0/ | Aditya-L1, India's first dedicated mission to study the Sun and its
atmosphere from the Sun-Earth Lagrangian L1 location was successfully launched
on 2023 September 2. It carries seven payloads. The Visible Emission Line
Coronagraph (VELC) is a major payload on Aditya-L1. VELC is designed to carry
out imaging and spectroscopic observations (the latter in three emission lines
of the corona), simultaneously. Images of the solar corona in the continuum at
5000 A, with a field of view (FoV) from 1.05 Ro to 3 Ro can be obtained at
variable intervals depending on the data volume that can be downloaded.
Spectroscopic observations of the solar corona in three emission lines, namely
5303 A Fexiv, 7892 A Fexi, and 10747 A Fexiii are possible simultaneously, with
different exposure times and cadence. Four slits, each of width 50 um,
separated by 3.75 mm help to simultaneously obtain spectra at four positions in
the solar corona at all the aforementioned lines. A Linear Scan Mechanism (LSM)
makes it possible to scan the solar corona up to +/-1.5 Ro. The instrument has
the facility to carry out spectropolarimetric observations at 10747 A also in
the FoV range 1.05 - 1.5 Ro. Various components of the instrument were tested
interferometrically on the optical bench before installation. The individual
components were aligned and performance of the payload was checked in the
laboratory using a laser source and tungsten lamp. Wavelength calibration of
the instrument was verified using Sun as a light source. All the detectors were
calibrated for different parameters such as dark current and its variation with
exposure time. Here, we discuss the various features of the VELC, alignment,
calibration, performance, possible observations, initial data analysis and
results of initial tests conducted in-orbit.
| [
{
"version": "v1",
"created": "Thu, 27 Mar 2025 06:06:34 GMT"
},
{
"version": "v2",
"created": "Tue, 1 Apr 2025 04:31:46 GMT"
},
{
"version": "v3",
"created": "Thu, 24 Apr 2025 12:18:16 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"Singh",
"Jagdev",
""
],
[
"Ramesh",
"R.",
""
],
[
"Prasad",
"B. Raghavendra",
""
],
[
"Priyal",
"V. Muthu",
""
],
[
"Raja",
"K. Sasikumar",
""
],
[
"Venkata",
"S. N.",
""
],
[
"Kamath",
"P. U.",
""
],
[
"Natarajan",
"V.",
""
],
[
"Pawankumar",
"S.",
""
],
[
"Krishnan",
"V. U. Sanal",
""
],
[
"Savarimuthu",
"P.",
""
],
[
"Mishra",
"Shalabh",
""
],
[
"Kumar",
"Varun",
""
],
[
"Sumana",
"Chavali",
""
],
[
"Hegde",
"S. Bhavana",
""
],
[
"Utkarsha",
"D.",
""
],
[
"Kumar",
"Amit",
""
],
[
"Nagabhushana",
"S.",
""
],
[
"Kathiravan",
"S.",
""
],
[
"Vemareddy",
"P.",
""
],
[
"Kathiravan",
"C.",
""
],
[
"Nagaraju",
"K.",
""
],
[
"Ravindra",
"Belur",
""
],
[
"Mishra",
"Wageesh",
""
]
] |
2503.21495 | Timo Budszuhn | Timo Budszuhn, Mark Joachim Krallmann, Daniel Horn | Adaptive Resampling with Bootstrap for Noisy Multi-Objective
Optimization Problems | 14 pages. 5 figures | null | null | null | cs.LG cs.AI stat.ML | http://creativecommons.org/licenses/by/4.0/ | The challenge of noisy multi-objective optimization lies in the constant
trade-off between exploring new decision points and improving the precision of
known points through resampling. This decision should take into account both
the variability of the objective functions and the current estimate of a point
in relation to the Pareto front. Since the amount and distribution of noise are
generally unknown, it is desirable for a decision function to be highly
adaptive to the properties of the optimization problem. This paper presents a
resampling decision function that incorporates the stochastic nature of the
optimization problem by using bootstrapping and the probability of dominance.
The distribution-free estimation of the probability of dominance is achieved
using bootstrap estimates of the means. To make the procedure applicable even
with very few observations, we transfer the distribution observed at other
decision points. The efficiency of this resampling approach is demonstrated by
applying it in the NSGA-II algorithm with a sequential resampling procedure
under multiple noise variations.
| [
{
"version": "v1",
"created": "Thu, 27 Mar 2025 13:32:42 GMT"
},
{
"version": "v2",
"created": "Thu, 24 Apr 2025 14:35:26 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"Budszuhn",
"Timo",
""
],
[
"Krallmann",
"Mark Joachim",
""
],
[
"Horn",
"Daniel",
""
]
] |
2503.21612 | Daniel Wachsmuth | Daniel Wachsmuth | A globalized inexact semismooth Newton method for strongly convex
optimal control problems | null | null | null | null | math.OC | http://creativecommons.org/licenses/by-sa/4.0/ | We investigate a globalized inexact semismooth Newton method applied to
strongly convex optimization problems in Hilbert spaces. Here, the semismooth
Newton method is appplied to the dual problem, which has a continuously
differentiable objective. We prove global strong convergence of iterates as
well as transition to local superlinear convergence. The latter needs a
second-order Taylor expansion involving semismooth derivative concepts. The
convergence of the globalized method is demonstrated in numerical examples, for
which the local unglobalized method diverges.
| [
{
"version": "v1",
"created": "Thu, 27 Mar 2025 15:33:48 GMT"
},
{
"version": "v2",
"created": "Thu, 24 Apr 2025 11:21:21 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"Wachsmuth",
"Daniel",
""
]
] |
2503.21870 | Sangjun Cha | Sangjun Cha, Boseong Young Cho, Hyungjin Joo, Wonki Lee, Kim
HyeongHan, Zachary P. Scofield, Kyle Finner, and M. James Jee | A High-Caliber View of the Bullet Cluster Through JWST Strong and Weak
Lensing Analyses | 11 pages, 5 figures, accepted in ApJL | null | null | null | astro-ph.GA | http://creativecommons.org/licenses/by/4.0/ | The Bullet Cluster (1E 0657-56) is a key astrophysical laboratory for
studying dark matter, galaxy cluster mergers, and shock propagation in extreme
environments. Using new JWST imaging, we present the highest-resolution mass
reconstruction to date, combining 146 strong lensing constraints from 37
systems with high-density (398 sources arcmin$^{-2}$) weak lensing data,
without assuming that light traces mass. The main cluster's mass distribution
is highly elongated (NW-SE) and consists of at least three subclumps aligned
with the brightest cluster galaxies. The subcluster is more compact but
elongated along the E-W direction, with a single dominant peak. We also detect
a possible mass and ICL trail extending from the subcluster's eastern side
toward the main cluster. Notably, these detailed features are closely traced by
the intracluster light, with a modified Hausdorff distance of $19.80 \pm 12.46$
kpc. Together with multi-wavelength data, the complex mass distribution
suggests that the merger history of the Bullet Cluster may be more complex than
previous binary cluster merger scenarios.
| [
{
"version": "v1",
"created": "Thu, 27 Mar 2025 18:00:01 GMT"
},
{
"version": "v2",
"created": "Mon, 21 Apr 2025 02:53:57 GMT"
},
{
"version": "v3",
"created": "Thu, 24 Apr 2025 02:07:57 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"Cha",
"Sangjun",
""
],
[
"Cho",
"Boseong Young",
""
],
[
"Joo",
"Hyungjin",
""
],
[
"Lee",
"Wonki",
""
],
[
"HyeongHan",
"Kim",
""
],
[
"Scofield",
"Zachary P.",
""
],
[
"Finner",
"Kyle",
""
],
[
"Jee",
"M. James",
""
]
] |
2503.22093 | Ximing Wen | Ximing Wen, Mallika Mainali, Anik Sen | How Well Can Vison-Language Models Understand Humans' Intention? An
Open-ended Theory of Mind Question Evaluation Benchmark | 4 pages, accepted by ToM@AAAI25 | null | null | null | cs.CV cs.AI | http://creativecommons.org/licenses/by-sa/4.0/ | Vision Language Models (VLMs) have demonstrated strong reasoning capabilities
in Visual Question Answering (VQA) tasks; however, their ability to perform
Theory of Mind (ToM) tasks, such as inferring human intentions, beliefs, and
mental states, remains underexplored. We propose an open-ended question
framework to evaluate VLMs' performance across diverse categories of ToM tasks.
We curated and annotated a benchmark dataset of 30 images and evaluated the
performance of four VLMs of varying sizes. Our results show that the GPT-4
model outperformed all the others, with only one smaller model, GPT-4o-mini,
achieving comparable performance. We observed that VLMs often struggle to infer
intentions in complex scenarios such as bullying or cheating. Our findings
reveal that smaller models can sometimes infer correct intentions despite
relying on incorrect visual cues. The dataset is available at
https://github.com/ximingwen/ToM-AAAI25-Multimodal.
| [
{
"version": "v1",
"created": "Fri, 28 Mar 2025 02:26:32 GMT"
},
{
"version": "v2",
"created": "Thu, 24 Apr 2025 03:20:58 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"Wen",
"Ximing",
""
],
[
"Mainali",
"Mallika",
""
],
[
"Sen",
"Anik",
""
]
] |
2503.23939 | Kaito Kishi | Kaito Kishi, Junpei Yamaguchi, Tetsuya Izu, Noboru Kunihiro | Simulation of Shor algorithm for discrete logarithm problems with
comprehensive pairs of modulo p and order q | null | null | null | null | quant-ph | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | The discrete logarithm problem (DLP) over finite fields, commonly used in
classical cryptography, has no known polynomial-time algorithm on classical
computers. However, Shor has provided its polynomial-time algorithm on quantum
computers. Nevertheless, there are only few examples simulating quantum
circuits that operate on general pairs of modulo $p$ and order $q$. In this
paper, we constructed such quantum circuits and solved DLPs for all 1,860
possible pairs of $p$ and $q$ up to 32 qubits using a quantum simulator with
PRIMEHPC FX700. From this, we obtained and verified values of the success
probabilities, which had previously been heuristically analyzed by Eker\r{a}.
As a result, the detailed waveform shape of the success probability of Shor's
algorithm for solving the DLP, known as a periodic function of order $q$, was
clarified. Additionally, we generated 1,015 quantum circuits for larger pairs
of $p$ and $q$, extrapolated the circuit sizes obtained, and compared them for
$p=2048$ bits between safe-prime groups and Schnorr groups. While in classical
cryptography, the cipher strength of safe-prime groups and Schnorr groups is
the same if $p$ is equal, we quantitatively demonstrated how much the strength
of the latter decreases to the bit length of $p$ in the former when using
Shor's quantum algorithm. In particular, it was experimentally and
theoretically shown that when a ripple carry adder is used in the addition
circuit, the cryptographic strength of a Schnorr group with $p=2048$ bits under
Shor's algorithm is almost equivalent to that of a safe-prime group with
$p=1024$ bits.
| [
{
"version": "v1",
"created": "Mon, 31 Mar 2025 10:39:10 GMT"
},
{
"version": "v2",
"created": "Thu, 24 Apr 2025 12:51:36 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"Kishi",
"Kaito",
""
],
[
"Yamaguchi",
"Junpei",
""
],
[
"Izu",
"Tetsuya",
""
],
[
"Kunihiro",
"Noboru",
""
]
] |
2503.24033 | Denis-Charles Cisinski | Denis-Charles Cisinski | Completion of motivic sheaves | This paper replaces a previous version that pretended to construct a
new non-trivial period isomorphism: this revised version proves that the
strategy was doomed to fail. The last version only fixes some typos and has a
slightly extended introduction | null | null | null | math.AG math.KT math.NT | http://creativecommons.org/licenses/by/4.0/ | We study the process of $\ell$-adic completion of motivic sheaves. We observe
that, in equal characteristic, when restricted to constructible objets, it is
compatible with the six operations. This implies that one can reconstruct
$\ell$-adic sheaves of geometric origin over a scheme of finite type over a
field from $\ell$-adic cohomology of smooth schemes. In the case of finite
fields, this includes perverse $\ell$-adic sheaves of geometric orgin. However,
the analogous behaviour fails systematically in mixed characteristic: the
reason is that it would imply strong independence of $\ell$ results that can be
proven to be too optimistic.
| [
{
"version": "v1",
"created": "Mon, 31 Mar 2025 12:59:51 GMT"
},
{
"version": "v2",
"created": "Wed, 2 Apr 2025 11:21:52 GMT"
},
{
"version": "v3",
"created": "Mon, 21 Apr 2025 17:57:11 GMT"
},
{
"version": "v4",
"created": "Tue, 22 Apr 2025 16:35:41 GMT"
},
{
"version": "v5",
"created": "Wed, 23 Apr 2025 19:59:12 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"Cisinski",
"Denis-Charles",
""
]
] |
2503.24294 | Michal \v{S}mejkal | Martin Hor\'ak, Michal \v{S}mejkal, Martin Kru\v{z}\'ik | Surface-Polyconvex Models for Soft Elastic Solids | null | null | null | null | math-ph math.MP | http://creativecommons.org/licenses/by-nc-nd/4.0/ | Soft solids with surface energy exhibit complex mechanical behavior,
necessitating advanced constitutive models to capture the interplay between
bulk and surface mechanics. This interplay has profound implications for
material design and emerging technologies. In this work, we set up variational
models for bulk-surface elasticity and explore a novel class of
surface-polyconvex constitutive models that account for surface energy while
ensuring the existence of minimizers. These models are implemented within a
finite element framework and validated through benchmark problems and
applications, including, e.g., the liquid bridge problem and the
Rayleigh-Plateau instability, for which the surface energy plays the dominant
role. The results demonstrate the ability of surface-polyconvex models to
accurately capture surface-driven phenomena, establishing them as a powerful
tool for advancing the mechanics of soft materials in both engineering and
biological applications.
| [
{
"version": "v1",
"created": "Mon, 31 Mar 2025 16:41:37 GMT"
},
{
"version": "v2",
"created": "Thu, 24 Apr 2025 12:14:56 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"Horák",
"Martin",
""
],
[
"Šmejkal",
"Michal",
""
],
[
"Kružík",
"Martin",
""
]
] |
2504.01351 | Ryou Ohsawa | Ryou Ohsawa, Daisuke Kawata, Takafumi Kamizuka, Yoshiyuki Yamada,
Wolfgang L\"offler, Michael Biermann | Demonstration of Plate Analysis, an algorithm for precise relative
astrometry | author's submit version, 31 pages, 11 figures, accepted for
publication in Journal of Astronomical Telescopes, Instruments, and Systems | J. Astron. Telesc. Instrum. Syst. 11(2), 028001 (2025) | 10.1117/1.JATIS.11.2.028001 | null | astro-ph.IM | http://creativecommons.org/licenses/by-nc-sa/4.0/ | Astrometry plays a crucial role in understanding the structure, dynamics, and
evolution of celestial objects by providing precise measurements of their
positions and motions. We propose a new approach to wide-field, relative
astrometry, Plate Analysis. Plate Analysis is an innovative algorithm that
estimates stellar coordinates and corrects geometric distortion based on
precise reference sources, without relying on dedicated calibration fields. It
is implemented as a probabilistic framework using Stochastic Variational
Inference to efficiently optimize the numerous parameters involved in the
model.
The methodology was tested through a simplified simulation designed after the
Galactic center survey of the JASMINE mission. This simulation, called the
JASMINE mini-mock survey, covered three years of observation with 100 satellite
orbits, providing a comprehensive dataset for evaluating the performance of
Plate Analysis. Although the observation model incorporated more than 30,000
parameters, the parameters were efficiently optimized through Plate Analysis.
The results showed that the estimated coordinates closely match the expected
stellar motions, with an average positional error of about $70\,\mathrm{\mu
as}$. These findings validate the potential of Plate Analysis for precise
wide-field astrometric applications, offering significant insights into
improving the accuracy of stellar dynamics measurements.
| [
{
"version": "v1",
"created": "Wed, 2 Apr 2025 04:51:21 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"Ohsawa",
"Ryou",
""
],
[
"Kawata",
"Daisuke",
""
],
[
"Kamizuka",
"Takafumi",
""
],
[
"Yamada",
"Yoshiyuki",
""
],
[
"Löffler",
"Wolfgang",
""
],
[
"Biermann",
"Michael",
""
]
] |
2504.01422 | Cheng Feng | L Zhang, C Feng and T Xia | Optimization of BLE Broadcast Mode in Offline Finding Network | null | null | null | null | cs.NI | http://creativecommons.org/licenses/by-nc-nd/4.0/ | In the Offline Finding Network(OFN), offline Bluetooth tags broadcast to the
surrounding area, the finder devices receiving the broadcast signal and upload
location information to the IoT(Internet of Things) cloud servers, thereby
achieving offline finding of lost items. This process is essentially a
Bluetooth low energy (BLE) neighbor discovery process(NDP). In the process, the
variety of Bluetooth scan modes caused by the scan interval and scan window
settings affects the discovery latency of finder devices finding the tag
broadcast packets. To optimize the experience of searching for lost devices, we
propose the CPBIS-mechanism, a certain proportion broadcast-intervals screening
mechanism that calculates the most suitable two broadcast intervals and their
proportion for offline tags. This reduces discovery latency in the BLE NDP,
improves the discovery success rate, further enhances the user experience. To
our knowledge, we are the first to propose a comprehensive solution for
configuring the broadcast interval parameters of advertisers in BLE NDP,
particularly for configurations involving two or more broadcast intervals. We
evaluated the results obtained by CPBIS on the nRF52832 chip. The data shows
that the CPBIS-mechanism achieves relatively low discovery latencies for
multiple scan modes.
| [
{
"version": "v1",
"created": "Wed, 2 Apr 2025 07:13:22 GMT"
},
{
"version": "v2",
"created": "Thu, 24 Apr 2025 02:48:13 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"Zhang",
"L",
""
],
[
"Feng",
"C",
""
],
[
"Xia",
"T",
""
]
] |
2504.01482 | Qihao Ye | Qihao Ye, Xiaochuan Tian, Yuhua Zhu | A Robust Model-Based Approach for Continuous-Time Policy Evaluation with
Unknown L\'evy Process Dynamics | 28 pages, 9 figures | null | null | null | cs.LG cs.NA math.NA | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | This paper develops a model-based framework for continuous-time policy
evaluation (CTPE) in reinforcement learning, incorporating both Brownian and
L\'evy noise to model stochastic dynamics influenced by rare and extreme
events. Our approach formulates the policy evaluation problem as solving a
partial integro-differential equation (PIDE) for the value function with
unknown coefficients. A key challenge in this setting is accurately recovering
the unknown coefficients in the stochastic dynamics, particularly when driven
by L\'evy processes with heavy tail effects. To address this, we propose a
robust numerical approach that effectively handles both unbiased and censored
trajectory datasets. This method combines maximum likelihood estimation with an
iterative tail correction mechanism, improving the stability and accuracy of
coefficient recovery. Additionally, we establish a theoretical bound for the
policy evaluation error based on coefficient recovery error. Through numerical
experiments, we demonstrate the effectiveness and robustness of our method in
recovering heavy-tailed L\'evy dynamics and verify the theoretical error
analysis in policy evaluation.
| [
{
"version": "v1",
"created": "Wed, 2 Apr 2025 08:37:14 GMT"
},
{
"version": "v2",
"created": "Thu, 24 Apr 2025 07:39:44 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"Ye",
"Qihao",
""
],
[
"Tian",
"Xiaochuan",
""
],
[
"Zhu",
"Yuhua",
""
]
] |
2504.01516 | Kevin Zwart | Kevin Zwart | An addendum on the Mathieu Conjecture for $SU(N)$, $Sp(N)$ and $G_2$ | 13 pages, no figures | null | null | null | math.GR | http://creativecommons.org/licenses/by-nc-nd/4.0/ | In this paper, we sharpen results obtained by the author in 2023. The new
results reduce the Mathieu Conjecture on $SU(N)$ (formulated for all compact
connected Lie groups by O. Mathieu in 1997) to a conjecture involving only
functions on $\mathbb{R}^n\times (S^1)^m$ with $n,m$ non-negative integers
instead of involving functions on $\mathbb{R}^n\times (S^1\setminus\{1\})^m$.
The proofs rely on a more recent work of the author (2024) and a specific $KAK$
decomposition. Finally, with these results we can also improve the results on
the groups $Sp(N)$ and $G_2$ in the latter paper, since they relied on the
construction introduced in the 2023 paper.
| [
{
"version": "v1",
"created": "Wed, 2 Apr 2025 09:00:47 GMT"
},
{
"version": "v2",
"created": "Thu, 24 Apr 2025 16:46:44 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"Zwart",
"Kevin",
""
]
] |
2504.01650 | Thomas Rochussen | Tommy Rochussen and Vincent Fortuin | Sparse Gaussian Neural Processes | Proceedings of the 7th Symposium on Advances in Approximate Bayesian
Inference, PMLR, 2025. 25 pages, 6 figures, 5 tables | null | null | null | stat.ML cs.LG | http://creativecommons.org/licenses/by/4.0/ | Despite significant recent advances in probabilistic meta-learning, it is
common for practitioners to avoid using deep learning models due to a
comparative lack of interpretability. Instead, many practitioners simply use
non-meta-models such as Gaussian processes with interpretable priors, and
conduct the tedious procedure of training their model from scratch for each
task they encounter. While this is justifiable for tasks with a limited number
of data points, the cubic computational cost of exact Gaussian process
inference renders this prohibitive when each task has many observations. To
remedy this, we introduce a family of models that meta-learn sparse Gaussian
process inference. Not only does this enable rapid prediction on new tasks with
sparse Gaussian processes, but since our models have clear interpretations as
members of the neural process family, it also allows manual elicitation of
priors in a neural process for the first time. In meta-learning regimes for
which the number of observed tasks is small or for which expert domain
knowledge is available, this offers a crucial advantage.
| [
{
"version": "v1",
"created": "Wed, 2 Apr 2025 12:00:09 GMT"
},
{
"version": "v2",
"created": "Thu, 24 Apr 2025 15:21:20 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"Rochussen",
"Tommy",
""
],
[
"Fortuin",
"Vincent",
""
]
] |
2504.01949 | Tristan Fauvel | Tristan Fauvel, Julien Tanniou, Pascal Godbillot, Marie G\'enin, Billy
Amzal | Comparison of Bayesian methods for extrapolation of treatment effects: a
large scale simulation study | null | null | null | null | stat.ME stat.AP | http://creativecommons.org/licenses/by/4.0/ | Extrapolating treatment effects from related studies is a promising strategy
for designing and analyzing clinical trials in situations where achieving an
adequate sample size is challenging. Bayesian methods are well-suited for this
purpose, as they enable the synthesis of prior information through the use of
prior distributions. While the operating characteristics of Bayesian approaches
for borrowing data from control arms have been extensively studied, methods
that borrow treatment effects -- quantities derived from the comparison between
two arms -- remain less well understood. In this paper, we present the findings
of an extensive simulation study designed to address this gap. We evaluate the
frequentist operating characteristics of these methods, including the
probability of success, mean squared error, bias, precision, and credible
interval coverage. Our results provide insights into the strengths and
limitations of existing methods in the context of confirmatory trials. In
particular, we show that the Conditional Power Prior and the Robust Mixture
Prior perform better overall, while the test-then-pool variants and the
p-value-based power prior display suboptimal performance.
| [
{
"version": "v1",
"created": "Wed, 2 Apr 2025 17:55:11 GMT"
},
{
"version": "v2",
"created": "Thu, 24 Apr 2025 12:05:13 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"Fauvel",
"Tristan",
""
],
[
"Tanniou",
"Julien",
""
],
[
"Godbillot",
"Pascal",
""
],
[
"Génin",
"Marie",
""
],
[
"Amzal",
"Billy",
""
]
] |
2504.02173 | Eric R. Bittner | Eric R. Bittner and Bhavay Tyagi | Statistical Control of Relaxation and Synchronization in Open Anyonic
Systems | null | null | null | null | quant-ph | http://creativecommons.org/licenses/by/4.0/ | Quantum statistics dictate how particles exchange and correlate-but in
two-dimensional systems, these rules extend beyond bosons and fermions to
anyons, quasiparticles with continuously tunable exchange phases. Here, we
develop a Lindblad framework for anyonic oscillators and show that fractional
statistics enable statistical control of decoherence in open quantum systems.
By varying the anyonic phase and environmental correlations, we demonstrate
tunable mode protection, identify exceptional points in the dissipative
spectrum, and reveal temperature-dependent coherence bifurcations. Using
coherent multidimensional spectroscopy as a probe, we show that statistical
phases leave distinct fingerprints in the third-order response, opening new
routes to detect and manipulate topological excitations. These results
establish the exchange phase as a functional control parameter for engineering
dissipation-resilient quantum states.
| [
{
"version": "v1",
"created": "Wed, 2 Apr 2025 23:16:18 GMT"
},
{
"version": "v2",
"created": "Thu, 24 Apr 2025 15:20:34 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"Bittner",
"Eric R.",
""
],
[
"Tyagi",
"Bhavay",
""
]
] |
2504.02218 | Arif Ullah | Arif Ullah | From short-sighted to far-sighted: A comparative study of recursive
machine learning approaches for open quantum systems | null | null | null | null | physics.chem-ph | http://creativecommons.org/licenses/by/4.0/ | Accurately modeling open quantum system dynamics is crucial for advancing
quantum technologies, yet traditional methods struggle to balance accuracy and
efficiency. Machine learning (ML) provides a promising alternative,
particularly through recursive models that predict system evolution based on
past history. While these models have shown success in predicting single
observables, their effectiveness in more complex tasks, such as forecasting the
full reduced density matrix (RDM), remains unclear. We extend history-based
recursive ML approaches to complex quantum systems, comparing four
physics-informed neural network (PINN) architectures: (i) single-RDM-predicting
PINN (SR-PINN), (ii) SR-PINN with simulation parameters (PSR-PINN), (iii)
multi-RDMs-predicting PINN (MR-PINN), and (iv) MR-PINN with simulation
parameters (PMR-PINN). These models are applied to the spin-boson (SB) model
and the Fenna-Matthews-Olson (FMO) complex. Our results show that SR-PINN and
PSR-PINN, constrained by a narrow history window, fail to capture complex
quantum evolution, leading to unstable long-term predictions, especially in
nonlinear and highly correlated dynamics. In contrast, MR-PINN and PMR-PINN
improve accuracy by extending the forecast horizon, incorporating long-range
correlations, and reducing error propagation. Surprisingly, explicitly
including simulation parameters such as temperature and reorganization energy
in PSR-PINN and PMR-PINN does not consistently enhance accuracy and can even
reduce performance, suggesting that these effects are already encoded in the
RDM evolution. These findings highlight the limitations of short-sighted
recursive forecasting and demonstrate the superior stability and accuracy of
far-sighted approaches for long-term predictions.
| [
{
"version": "v1",
"created": "Thu, 3 Apr 2025 02:14:47 GMT"
},
{
"version": "v2",
"created": "Thu, 24 Apr 2025 07:01:01 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"Ullah",
"Arif",
""
]
] |
2504.02269 | Hanqi Su | Jay Lee, Hanqi Su, Dai-Yan Ji, and Takanobu Minami | Engineering Artificial Intelligence: Framework, Challenges, and Future
Direction | The paper submitted to the Journal Machine Learning: Engineering has
been accepted | null | null | null | cs.AI cs.LG | http://creativecommons.org/licenses/by/4.0/ | Over the past ten years, the application of artificial intelligence (AI) and
machine learning (ML) in engineering domains has gained significant popularity,
showcasing their potential in data-driven contexts. However, the complexity and
diversity of engineering problems often require the development of
domain-specific AI approaches, which are frequently hindered by a lack of
systematic methodologies, scalability, and robustness during the development
process. To address this gap, this paper introduces the "ABCDE" as the key
elements of Engineering AI and proposes a unified, systematic engineering AI
ecosystem framework, including eight essential layers, along with attributes,
goals, and applications, to guide the development and deployment of AI
solutions for specific engineering needs. Additionally, key challenges are
examined, and eight future research directions are highlighted. By providing a
comprehensive perspective, this paper aims to advance the strategic
implementation of AI, fostering the development of next-generation engineering
AI solutions.
| [
{
"version": "v1",
"created": "Thu, 3 Apr 2025 04:30:10 GMT"
},
{
"version": "v2",
"created": "Thu, 17 Apr 2025 03:14:31 GMT"
},
{
"version": "v3",
"created": "Wed, 23 Apr 2025 18:36:36 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"Lee",
"Jay",
""
],
[
"Su",
"Hanqi",
""
],
[
"Ji",
"Dai-Yan",
""
],
[
"Minami",
"Takanobu",
""
]
] |
2504.02441 | Lianlei Shan | Lianlei Shan, Shixian Luo, Zezhou Zhu, Yu Yuan, Yong Wu | Cognitive Memory in Large Language Models | 37 pages, 9 figures | null | null | null | cs.CL cs.AI | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | This paper examines memory mechanisms in Large Language Models (LLMs),
emphasizing their importance for context-rich responses, reduced
hallucinations, and improved efficiency. It categorizes memory into sensory,
short-term, and long-term, with sensory memory corresponding to input prompts,
short-term memory processing immediate context, and long-term memory
implemented via external databases or structures. The text-based memory section
covers acquisition (selection and summarization), management (updating,
accessing, storing, and resolving conflicts), and utilization (full-text
search, SQL queries, semantic search). The KV cache-based memory section
discusses selection methods (regularity-based summarization, score-based
approaches, special token embeddings) and compression techniques (low-rank
compression, KV merging, multimodal compression), along with management
strategies like offloading and shared attention mechanisms. Parameter-based
memory methods (LoRA, TTT, MoE) transform memories into model parameters to
enhance efficiency, while hidden-state-based memory approaches (chunk
mechanisms, recurrent transformers, Mamba model) improve long-text processing
by combining RNN hidden states with current methods. Overall, the paper offers
a comprehensive analysis of LLM memory mechanisms, highlighting their
significance and future research directions.
| [
{
"version": "v1",
"created": "Thu, 3 Apr 2025 09:58:19 GMT"
},
{
"version": "v2",
"created": "Thu, 24 Apr 2025 01:47:25 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"Shan",
"Lianlei",
""
],
[
"Luo",
"Shixian",
""
],
[
"Zhu",
"Zezhou",
""
],
[
"Yuan",
"Yu",
""
],
[
"Wu",
"Yong",
""
]
] |
2504.03515 | Shan An | Shan An, Ziyu Meng, Chao Tang, Yuning Zhou, Tengyu Liu, Fangqiang
Ding, Shufang Zhang, Yao Mu, Ran Song, Wei Zhang, Zeng-Guang Hou, and Hong
Zhang | Dexterous Manipulation through Imitation Learning: A Survey | 22pages, 5 figures | null | null | null | cs.RO cs.LG | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Dexterous manipulation, which refers to the ability of a robotic hand or
multi-fingered end-effector to skillfully control, reorient, and manipulate
objects through precise, coordinated finger movements and adaptive force
modulation, enables complex interactions similar to human hand dexterity. With
recent advances in robotics and machine learning, there is a growing demand for
these systems to operate in complex and unstructured environments. Traditional
model-based approaches struggle to generalize across tasks and object
variations due to the high dimensionality and complex contact dynamics of
dexterous manipulation. Although model-free methods such as reinforcement
learning (RL) show promise, they require extensive training, large-scale
interaction data, and carefully designed rewards for stability and
effectiveness. Imitation learning (IL) offers an alternative by allowing robots
to acquire dexterous manipulation skills directly from expert demonstrations,
capturing fine-grained coordination and contact dynamics while bypassing the
need for explicit modeling and large-scale trial-and-error. This survey
provides an overview of dexterous manipulation methods based on imitation
learning, details recent advances, and addresses key challenges in the field.
Additionally, it explores potential research directions to enhance IL-driven
dexterous manipulation. Our goal is to offer researchers and practitioners a
comprehensive introduction to this rapidly evolving domain.
| [
{
"version": "v1",
"created": "Fri, 4 Apr 2025 15:14:38 GMT"
},
{
"version": "v2",
"created": "Thu, 24 Apr 2025 13:15:13 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"An",
"Shan",
""
],
[
"Meng",
"Ziyu",
""
],
[
"Tang",
"Chao",
""
],
[
"Zhou",
"Yuning",
""
],
[
"Liu",
"Tengyu",
""
],
[
"Ding",
"Fangqiang",
""
],
[
"Zhang",
"Shufang",
""
],
[
"Mu",
"Yao",
""
],
[
"Song",
"Ran",
""
],
[
"Zhang",
"Wei",
""
],
[
"Hou",
"Zeng-Guang",
""
],
[
"Zhang",
"Hong",
""
]
] |
2504.03645 | Karima Benali | Karima Benali | Bialgebra theory and $\mathcal O$-operators of admissible Hom-Poisson
algebras | null | null | null | null | math.RA | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | In this paper, we present and explore several key concepts within the
framework of Hom-Poisson algebras. Specifically, we introduce the notions of
admissible Hom-Poisson algebras, along with the related ideas of matched pairs
and Manin triples for such algebras. We then define the concept of a purely
admissible Hom-Poisson bialgebra, placing particular emphasis on its
compatibility with the Manin triple structure associated with a nondegenerate
symmetric bilinear form. This compatibility is crucial for understanding the
structural interplay between these algebraic objects. Additionally, we
investigate the notion of Hom-$ \mathcal O$-operators acting on admissible
Hom-Poisson algebras. We analyze their properties and establish a connection
with admissible Hom-pre-Poisson algebras, shedding light on the relationship
between these two structures.
| [
{
"version": "v1",
"created": "Fri, 7 Feb 2025 11:08:23 GMT"
},
{
"version": "v2",
"created": "Thu, 24 Apr 2025 09:49:32 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"Benali",
"Karima",
""
]
] |
2504.03816 | Nikolai Terekhov | Nikolai Terekhov | A short proof of Tuza's conjecture for weak saturation in hypergraphs | null | null | null | null | math.CO | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Given an $r$-uniform hypergraph $H$ and a positive integer $n$, the weak
saturation number $\mathrm{wsat}(n,H)$ is the minimum number of edges in an
$r$-uniform hypergraph $F$ on $n$ vertices such that the missing edges in $F$
can be added, one at a time, so that each added edge creates a copy of $H$.
Shapira and Tyomkyn (Proceedings of the American Mathematical Society, 2023)
proved Tuza's conjecture on asymptotic behaviour of $\mathrm{wsat}(n, H)$. In
this paper we provide a significantly shorter proof of the conjecture.
| [
{
"version": "v1",
"created": "Fri, 4 Apr 2025 15:39:40 GMT"
},
{
"version": "v2",
"created": "Thu, 24 Apr 2025 08:09:50 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"Terekhov",
"Nikolai",
""
]
] |
2504.03985 | Tanay Raj Bhatt | Tanay Raj Bhatt | Uncertainty, Single Crossing Property, and Stochastic Choice Data | null | null | null | null | econ.TH | http://creativecommons.org/licenses/by/4.0/ | In a typical model of private information and choice under uncertainty, a
decision maker observes a signal, updates her prior beliefs using Bayes rule,
and maximizes her expected utility. If the decision maker's utility function
satisfies the single crossing property, and the information structure is
ordered according to the monotone likelihood ratio, then the comparative
statics exhibit monotonicity with respect to signals. We consider the
restrictions placed by this model of signal processing on state conditional
stochastic choice data. In particular, we show that this model rationalizes a
state conditional stochastic choice dataset if and only if the dataset itself
is ordered according to the monotone likelihood ratio.
| [
{
"version": "v1",
"created": "Fri, 4 Apr 2025 22:56:45 GMT"
},
{
"version": "v2",
"created": "Thu, 24 Apr 2025 17:20:20 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"Bhatt",
"Tanay Raj",
""
]
] |
2504.04002 | Abigail Hering | Abigail R. Hering, Mansha Dubey, Elahe Hosseini, Meghna Srivastava, Yu
An, Juan-Pablo Correa-Baena, Houman Homayoun, and Marina S. Leite | Machine Learning Reveals Composition Dependent Thermal Stability in
Halide Perovskites | 21 pages, 5 figures | null | null | null | cond-mat.mtrl-sci cs.LG | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Halide perovskites exhibit unpredictable properties in response to
environmental stressors, due to several composition-dependent degradation
mechanisms. In this work, we apply data visualization and machine learning (ML)
techniques to reveal unexpected correlations between composition, temperature,
and material properties while using high throughput, in situ environmental
photoluminescence (PL) experiments. Correlation heatmaps show the strong
influence of Cs content on film degradation, and dimensionality reduction
visualization methods uncover clear composition-based data clusters. An extreme
gradient boosting algorithm (XGBoost) effectively forecasts PL features for ten
perovskite films with both composition-agnostic (>85% accuracy) and
composition-dependent (>75% accuracy) model approaches, while elucidating the
relative feature importance of composition (up to 99%). This model validates a
previously unseen anti-correlation between Cs content and material thermal
stability. Our ML-based framework can be expanded to any perovskite family,
significantly reducing the analysis time currently employed to identify stable
options for photovoltaics.
| [
{
"version": "v1",
"created": "Sat, 5 Apr 2025 00:13:18 GMT"
},
{
"version": "v2",
"created": "Thu, 24 Apr 2025 02:02:44 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"Hering",
"Abigail R.",
""
],
[
"Dubey",
"Mansha",
""
],
[
"Hosseini",
"Elahe",
""
],
[
"Srivastava",
"Meghna",
""
],
[
"An",
"Yu",
""
],
[
"Correa-Baena",
"Juan-Pablo",
""
],
[
"Homayoun",
"Houman",
""
],
[
"Leite",
"Marina S.",
""
]
] |
2504.05058 | Aravind Krishnan | Aravind Krishnan and Siva Reddy and Marius Mosbach | Not All Data Are Unlearned Equally | null | null | null | null | cs.CL | http://creativecommons.org/licenses/by/4.0/ | Machine unlearning is concerned with the task of removing knowledge learned
from particular data points from a trained model. In the context of large
language models (LLMs), unlearning has recently received increased attention,
particularly for removing knowledge about named entities from models for
privacy purposes. While various approaches have been proposed to address the
unlearning problem, most existing approaches treat all data points to be
unlearned equally, i.e., unlearning that Montreal is a city in Canada is
treated exactly the same as unlearning the phone number of the first author of
this paper. In this work, we show that this all data is equal assumption does
not hold for LLM unlearning. We study how the success of unlearning depends on
the frequency of the knowledge we want to unlearn in the pre-training data of a
model and find that frequency strongly affects unlearning, i.e., more frequent
knowledge is harder to unlearn. Additionally, we uncover a misalignment between
probability and generation-based evaluations of unlearning and show that this
problem worsens as models become larger. Overall, our experiments highlight the
need for better evaluation practices and novel methods for LLM unlearning that
take the training data of models into account.
| [
{
"version": "v1",
"created": "Mon, 7 Apr 2025 13:29:02 GMT"
},
{
"version": "v2",
"created": "Tue, 8 Apr 2025 08:18:14 GMT"
},
{
"version": "v3",
"created": "Fri, 11 Apr 2025 17:03:04 GMT"
},
{
"version": "v4",
"created": "Thu, 24 Apr 2025 13:16:26 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"Krishnan",
"Aravind",
""
],
[
"Reddy",
"Siva",
""
],
[
"Mosbach",
"Marius",
""
]
] |
2504.05436 | Bishnu Prasad Belbase | Bishnu P. Belbase, Arjun Unnikrishnan, Eun Sang Choi, and Arnab
Banerjee | Field-tunable spin disordered phase in the triangular-lattice
delafossite TlYbSe2 | null | null | null | null | cond-mat.str-el cond-mat.mtrl-sci cond-mat.other quant-ph | http://creativecommons.org/licenses/by/4.0/ | We introduce a relatively disorder-free frustrated triangular lattice
compound TlYbSe2 belonging to the rare-earth delafossite family - extending the
search for the quintessential chiral quantum spin liquid state. While DC
magnetization suggests magnetic exchange interactions in the order of several
kelvin, the zero-field AC magnetization and heat capacity measurements reveal
no signs of long-range magnetic order down to 20 mK, indicating a highly
frustrated quantum-disordered ground state. The high-field AC magnetization
reveals a phase diagram generally consistent with a large family of Yb
delafossites. We observe a spin glass transition around ~ 30 mK at zero field,
which we argue is due to free spins. A broad anomaly in the heat capacity
measurements between 2-5 K - indicative of short-range spin correlations -
along with a linear temperature dependence at low temperatures and the complete
absence of long-range order at low fields, establishes the low-temperature,
low-field regime of TlYbSe2 as a prime location for exploring field-tunable QSL
behavior.
| [
{
"version": "v1",
"created": "Mon, 7 Apr 2025 18:59:51 GMT"
},
{
"version": "v2",
"created": "Thu, 24 Apr 2025 15:17:34 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"Belbase",
"Bishnu P.",
""
],
[
"Unnikrishnan",
"Arjun",
""
],
[
"Choi",
"Eun Sang",
""
],
[
"Banerjee",
"Arnab",
""
]
] |
2504.06290 | Kevin E. Cahill | Kevin Cahill | Is the universe 15.5 Gyr old with time-dependent dark energy and galaxy
formation at redshift 14? | 6 pages, 3 figures | null | null | null | physics.gen-ph | http://creativecommons.org/licenses/by/4.0/ | On the basis of their redshift and Lyman-$\alpha$ measurements, the DESI
Collaboration suggest that dark energy varies with the scale factor $a$ as $
\rho_{de}(a)={} \, a^{-3(1 + w_0 + w_a)} e^{-3 w_a (1-a)}\rho_{de,0} $ in which
$w_0 = - 0.752 \pm 0.057$ and $w_a ={} - 0.86^{+0.23}_{-0.20} $. A DESI
universe would then be 15.5 Gyr old and would have been older than a
$\Lambda$CDM universe at $z=8$ and $z=14$ by 132 and 61 Myr respectively. These
age differences make a DESI universe slightly more compatible than a
$\Lambda$CDM universe with JWST observations of bright and well-formed galaxies
in the very early universe.
| [
{
"version": "v1",
"created": "Fri, 4 Apr 2025 03:46:52 GMT"
},
{
"version": "v2",
"created": "Thu, 24 Apr 2025 04:34:27 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"Cahill",
"Kevin",
""
]
] |
2504.06314 | Abdelghani Maddi | Abdelghani Maddi and Jaime A. Teixeira da Silva | Beyond authorship: Analyzing contributions in PLOS ONE and the
challenges of appropriate attribution | null | Abdelghani Maddi, Jaime A. Teixeira da Silva. Beyond authorship:
Analyzing contributions in PLOS ONE and the challenges of appropriate
attribution[J]. Journal of Data and Information Science, 2024 | 10.2478/jdis-2024-0015 | null | cs.DL cs.CY | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | This study aims to evaluate the accuracy of authorship attributions in
scientific publications, focusing on the fairness and precision of individual
contributions within academic works. The study analyzes 81,823 publications
from the journal PLOS ONE, covering the period from January 2018 to June 2023.
It examines the authorship attributions within these publications to try and
determine the prevalence of inappropriate authorship. It also investigates the
demographic and professional profiles of affected authors, exploring trends and
potential factors contributing to inaccuracies in authorship. Surprisingly,
9.14% of articles feature at least one author with inappropriate authorship,
affecting over 14,000 individuals (2.56% of the sample). Inappropriate
authorship is more concentrated in Asia, Africa, and specific European
countries like Italy. Established researchers with significant publication
records and those affiliated with companies or nonprofits show higher instances
of potential monetary authorship. Our findings are based on contributions as
declared by the authors, which implies a degree of trust in their transparency.
However, this reliance on self-reporting may introduce biases or inaccuracies
into the dataset. Further research could employ additional verification methods
to enhance the reliability of the findings. These findings have significant
implications for journal publishers, highlighting the necessity for robust
control mechanisms to ensure the integrity of authorship attributions.
Moreover, researchers must exercise discernment in determining when to
acknowledge a contributor and when to include them in the author list.
Addressing these issues is crucial for maintaining the credibility and fairness
of academic publications.
| [
{
"version": "v1",
"created": "Tue, 8 Apr 2025 06:47:52 GMT"
},
{
"version": "v2",
"created": "Thu, 24 Apr 2025 07:33:17 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"Maddi",
"Abdelghani",
""
],
[
"da Silva",
"Jaime A. Teixeira",
""
]
] |
2504.06398 | Saber Malekmohammadi | Saber Malekmohammadi, Hong kyu Lee, Li Xiong | Sharpness-Aware Parameter Selection for Machine Unlearning | null | null | null | null | cs.LG | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | It often happens that some sensitive personal information, such as credit
card numbers or passwords, are mistakenly incorporated in the training of
machine learning models and need to be removed afterwards. The removal of such
information from a trained model is a complex task that needs to partially
reverse the training process. There have been various machine unlearning
techniques proposed in the literature to address this problem. Most of the
proposed methods revolve around removing individual data samples from a trained
model. Another less explored direction is when features/labels of a group of
data samples need to be reverted. While the existing methods for these tasks do
the unlearning task by updating the whole set of model parameters or only the
last layer of the model, we show that there are a subset of model parameters
that have the largest contribution in the unlearning target features. More
precisely, the model parameters with the largest corresponding diagonal value
in the Hessian matrix (computed at the learned model parameter) have the most
contribution in the unlearning task. By selecting these parameters and updating
them during the unlearning stage, we can have the most progress in unlearning.
We provide theoretical justifications for the proposed strategy by connecting
it to sharpness-aware minimization and robust unlearning. We empirically show
the effectiveness of the proposed strategy in improving the efficacy of
unlearning with a low computational cost.
| [
{
"version": "v1",
"created": "Tue, 8 Apr 2025 19:41:07 GMT"
},
{
"version": "v2",
"created": "Thu, 24 Apr 2025 14:14:37 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"Malekmohammadi",
"Saber",
""
],
[
"Lee",
"Hong kyu",
""
],
[
"Xiong",
"Li",
""
]
] |
2504.06768 | Shutong Chen | Shutong Chen, Tianyi Zhou, Guodong Long, Jing Jiang, Chengqi Zhang | FedMerge: Federated Personalization via Model Merging | null | null | null | null | cs.LG | http://creativecommons.org/licenses/by/4.0/ | One global model in federated learning (FL) might not be sufficient to serve
many clients with non-IID tasks and distributions. While there has been
advances in FL to train multiple global models for better personalization, they
only provide limited choices to clients so local finetuning is still
indispensable. In this paper, we propose a novel ``FedMerge'' approach that can
create a personalized model per client by simply merging multiple global models
with automatically optimized and customized weights. In FedMerge, a few global
models can serve many non-IID clients, even without further local finetuning.
We formulate this problem as a joint optimization of global models and the
merging weights for each client. Unlike existing FL approaches where the server
broadcasts one or multiple global models to all clients, the server only needs
to send a customized, merged model to each client. Moreover, instead of
periodically interrupting the local training and re-initializing it to a global
model, the merged model aligns better with each client's task and data
distribution, smoothening the local-global gap between consecutive rounds
caused by client drift. We evaluate FedMerge on three different non-IID
settings applied to different domains with diverse tasks and data types, in
which FedMerge consistently outperforms existing FL approaches, including
clustering-based and mixture-of-experts (MoE) based methods.
| [
{
"version": "v1",
"created": "Wed, 9 Apr 2025 10:44:14 GMT"
},
{
"version": "v2",
"created": "Thu, 24 Apr 2025 11:12:13 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"Chen",
"Shutong",
""
],
[
"Zhou",
"Tianyi",
""
],
[
"Long",
"Guodong",
""
],
[
"Jiang",
"Jing",
""
],
[
"Zhang",
"Chengqi",
""
]
] |
2504.07315 | Claire Bowern | Alessio Tosolini and Claire Bowern | Multilingual MFA: Forced Alignment on Low-Resource Related Languages | null | ComputEl8, 2025 | null | null | cs.CL | http://creativecommons.org/licenses/by-nc-sa/4.0/ | We compare the outcomes of multilingual and crosslingual training for related
and unrelated Australian languages with similar phonological inventories. We
use the Montreal Forced Aligner to train acoustic models from scratch and adapt
a large English model, evaluating results against seen data, unseen data (seen
language), and unseen data and language. Results indicate benefits of adapting
the English baseline model for previously unseen languages.
| [
{
"version": "v1",
"created": "Wed, 9 Apr 2025 22:32:57 GMT"
},
{
"version": "v2",
"created": "Wed, 23 Apr 2025 21:43:01 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"Tosolini",
"Alessio",
""
],
[
"Bowern",
"Claire",
""
]
] |
2504.07347 | Yueying Li | Yueying Li and Jim Dai and Tianyi Peng | Throughput-Optimal Scheduling Algorithms for LLM Inference and AI Agents | null | null | null | null | stat.ML cs.LG math.PR | http://creativecommons.org/licenses/by/4.0/ | As demand for Large Language Models (LLMs) and AI agents rapidly grows,
optimizing systems for efficient LLM inference becomes critical. While
significant efforts have focused on system-level engineering, little is
explored from a mathematical modeling and queuing perspective.
In this paper, we aim to develop the queuing fundamentals for large language
model (LLM) inference, bridging the gap between the queueing theory and LLM
system communities. In particular, we study the throughput aspect in LLM
inference systems. We prove that a large class of 'work-conserving' scheduling
algorithms can achieve maximum throughput for individual inference LLM engine,
highlighting 'work-conserving' as a key design principle in practice. In a
network of LLM agents, work-conserving scheduling alone is insufficient,
particularly when facing specific workload structures and multi-class workflows
that require more sophisticated scheduling strategies. Evaluations of
real-world systems show that Orca and Sarathi-serve are throughput-optimal,
reassuring practitioners, while FasterTransformer and vanilla vLLM are not
maximally stable and should be used with caution. Our results highlight the
substantial benefits that the queueing community can offer in improving LLM
inference systems and call for more interdisciplinary development.
| [
{
"version": "v1",
"created": "Thu, 10 Apr 2025 00:12:12 GMT"
},
{
"version": "v2",
"created": "Thu, 24 Apr 2025 14:10:22 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"Li",
"Yueying",
""
],
[
"Dai",
"Jim",
""
],
[
"Peng",
"Tianyi",
""
]
] |
2504.07687 | Yihao Wang | Yihao Wang, Zhong Qian, Peifeng Li | FMNV: A Dataset of Media-Published News Videos for Fake News Detection | null | null | null | null | cs.CV cs.MM | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | News media, particularly video-based platforms, have become deeply embedded
in daily life, concurrently amplifying risks of misinformation dissemination.
Consequently, multimodal fake news detection has garnered significant research
attention. However, existing datasets predominantly comprise user-generated
videos characterized by crude editing and limited public engagement, whereas
professionally crafted fake news videos disseminated by media outlets, often
politically or virally motivated-pose substantially greater societal harm. To
address this gap, we construct FMNV, a novel dataset exclusively composed of
news videos published by media organizations. Through empirical analysis of
existing datasets and our curated collection, we categorize fake news videos
into four distinct types. Building upon this taxonomy, we employ Large Language
Models (LLMs) to automatically generate deceptive content by manipulating
authentic media-published news videos. Furthermore, we propose FMNVD, a
baseline model featuring a dual-stream architecture integrating CLIP and Faster
R-CNN for video feature extraction, enhanced by co-attention mechanisms for
feature refinement and multimodal aggregation. Comparative experiments
demonstrate both the generalization capability of FMNV across multiple
baselines and the superior detection efficacy of FMNVD. This work establishes
critical benchmarks for detecting high-impact fake news in media ecosystems
while advancing methodologies for cross-modal inconsistency analysis.
| [
{
"version": "v1",
"created": "Thu, 10 Apr 2025 12:16:32 GMT"
},
{
"version": "v2",
"created": "Thu, 24 Apr 2025 13:53:07 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"Wang",
"Yihao",
""
],
[
"Qian",
"Zhong",
""
],
[
"Li",
"Peifeng",
""
]
] |
2504.07804 | Charul Rajput | Charul Rajput, B. Sundar Rajan, Ragnar Freij-Hollanti and Camilla
Hollanti | Function-Correcting Codes for Locally Bounded Functions | The title has been updated | null | null | null | cs.IT math.IT | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | In this paper, we introduce a class of functions that assume only a limited
number $\lambda$ of values within a given Hamming $\rho$-ball and call them
locally $(\rho, \lambda)$-bounded functions. We develop function-correcting
codes (FCCs) for these functions and propose an upper bound on the redundancy
of FCCs. The bound is based on the minimum length of an error-correcting code
with a given number of codewords and a minimum distance. Furthermore, we
provide a sufficient optimality condition for FCCs when $\lambda =4$. We also
demonstrate that any function can be represented as a locally $(\rho,
\lambda)$-bounded function, illustrating this with a representation of Hamming
weight distribution functions. Furthermore, we present another construction of
function-correcting codes for Hamming weight distribution functions.
| [
{
"version": "v1",
"created": "Thu, 10 Apr 2025 14:41:51 GMT"
},
{
"version": "v2",
"created": "Thu, 24 Apr 2025 14:30:44 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"Rajput",
"Charul",
""
],
[
"Rajan",
"B. Sundar",
""
],
[
"Freij-Hollanti",
"Ragnar",
""
],
[
"Hollanti",
"Camilla",
""
]
] |
2504.08469 | Khrystyna Semkiv | Khrystyna Semkiv, Jia Zhang, Maria Laura Ferster, and Walter Karlen | Artifact detection and localization in single-channel mobile EEG for
sleep research using deep learning and attention mechanisms | null | null | null | null | eess.SP cs.LG | http://creativecommons.org/licenses/by-nc-sa/4.0/ | Artifacts in the electroencephalogram (EEG) degrade signal quality and impact
the analysis of brain activity. Current methods for detecting artifacts in
sleep EEG rely on simple threshold-based algorithms that require manual
intervention, which is time-consuming and impractical due to the vast volume of
data that novel mobile recording systems generate. We propose a convolutional
neural network (CNN) model incorporating a convolutional block attention module
(CNN-CBAM) to detect and identify the location of artifacts in the sleep EEG
with attention maps. We benchmarked this model against six other machine
learning and signal processing approaches. We trained/tuned all models on 72
manually annotated EEG recordings obtained during home-based monitoring from 18
healthy participants with a mean (SD) age of 68.05 y ($\pm$5.02). We tested
them on 26 separate recordings from 6 healthy participants with a mean (SD) age
of 68.33 y ($\pm$4.08), with contained artifacts in 4\% of epochs. CNN-CBAM
achieved the highest area under the receiver operating characteristic curve
(0.88), sensitivity (0.81), and specificity (0.86) when compared to the other
approaches. The attention maps from CNN-CBAM localized artifacts within the
epoch with a sensitivity of 0.71 and specificity of 0.67. This work
demonstrates the feasibility of automating the detection and localization of
artifacts in wearable sleep EEG.
| [
{
"version": "v1",
"created": "Fri, 11 Apr 2025 11:57:06 GMT"
},
{
"version": "v2",
"created": "Thu, 24 Apr 2025 13:33:33 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"Semkiv",
"Khrystyna",
""
],
[
"Zhang",
"Jia",
""
],
[
"Ferster",
"Maria Laura",
""
],
[
"Karlen",
"Walter",
""
]
] |
2504.08703 | Muhammad Shihab Rashid | Muhammad Shihab Rashid, Christian Bock, Yuan Zhuang, Alexander
Buchholz, Tim Esler, Simon Valentin, Luca Franceschi, Martin Wistuba, Prabhu
Teja Sivaprasad, Woo Jung Kim, Anoop Deoras, Giovanni Zappella, Laurent
Callot | SWE-PolyBench: A multi-language benchmark for repository level
evaluation of coding agents | 20 pages, 6 figures, corrected author name spelling | null | null | null | cs.SE | http://creativecommons.org/licenses/by-nc-nd/4.0/ | Coding agents powered by large language models have shown impressive
capabilities in software engineering tasks, but evaluating their performance
across diverse programming languages and real-world scenarios remains
challenging. We introduce SWE-PolyBench, a new multi-language benchmark for
repository-level, execution-based evaluation of coding agents. SWE-PolyBench
contains 2110 instances from 21 repositories and includes tasks in Java (165),
JavaScript (1017), TypeScript (729) and Python (199), covering bug fixes,
feature additions, and code refactoring. We provide a task and
repository-stratified subsample (SWE-PolyBench500) and release an evaluation
harness allowing for fully automated evaluation. To enable a more comprehensive
comparison of coding agents, this work also presents a novel set of metrics
rooted in syntax tree analysis. We evaluate leading open source coding agents
on SWE-PolyBench, revealing their strengths and limitations across languages,
task types, and complexity classes. Our experiments show that current agents
exhibit uneven performances across languages and struggle with complex problems
while showing higher performance on simpler tasks. SWE-PolyBench aims to drive
progress in developing more versatile and robust AI coding assistants for
real-world software engineering. Our datasets and code are available at:
https://github.com/amazon-science/SWE-PolyBench
| [
{
"version": "v1",
"created": "Fri, 11 Apr 2025 17:08:02 GMT"
},
{
"version": "v2",
"created": "Mon, 14 Apr 2025 20:52:04 GMT"
},
{
"version": "v3",
"created": "Wed, 23 Apr 2025 20:22:37 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"Rashid",
"Muhammad Shihab",
""
],
[
"Bock",
"Christian",
""
],
[
"Zhuang",
"Yuan",
""
],
[
"Buchholz",
"Alexander",
""
],
[
"Esler",
"Tim",
""
],
[
"Valentin",
"Simon",
""
],
[
"Franceschi",
"Luca",
""
],
[
"Wistuba",
"Martin",
""
],
[
"Sivaprasad",
"Prabhu Teja",
""
],
[
"Kim",
"Woo Jung",
""
],
[
"Deoras",
"Anoop",
""
],
[
"Zappella",
"Giovanni",
""
],
[
"Callot",
"Laurent",
""
]
] |
2504.08824 | Natalia Sikora | Natalia Sikora, Robert L. Manschke, Alethea M. Tang, Peter Dunstan,
Dean A. Harris, Su Yang | ColonScopeX: Leveraging Explainable Expert Systems with Multimodal Data
for Improved Early Diagnosis of Colorectal Cancer | Published to AAAI-25 Bridge Program | null | null | null | cs.LG cs.AI cs.CV cs.HC stat.AP | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Colorectal cancer (CRC) ranks as the second leading cause of cancer-related
deaths and the third most prevalent malignant tumour worldwide. Early detection
of CRC remains problematic due to its non-specific and often embarrassing
symptoms, which patients frequently overlook or hesitate to report to
clinicians. Crucially, the stage at which CRC is diagnosed significantly
impacts survivability, with a survival rate of 80-95\% for Stage I and a stark
decline to 10\% for Stage IV. Unfortunately, in the UK, only 14.4\% of cases
are diagnosed at the earliest stage (Stage I).
In this study, we propose ColonScopeX, a machine learning framework utilizing
explainable AI (XAI) methodologies to enhance the early detection of CRC and
pre-cancerous lesions. Our approach employs a multimodal model that integrates
signals from blood sample measurements, processed using the Savitzky-Golay
algorithm for fingerprint smoothing, alongside comprehensive patient metadata,
including medication history, comorbidities, age, weight, and BMI. By
leveraging XAI techniques, we aim to render the model's decision-making process
transparent and interpretable, thereby fostering greater trust and
understanding in its predictions. The proposed framework could be utilised as a
triage tool or a screening tool of the general population.
This research highlights the potential of combining diverse patient data
sources and explainable machine learning to tackle critical challenges in
medical diagnostics.
| [
{
"version": "v1",
"created": "Wed, 9 Apr 2025 20:45:11 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"Sikora",
"Natalia",
""
],
[
"Manschke",
"Robert L.",
""
],
[
"Tang",
"Alethea M.",
""
],
[
"Dunstan",
"Peter",
""
],
[
"Harris",
"Dean A.",
""
],
[
"Yang",
"Su",
""
]
] |
2504.09602 | Zhen Lu | Zhehao Dong and Zhen Lu and Yue Yang | Fine-tuning a Large Language Model for Automating Computational Fluid
Dynamics Simulations | null | null | 10.1016/j.taml.2025.100594 | null | physics.flu-dyn cs.AI cs.CL | http://creativecommons.org/licenses/by-nc-nd/4.0/ | Configuring computational fluid dynamics (CFD) simulations typically demands
extensive domain expertise, limiting broader access. Although large language
models (LLMs) have advanced scientific computing, their use in automating CFD
workflows is underdeveloped. We introduce a novel approach centered on
domain-specific LLM adaptation. By fine-tuning Qwen2.5-7B-Instruct on NL2FOAM,
our custom dataset of 28716 natural language-to-OpenFOAM configuration pairs
with chain-of-thought (CoT) annotations, we enable direct translation from
natural language descriptions to executable CFD setups. A multi-agent framework
orchestrates the process, autonomously verifying inputs, generating
configurations, running simulations, and correcting errors. Evaluation on a
benchmark of 21 diverse flow cases demonstrates state-of-the-art performance,
achieving 88.7% solution accuracy and 82.6% first-attempt success rate. This
significantly outperforms larger general-purpose models like
Qwen2.5-72B-Instruct, DeepSeek-R1, and Llama3.3-70B-Instruct, while also
requiring fewer correction iterations and maintaining high computational
efficiency. The results highlight the critical role of domain-specific
adaptation in deploying LLM assistants for complex engineering workflows. Our
code and fine-tuned model have been deposited at
https://github.com/YYgroup/AutoCFD.
| [
{
"version": "v1",
"created": "Sun, 13 Apr 2025 14:35:30 GMT"
},
{
"version": "v2",
"created": "Mon, 21 Apr 2025 07:04:57 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"Dong",
"Zhehao",
""
],
[
"Lu",
"Zhen",
""
],
[
"Yang",
"Yue",
""
]
] |
2504.09610 | Sergei Mukhin I | S. I. Mukhin | Q-ball mechanism of electron transport properties of high-T$_c$
superconductors | null | null | null | null | cond-mat.supr-con quant-ph | http://creativecommons.org/licenses/by/4.0/ | Proposed recently by the author Q-ball mechanism of the pseudogap state and
high-Tc superconductivity in cuprates (2022) was supported by micro X-ray
diffraction data in HgBa$_2$CuO$_{4+y}$ (2023). In the present paper it is
demonstrated that T-linear temperature dependence of electrical resistivity
arises naturally in the Q-ball gas phase, that may explain corresponding
experimental data in the "strange metal" phase of high-T$_c$ cuprates, as
reviewed by Barisic et al. (2013). In the present theory it arises due to
scattering of electrons on the Q-balls gas of condensed charge/spin
fluctuations. Close to the lowest temperature boundary of the "strange metal"
phase, at which Q-ball radius diverges, electrical resistivity caused by a
slide of the Q-balls as a whole is calculated using fluctuation
paraconductivity calculation method by Alex Abrikosov (1987). The diamagnetic
response of Q-balls gas is calculated as well and shows good accord with
experimental data by L.Li et al. (2010) in the "strange metal" phase. In total,
obtained results demonstrate different properties of the correlated electrons
systems that arise due to formation of Q-balls possessing internal bosonic
frequency $\Omega=2\pi nT$ in Matsubara time and, thus, forming the quantum
thermodynamic time polycrystals. Presented theory may give a clue concerning a
possible mechanism of the experimentally measured properties of high-T$_c$
cuprates in the "strange metal" phase of their phase diagram. We believe ,
these results provide support to the quantum thermodynamic time crystal model
of the Euclidean Q-balls considered in the present paper.
| [
{
"version": "v1",
"created": "Sun, 13 Apr 2025 15:03:17 GMT"
},
{
"version": "v2",
"created": "Thu, 24 Apr 2025 09:54:19 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"Mukhin",
"S. I.",
""
]
] |
2504.09655 | Justin Kim | Justin Namuk Kim, Yiqiao Liu, Rajath Soans, Keith Persson, Sarah
Halek, Michal Tomaszewski, Jianda Yuan, Gregory Goldmacher, Antong Chen | OmniMamba4D: Spatio-temporal Mamba for longitudinal CT lesion
segmentation | Accepted at IEEE International Symposium on Biomedical Imaging (ISBI)
2025 | null | null | null | eess.IV cs.CV | http://creativecommons.org/publicdomain/zero/1.0/ | Accurate segmentation of longitudinal CT scans is important for monitoring
tumor progression and evaluating treatment responses. However, existing 3D
segmentation models solely focus on spatial information. To address this gap,
we propose OmniMamba4D, a novel segmentation model designed for 4D medical
images (3D images over time). OmniMamba4D utilizes a spatio-temporal
tetra-orientated Mamba block to effectively capture both spatial and temporal
features. Unlike traditional 3D models, which analyze single-time points,
OmniMamba4D processes 4D CT data, providing comprehensive spatio-temporal
information on lesion progression. Evaluated on an internal dataset comprising
of 3,252 CT scans, OmniMamba4D achieves a competitive Dice score of 0.682,
comparable to state-of-the-arts (SOTA) models, while maintaining computational
efficiency and better detecting disappeared lesions. This work demonstrates a
new framework to leverage spatio-temporal information for longitudinal CT
lesion segmentation.
| [
{
"version": "v1",
"created": "Sun, 13 Apr 2025 17:02:54 GMT"
},
{
"version": "v2",
"created": "Thu, 24 Apr 2025 04:25:34 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"Kim",
"Justin Namuk",
""
],
[
"Liu",
"Yiqiao",
""
],
[
"Soans",
"Rajath",
""
],
[
"Persson",
"Keith",
""
],
[
"Halek",
"Sarah",
""
],
[
"Tomaszewski",
"Michal",
""
],
[
"Yuan",
"Jianda",
""
],
[
"Goldmacher",
"Gregory",
""
],
[
"Chen",
"Antong",
""
]
] |
2504.09698 | Tomoi Koide | Tomoi Koide and Armin van de Venn | Beyond Spin: Torsion-Driven Nonlinearity in Spinless Quantum Mechanics | 7 pages, no figure, the abstract was modified | null | null | null | gr-qc hep-th quant-ph | http://creativecommons.org/licenses/by/4.0/ | We investigate the previously unexplored quantum dynamics of
non-relativistic, spinless particles propagating in curved spaces with torsion.
Our findings demonstrate that while torsion has been predominantly associated
with spin, it can also influence the quantum behavior of spinless particles by
inducing a logarithmic nonlinearity in the Schroedinger equation through
quantum fluctuations, even in flat space. To facilitate quantization in curved
spaces, we introduce a novel stochastic variational method. Unlike canonical
quantization, this approach is naturally suited to general coordinate systems,
with quantum fluctuations arising from a noise term in the stochastic process
that is directly influenced by torsion. By requiring consistency with quantum
dynamics, we ultimately derive an upper bound on the magnitude of torsion. Our
results reveal a previously unrecognized mechanism by which torsion, as
predicted in certain extensions of general relativity, can influence quantum
systems, with potential implications for early-universe physics and dark matter
or energy models.
| [
{
"version": "v1",
"created": "Sun, 13 Apr 2025 19:13:35 GMT"
},
{
"version": "v2",
"created": "Thu, 24 Apr 2025 16:26:44 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"Koide",
"Tomoi",
""
],
[
"van de Venn",
"Armin",
""
]
] |
2504.09749 | Daniele Celoria | Agnese Barbensi, Daniele Celoria, Christopher Ktenidis, Kazuhiro
Ichihara, In Dae Jong, Masakazu Teragaito | An exploration of low crossing and chiral cosmetic bands with grid
diagrams | New version includes an appendix by Kazuhiro Ichihara, In Dae Jong
and Masakazu Teragaito | null | null | null | math.GT | http://creativecommons.org/licenses/by/4.0/ | We computationally explore non-coherent band attachments between low crossing
number knots, using grid diagrams. We significantly improve the current
H(2)-distance table. In particular, we find two new distance one pairs with
fewer than seven crossings: one between $3_1\#3_1$ and $7_4m$, and a chirally
cosmetic one for $7_3$. We further determine a total of 33 previously unknown
$H(2)$-distance one pairs for knots with up to $8$ crossings. The appendix by
Kazuhiro Ichihara, In Dae Jong and Masakazu Teragaito contains a construction
explaining the existence of chirally cosmetic bands for an infinite family of
knots, including $5_1,\, 7_3$ and $8_8$.
| [
{
"version": "v1",
"created": "Sun, 13 Apr 2025 22:57:50 GMT"
},
{
"version": "v2",
"created": "Thu, 24 Apr 2025 03:12:31 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"Barbensi",
"Agnese",
""
],
[
"Celoria",
"Daniele",
""
],
[
"Ktenidis",
"Christopher",
""
],
[
"Ichihara",
"Kazuhiro",
""
],
[
"Jong",
"In Dae",
""
],
[
"Teragaito",
"Masakazu",
""
]
] |
2504.09818 | Rong Yao | Rong Yao, Hailin Hu, Yifei Fu, Hanting Chen, Wenyi Fang, Fanyi Du, Kai
Han, Yunhe Wang | Transferable text data distillation by trajectory matching | null | null | null | null | cs.CL | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | In the realm of large language model (LLM), as the size of large models
increases, it also brings higher training costs. There is a urgent need to
minimize the data size in LLM training. Compared with data selection method,
the data distillation method aims to synthesize a small number of data samples
to achieve the training effect of the full data set and has better flexibility.
Despite its successes in computer vision, the discreteness of text data has
hitherto stymied its exploration in natural language processing (NLP). In this
work, we proposed a method that involves learning pseudo prompt data based on
trajectory matching and finding its nearest neighbor ID to achieve
cross-architecture transfer. During the distillation process, we introduce a
regularization loss to improve the robustness of our distilled data. To our
best knowledge, this is the first data distillation work suitable for text
generation tasks such as instruction tuning. Evaluations on two benchmarks,
including ARC-Easy and MMLU instruction tuning datasets, established the
superiority of our distillation approach over the SOTA data selection method
LESS. Furthermore, our method demonstrates a good transferability over LLM
structures (i.e., OPT to Llama).
| [
{
"version": "v1",
"created": "Mon, 14 Apr 2025 02:39:26 GMT"
},
{
"version": "v2",
"created": "Thu, 24 Apr 2025 12:46:05 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"Yao",
"Rong",
""
],
[
"Hu",
"Hailin",
""
],
[
"Fu",
"Yifei",
""
],
[
"Chen",
"Hanting",
""
],
[
"Fang",
"Wenyi",
""
],
[
"Du",
"Fanyi",
""
],
[
"Han",
"Kai",
""
],
[
"Wang",
"Yunhe",
""
]
] |
2504.09924 | Florian Euchner | Florian Euchner, David Kellner, Phillip Stephan, Stephan ten Brink | Passive Channel Charting: Locating Passive Targets using Wi-Fi Channel
State Information | null | null | null | null | cs.IT eess.SP math.IT | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We propose passive channel charting, an extension of channel charting to
passive target localization. As in conventional channel charting, we follow a
dimensionality reduction approach to reconstruct a physically interpretable map
of target positions from similarities in high-dimensional channel state
information. We show that algorithms and neural network architectures developed
in the context of channel charting with active mobile transmitters can be
straightforwardly applied to the passive case, where we assume a scenario with
static transmitters and receivers and a mobile target. We evaluate our method
on a channel state information dataset collected indoors with a distributed
setup of ESPARGOS Wi-Fi sensing antenna arrays. This scenario can be
interpreted as either a multi-static or passive radar system. We demonstrate
that passive channel charting outperforms a baseline based on classical
triangulation in terms of localization accuracy. We discuss our results and
highlight some unsolved issues related to the proposed concept.
| [
{
"version": "v1",
"created": "Mon, 14 Apr 2025 06:33:07 GMT"
},
{
"version": "v2",
"created": "Thu, 24 Apr 2025 15:28:39 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"Euchner",
"Florian",
""
],
[
"Kellner",
"David",
""
],
[
"Stephan",
"Phillip",
""
],
[
"Brink",
"Stephan ten",
""
]
] |
2504.10128 | Anna Moreso Serra | Anna Moreso Serra and Oleg Bulashenko | Probing Binary Lens Caustics with Gravitational Waves: A Uniform
Approximation Approach | 14 pages, 8 figures, updated references | null | null | VIR-0438A-25 | gr-qc astro-ph.CO astro-ph.GA astro-ph.HE | http://creativecommons.org/licenses/by/4.0/ | We present a new framework for modeling gravitational wave diffraction near
fold caustics using the Uniform Approximation (UA), focusing on binary mass
lenses - axially asymmetric systems with complex caustic structures. Full-wave
methods based on the Kirchhoff integral become impractical in this regime due
to highly oscillatory integrands. The UA provides a robust and accurate
description of the wave field near folds, resolving the breakdown of
Geometrical Optics at caustics and improving upon Transitional Asymptotics -
based on Airy function approximations - which lack global validity. Central to
our approach is the concept of the caustic width, $d_c$, a characteristic
length scale defining the region where diffraction significantly alters wave
propagation. We find that $d_c$ scales universally with the gravitational
wavelength as ~ $ \lambda^{2/3}$ and inversely with the redshifted lens mass as
~ $ M_{Lz}^{-2/3}$. The wave amplification near the fold grows as ~ $
d_c^{-1/4}$, substantially enhancing the signal and potentially playing a key
role in the detection of gravitational waves lensed near caustics. Notably, for
lens masses below the galactic scale, the caustic width for gravitational waves
is not negligible compared to the Einstein radius - as it is in electromagnetic
lensing - making the UA essential for accurately capturing wave effects.
| [
{
"version": "v1",
"created": "Mon, 14 Apr 2025 11:36:36 GMT"
},
{
"version": "v2",
"created": "Thu, 24 Apr 2025 09:42:05 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"Serra",
"Anna Moreso",
""
],
[
"Bulashenko",
"Oleg",
""
]
] |
2504.10323 | Wim Martens | Molham Aref and Paolo Guagliardo and George Kastrinis and Leonid
Libkin and Victor Marsault and Wim Martens and Mary McGrath and Filip Murlak
and Nathaniel Nystrom and Liat Peterfreund and Allison Rogers and Cristina
Sirangelo and Domagoj Vrgoc and David Zhao and Abdul Zreika | Rel: A Programming Language for Relational Data | null | null | 10.1145/3722212.3724450 | null | cs.DB cs.PL | http://creativecommons.org/licenses/by/4.0/ | From the moment of their inception, languages for relational data have been
described as sublanguages embedded in a host programming language. Rel is a new
relational language whose key design goal is to go beyond this paradigm with
features that allow for programming in the large, making it possible to fully
describe end to end application semantics. With the new approach we can model
the semantics of entire enterprise applications relationally, which helps
significantly reduce architecture complexity and avoid the well-known impedance
mismatch problem. This paradigm shift is enabled by 50 years of database
research, making it possible to revisit the sublanguage/host language paradigm,
starting from the fundamental principles. We present the main features of Rel:
those that give it the power to express traditional query language operations
and those that are designed to grow the language and allow programming in the
large.
| [
{
"version": "v1",
"created": "Mon, 14 Apr 2025 15:32:47 GMT"
},
{
"version": "v2",
"created": "Thu, 24 Apr 2025 14:33:37 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"Aref",
"Molham",
""
],
[
"Guagliardo",
"Paolo",
""
],
[
"Kastrinis",
"George",
""
],
[
"Libkin",
"Leonid",
""
],
[
"Marsault",
"Victor",
""
],
[
"Martens",
"Wim",
""
],
[
"McGrath",
"Mary",
""
],
[
"Murlak",
"Filip",
""
],
[
"Nystrom",
"Nathaniel",
""
],
[
"Peterfreund",
"Liat",
""
],
[
"Rogers",
"Allison",
""
],
[
"Sirangelo",
"Cristina",
""
],
[
"Vrgoc",
"Domagoj",
""
],
[
"Zhao",
"David",
""
],
[
"Zreika",
"Abdul",
""
]
] |
2504.10528 | Usman Zafar | Hamza Tariq, Usman Zafar, Shahid Chaudhary, Kazuharu Bamba, Abdul
Jawad, Sanjar Shaymatov | Exploring the Effects of Generalized Entropy onto Bardeen Black Hole
Surrounded by Cloud of Strings | 15 pages, 7 figures, version published in Nuclear physics B | null | 10.1016/j.nuclphysb.2025.116906 | FU-PCG-147 | gr-qc hep-th | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | This work explores the thermodynamic characteristics and geothermodynamics of
a Bardeen black hole (BH) that interacts with a string cloud and is minimally
connected to nonlinear electrodynamics. To avoid the singularities throughout
the cosmic evolution, we consider an entropy function which comprises five
parameters. In addition, by employing this entropy function for the specific
range of parameters, we obtain the representations of BH entropy based on the
holographic principle. Moreover, we employ this entropy function to investigate
its impact on the thermodynamics of the BH by studying various thermodynamic
properties like mass, temperature, heat capacity, and Gibbs free energy for
numerous scalar charge and string cloud values. To support our investigation,
we use various geothermodynamics formalisms to evaluate the stable behavior and
identify different physical scenarios. Furthermore, in this analysis, we
observe that only one entropy formalism provides us with better results
regarding the thermodynamic behavior of the BH. Moreover, it is shown that one
of the entropy models provides a thermodynamic geometric behavior compared to
the other entropy models.
| [
{
"version": "v1",
"created": "Sat, 12 Apr 2025 12:00:22 GMT"
},
{
"version": "v2",
"created": "Thu, 24 Apr 2025 02:53:58 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"Tariq",
"Hamza",
""
],
[
"Zafar",
"Usman",
""
],
[
"Chaudhary",
"Shahid",
""
],
[
"Bamba",
"Kazuharu",
""
],
[
"Jawad",
"Abdul",
""
],
[
"Shaymatov",
"Sanjar",
""
]
] |
2504.10814 | David P\'erez Pi\~neiro | Eric Luxenberg, David P\'erez-Pi\~neiro, Steven Diamond, and Stephen
Boyd | An Operator Splitting Method for Large-Scale CVaR-Constrained Quadratic
Programs | null | null | null | null | math.OC | http://creativecommons.org/licenses/by/4.0/ | We introduce a fast and scalable method for solving quadratic programs with
conditional value-at-risk (CVaR) constraints. While these problems can be
formulated as standard quadratic programs, the number of variables and
constraints grows linearly with the number of scenarios, making general-purpose
solvers impractical for large-scale problems. Our method combines operator
splitting with a specialized $O(m\log m)$ algorithm for projecting onto CVaR
constraints, where $m$ is the number of scenarios. The method alternates
between solving a linear system and performing parallel projections: onto CVaR
constraints using our specialized algorithm and onto box constraints with a
closed-form solution. Numerical examples from several application domains
demonstrate that our method outperforms general-purpose solvers by several
orders of magnitude on problems with up to millions of scenarios. Our method is
implemented in an open-source package called CVQP.
| [
{
"version": "v1",
"created": "Tue, 15 Apr 2025 02:28:55 GMT"
},
{
"version": "v2",
"created": "Thu, 24 Apr 2025 04:20:47 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"Luxenberg",
"Eric",
""
],
[
"Pérez-Piñeiro",
"David",
""
],
[
"Diamond",
"Steven",
""
],
[
"Boyd",
"Stephen",
""
]
] |
2504.10847 | Saikat Das | Saikat Das, Bing Zhang, Soebur Razzaque, Siyao Xu | Cosmic-Ray Constraints on the Flux of Ultra-High-Energy Neutrino Event
KM3-230213A | 7 pages, 4 figures; comments are welcome! | null | null | null | astro-ph.HE | http://creativecommons.org/licenses/by/4.0/ | The detection of a $\simeq220$~PeV muon neutrino by the KM3NeT neutrino
telescope offers an unprecedented opportunity to probe the Universe at extreme
energies. We analyze the origin of this event under three scenarios, viz., a
transient point source, a diffuse astrophysical emission, and line-of-sight
interaction of ultrahigh-energy cosmic rays (UHECR; $E \gtrsim 0.1$~EeV). Our
analysis includes the flux from both a KM3NeT-only fit and a joint fit,
incorporating data from KM3NeT, IceCube, and Pierre Auger Observatory. If the
neutrino event originates from transients, it requires a new population of
transient that is energetic, gamma-ray dark, and more abundant than known ones.
In the framework of diffuse astrophysical emission, we compare the required
local UHECR energy injection rate at $\gtrsim4$ EeV, assuming a proton primary,
with the rate derived from the flux measurements by Auger. This disfavors the
KM3NeT-only fit at all redshifts, while the joint fit remains viable for
$z\gtrsim 1$, based on redshift evolution models of known source populations.
For cosmogenic origin from point sources, our results suggest that the
luminosity obtained at redshifts $z \lesssim 1$ from the joint fit is
compatible with the Eddington luminosity of supermassive black holes in active
galactic nuclei.
| [
{
"version": "v1",
"created": "Tue, 15 Apr 2025 04:07:13 GMT"
},
{
"version": "v2",
"created": "Wed, 23 Apr 2025 20:39:29 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"Das",
"Saikat",
""
],
[
"Zhang",
"Bing",
""
],
[
"Razzaque",
"Soebur",
""
],
[
"Xu",
"Siyao",
""
]
] |
2504.11029 | Vittorio De Falco Dr | Vittorio De Falco, Marco Gallo | Analytical coordinate time at the second post-Newtonian order | 5 pages; 1 figure; 24 references; letter accepted for publication on
Physics Letters B | null | 10.1016/j.physletb.2025.139484 | null | gr-qc astro-ph.HE hep-th math-ph math.MP | http://creativecommons.org/licenses/by/4.0/ | We derive the analytical expression of the coordinate time $t$ in terms of
the eccentric anomaly $u$ at the second post-Newtonian order in General
Relativity for a compact binary system moving on eccentric orbits. The
parametrization of $t$ with $u$ permits to reduce at the minimum the presence
of discontinuous trigonometric functions. This is helpful as they must be
properly connected via accumulation functions to finally have a smooth
coordinate time $t(u)$. Another difficulty relies on the presence of an
infinite sum, about which we derive a compact form. This effort reveals to be
extremely useful for application purposes. Indeed, we need to truncate the
aforementioned sum to a certain finite threshold, which strongly depends on the
selected parameter values and the accuracy error we would like to achieve.
Thanks to our work, this analysis can be easily carried out.
| [
{
"version": "v1",
"created": "Tue, 15 Apr 2025 09:53:26 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"De Falco",
"Vittorio",
""
],
[
"Gallo",
"Marco",
""
]
] |
2504.11219 | Hua-Xing Chen | Xuan Luo, Yi-Jie Wang and Hua-Xing Chen | $P$-wave single charmed baryons of the $SU(3)$ flavor $\bf\bar3_F$ | 16 pages, 3 figures, 5 tables, revised version | null | null | null | hep-ph hep-ex | http://creativecommons.org/licenses/by/4.0/ | We study the $P$-wave single charmed baryons of the $SU(3)$ flavor
$\bf\bar3_F$ within the framework of heavy quark effective theory. We
systematically calculate their strong and radiative decay properties using the
light-cone sum rule method. Besides the $\Lambda_c(2595)$, $\Lambda_c(2625)$,
$\Xi_c(2790)$, and $\Xi_c(2815)$, our results suggest the existence of two
additional $\Lambda_c$ baryons and two additional $\Xi_c$ baryons. Their
masses, mass splittings within the same multiplets, and decay properties are
summarized in Table V for future experimental searches.
| [
{
"version": "v1",
"created": "Tue, 15 Apr 2025 14:22:57 GMT"
},
{
"version": "v2",
"created": "Thu, 24 Apr 2025 11:30:05 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"Luo",
"Xuan",
""
],
[
"Wang",
"Yi-Jie",
""
],
[
"Chen",
"Hua-Xing",
""
]
] |
2504.11336 | Abitha Thankaraj | Abitha Thankaraj, Yiding Jiang, J. Zico Kolter and Yonatan Bisk | Looking beyond the next token | null | null | null | null | cs.LG cs.AI cs.CL | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | The structure of causal language model training assumes that each token can
be accurately predicted from the previous context. This contrasts with humans'
natural writing and reasoning process, where goals are typically known before
the exact argument or phrasings. While this mismatch has been well studied in
the literature, the working assumption has been that architectural changes are
needed to address this mismatch. We argue that rearranging and processing the
training data sequences can allow models to more accurately imitate the true
data-generating process, and does not require any other changes to the
architecture or training infrastructure. We demonstrate that this technique,
Trelawney, and the inference algorithms derived from it allow us to improve
performance on several key benchmarks that span planning, algorithmic
reasoning, and story generation tasks. Finally, our method naturally enables
the generation of long-term goals at no additional cost. We investigate how
using the model's goal-generation capability can further improve planning and
reasoning. Additionally, we believe Trelawney could potentially open doors to
new capabilities beyond the current language modeling paradigm.
| [
{
"version": "v1",
"created": "Tue, 15 Apr 2025 16:09:06 GMT"
},
{
"version": "v2",
"created": "Thu, 24 Apr 2025 03:13:28 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"Thankaraj",
"Abitha",
""
],
[
"Jiang",
"Yiding",
""
],
[
"Kolter",
"J. Zico",
""
],
[
"Bisk",
"Yonatan",
""
]
] |
2504.11341 | Giuseppe Destefanis | Silvio Meneguzzo, Claudio Schifanella, Valentina Gatteschi, Giuseppe
Destefanis | Evaluating DAO Sustainability and Longevity Through On-Chain Governance
Metrics | null | null | null | null | cs.CY cs.ET cs.SI | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Decentralised Autonomous Organisations (DAOs) automate governance and
resource allocation through smart contracts, aiming to shift decision-making to
distributed token holders. However, many DAOs face sustainability challenges
linked to limited user participation, concentrated voting power, and technical
design constraints. This paper addresses these issues by identifying research
gaps in DAO evaluation and introducing a framework of Key Performance
Indicators (KPIs) that capture governance efficiency, financial robustness,
decentralisation, and community engagement. We apply the framework to a
custom-built dataset of real-world DAOs constructed from on-chain data and
analysed using non-parametric methods. The results reveal recurring governance
patterns, including low participation rates and high proposer concentration,
which may undermine long-term viability. The proposed KPIs offer a replicable,
data-driven method for assessing DAO governance structures and identifying
potential areas for improvement. These findings support a multidimensional
approach to evaluating decentralised systems and provide practical tools for
researchers and practitioners working to improve the resilience and
effectiveness of DAO-based governance models.
| [
{
"version": "v1",
"created": "Tue, 15 Apr 2025 16:13:20 GMT"
},
{
"version": "v2",
"created": "Thu, 24 Apr 2025 13:26:16 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"Meneguzzo",
"Silvio",
""
],
[
"Schifanella",
"Claudio",
""
],
[
"Gatteschi",
"Valentina",
""
],
[
"Destefanis",
"Giuseppe",
""
]
] |
2504.11364 | Tianwei Ni | Tianwei Ni, Allen Nie, Sapana Chaudhary, Yao Liu, Huzefa Rangwala,
Rasool Fakoor | Teaching Large Language Models to Reason through Learning and Forgetting | Code: https://github.com/twni2016/llm-reasoning-uft | null | null | null | cs.LG cs.AI cs.CL | http://creativecommons.org/licenses/by/4.0/ | Leveraging inference-time search in large language models has proven
effective in further enhancing a trained model's capability to solve complex
mathematical and reasoning problems. However, this approach significantly
increases computational costs and inference time, as the model must generate
and evaluate multiple candidate solutions to identify a viable reasoning path.
To address this, we propose an effective approach that integrates search
capabilities directly into the model by fine-tuning it using both successful
(learning) and failed reasoning paths (forgetting) derived from diverse search
methods. While fine-tuning the model with these data might seem
straightforward, we identify a critical issue: the model's search capability
tends to degrade rapidly if fine-tuning is performed naively. We show that this
degradation can be substantially mitigated by employing a smaller learning
rate. Extensive experiments on the challenging Game-of-24 and Countdown
mathematical reasoning benchmarks show that our approach not only outperforms
both standard fine-tuning and inference-time search baselines but also
significantly reduces inference time by 180$\times$.
| [
{
"version": "v1",
"created": "Tue, 15 Apr 2025 16:30:02 GMT"
},
{
"version": "v2",
"created": "Wed, 23 Apr 2025 21:27:59 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"Ni",
"Tianwei",
""
],
[
"Nie",
"Allen",
""
],
[
"Chaudhary",
"Sapana",
""
],
[
"Liu",
"Yao",
""
],
[
"Rangwala",
"Huzefa",
""
],
[
"Fakoor",
"Rasool",
""
]
] |
2504.12068 | Philip D. Mannheim | Philip D. Mannheim | Time Advance and Probability Conservation in $PT$-Symmetric Quantum
Mechanics | 5 pages, revtex4. Emphasis now on probability conservation for
non-Hermitian Hamiltonians. Change in title of paper | null | null | null | quant-ph hep-th | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | When excited states decay the time evolution operator $U(t)=e^{-iHt}$ does
not obey $U^{\dagger}(t)U(t)=I$. Nonetheless, probability conservation is not
lost if one includes both excitation and decay, though it takes a different
form. Specifically, if the eigenspectrum of a Hamiltonian is complete, then due
to $CPT$ symmetry, a symmetry that holds for all physical systems, there must
exist an operator $V$ that effects $VHV^{-1}=H^{\dagger}$, so that
$V^{-1}U^{\dagger}(t)VU(t)=I$. In consequence, the time delay associated with
decay must be accompanied by an equal and opposite time advance for excitation.
Thus when a photon excites an atom the spontaneous emission of a photon from
the excited state must occur without any decay time delay at all.
| [
{
"version": "v1",
"created": "Wed, 16 Apr 2025 13:24:07 GMT"
},
{
"version": "v2",
"created": "Thu, 24 Apr 2025 17:51:37 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"Mannheim",
"Philip D.",
""
]
] |
2504.12176 | Harold Berjamin | Harold Berjamin and Stephan Rudykh | Elastic wave propagation in magneto-active fibre composites | null | null | 10.1016/j.ijsolstr.2025.113373 | null | cond-mat.soft physics.app-ph | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Fibre-reinforced elastomers are lightweight and strong materials that can
sustain large deformations. When filled with magnetic particles, their
effective mechanical response can be modified by an external magnetic field. In
the present study, we propose an effective theory of fibre-reinforced
composite, based on a neo-Hookean elastic response and a linear magnetic law in
each phase. The theory is shown suitable to describe the motion of composite
cylinders. Furthermore, it is found appropriate for the modelling of
fibre-reinforced composites subjected to a permanent magnetic field aligned
with the fibres. To reach this result, we use the incremental theory ('small on
large'), in combination with homogenisation theory and the Bloch-Floquet
method. This way, we show that wave directivity is sensitive to the application
of a permanent magnetic field, whereas the frequency range in which wave
propagation is forbidden is not modified by such a load (the band gaps are
invariant). In passing, we describe a method to deduce the total stress in the
material based on the measurement of two wave speeds. Furthermore, we propose
an effective energy function for the description of nonlinear composites made
of Yeoh-type generalised neo-Hookean fibres within a neo-Hookean matrix.
| [
{
"version": "v1",
"created": "Wed, 16 Apr 2025 15:26:24 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"Berjamin",
"Harold",
""
],
[
"Rudykh",
"Stephan",
""
]
] |
2504.12352 | Ruijie Wang | Ruijie Wang, Luca Rossetto, Susan M\'erillat, Christina R\"ocke, Mike
Martin, Abraham Bernstein | Deep Generative Model-Based Generation of Synthetic Individual-Specific
Brain MRI Segmentations | null | null | null | null | q-bio.NC cs.AI eess.IV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | To the best of our knowledge, all existing methods that can generate
synthetic brain magnetic resonance imaging (MRI) scans for a specific
individual require detailed structural or volumetric information about the
individual's brain. However, such brain information is often scarce, expensive,
and difficult to obtain. In this paper, we propose the first approach capable
of generating synthetic brain MRI segmentations -- specifically, 3D white
matter (WM), gray matter (GM), and cerebrospinal fluid (CSF) segmentations --
for individuals using their easily obtainable and often readily available
demographic, interview, and cognitive test information. Our approach features a
novel deep generative model, CSegSynth, which outperforms existing prominent
generative models, including conditional variational autoencoder (C-VAE),
conditional generative adversarial network (C-GAN), and conditional latent
diffusion model (C-LDM). We demonstrate the high quality of our synthetic
segmentations through extensive evaluations. Also, in assessing the
effectiveness of the individual-specific generation, we achieve superior volume
prediction, with mean absolute errors of only 36.44mL, 29.20mL, and 35.51mL
between the ground-truth WM, GM, and CSF volumes of test individuals and those
volumes predicted based on generated individual-specific segmentations,
respectively.
| [
{
"version": "v1",
"created": "Tue, 15 Apr 2025 21:25:36 GMT"
},
{
"version": "v2",
"created": "Wed, 23 Apr 2025 18:53:03 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"Wang",
"Ruijie",
""
],
[
"Rossetto",
"Luca",
""
],
[
"Mérillat",
"Susan",
""
],
[
"Röcke",
"Christina",
""
],
[
"Martin",
"Mike",
""
],
[
"Bernstein",
"Abraham",
""
]
] |
2504.12597 | Liangyu Xu | Liangyu Xu, Yingxiu Zhao, Jingyun Wang, Yingyao Wang, Bu Pi, Chen
Wang, Mingliang Zhang, Jihao Gu, Xiang Li, Xiaoyong Zhu, Jun Song, Bo Zheng | GeoSense: Evaluating Identification and Application of Geometric
Principles in Multimodal Reasoning | 10 pages, 8 figures | null | null | null | cs.CL | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Geometry problem-solving (GPS), a challenging task requiring both visual
comprehension and symbolic reasoning, effectively measures the reasoning
capabilities of multimodal large language models (MLLMs). Humans exhibit strong
reasoning ability in this task through accurate identification and adaptive
application of geometric principles within visual contexts. However, existing
benchmarks fail to jointly assess both dimensions of the human-like geometric
reasoning mechanism in MLLMs, remaining a critical gap in assessing their
ability to tackle GPS. To this end, we introduce GeoSense, the first
comprehensive bilingual benchmark designed to systematically evaluate the
geometric reasoning abilities of MLLMs through the lens of geometric
principles. GeoSense features a five-level hierarchical framework of geometric
principles spanning plane and solid geometry, an intricately annotated dataset
of 1,789 problems, and an innovative evaluation strategy. Through extensive
experiments on GeoSense with various open-source and closed-source MLLMs, we
observe that Gemini-2.0-pro-flash performs best, achieving an overall score of
$65.3$. Our in-depth analysis reveals that the identification and application
of geometric principles remain a bottleneck for leading MLLMs, jointly
hindering their reasoning abilities. These findings underscore GeoSense's
potential to guide future advancements in MLLMs' geometric reasoning
capabilities, paving the way for more robust and human-like reasoning in
artificial intelligence.
| [
{
"version": "v1",
"created": "Thu, 17 Apr 2025 02:46:27 GMT"
},
{
"version": "v2",
"created": "Thu, 24 Apr 2025 02:21:05 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"Xu",
"Liangyu",
""
],
[
"Zhao",
"Yingxiu",
""
],
[
"Wang",
"Jingyun",
""
],
[
"Wang",
"Yingyao",
""
],
[
"Pi",
"Bu",
""
],
[
"Wang",
"Chen",
""
],
[
"Zhang",
"Mingliang",
""
],
[
"Gu",
"Jihao",
""
],
[
"Li",
"Xiang",
""
],
[
"Zhu",
"Xiaoyong",
""
],
[
"Song",
"Jun",
""
],
[
"Zheng",
"Bo",
""
]
] |
2504.12745 | Samuel Skirvin | Samuel Skirvin, Viktor Fedun, Gary Verth, Istvan Ballai | 3D MHD wave propagation and energy transport in a simulated solar vortex | 19 pages, 12 figures, accepted for publication in ApJ | null | null | null | astro-ph.SR | http://creativecommons.org/licenses/by/4.0/ | Magnetic flux tubes in the presence of background rotational flows are
abundant throughout the solar atmosphere and may act as conduits for MHD waves
to transport energy throughout the solar atmosphere. Here we investigate the
contribution from MHD waves to the Poynting flux in a 3D numerical simulation
of a realistic solar atmosphere, modelling a structure resembling a solar
vortex tube, using the PLUTO code in the presence of different plasma flow
configurations. These simulations feature a closed magnetic loop system where a
rotational flow is imposed at one foot-point in addition to photospheric
perturbations acting as a wave driver mimicking those of p-modes. We find that
a variety of MHD waves exist within the vortex tube, including sausage, kink
and torsional Alfv\'{e}n waves, owing to the photospheric wave driver and the
nature of the rotational flow itself. We demonstrate how the visual
interpretation of different MHD modes becomes non-trivial when a background
rotational flow is present compared to a static flux tube. By conducting a
simulation both with and without the rotational plasma flow, we demonstrate how
the perturbed Poynting flux increases in the presence of the rotational flow as
the waves transport increased magnetic energy. We attribute this increase to
the dynamical pressure from the rotational flow increasing the plasma density
at the tube boundary, which acts to trap the wave energy more effectively
inside the vortex. Moreover, we demonstrate how the Poynting flux is always
directed upwards in weakly twisted magnetic flux tubes.
| [
{
"version": "v1",
"created": "Thu, 17 Apr 2025 08:36:31 GMT"
},
{
"version": "v2",
"created": "Thu, 24 Apr 2025 14:54:12 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"Skirvin",
"Samuel",
""
],
[
"Fedun",
"Viktor",
""
],
[
"Verth",
"Gary",
""
],
[
"Ballai",
"Istvan",
""
]
] |
2504.13192 | Liangbo Ning | Liang-bo Ning, Shijie Wang, Wenqi Fan, Qing Li, Xin Xu, Hao Chen,
Feiran Huang | CheatAgent: Attacking LLM-Empowered Recommender Systems via LLM Agent | Accepted by KDD 2024; | null | 10.1145/3637528.3671837 | null | cs.CR cs.AI | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Recently, Large Language Model (LLM)-empowered recommender systems (RecSys)
have brought significant advances in personalized user experience and have
attracted considerable attention. Despite the impressive progress, the research
question regarding the safety vulnerability of LLM-empowered RecSys still
remains largely under-investigated. Given the security and privacy concerns, it
is more practical to focus on attacking the black-box RecSys, where attackers
can only observe the system's inputs and outputs. However, traditional attack
approaches employing reinforcement learning (RL) agents are not effective for
attacking LLM-empowered RecSys due to the limited capabilities in processing
complex textual inputs, planning, and reasoning. On the other hand, LLMs
provide unprecedented opportunities to serve as attack agents to attack RecSys
because of their impressive capability in simulating human-like decision-making
processes. Therefore, in this paper, we propose a novel attack framework called
CheatAgent by harnessing the human-like capabilities of LLMs, where an
LLM-based agent is developed to attack LLM-Empowered RecSys. Specifically, our
method first identifies the insertion position for maximum impact with minimal
input modification. After that, the LLM agent is designed to generate
adversarial perturbations to insert at target positions. To further improve the
quality of generated perturbations, we utilize the prompt tuning technique to
improve attacking strategies via feedback from the victim RecSys iteratively.
Extensive experiments across three real-world datasets demonstrate the
effectiveness of our proposed attacking method.
| [
{
"version": "v1",
"created": "Sun, 13 Apr 2025 05:31:37 GMT"
},
{
"version": "v2",
"created": "Thu, 24 Apr 2025 02:16:04 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"Ning",
"Liang-bo",
""
],
[
"Wang",
"Shijie",
""
],
[
"Fan",
"Wenqi",
""
],
[
"Li",
"Qing",
""
],
[
"Xu",
"Xin",
""
],
[
"Chen",
"Hao",
""
],
[
"Huang",
"Feiran",
""
]
] |
2504.13340 | Oliver Mills | Oliver Mills, Philip Conaghan, Nishant Ravikumar, Samuel Relton | Putting the Segment Anything Model to the Test with 3D Knee MRI - A
Comparison with State-of-the-Art Performance | Work accepted at BMVC 2024. Minor changes to the camera-ready version
since acceptance include a corrected running header and the addition of an
Acknowledgments section (including code availability) | null | null | null | eess.IV cs.AI cs.CV | http://creativecommons.org/licenses/by/4.0/ | Menisci are cartilaginous tissue found within the knee that contribute to
joint lubrication and weight dispersal. Damage to menisci can lead to onset and
progression of knee osteoarthritis (OA), a condition that is a leading cause of
disability, and for which there are few effective therapies. Accurate automated
segmentation of menisci would allow for earlier detection and treatment of
meniscal abnormalities, as well as shedding more light on the role the menisci
play in OA pathogenesis. Focus in this area has mainly used variants of
convolutional networks, but there has been no attempt to utilise recent large
vision transformer segmentation models. The Segment Anything Model (SAM) is a
so-called foundation segmentation model, which has been found useful across a
range of different tasks due to the large volume of data used for training the
model. In this study, SAM was adapted to perform fully-automated segmentation
of menisci from 3D knee magnetic resonance images. A 3D U-Net was also trained
as a baseline. It was found that, when fine-tuning only the decoder, SAM was
unable to compete with 3D U-Net, achieving a Dice score of $0.81\pm0.03$,
compared to $0.87\pm0.03$, on a held-out test set. When fine-tuning SAM
end-to-end, a Dice score of $0.87\pm0.03$ was achieved. The performance of both
the end-to-end trained SAM configuration and the 3D U-Net were comparable to
the winning Dice score ($0.88\pm0.03$) in the IWOAI Knee MRI Segmentation
Challenge 2019. Performance in terms of the Hausdorff Distance showed that both
configurations of SAM were inferior to 3D U-Net in matching the meniscus
morphology. Results demonstrated that, despite its generalisability, SAM was
unable to outperform a basic 3D U-Net in meniscus segmentation, and may not be
suitable for similar 3D medical image segmentation tasks also involving fine
anatomical structures with low contrast and poorly-defined boundaries.
| [
{
"version": "v1",
"created": "Thu, 17 Apr 2025 21:18:58 GMT"
},
{
"version": "v2",
"created": "Wed, 23 Apr 2025 10:50:09 GMT"
},
{
"version": "v3",
"created": "Thu, 24 Apr 2025 14:18:18 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"Mills",
"Oliver",
""
],
[
"Conaghan",
"Philip",
""
],
[
"Ravikumar",
"Nishant",
""
],
[
"Relton",
"Samuel",
""
]
] |
2504.13471 | Jiliang Ni | Jiliang Ni, Jiachen Pu, Zhongyi Yang, Kun Zhou, Hui Wang, Xiaoliang
Xiao, Dakui Wang, Xin Li, Jingfeng Luo, Conggang Hu | From Large to Super-Tiny: End-to-End Optimization for Cost-Efficient
LLMs | null | null | null | null | cs.CL | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | In recent years, Large Language Models (LLMs) have significantly advanced
artificial intelligence by optimizing traditional Natural Language Processing
(NLP) pipelines, improving performance and generalization. This has spurred
their integration into various systems. Many NLP systems, including ours,
employ a "one-stage" pipeline directly incorporating LLMs. While effective,
this approach incurs substantial costs and latency due to the need for large
model parameters to achieve satisfactory outcomes. This paper introduces a
three-stage cost-efficient end-to-end LLM deployment pipeline-including
prototyping, knowledge transfer, and model compression-to tackle the
cost-performance dilemma in LLM-based frameworks. Our approach yields a super
tiny model optimized for cost and performance in online systems, simplifying
the system architecture. Initially, by transforming complex tasks into a
function call-based LLM-driven pipeline, an optimal performance prototype
system is constructed to produce high-quality data as a teacher model. The
second stage combines techniques like rejection fine-tuning, reinforcement
learning, and knowledge distillation to transfer knowledge to a smaller 0.5B
student model, delivering effective performance at minimal cost. The final
stage applies quantization and pruning to extremely compress models to 0.4B,
achieving ultra-low latency and cost. The framework's modular design and
cross-domain capabilities suggest potential applicability in other NLP areas.
| [
{
"version": "v1",
"created": "Fri, 18 Apr 2025 05:25:22 GMT"
},
{
"version": "v2",
"created": "Thu, 24 Apr 2025 07:30:24 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"Ni",
"Jiliang",
""
],
[
"Pu",
"Jiachen",
""
],
[
"Yang",
"Zhongyi",
""
],
[
"Zhou",
"Kun",
""
],
[
"Wang",
"Hui",
""
],
[
"Xiao",
"Xiaoliang",
""
],
[
"Wang",
"Dakui",
""
],
[
"Li",
"Xin",
""
],
[
"Luo",
"Jingfeng",
""
],
[
"Hu",
"Conggang",
""
]
] |
2504.13611 | Tobias Dornheim | Sebastian Schwalbe and Hannah Bellenbaum and Tilo D\"oppner and
Maximilian B\"ohme and Thomas Gawne and Dominik Kraus and Michael J.
MacDonald and Zhandos Moldabekov and Panagiotis Tolias and Jan Vorberger and
Tobias Dornheim | Static linear density response from X-ray Thomson scattering
measurements: a case study of warm dense beryllium | null | null | null | null | physics.plasm-ph physics.chem-ph | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Linear response theory is ubiquitous throughout physics and plays a central
role in the theoretical description of warm dense matter -- an extreme state
that occurs within compact astrophysical objects and that is traversed on the
compression path of a fuel capsule in inertial confinement fusion applications.
Here we show how one can relate the static linear density response function to
X-ray Thomson scattering (XRTS) measurements, which opens up new possibilities
for the diagnostics of extreme states of matter, and for the rigorous
assessment and verification of theoretical models and approximations. As a
practical example, we consider an XRTS data set of warm dense beryllium taken
at the National Ignition Facility [T.~D\"oppner \emph{et al.}, \textit{Nature}
\textbf{618}, 270-275 (2023)]. The comparison with state-of-the-art \emph{ab
initio} path integral Monte Carlo (PIMC) simulations [T.~Dornheim \emph{et
al.}, \textit{Nature Commun.}~(in print), arXiv:2402.19113] gives us a best
estimate of the mass density of $\rho=18\pm6\,$g/cc, which is consistent with
previous PIMC and density functional theory based studies, but rules out the
original estimate of $\rho=34\pm4\,$g/cc based on a Chihara model fit.
| [
{
"version": "v1",
"created": "Fri, 18 Apr 2025 10:32:35 GMT"
},
{
"version": "v2",
"created": "Thu, 24 Apr 2025 12:14:15 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"Schwalbe",
"Sebastian",
""
],
[
"Bellenbaum",
"Hannah",
""
],
[
"Döppner",
"Tilo",
""
],
[
"Böhme",
"Maximilian",
""
],
[
"Gawne",
"Thomas",
""
],
[
"Kraus",
"Dominik",
""
],
[
"MacDonald",
"Michael J.",
""
],
[
"Moldabekov",
"Zhandos",
""
],
[
"Tolias",
"Panagiotis",
""
],
[
"Vorberger",
"Jan",
""
],
[
"Dornheim",
"Tobias",
""
]
] |
2504.13941 | Syeda Nahida Akter | Syeda Nahida Akter, Shrimai Prabhumoye, Matvei Novikov, Seungju Han,
Ying Lin, Evelina Bakhturina, Eric Nyberg, Yejin Choi, Mostofa Patwary,
Mohammad Shoeybi, Bryan Catanzaro | Nemotron-CrossThink: Scaling Self-Learning beyond Math Reasoning | 18 pages, 7 figures | null | null | null | cs.LG cs.AI | http://creativecommons.org/licenses/by/4.0/ | Large Language Models (LLMs) have shown strong reasoning capabilities,
particularly when enhanced through Reinforcement Learning (RL). While prior
work has successfully applied RL to mathematical reasoning -- where rules and
correctness are well-defined -- generalizing these methods to broader reasoning
domains remains challenging due to limited data, the lack of verifiable reward
structures, and diverse task requirements. In this work, we propose
NEMOTRON-CROSSTHINK, a framework that systematically incorporates multi-domain
corpora, including both synthetic and real-world question-answer pairs, into RL
training to improve generalization across diverse reasoning tasks.
NEMOTRON-CROSSTHINK addresses key challenges by (1) incorporating data from
varied sources spanning STEM, humanities, social sciences, etc.; (2) applying
structured templates (e.g., multiple-choice and open-ended) to control
answer-space complexity; (3) filtering for verifiable answers; and (4)
optimizing data blending strategies that utilizes data from multiple sources
effectively. Our approach enables scalable and verifiable reward modeling
beyond mathematics and demonstrates improved accuracies on both math (MATH-500:
+30.1%, AMC23:+27.5%) and non-math reasoning benchmarks (MMLU-PRO: +12.8%,
GPQA-DIAMOND: +11.3%, AGIEVAL: +15.1%, SUPERGPQA: +3.8%). Moreover,
NEMOTRON-CROSSTHINK exhibits significantly improved response efficiency --
using 28% fewer tokens for correct answers -- highlighting more focused and
effective reasoning. Through NEMOTRON-CROSSTHINK, we demonstrate that
integrating multi-domain, multi-format data in RL leads to more accurate,
efficient, and generalizable LLMs.
| [
{
"version": "v1",
"created": "Tue, 15 Apr 2025 21:37:13 GMT"
},
{
"version": "v2",
"created": "Thu, 24 Apr 2025 02:38:53 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"Akter",
"Syeda Nahida",
""
],
[
"Prabhumoye",
"Shrimai",
""
],
[
"Novikov",
"Matvei",
""
],
[
"Han",
"Seungju",
""
],
[
"Lin",
"Ying",
""
],
[
"Bakhturina",
"Evelina",
""
],
[
"Nyberg",
"Eric",
""
],
[
"Choi",
"Yejin",
""
],
[
"Patwary",
"Mostofa",
""
],
[
"Shoeybi",
"Mohammad",
""
],
[
"Catanzaro",
"Bryan",
""
]
] |
2504.14128 | Christopher Cui | Christopher Zhang Cui, Xingdi Yuan, Ziang Xiao, Prithviraj
Ammanabrolu, Marc-Alexandre C\^ot\'e | TALES: Text Adventure Learning Environment Suite | null | null | null | null | cs.AI cs.CL | http://creativecommons.org/licenses/by/4.0/ | Reasoning is an essential skill to enable Large Language Models (LLMs) to
interact with the world. As tasks become more complex, they demand increasingly
sophisticated and diverse reasoning capabilities for sequential
decision-making, requiring structured reasoning over the context history to
determine the next best action. We introduce TALES, a diverse collection of
synthetic and human-written text-adventure games designed to challenge and
evaluate diverse reasoning capabilities. We present results over a range of
LLMs, open- and closed-weights, performing a qualitative analysis on the top
performing models. Despite an impressive showing on synthetic games, even the
top LLM-driven agents fail to achieve 15% on games designed for human
enjoyment. Code and visualization of the experiments can be found at
https://microsoft.github.io/tale-suite.
| [
{
"version": "v1",
"created": "Sat, 19 Apr 2025 01:02:42 GMT"
},
{
"version": "v2",
"created": "Tue, 22 Apr 2025 02:01:49 GMT"
},
{
"version": "v3",
"created": "Wed, 23 Apr 2025 16:31:11 GMT"
},
{
"version": "v4",
"created": "Thu, 24 Apr 2025 02:50:28 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"Cui",
"Christopher Zhang",
""
],
[
"Yuan",
"Xingdi",
""
],
[
"Xiao",
"Ziang",
""
],
[
"Ammanabrolu",
"Prithviraj",
""
],
[
"Côté",
"Marc-Alexandre",
""
]
] |
2504.14291 | Ziwei Hong | Ziwei Hong and Zhiyiong Zheng | The first moment of central value of primitive quartic $L$-functions
with fixed genus | null | null | null | null | math.NT | http://creativecommons.org/licenses/by-nc-nd/4.0/ | We investigate the mean value of the first moment of primitive quartic
$L$-functions over $\mathbb{F}_q(T)$ in the non-Kummer setting. Specifically,
we study the sum
\begin{equation*}
\sum_{\substack{\chi\ primitive\ quartic\\ \chi^2 primitive\\
genus(\chi)=g}}L_q(\frac{1}{2}, \chi),
\end{equation*} where $L_q(s,\chi)$ denotes the $L$-function associated with
primitive quartic character $\chi$. Using double Dirichlet series, we derive an
error term of size $q^{(\frac{3}{5}+\varepsilon)g}$.
| [
{
"version": "v1",
"created": "Sat, 19 Apr 2025 13:33:31 GMT"
},
{
"version": "v2",
"created": "Thu, 24 Apr 2025 04:14:52 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"Hong",
"Ziwei",
""
],
[
"Zheng",
"Zhiyiong",
""
]
] |
2504.14317 | Anna Nickolaevna Morozovska | Anna N. Morozovska, Salia Cherifi-Hertel, Eugene A. Eliseev, Victoria
V. Khist, Riccardo Hertel, and Dean R. Evans | The Role of Flexoelectric Coupling and Chemical Strains in the Emergence
of Polar Chiral Nano-Structures | 37 pages, 9 figures, 128 references. Submitted to the Journal of
Applied Physics "Special Topic on Ferroic Materials, Domains, and Domain
Walls: Bridging Fundamentals with Next-Generation Technology" | null | null | null | cond-mat.mtrl-sci cond-mat.mes-hall | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | This review examines the conditions that lead to the formation of
flexo-sensitive chiral polar structures in thin films and core-shell
ferroelectric nanoparticles. It also analyzes possible mechanisms by which the
flexoelectric effect impacts the polarization structure in core-shell
ferroelectric nanoparticles. Special attention is given to the role of the
anisotropic flexoelectric effect in forming a unique type of polarization
states with distinct chiral properties, referred to as "flexons". In the first
part of the review, we study the influence of the flexoelectric coupling on the
polarity, chirality and branching of metastable labyrinthine domain structures
in uniaxial ferroelectric core-shell nanoparticles. We reveal that the
transition from sinuous branched domain stripes to spiral-like domains occurs
gradually as the flexoelectric coupling strength is increased. Our findings
indicate that the joint action of flexoelectric effect and chemical strains,
termed as "flexo-chemical" coupling, can significantly influence the effective
Curie temperature, polarization distribution, domain morphology, and chirality
in multiaxial ferroelectric core-shell nanoparticles. Furthermore, we
demonstrate that the combination of flexo-chemical coupling and screening
effects leads to the appearance and stabilization of a chiral polarization
morphology in nanoflakes of van der Waals ferrielectrics. In the second part of
the review, we discuss several advanced applications of flexo-sensitive chiral
polar structures in core-shell ferroelectric nanoparticles for nanoelectronics
elements and cryptography. We underline the possibilities of the flexoelectric
control of multiple-degenerated labyrinthine states, which may correspond to a
differential negative capacitance (NC) state stabilized in the uniaxial
ferroelectric core by the presence of a screening shell.
| [
{
"version": "v1",
"created": "Sat, 19 Apr 2025 14:37:38 GMT"
},
{
"version": "v2",
"created": "Thu, 24 Apr 2025 10:39:38 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"Morozovska",
"Anna N.",
""
],
[
"Cherifi-Hertel",
"Salia",
""
],
[
"Eliseev",
"Eugene A.",
""
],
[
"Khist",
"Victoria V.",
""
],
[
"Hertel",
"Riccardo",
""
],
[
"Evans",
"Dean R.",
""
]
] |
2504.14450 | Weijie Wang | Weizhi Nie, Zichun Zhang, Weijie Wang, Bruno Lepri, Anan Liu, Nicu
Sebe | Causal Disentanglement for Robust Long-tail Medical Image Generation | null | null | null | null | cs.CV | http://creativecommons.org/licenses/by/4.0/ | Counterfactual medical image generation effectively addresses data scarcity
and enhances the interpretability of medical images. However, due to the
complex and diverse pathological features of medical images and the imbalanced
class distribution in medical data, generating high-quality and diverse medical
images from limited data is significantly challenging. Additionally, to fully
leverage the information in limited data, such as anatomical structure
information and generate more structurally stable medical images while avoiding
distortion or inconsistency. In this paper, in order to enhance the clinical
relevance of generated data and improve the interpretability of the model, we
propose a novel medical image generation framework, which generates independent
pathological and structural features based on causal disentanglement and
utilizes text-guided modeling of pathological features to regulate the
generation of counterfactual images. First, we achieve feature separation
through causal disentanglement and analyze the interactions between features.
Here, we introduce group supervision to ensure the independence of pathological
and identity features. Second, we leverage a diffusion model guided by
pathological findings to model pathological features, enabling the generation
of diverse counterfactual images. Meanwhile, we enhance accuracy by leveraging
a large language model to extract lesion severity and location from medical
reports. Additionally, we improve the performance of the latent diffusion model
on long-tailed categories through initial noise optimization.
| [
{
"version": "v1",
"created": "Sun, 20 Apr 2025 01:54:18 GMT"
},
{
"version": "v2",
"created": "Thu, 24 Apr 2025 10:23:24 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"Nie",
"Weizhi",
""
],
[
"Zhang",
"Zichun",
""
],
[
"Wang",
"Weijie",
""
],
[
"Lepri",
"Bruno",
""
],
[
"Liu",
"Anan",
""
],
[
"Sebe",
"Nicu",
""
]
] |
2504.14539 | Linkun Liu | Linkun Liu, Jian Sun, and Ye Tian | Should Benevolent Deception be Allowed in EHMI? A Mechanism Explanation
Based on Game Theory | null | null | null | null | cs.HC | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | The application of external human-machine interface (EHMI) on autonomous
vehicles (AVs) facilitates information exchange. Existing research fails to
consider the impact of the sequence of actions, as well as the effects of EHMI
applications and deception, raising the question of whether benevolent,
well-intentioned deception should be permitted (i.e., misleading statements
that are intended to benefit both parties). We established a game theory based
EHMI information disclosure framework for AVs in this study. In considering
benevolent deception, this framework divided the decision-making process into
three stages, respectively encompassing three key questions: whether to
disclose, when to disclose, and what type of intention information to disclose.
The results show that theoretical advantages of deception exist in certain
cases when AV expects to maximize the safety of the interaction. In 40 out of
484 cases (8.3%), safety can be enhanced through successful deception. Those
successful deceptions fall into two categories: 1) In 28 of these cases, the
straight-going AV expected the left-turning HV to yield, while HV exhibited
lower speed and higher acceleration; 2) In 12 of these cases, AV expected HV to
proceed first, while HV exhibited higher speed and lower acceleration. We also
conducted a VR-based driving simulation experiment, and the results confirmed
our conclusion. Additionally, we found that when participants had low trust in
the EHMI, its use negatively impacted interaction efficiency instead. This
study aims to analyze the mechanisms of EHMI information disclosure and
contribute to the ongoing discourse on the ethical framework governing
autonomous driving systems.
| [
{
"version": "v1",
"created": "Sun, 20 Apr 2025 09:03:28 GMT"
},
{
"version": "v2",
"created": "Thu, 24 Apr 2025 14:23:01 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"Liu",
"Linkun",
""
],
[
"Sun",
"Jian",
""
],
[
"Tian",
"Ye",
""
]
] |
2504.14543 | Kyriaki-Evangelia Aslani Dr. | Kyriaki-Evangelia Aslani, Ioannis E. Sarris, Efstratios Tzirtzilakis | On the development of OpenFOAM solvers for simulating MHD micropolar
fluid flows with or without the effect of micromagnetorotation | 29 pages, 13 Figures, 1 Table | null | null | null | physics.flu-dyn | http://creativecommons.org/licenses/by/4.0/ | The paper introduces two new OpenFOAM solvers, epotMicropolarFoam and
epotMMRFoam, developed for simulating magnetohydrodynamic (MHD) micropolar
flows with magnetic particles, such as blood, without or with the effect of
micromagnetorotation (MMR). MMR refers to the magnetic torque resulting from
the misalignment between the magnetization of magnetic particles within the
fluid and the magnetic field, influencing the internal rotation (microrotation)
of these particles. Blood can be represented as a micropolar fluid containing
magnetic particles, owing to the magnetization of erythrocytes.
EpotMicropolarFoam utilizes a transient approach and the PISO algorithm for
pressure-velocity coupling, adopting the low-magnetic-Reynolds-number
approximation for the MHD phenomena. It also accounts for micropolar effects,
incorporating the force term arising from the microrotation-vorticity
difference in the momentum equation, and solves the internal angular momentum
equation. EpotMMRFoam is a modification of epotMicropolarFoam to include the
MMR term in the internal angular momentum equation, while a constitutive
magnetization equation is also solved. Validation was conducted against
analytical solutions of MHD micropolar Poiseuille blood flow, demonstrating
excellent accuracy (error < 2%) for both solvers. When the MMR term is
included, velocity and microrotation were significantly reduced (up to 40% and
99.9%, respectively), particularly under strong magnetic fields and high
hematocrit values. In the absence of MMR, magnetic effects were minimal due to
the blood's relatively low conductivity and small vessel size. MHD 3D artery
blood flow simulations confirmed similar trends. The solvers exhibit strong
potential for biomedical applications such as magnetic hyperthermia and
targeted drug delivery.
| [
{
"version": "v1",
"created": "Sun, 20 Apr 2025 09:15:05 GMT"
},
{
"version": "v2",
"created": "Thu, 24 Apr 2025 10:28:42 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"Aslani",
"Kyriaki-Evangelia",
""
],
[
"Sarris",
"Ioannis E.",
""
],
[
"Tzirtzilakis",
"Efstratios",
""
]
] |
2504.14555 | Piet Groeneboom | Piet Groeneboom, Geurt Jongbloed | Nonparametric Estimation in Uniform Deconvolution and Interval Censoring | 16 pages, 4 figures | null | null | null | math.ST stat.TH | http://creativecommons.org/licenses/by/4.0/ | In the uniform deconvolution problem one is interested in estimating the
distribution function $F_0$ of a nonnegative random variable, based on a sample
with additive uniform noise. A peculiar and not well understood phenomenon of
the nonparametric maximum likelihood estimator in this setting is the dichotomy
between the situations where $F_0(1)=1$ and $F_0(1)<1$. If $F_0(1)=1$, the MLE
can be computed in a straightforward way and its asymptotic pointwise behavior
can be derived using the connection to the so-called current status problem.
However, if $F_0(1)<1$, one needs an iterative procedure to compute it and the
asymptotic pointwise behavior of the nonparametric maximum likelihood estimator
is not known. In this paper we describe the problem, connect it to interval
censoring problems and a more general model studied in Groeneboom (2024) to
state two competing naturally occurring conjectures for the case $F_0(1)<1$.
Asymptotic arguments related to smooth functional theory and extensive
simulations lead us to to bet on one of these two conjectures.
| [
{
"version": "v1",
"created": "Sun, 20 Apr 2025 10:03:08 GMT"
},
{
"version": "v2",
"created": "Wed, 23 Apr 2025 19:26:29 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"Groeneboom",
"Piet",
""
],
[
"Jongbloed",
"Geurt",
""
]
] |
2504.14614 | Jiajv Deng | Jia-Jv Deng (1 and 2), Feng-Yu Lu (1, 2 and 3), Zhen-Qiu Zhong (1, 2
and 3), Xiao-Hai Zhan (1, 2 and 3), Zhen-Qiang Yin (1, 2 and 3), Shuang Wang
(1, 2 and 3), Wei Chen (1, 2 and 3), De-Yong He (1, 2 and 3), Guang-Can Guo
(1, 2 and 3), Zheng-Fu Han (1, 2 and 3) ((1) CAS Key Laboratory of Quantum
Information, University of Science and Technology of China, Hefei, Anhui, P.
R. China, (2) CAS Center for Excellence in Quantum Information and Quantum
Physics, University of Science and Technology of China, Hefei, Anhui, P. R.
China (3) Hefei National Laboratory, University of Science and Technology of
China, Hefei, China) | Measurement-device-independent quantum key distribution with asymmetric
sources | null | null | null | null | quant-ph physics.optics | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Measurement-device-independent quantum key distribution (MDI-QKD), which
eliminates all the attacks from the eavesdropper to the measurement party, has
been one of the most promising technology for the implementation of end-to-end
quantum networks. In practice, the asymmetry of both sources and channels is
generally inevitable. Therefore, we propose a theory to analyze the performance
when any two MDI users in networks communicates using asymmetric sources in
distinct single or multiple temporal modes. As a specific application, we model
to obtain the key rate of MDI-QKD with weak coherent pulse source and
spontaneous parametric down-conversion source, and compare the performance to
the cases with symmetric (i.e. identical) sources. The result demonstrates that
the actual performance does not degrade due to the asymmetry of sources. In
contrary, it maintains at a good level over the entire distance we study. This
work provides a theoretical basis for analyzing and optimizing MDI-QKD networks
with asymmetric sources, and thus paving the way for the practical deployment
of completely asymmetric MDI-QKD networks.
| [
{
"version": "v1",
"created": "Sun, 20 Apr 2025 13:44:39 GMT"
},
{
"version": "v2",
"created": "Thu, 24 Apr 2025 12:24:32 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"Deng",
"Jia-Jv",
"",
"1 and 2"
],
[
"Lu",
"Feng-Yu",
"",
"1, 2 and 3"
],
[
"Zhong",
"Zhen-Qiu",
"",
"1, 2\n and 3"
],
[
"Zhan",
"Xiao-Hai",
"",
"1, 2 and 3"
],
[
"Yin",
"Zhen-Qiang",
"",
"1, 2 and 3"
],
[
"Wang",
"Shuang",
"",
"1, 2 and 3"
],
[
"Chen",
"Wei",
"",
"1, 2 and 3"
],
[
"He",
"De-Yong",
"",
"1, 2 and 3"
],
[
"Guo",
"Guang-Can",
"",
"1, 2 and 3"
],
[
"Han",
"Zheng-Fu",
"",
"1, 2 and 3"
]
] |
2504.14634 | Ladislau B\"ol\"oni | Sahara Sheikholeslami and Ladislau B\"ol\"oni | Latent Representations for Visual Proprioception in Inexpensive Robots | null | null | null | null | cs.RO cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Robotic manipulation requires explicit or implicit knowledge of the robot's
joint positions. Precise proprioception is standard in high-quality industrial
robots but is often unavailable in inexpensive robots operating in unstructured
environments. In this paper, we ask: to what extent can a fast, single-pass
regression architecture perform visual proprioception from a single external
camera image, available even in the simplest manipulation settings? We explore
several latent representations, including CNNs, VAEs, ViTs, and bags of
uncalibrated fiducial markers, using fine-tuning techniques adapted to the
limited data available. We evaluate the achievable accuracy through experiments
on an inexpensive 6-DoF robot.
| [
{
"version": "v1",
"created": "Sun, 20 Apr 2025 14:24:54 GMT"
},
{
"version": "v2",
"created": "Thu, 24 Apr 2025 12:36:35 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"Sheikholeslami",
"Sahara",
""
],
[
"Bölöni",
"Ladislau",
""
]
] |
2504.14644 | Sikarin Yoo-Kong | Umpon Jairuk, Thanadon Kongkoom and Sikarin Yoo-Kong | The $q$-deformed Calogero's Goldfish Systems | 14 pages | null | null | null | nlin.SI | http://creativecommons.org/licenses/by/4.0/ | Searching for integrable models is a central theme in theoretical and
mathematical physics, as such systems offer valuable insights into the
underlying structure and symmetries of complex physical phenomena. In this
work, we contribute to this pursuit by proposing a new class of one-dimensional
many-body integrable systems, which we refer to as the $q$-deformed Calogero's
Goldfish system. Our construction employs $q$-deformation of logarithmic and
exponential functions inspired by Tsallis' formalism in non-extensive
statistical mechanics. Notably, the model satisfies the double-zero condition
on its solutions, underscoring its integrable nature and offering a novel
perspective on deformation techniques within exactly solvable systems.
| [
{
"version": "v1",
"created": "Sun, 20 Apr 2025 14:59:28 GMT"
},
{
"version": "v2",
"created": "Thu, 24 Apr 2025 06:08:50 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"Jairuk",
"Umpon",
""
],
[
"Kongkoom",
"Thanadon",
""
],
[
"Yoo-Kong",
"Sikarin",
""
]
] |
2504.14971 | Yanfang Lyu | Yanfang Lyu, Yunyang Zhang, Zhiqiang Gong, Xiao Kang, Wen Yao and
Yongmao Pei | A novel hybrid neural network of fluid-structure interaction prediction
for two cylinders in tandem arrangement | null | null | null | null | physics.flu-dyn | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Deep learning has shown promise in improving computing efficiency while
ensuring modeling accuracy in fluid-structure interaction (FSI) analysis.
However, its current capabilities are limited when it comes to constructing
multi-object coupling systems with dynamic boundaries. To address such
limitation, a novel FSI neural solver integrated by a fluid deep learning model
with multi-time steps and a structural dynamic solver is proposed to accurately
and reliably predict the vortex-induced vibration (VIV) evolution for two
cylinders in tandem. This well-designed model in the form of end-to-end can
precisely predict the instantaneous flow field state at the subsequent time by
coupling the temporal flow fields of historical multi-time sequences and the
current structural responses, moreover, derives the structural state at the
next time. Furthermore, the novel fluid deep learning model consists of a wall
shear model utilizing a multilayer perception network and flow field model with
U-shaped architecture jointing the Fourier neural operator and modified
convolution long-short term memory model. Both models effectively capture
coupling transfer forces and predict instantaneous flow fields, with the latter
demonstrating superior accuracy compared to Convolutional Neural Network- or
Unet- based models with similar parameters. The prediction speed of the
proposed models realizes an improvement of over 1000 times compared with the
numerical simulation. Significantly, the proposed FSI neural model demonstrates
exceptional capability in constructing the nonlinear complex multi- vibration
systems and has substantial potential for advancing FSI modeling of flexible
structures featuring pronounced nonlinear deformation boundaries.
| [
{
"version": "v1",
"created": "Mon, 21 Apr 2025 08:58:00 GMT"
},
{
"version": "v2",
"created": "Thu, 24 Apr 2025 05:22:10 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"Lyu",
"Yanfang",
""
],
[
"Zhang",
"Yunyang",
""
],
[
"Gong",
"Zhiqiang",
""
],
[
"Kang",
"Xiao",
""
],
[
"Yao",
"Wen",
""
],
[
"Pei",
"Yongmao",
""
]
] |
2504.14992 | Bohong Wu | Bohong Wu, Shen Yan, Sijun Zhang, Jianqiao Lu, Yutao Zeng, Ya Wang,
Xun Zhou | Efficient Pretraining Length Scaling | null | null | null | null | cs.CL | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Recent advances in large language models have demonstrated the effectiveness
of length scaling during post-training, yet its potential in pre-training
remains underexplored. We present the Parallel Hidden Decoding Transformer
(\textit{PHD}-Transformer), a novel framework that enables efficient length
scaling during pre-training while maintaining inference efficiency.
\textit{PHD}-Transformer achieves this through an innovative KV cache
management strategy that distinguishes between original tokens and hidden
decoding tokens. By retaining only the KV cache of original tokens for
long-range dependencies while immediately discarding hidden decoding tokens
after use, our approach maintains the same KV cache size as the vanilla
transformer while enabling effective length scaling. To further enhance
performance, we introduce two optimized variants: \textit{PHD-SWA} employs
sliding window attention to preserve local dependencies, while
\textit{PHD-CSWA} implements chunk-wise sliding window attention to eliminate
linear growth in pre-filling time. Extensive experiments demonstrate consistent
improvements across multiple benchmarks.
| [
{
"version": "v1",
"created": "Mon, 21 Apr 2025 09:41:26 GMT"
},
{
"version": "v2",
"created": "Thu, 24 Apr 2025 04:13:49 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"Wu",
"Bohong",
""
],
[
"Yan",
"Shen",
""
],
[
"Zhang",
"Sijun",
""
],
[
"Lu",
"Jianqiao",
""
],
[
"Zeng",
"Yutao",
""
],
[
"Wang",
"Ya",
""
],
[
"Zhou",
"Xun",
""
]
] |
2504.15007 | Mohamed Amine Kerkouri | David C Wong, Bin Wang, Gorkem Durak, Marouane Tliba, Mohamed Amine
Kerkouri, Aladine Chetouani, Ahmet Enis Cetin, Cagdas Topel, Nicolo Gennaro,
Camila Vendrami, Tugce Agirlar Trabzonlu, Amir Ali Rahsepar, Laetitia
Perronne, Matthew Antalek, Onural Ozturk, Gokcan Okur, Andrew C. Gordon, Ayis
Pyrros, Frank H Miller, Amir A Borhani, Hatice Savas, Eric M. Hart, Elizabeth
A Krupinski, Ulas Bagci | Shifts in Doctors' Eye Movements Between Real and AI-Generated Medical
Images | This paper was accepted at ETRA 2025 Japan | null | null | null | cs.CV cs.HC | http://creativecommons.org/licenses/by-nc-nd/4.0/ | Eye-tracking analysis plays a vital role in medical imaging, providing key
insights into how radiologists visually interpret and diagnose clinical cases.
In this work, we first analyze radiologists' attention and agreement by
measuring the distribution of various eye-movement patterns, including saccades
direction, amplitude, and their joint distribution. These metrics help uncover
patterns in attention allocation and diagnostic strategies. Furthermore, we
investigate whether and how doctors' gaze behavior shifts when viewing
authentic (Real) versus deep-learning-generated (Fake) images. To achieve this,
we examine fixation bias maps, focusing on first, last, short, and longest
fixations independently, along with detailed saccades patterns, to quantify
differences in gaze distribution and visual saliency between authentic and
synthetic images.
| [
{
"version": "v1",
"created": "Mon, 21 Apr 2025 10:13:59 GMT"
},
{
"version": "v2",
"created": "Thu, 24 Apr 2025 09:52:55 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"Wong",
"David C",
""
],
[
"Wang",
"Bin",
""
],
[
"Durak",
"Gorkem",
""
],
[
"Tliba",
"Marouane",
""
],
[
"Kerkouri",
"Mohamed Amine",
""
],
[
"Chetouani",
"Aladine",
""
],
[
"Cetin",
"Ahmet Enis",
""
],
[
"Topel",
"Cagdas",
""
],
[
"Gennaro",
"Nicolo",
""
],
[
"Vendrami",
"Camila",
""
],
[
"Trabzonlu",
"Tugce Agirlar",
""
],
[
"Rahsepar",
"Amir Ali",
""
],
[
"Perronne",
"Laetitia",
""
],
[
"Antalek",
"Matthew",
""
],
[
"Ozturk",
"Onural",
""
],
[
"Okur",
"Gokcan",
""
],
[
"Gordon",
"Andrew C.",
""
],
[
"Pyrros",
"Ayis",
""
],
[
"Miller",
"Frank H",
""
],
[
"Borhani",
"Amir A",
""
],
[
"Savas",
"Hatice",
""
],
[
"Hart",
"Eric M.",
""
],
[
"Krupinski",
"Elizabeth A",
""
],
[
"Bagci",
"Ulas",
""
]
] |
2504.15036 | Ori Lahav | Roy Margalit, Michalis Kokologiannakis, Shachar Itzhaky, Ori Lahav | Dynamic Robustness Verification Against Weak Memory (Extended Version) | Extended version for PLDI'25 paper | null | null | null | cs.PL | http://creativecommons.org/licenses/by/4.0/ | Dynamic race detection is a highly effective runtime verification technique
for identifying data races by instrumenting and monitoring concurrent program
runs. However, standard dynamic race detection is incompatible with practical
weak memory models; the added instrumentation introduces extra synchronization,
which masks weakly consistent behaviors and inherently misses certain data
races. In response, we propose to dynamically verify program robustness-a
property ensuring that a program exhibits only strongly consistent behaviors.
Building on an existing static decision procedures, we develop an algorithm for
dynamic robustness verification under a C11-style memory model. The algorithm
is based on "location clocks", a variant of vector clocks used in standard race
detection. It allows effective and easy-to-apply defense against weak memory on
a per-program basis, which can be combined with race detection that assumes
strong consistency. We implement our algorithm in a tool, called RSAN, and
evaluate it across various settings. To our knowledge, this work is the first
to propose and develop dynamic verification of robustness against weak memory
models.
| [
{
"version": "v1",
"created": "Mon, 21 Apr 2025 11:44:58 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"Margalit",
"Roy",
""
],
[
"Kokologiannakis",
"Michalis",
""
],
[
"Itzhaky",
"Shachar",
""
],
[
"Lahav",
"Ori",
""
]
] |
2504.15103 | Rafael Armando Martinez Rivero Sr | Rafael Armando Mart\'inez-Rivero, Christian Sarmiento-Cano, Diego
Catillo-Morales, Jorge Perea-P\'erez, Victor Baldovino-Medrano, Jose David
Sanabria-G\'omez, Luis Alberto N\'u\~nez | Muon Imaging of Hydrotreatment Reactors | null | null | null | null | physics.ins-det physics.app-ph | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | This study presents the design and simulation-based validation of a muon
imaging system tailored for potential applications in industrial hydrotreatment
units. The system is built around a two-panel plastic scintillator hodoscope,
equipped with silicon photomultipliers and read-out via a CAEN FERS-A5202
acquisition system. The detector was calibrated using a stepwise ``staircase''
method and characterized under open-sky and controlled conditions. We conducted
muon flux attenuation measurements to validate its response using variable lead
shielding. We found agreement with simulations generated using the MEIGA
framework and realistic cosmic ray spectra from the ARTI simulation chain. With
the detector response validated, we modelled muon transmission through a
realistic 3D representation of a hydrotreatment tower, incorporating internal
variations in catalyst bed density. By reconstructing angular muon fluxes and
computing relative attenuation maps, we demonstrated the system's capability to
detect internal density contrasts. Simulation results indicate that 20~hours of
exposure to vertical muon flux is sufficient to retrieve structural
information. In comparison, inclined configurations (30$^\circ$ and 60$^\circ$
from vertical) require extended exposure times--up to 8~days--yet remain
feasible within industrial monitoring schedules. These findings highlight the
feasibility of muography as a non-invasive diagnostic tool for complex
industrial infrastructure. The proposed system shows strong potential for
real-time monitoring of catalyst bed integrity and long-term structural
analysis in high-pressure chemical reactors.
| [
{
"version": "v1",
"created": "Mon, 21 Apr 2025 13:53:17 GMT"
},
{
"version": "v2",
"created": "Wed, 23 Apr 2025 21:09:53 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"Martínez-Rivero",
"Rafael Armando",
""
],
[
"Sarmiento-Cano",
"Christian",
""
],
[
"Catillo-Morales",
"Diego",
""
],
[
"Perea-Pérez",
"Jorge",
""
],
[
"Baldovino-Medrano",
"Victor",
""
],
[
"Sanabria-Gómez",
"Jose David",
""
],
[
"Núñez",
"Luis Alberto",
""
]
] |
2504.15190 | Bikash R. Dinda | Bikash R. Dinda and Roy Maartens | Physical vs phantom dark energy after DESI: thawing quintessence in a
curved background | references added | null | null | null | astro-ph.CO gr-qc | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Recent data from DESI, in combination with other data, provide moderate
evidence of dynamical dark energy, $w\neq-1$. In the $w_0, w_a$ parametrization
of $w$, there is a preference for a phantom crossing, $w<-1$, at redshift
$z\sim0.5$. In general relativity, the phantom equation of state is unphysical.
Thus it is important to check whether phantom crossing is present in other
physically self-consistent models of dark energy that have equivalent evidence
to the $w_0, w_a$ parametrization. We find that thawing quintessence with
nonzero cosmic curvature can fit the recent data as well as $w_0, w_a$ in a
flat background, based on both parametric and realistic scalar field
evolutions. Although the realistic model does not allow $w<-1$, the
parametrizations do allow it. However even if we allow $w<-1$ the data do not
enforce phantom crossing. Thus, the phantom crossing is an artifact of a
parametrization that is not based on a physical model.
| [
{
"version": "v1",
"created": "Mon, 21 Apr 2025 16:00:28 GMT"
},
{
"version": "v2",
"created": "Thu, 24 Apr 2025 16:41:57 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"Dinda",
"Bikash R.",
""
],
[
"Maartens",
"Roy",
""
]
] |
2504.15203 | Aditya Banerjee | Aditya Banerjee | Fine features of entanglement dynamics in quenches across the Ising
quantum critical point | 14 pages, 10 figures. Made some edits, added some discussion and a
few references | null | null | null | quant-ph cond-mat.quant-gas cond-mat.stat-mech cond-mat.str-el | http://creativecommons.org/licenses/by/4.0/ | The task of exploring and understanding important aspects of
far-from-equilibrium dynamics of closed and generic quantum many-body systems
has received a thrust of attention in recent years, driven partly by remarkable
advances in ultracold experimental technologies. In this work, for the
paradigmatic Ising spin chain with transverse and longitudinal fields, we
present numerical observations of several "fine-grained" features of
far-from-equilibrium dynamics from a quantum informational point of view that
have hitherto escaped notice, induced by quantum quenches across the Ising
critical point between states deep inside the para- and ferromagnetic regimes.
Rather featureless dynamics is seen for ferromagnetic to paramagnetic quenches,
but paramagnetic to ferromagnetic quenches exhibit rich behaviour, including a
series of sudden deaths and revivals of entanglement between two spins in the
system's bulk, periodic but short-lived occurrences of approximately
$1-$uniform states and recurrences of an approximately Page-like dynamics of
entanglement entropies of one- and two-spin subsystems, non-analytic cusps in
single-copy entanglement entropy for sufficiently big subsystems, insufficient
mixedness and a series of scrambling-$\textit{un}$scrambling of local
information between neighboring spins. Moreover, essentially indistinguishable
dynamics is seen at very early times between the integrable limit (zero
longitudinal field) and non-integrable cases, with the former eventually
showing signatures of better mixing and faster approach to equilibration than
the latter. These features are expected to hold for quench dynamics across
Ising quantum critical points in more complicated systems.
| [
{
"version": "v1",
"created": "Mon, 21 Apr 2025 16:19:17 GMT"
},
{
"version": "v2",
"created": "Thu, 24 Apr 2025 01:57:02 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"Banerjee",
"Aditya",
""
]
] |
2504.15284 | Weichen Li | Weichen Li, Albert Jan, Baishakhi Ray, Chengzhi Mao, Junfeng Yang,
Kexin Pei | EditLord: Learning Code Transformation Rules for Code Editing | null | null | null | null | cs.SE cs.CR cs.LG | http://creativecommons.org/licenses/by/4.0/ | Code editing is a foundational task in software development, where its
effectiveness depends on whether it introduces desired code property changes
without changing the original code's intended functionality. Existing
approaches often formulate code editing as an implicit end-to-end task,
omitting the fact that code-editing procedures inherently consist of discrete
and explicit steps. Thus, they suffer from suboptimal performance and lack of
robustness and generalization. We introduce EditLord, a code editing framework
that makes the code transformation steps explicit. Our key insight is to employ
a language model (LM) as an inductive learner to extract code editing rules
from the training code pairs as concise meta-rule sets. Such rule sets will be
manifested for each training sample to augment them for finetuning or assist in
prompting- and iterative-based code editing. EditLordoutperforms the
state-of-the-art by an average of 22.7% in editing performance and 58.1% in
robustness while achieving 20.2% higher functional correctness across critical
software engineering and security applications, LM models, and editing modes.
| [
{
"version": "v1",
"created": "Mon, 10 Mar 2025 16:33:59 GMT"
},
{
"version": "v2",
"created": "Wed, 23 Apr 2025 18:37:18 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"Li",
"Weichen",
""
],
[
"Jan",
"Albert",
""
],
[
"Ray",
"Baishakhi",
""
],
[
"Mao",
"Chengzhi",
""
],
[
"Yang",
"Junfeng",
""
],
[
"Pei",
"Kexin",
""
]
] |
2504.15364 | Junyoung Park | Junyoung Park, Dalton Jones, Matt J Morse, Raghavv Goel, Mingu Lee,
and Chris Lott | KeyDiff: Key Similarity-Based KV Cache Eviction for Long-Context LLM
Inference in Resource-Constrained Environments | 8 pages, 14 figures | null | null | null | cs.AI | http://creativecommons.org/licenses/by/4.0/ | In this work, we demonstrate that distinctive keys during LLM inference tend
to have high attention scores. We explore this phenomenon and propose KeyDiff,
a training-free KV cache eviction method based on key similarity. This method
facilitates the deployment of LLM-based application requiring long input
prompts in resource-constrained environments with limited memory and compute
budgets. Unlike other KV cache eviction methods, KeyDiff can process
arbitrarily long prompts within strict resource constraints and efficiently
generate responses. We demonstrate that KeyDiff computes the optimal solution
to a KV cache selection problem that maximizes key diversity, providing a
theoretical understanding of KeyDiff. Notably,KeyDiff does not rely on
attention scores, allowing the use of optimized attention mechanisms like
FlashAttention. We demonstrate the effectiveness of KeyDiff across diverse
tasks and models, illustrating a performance gap of less than 0.04\% with 8K
cache budget ($\sim$ 23\% KV cache reduction) from the non-evicting baseline on
the LongBench benchmark for Llama 3.1-8B and Llama 3.2-3B.
| [
{
"version": "v1",
"created": "Mon, 21 Apr 2025 18:12:46 GMT"
},
{
"version": "v2",
"created": "Wed, 23 Apr 2025 18:02:55 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"Park",
"Junyoung",
""
],
[
"Jones",
"Dalton",
""
],
[
"Morse",
"Matt J",
""
],
[
"Goel",
"Raghavv",
""
],
[
"Lee",
"Mingu",
""
],
[
"Lott",
"Chris",
""
]
] |
2504.15396 | Igor Ladnik | Igor Ladnik | A Quadratic Control Framework for Dynamic Systems | 16 pages, 10 figures, 16 tables | null | null | null | eess.SY cs.SY | http://creativecommons.org/licenses/by/4.0/ | This article presents a unified approach to quadratic optimal control for
both linear and nonlinear discrete-time systems, with a focus on trajectory
tracking. The control strategy is based on minimizing a quadratic cost function
that penalizes deviations of system states and control inputs from their
desired trajectories.
For linear systems, the classical Linear Quadratic Regulator (LQR) solution
is derived using dynamic programming, resulting in recursive equations for
feedback and feedforward terms. For nonlinear dynamics, the Iterative Linear
Quadratic Regulator (iLQR) method is employed, which iteratively linearizes the
system and solves a sequence of LQR problems to converge to an optimal policy.
To implement this approach, a software service was developed and tested on
several canonical models, including: Rayleigh oscillator, inverted pendulum on
a moving cart, two-link manipulator, and quadcopter. The results confirm that
iLQR enables efficient and accurate trajectory tracking in the presence of
nonlinearities.
To further enhance performance, it can be seamlessly integrated with Model
Predictive Control (MPC), enabling online adaptation and improved robustness to
constraints and system uncertainties.
| [
{
"version": "v1",
"created": "Mon, 21 Apr 2025 19:00:53 GMT"
},
{
"version": "v2",
"created": "Thu, 24 Apr 2025 15:11:09 GMT"
}
] | 2025-04-25T00:00:00 | [
[
"Ladnik",
"Igor",
""
]
] |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.