id
stringlengths 9
16
| submitter
stringlengths 1
64
⌀ | authors
stringlengths 3
62.8k
| title
stringlengths 1
382
| comments
stringlengths 1
1.15k
⌀ | journal-ref
stringlengths 1
557
⌀ | doi
stringlengths 7
153
⌀ | report-no
stringlengths 1
509
⌀ | categories
stringlengths 5
125
| license
stringclasses 9
values | abstract
stringlengths 6
6.09k
| versions
listlengths 1
187
| update_date
timestamp[s]date 2007-05-23 00:00:00
2025-04-25 00:00:00
| authors_parsed
sequencelengths 1
2.83k
|
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
2409.07110 | Ali Forootani | Ali Forootani, Danial Esmaeili Aliabadi, Daniela Thraen | Bio-Eng-LMM AI Assist chatbot: A Comprehensive Tool for Research and
Education | null | null | null | null | eess.SY cs.SY | http://creativecommons.org/licenses/by-nc-nd/4.0/ | This article introduces Bio-Eng-LMM AI chatbot, a versatile platform designed
to enhance user interaction for educational and research purposes. Leveraging
cutting-edge open-source Large Language Models (LLMs), Bio-Eng-LMM operates as
a sophisticated AI assistant, exploiting the capabilities of traditional models
like ChatGPT. Central to Bio-Eng-LMM is its implementation of Retrieval
Augmented Generation (RAG) through three primary methods: integration of
preprocessed documents, real-time processing of user-uploaded files, and
information retrieval from any specified website. Additionally, the chatbot
incorporates image generation via a Stable Diffusion Model (SDM), image
understanding and response generation through LLAVA, and search functionality
on the internet powered by secure search engine such as DuckDuckGo. To provide
comprehensive support, Bio-Eng-LMM offers text summarization, website content
summarization, and both text and voice interaction. The chatbot maintains
session memory to ensure contextually relevant and coherent responses. This
integrated platform builds upon the strengths of RAG-GPT and Web-Based RAG
Query (WBRQ) where the system fetches relevant information directly from the
web to enhance the LLMs response generation.
| [
{
"version": "v1",
"created": "Wed, 11 Sep 2024 08:56:27 GMT"
},
{
"version": "v2",
"created": "Sat, 19 Apr 2025 19:30:20 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Forootani",
"Ali",
""
],
[
"Aliabadi",
"Danial Esmaeili",
""
],
[
"Thraen",
"Daniela",
""
]
] |
2409.07602 | Johannes B\"urger | Johannes B\"urger, Jisoo Kim, Thomas Weiss, Stefan A. Maier, Markus A.
Schmidt | On-chip twisted hollow-core light cages: enhancing planar photonics with
3D nanoprinting | Main text (51 pages, 7 figures) and Supplemental Material (32 pages,
21 figures) | null | null | null | physics.optics | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Twisted optical fibers are a promising platform for manipulating circularly
polarized light and orbital angular momentum beams for applications such as
nonlinear frequency conversion, optical communication, or chiral sensing.
However, integration into chip-scale technology is challenging because twisted
fibers are incompatible with planar photonics and the achieved twist rates are
limited. Here, we address these challenges by introducing the concept of
3D-nanoprinted on-chip twisted hollow-core light cages. We show theoretically
and experimentally that geometrical twisting of light cages forces the
fundamental core mode of a given handedness to couple with selected
higher-order core modes, resulting in strong circular dichroism (CD). These
chiral resonances result from the angular momentum harmonics of the fundamental
mode, allowing us to predict their spectral locations and the occurrence of
circular birefringence. Twisted light cages enable very high twist rates and
CD, exceeding those of twisted hollow-core fibers by more than two orders of
magnitude (twist period: 90 $\mu$m, CD: 0.8 dB/mm). Moreover, the unique cage
design provides lateral access to the central core region, enabling future
applications in chiral spectroscopy. Therefore, the presented concept opens a
path for translating twisted fiber research to on-chip technology, resulting in
a new platform for integrated chiral photonics.
| [
{
"version": "v1",
"created": "Wed, 11 Sep 2024 20:30:18 GMT"
},
{
"version": "v2",
"created": "Sun, 20 Apr 2025 13:09:02 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Bürger",
"Johannes",
""
],
[
"Kim",
"Jisoo",
""
],
[
"Weiss",
"Thomas",
""
],
[
"Maier",
"Stefan A.",
""
],
[
"Schmidt",
"Markus A.",
""
]
] |
2409.07759 | Bangya Liu | Bangya Liu and Suman Banerjee | SwinGS: Sliding Window Gaussian Splatting for Volumetric Video Streaming
with Arbitrary Length | null | null | null | null | cs.MM cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Recent advances in 3D Gaussian Splatting (3DGS) have garnered significant
attention in computer vision and computer graphics due to its high rendering
speed and remarkable quality. While extant research has endeavored to extend
the application of 3DGS from static to dynamic scenes, such efforts have been
consistently impeded by excessive model sizes, constraints on video duration,
and content deviation. These limitations significantly compromise the
streamability of dynamic 3D Gaussian models, thereby restricting their utility
in downstream applications, including volumetric video, autonomous vehicle, and
immersive technologies such as virtual, augmented, and mixed reality.
This paper introduces SwinGS, a novel framework for training, delivering, and
rendering volumetric video in a real-time streaming fashion. To address the
aforementioned challenges and enhance streamability, SwinGS integrates
spacetime Gaussian with Markov Chain Monte Carlo (MCMC) to adapt the model to
fit various 3D scenes across frames, in the meantime employing a sliding window
captures Gaussian snapshots for each frame in an accumulative way. We implement
a prototype of SwinGS and demonstrate its streamability across various datasets
and scenes. Additionally, we develop an interactive WebGL viewer enabling
real-time volumetric video playback on most devices with modern browsers,
including smartphones and tablets. Experimental results show that SwinGS
reduces transmission costs by 83.6% compared to previous work and could be
easily scaled to volumetric videos with arbitrary length with no increasing of
required GPU resources.
| [
{
"version": "v1",
"created": "Thu, 12 Sep 2024 05:33:15 GMT"
},
{
"version": "v2",
"created": "Mon, 23 Dec 2024 20:03:22 GMT"
},
{
"version": "v3",
"created": "Sat, 19 Apr 2025 05:21:50 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Liu",
"Bangya",
""
],
[
"Banerjee",
"Suman",
""
]
] |
2409.08246 | Takuto Ishikawa | Takuto Ishikawa, Wolfgang Belzig, and Takeo Kato | Spin pumping into quantum spin nematic states | 13 pages, 6 figures | null | null | null | cond-mat.mes-hall | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We theoretically study spin pumping into a spin-nematic state in a junction
system composed of a ferromagnetic insulator and a spin-nematic insulator,
described by using the spin-1 bilinear-biquadratic model. We analyze an
increase of the Gilbert damping in ferromagnetic resonance (FMR) due to an
interfacial exchange coupling within a mean-field theory based on the Schwinger
boson method. We find that the two Schwinger bosons contribute in distinct ways
to spin pumping. We report a detailed dependence of the spin pumping on a
resonant frequency, a magnetic field, and an interface type.
| [
{
"version": "v1",
"created": "Thu, 12 Sep 2024 17:44:44 GMT"
},
{
"version": "v2",
"created": "Sat, 19 Apr 2025 11:26:24 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Ishikawa",
"Takuto",
""
],
[
"Belzig",
"Wolfgang",
""
],
[
"Kato",
"Takeo",
""
]
] |
2409.09275 | Yida Yang | Yida Yang, Chen Wu, Ji-Feng Yang | Impact of dark matter on strange quark stars described by different
quark models | 11 pages, 11 figures, revised version published in The European
Physical Journal C | Eur. Phys. J. C 85, 426 (2025) | 10.1140/epjc/s10052-025-14150-9 | null | astro-ph.HE | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Dark matter is hypothesized to interact with ordinary matter solely through
gravity and may be present in compact objects such as strange quark stars. We
treat strange quark stars admixed with dark matter as two-fluid systems to
investigate the potential effects of dark matter on strange quark stars. Quark
matter is described by the quasiparticle model and the extended MIT bag model
for comparison. Dark matter is treated as asymmetric, self-interacting, and
composed of massive fermionic particles. The two-fluid
Tolman-Oppenheimer-Volkoff (TOV) equations are employed to solve for specific
stellar properties. Our analysis yields relations between central energy
density and mass, radius and mass, as well as tidal deformability and mass. The
calculated curves generally align with observational data. In particular, we
find that the pattern in which fermionic asymmetric dark matter affects the
properties of strange quark stars may not be influenced by the equation of
state (EOS) of strange quark matter.
| [
{
"version": "v1",
"created": "Sat, 14 Sep 2024 02:57:28 GMT"
},
{
"version": "v2",
"created": "Sat, 19 Apr 2025 08:48:38 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Yang",
"Yida",
""
],
[
"Wu",
"Chen",
""
],
[
"Yang",
"Ji-Feng",
""
]
] |
2409.09542 | Brighton Ancelin | Brighton Ancelin, Yenho Chen, Peimeng Guan, Chiraag Kaushik, Belen
Martin-Urcelay, Alex Saad-Falcon, Nakul Singh | MANGO: Learning Disentangled Image Transformation Manifolds with Grouped
Operators | Submitted to SampTA 2025. This work has been submitted to the IEEE
for possible publication | null | null | null | eess.IV cs.CV cs.LG | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Learning semantically meaningful image transformations (i.e. rotation,
thickness, blur) directly from examples can be a challenging task. Recently,
the Manifold Autoencoder (MAE) proposed using a set of Lie group operators to
learn image transformations directly from examples. However, this approach has
limitations, as the learned operators are not guaranteed to be disentangled and
the training routine is prohibitively expensive when scaling up the model. To
address these limitations, we propose MANGO (transformation Manifolds with
Grouped Operators) for learning disentangled operators that describe image
transformations in distinct latent subspaces. Moreover, our approach allows
practitioners the ability to define which transformations they aim to model,
thus improving the semantic meaning of the learned operators. Through our
experiments, we demonstrate that MANGO enables composition of image
transformations and introduces a one-phase training routine that leads to a
100x speedup over prior works.
| [
{
"version": "v1",
"created": "Sat, 14 Sep 2024 21:42:38 GMT"
},
{
"version": "v2",
"created": "Sun, 20 Apr 2025 18:49:55 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Ancelin",
"Brighton",
""
],
[
"Chen",
"Yenho",
""
],
[
"Guan",
"Peimeng",
""
],
[
"Kaushik",
"Chiraag",
""
],
[
"Martin-Urcelay",
"Belen",
""
],
[
"Saad-Falcon",
"Alex",
""
],
[
"Singh",
"Nakul",
""
]
] |
2409.09793 | Andrei Caragea | Andrei Caragea and Dae Gwan Lee | On the principal minors of Fourier matrices | 7 pages, 0 figures | null | null | null | math.FA | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | For the $N$-dimensional Fourier matrix $\mathcal{F}_N$, we prove that if
$N\geq 4$ is square-free, then every $2 \times 2$ and $3\times 3$ principal
minor of $\mathcal{F}_N$ is nonzero. We also show that if $N\geq 4$ is not
square-free, then $\mathcal{F}_N$ has zero principal minors of all sizes.
Moreover, based on numerical experiments, we conjecture that if $N$ is
square-free, then all principal minors of $\mathcal{F}_N$ are nonzero.
| [
{
"version": "v1",
"created": "Sun, 15 Sep 2024 16:59:47 GMT"
},
{
"version": "v2",
"created": "Fri, 20 Sep 2024 09:15:04 GMT"
},
{
"version": "v3",
"created": "Sun, 20 Oct 2024 23:53:07 GMT"
},
{
"version": "v4",
"created": "Sun, 20 Apr 2025 13:48:39 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Caragea",
"Andrei",
""
],
[
"Lee",
"Dae Gwan",
""
]
] |
2409.10619 | Qiang Gao | Qiang Gao, Ryan A. Lanzetta, Patrick Ledwith, Jie Wang, and Eslam
Khalaf | Bootstrapping the Quantum Hall problem | Total 30 pages. Main text: 22 pages, 9 figures | null | null | null | cond-mat.str-el | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | The bootstrap method aims to solve problems by imposing constraints on the
space of physical observables, which often follow from physical assumptions
such as positivity and symmetry. Here, we employ a bootstrap approach to study
interacting electrons in the lowest Landau level by minimizing the energy as a
function of the static structure factor subject to a set of constraints,
bypassing the need to construct the full many-body wavefunction. This approach
rigorously lower bounds the ground state energy, making it complementary to
conventional variational upper bounds. We show that the lower bound we obtain
is relatively tight, within at most 5\% from the ground state energy computed
with exact diagonalization (ED) at small system sizes, and generally gets
tighter as we include more constraints. In addition to energetics, our results
reproduce the correct power law dependence of the pair correlation function at
short distances and the existence of a large entanglement gap in the
two-particle entanglement spectra for the Laughlin states at $\nu = 1/3$. We
further identify signatures of the composite Fermi liquid state close to
half-filling. This shows that the bootstrap approach is capable, in principle,
of describing non-trivial gapped topologically ordered, as well as gapless,
phases. At the end, we will discuss possible extensions and limitations of this
approach. Our work establishes numerical bootstrap as a promising method to
study many-body phases in topological bands, paving the way to its application
in moir\'e platforms where the energetic competition between fractional quantum
anomalous Hall, symmetry broken, and gapless states remains poorly understood.
| [
{
"version": "v1",
"created": "Mon, 16 Sep 2024 18:00:18 GMT"
},
{
"version": "v2",
"created": "Fri, 18 Apr 2025 22:09:50 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Gao",
"Qiang",
""
],
[
"Lanzetta",
"Ryan A.",
""
],
[
"Ledwith",
"Patrick",
""
],
[
"Wang",
"Jie",
""
],
[
"Khalaf",
"Eslam",
""
]
] |
2409.11686 | Asad Aali | Asad Aali, Andrew Johnston, Louis Blankemeier, Dave Van Veen, Laura T
Derry, David Svec, Jason Hom, Robert D. Boutin, Akshay S. Chaudhari | Detecting underdiagnosed medical conditions with opportunistic imaging | null | null | null | null | cs.CV cs.AI cs.LG | http://creativecommons.org/licenses/by/4.0/ | Abdominal computed tomography (CT) scans are frequently performed in clinical
settings. Opportunistic CT involves repurposing routine CT images to extract
diagnostic information and is an emerging tool for detecting underdiagnosed
conditions such as sarcopenia, hepatic steatosis, and ascites. This study
utilizes deep learning methods to promote accurate diagnosis and clinical
documentation. We analyze 2,674 inpatient CT scans to identify discrepancies
between imaging phenotypes (characteristics derived from opportunistic CT
scans) and their corresponding documentation in radiology reports and ICD
coding. Through our analysis, we find that only 0.5%, 3.2%, and 30.7% of scans
diagnosed with sarcopenia, hepatic steatosis, and ascites (respectively)
through either opportunistic imaging or radiology reports were ICD-coded. Our
findings demonstrate opportunistic CT's potential to enhance diagnostic
precision and accuracy of risk adjustment models, offering advancements in
precision medicine.
| [
{
"version": "v1",
"created": "Wed, 18 Sep 2024 03:56:56 GMT"
},
{
"version": "v2",
"created": "Mon, 21 Apr 2025 00:32:35 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Aali",
"Asad",
""
],
[
"Johnston",
"Andrew",
""
],
[
"Blankemeier",
"Louis",
""
],
[
"Van Veen",
"Dave",
""
],
[
"Derry",
"Laura T",
""
],
[
"Svec",
"David",
""
],
[
"Hom",
"Jason",
""
],
[
"Boutin",
"Robert D.",
""
],
[
"Chaudhari",
"Akshay S.",
""
]
] |
2409.13313 | Yuki Uchino | Yuki Uchino, Katsuhisa Ozaki, Toshiyuki Imamura | Performance Enhancement of the Ozaki Scheme on Integer Matrix
Multiplication Unit | null | null | 10.1177/10943420241313064 | null | cs.DC | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | This study was aimed at simultaneously achieving sufficient accuracy and high
performance for general matrix multiplications. Recent architectures, such as
NVIDIA GPUs, feature high-performance units designed for low-precision matrix
multiplications in machine learning models, and next-generation architectures
are expected to follow the same design principle. The key to achieving superior
performance is to fully leverage such architectures. The Ozaki scheme, a highly
accurate matrix multiplication algorithm using error-free transformations,
enables higher-precision matrix multiplication to be performed through multiple
lower-precision matrix multiplications and higher-precision matrix additions.
Ootomo et al. implemented the Ozaki scheme on high-performance matrix
multiplication units with the aim of achieving both sufficient accuracy and
high performance. This paper proposes alternative approaches to improving
performance by reducing the numbers of lower-precision matrix multiplications
and higher-precision matrix additions. Numerical experiments demonstrate the
accuracy of the results and conduct performance benchmarks of the proposed
approaches. These approaches are expected to yield more efficient results in
next-generation architectures.
| [
{
"version": "v1",
"created": "Fri, 20 Sep 2024 08:15:42 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Uchino",
"Yuki",
""
],
[
"Ozaki",
"Katsuhisa",
""
],
[
"Imamura",
"Toshiyuki",
""
]
] |
2409.13998 | Xiaotong Zhang | Xiaotong Zhang, Dingcheng Huang, Kamal Youcef-Toumi | Relevance-driven Decision Making for Safer and More Efficient Human
Robot Collaboration | null | null | null | null | cs.RO cs.AI | http://creativecommons.org/licenses/by/4.0/ | Human brain possesses the ability to effectively focus on important
environmental components, which enhances perception, learning, reasoning, and
decision-making. Inspired by this cognitive mechanism, we introduced a novel
concept termed relevance for Human-Robot Collaboration (HRC). Relevance is a
dimensionality reduction process that incorporates a continuously operating
perception module, evaluates cue sufficiency within the scene, and applies a
flexible formulation and computation framework. In this paper, we present an
enhanced two-loop framework that integrates real-time and asynchronous
processing to quantify relevance and leverage it for safer and more efficient
human-robot collaboration (HRC). The two-loop framework integrates an
asynchronous loop, which leverages LLM world knowledge to quantify relevance,
and a real-time loop, which performs scene understanding, human intent
prediction, and decision-making based on relevance. HRC decision-making is
enhanced by a relevance-based task allocation method, as well as a motion
generation and collision avoidance approach that incorporates human trajectory
prediction. Simulations and experiments show that our methodology for relevance
quantification can accurately and robustly predict the human objective and
relevance, with an average accuracy of up to 0.90 for objective prediction and
up to 0.96 for relevance prediction. Moreover, our motion generation
methodology reduces collision cases by 63.76% and collision frames by 44.74%
when compared with a state-of-the-art (SOTA) collision avoidance method. Our
framework and methodologies, with relevance, guide the robot on how to best
assist humans and generate safer and more efficient actions for HRC.
| [
{
"version": "v1",
"created": "Sat, 21 Sep 2024 03:20:53 GMT"
},
{
"version": "v2",
"created": "Fri, 18 Apr 2025 18:40:16 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Zhang",
"Xiaotong",
""
],
[
"Huang",
"Dingcheng",
""
],
[
"Youcef-Toumi",
"Kamal",
""
]
] |
2409.15318 | Micah Adler Ph.D. | Micah Adler and Nir Shavit | On the Complexity of Neural Computation in Superposition | 32 pages, 6 figures | null | null | null | cs.CC cs.AI cs.DS cs.NE | http://creativecommons.org/licenses/by/4.0/ | Superposition, the ability of neural networks to represent more features than
neurons, is increasingly seen as key to the efficiency of large models. This
paper investigates the theoretical foundations of computing in superposition,
establishing complexity bounds for explicit, provably correct algorithms.
We present the first lower bounds for a neural network computing in
superposition, showing that for a broad class of problems, including
permutations and pairwise logical operations, computing $m'$ features in
superposition requires at least $\Omega(\sqrt{m' \log m'})$ neurons and
$\Omega(m' \log m')$ parameters. This implies the first subexponential upper
bound on superposition capacity: a network with $n$ neurons can compute at most
$O(n^2 / \log n)$ features. Conversely, we provide a nearly tight constructive
upper bound: logical operations like pairwise AND can be computed using
$O(\sqrt{m'} \log m')$ neurons and $O(m' \log^2 m')$ parameters. There is thus
an exponential gap between the complexity of computing in superposition (the
subject of this work) versus merely representing features, which can require as
little as $O(\log m')$ neurons based on the Johnson-Lindenstrauss Lemma.
Our hope is that our results open a path for using complexity theoretic
techniques in neural network interpretability research.
| [
{
"version": "v1",
"created": "Thu, 5 Sep 2024 18:58:59 GMT"
},
{
"version": "v2",
"created": "Fri, 18 Apr 2025 18:13:39 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Adler",
"Micah",
""
],
[
"Shavit",
"Nir",
""
]
] |
2409.15597 | Tingnan Gong | Tingnan Gong, Alon Kipnis, Yao Xie | Higher-criticism for sparse multi-stream change-point detection | Authors are listed in alphabetical order | null | null | null | stat.ME math.ST stat.TH | http://creativecommons.org/licenses/by/4.0/ | We study a statistical procedure based on higher criticism (HC) to address
the sparse multi-stream quickest change-point detection problem. Namely, we aim
to detect a potential change in the distribution of multiple data streams at
some unknown time. If a change occurs, only a few streams are affected, whereas
the identity of the affected streams is unknown. The HC-based procedure
involves testing for a change point in individual streams and combining
multiple tests using higher criticism. Relying on HC thresholding, the
procedure also indicates a set of streams suspected to be affected by the
change. We provide a theoretical analysis under a sparse heteroscedastic normal
change-point model. We establish an information-theoretic detection delay lower
bound when individual tests are based on the likelihood ratio or the
generalized likelihood ratio statistics and show that the delay of the HC-based
method converges in distribution to this bound. In the special case of constant
variance, our bound coincides with known results in (Chan, 2017). We
demonstrate the effectiveness of the HC-based method compared to other methods
in detecting sparse changes through extensive numerical evaluations.
| [
{
"version": "v1",
"created": "Mon, 23 Sep 2024 23:04:47 GMT"
},
{
"version": "v2",
"created": "Sat, 19 Apr 2025 14:53:49 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Gong",
"Tingnan",
""
],
[
"Kipnis",
"Alon",
""
],
[
"Xie",
"Yao",
""
]
] |
2409.16087 | Dev Prakash Jha | Dev Prakash Jha, Raju K. George | Exact Null Controllability of Non-Autonomous Conformable Fractional
Semi-Linear Systems with Nonlocal Conditions | 24 pages, 0 figure. arXiv admin note: text overlap with
arXiv:2408.13814 | null | null | null | math.OC | http://creativecommons.org/licenses/by/4.0/ | We study the exact null controllability of a class of non-autonomous
conformable fractional semi-linear evolution systems with nonlocal initial
conditions in Hilbert spaces. The analysis is carried out within the framework
of conformable fractional calculus and linear evolution operator theory. Under
suitable assumptions, we establish the existence of mild solutions and provide
sufficient conditions for exact null controllability. Notably, the nonlocal
term is allowed to be continuous without requiring compactness or
Lipschitz-type conditions. An example is included to illustrate the
applicability of the main results.
| [
{
"version": "v1",
"created": "Tue, 24 Sep 2024 13:40:05 GMT"
},
{
"version": "v2",
"created": "Sun, 20 Apr 2025 18:10:42 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Jha",
"Dev Prakash",
""
],
[
"George",
"Raju K.",
""
]
] |
2409.17269 | Tadayoshi Kano | Tadayoshi Kano | Tunamis on a deep open sea and on a sloping beach -- a mathematical
theory | null | null | null | null | math.AP physics.ao-ph | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | I describe in this paper how the shallow water waves of Airy, produced by
submarine earthquake, gentle on a deep open sea traveling by the cruising
speed, get instantaneous $ +\infty$ or $-\infty$ traveling speed in approaching
the shallow water near the coasts at the points of the surface on x=X, near the
crest or the trough, where the tangent of itself and the tangent of the seabed
coincide, rushing as inland tunamis ( near the crest) or as offshore tunamis
(near the trough).
| [
{
"version": "v1",
"created": "Sun, 1 Sep 2024 05:35:47 GMT"
},
{
"version": "v2",
"created": "Wed, 23 Oct 2024 10:06:35 GMT"
},
{
"version": "v3",
"created": "Sat, 19 Apr 2025 02:58:40 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Kano",
"Tadayoshi",
""
]
] |
2409.18187 | Mona Merling | Gabriel Angelini-Knoll, Mona Merling, Maximilien P\'eroux | Topological $\Delta G$ homology of rings with twisted $G$-action | Fixed some broken references | null | null | MPIM-Bonn-2024 | math.AT math.KT | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We provide a unifying framework for studying variants of topological
Hochschild homology such as Real topological Hochschild homology. Associated to
a crossed simplicial group $\Delta G$, a category that generalizes Connes'
cyclic category, we introduce an invariant of rings with twisted $G$-action,
which we call topological $\Delta G$-homology. Here a ring with twisted
$G$-action is a common generalization of a ring with involution and a ring with
$G$-action.
Our construction provides homotopical analogues of quaternionic homology,
symmetric homology, braid homology, and hyperoctahedral homology defined by
Fiedorowicz--Loday. In the quaternionic case our construction, called
quaternionic topological Hochschild homology, is equipped with a left
Pin(2)-action and we compute it for loop spaces with a twisted $C_4$-action. As
an important step along the way, we introduce a new family of crossed
simplicial groups called the twisted symmetric crossed simplicial groups and we
prove that they corresponds to operads for algebras with twisted $G$-action.
| [
{
"version": "v1",
"created": "Thu, 26 Sep 2024 18:01:01 GMT"
},
{
"version": "v2",
"created": "Fri, 4 Apr 2025 05:15:54 GMT"
},
{
"version": "v3",
"created": "Sat, 19 Apr 2025 06:50:07 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Angelini-Knoll",
"Gabriel",
""
],
[
"Merling",
"Mona",
""
],
[
"Péroux",
"Maximilien",
""
]
] |
2409.18757 | Takashi Goda | Mou Cai, Takashi Goda, Yoshihito Kazashi | $L_2$-approximation using randomized lattice algorithms | major revision, 24 pages | null | null | null | math.NA cs.NA | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We propose a randomized lattice algorithm for approximating multivariate
periodic functions over the $d$-dimensional unit cube from the weighted Korobov
space with mixed smoothness $\alpha > 1/2$ and product weights
$\gamma_1,\gamma_2,\ldots\in [0,1]$. Building upon the deterministic lattice
algorithm by Kuo, Sloan, and Wo\'{z}niakowski (2006), we incorporate a
randomized quadrature rule by Dick, Goda, and Suzuki (2022) to accelerate the
convergence rate. This randomization involves drawing the number of points for
function evaluations randomly, and selecting a good generating vector for
rank-1 lattice points using the randomized component-by-component algorithm. We
prove that our randomized algorithm achieves a worst-case root mean squared
$L_2$-approximation error of order $M^{-\alpha/2 - 1/8 + \varepsilon}$ for an
arbitrarily small $\varepsilon > 0$, where $M$ denotes the maximum number of
function evaluations, and that the error bound is independent of the dimension
$d$ if the weights satisfy $\sum_{j=1}^\infty \gamma_j^{1/\alpha} < \infty$.
Our upper bound converges faster than a lower bound on the worst-case
$L_2$-approximation error for deterministic rank-1 lattice-based approximation
proved by Byrenheid, K\"{a}mmerer, Ullrich, and Volkmer (2017). We also show a
lower error bound of order $M^{-\alpha/2-1/2}$ for our randomized algorithm,
leaving a slight gap between the upper and lower bounds open for future
research.
| [
{
"version": "v1",
"created": "Fri, 27 Sep 2024 13:49:16 GMT"
},
{
"version": "v2",
"created": "Mon, 21 Apr 2025 05:07:47 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Cai",
"Mou",
""
],
[
"Goda",
"Takashi",
""
],
[
"Kazashi",
"Yoshihito",
""
]
] |
2409.18952 | Andr\'e Silva | Andr\'e Silva, Martin Monperrus | RepairBench: Leaderboard of Frontier Models for Program Repair | null | Proceedings of the International Workshop on Large Language Models
for Code (LLM4Code), 2025 | null | null | cs.SE cs.LG | http://creativecommons.org/licenses/by/4.0/ | AI-driven program repair uses AI models to repair buggy software by producing
patches. Rapid advancements in AI surely impact state-of-the-art performance of
program repair. Yet, grasping this progress requires frequent and standardized
evaluations. We propose RepairBench, a novel leaderboard for AI-driven program
repair. The key characteristics of RepairBench are: 1) it is execution-based:
all patches are compiled and executed against a test suite, 2) it assesses
frontier models in a frequent and standardized way. RepairBench leverages two
high-quality benchmarks, Defects4J and GitBug-Java, to evaluate frontier models
against real-world program repair tasks. We publicly release the evaluation
framework of RepairBench. We will update the leaderboard as new frontier models
are released.
| [
{
"version": "v1",
"created": "Fri, 27 Sep 2024 17:52:34 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Silva",
"André",
""
],
[
"Monperrus",
"Martin",
""
]
] |
2409.19427 | Yuan Zhang | Yuan Zhang | Root group data (RGD) systems of affine type for significant subgroups
of isotropic reductive groups over $k[t,t^{-1}]$ | 25 pages | null | null | null | math.GR | http://creativecommons.org/licenses/by-nc-sa/4.0/ | Given a connected isotropic reductive not necessarily split $k$-group
$\mathcal{G}$ with irreducible relative root system, we construct root group
data (RGD) system of affine type for significant subgroups of
$\mathcal{G}(k[t,t^{-1}])$, which can be extended to the whole group
$\mathcal{G}(k[t,t^{-1}])$ under certain additional requirements. We rely on
the relative pinning maps from paper "Elementary subgroups of isotropic
reductive groups" by V. Petrov and A. Stavrova to construct the affine root
groups. To verify the RGD axioms, we utilize the properties of the affine root
groups, and the properties of reflections associated with the $k$-roots of
$\mathcal{G}$.
| [
{
"version": "v1",
"created": "Sat, 28 Sep 2024 18:00:05 GMT"
},
{
"version": "v2",
"created": "Sat, 19 Apr 2025 22:47:19 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Zhang",
"Yuan",
""
]
] |
2409.19789 | Danny Nackan | Danny Nackan | Interpreting the Ooguri-Vafa symplectic form \`a la Atiyah-Bott | 81 pages, 28 figures; v2: new section about metric, expanded
appendices, and a few clarifications | null | null | null | math.DG hep-th math.AG | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Gaiotto, Moore, and Neitzke predicted that the hyperk\"ahler Ooguri-Vafa
space $\mathcal{M}^{\rm ov}$ should provide a local model for Hitchin moduli
spaces near the discriminant locus. To this end, Tulli identified
$\mathcal{M}^{\rm ov}$ with a certain space of framed Higgs bundles with an
irregular singularity. We extend this result by identifying the Ooguri-Vafa
holomorphic symplectic form with a regularized version of the Atiyah-Bott form
on the associated space of framed connections. We also prove the analogous
statement for the corresponding semiflat forms. Finally, restricting to the
Hitchin section, we identify a regularized version of Hitchin's $L^2$-metric
with the Ooguri-Vafa metric.
| [
{
"version": "v1",
"created": "Sun, 29 Sep 2024 21:23:00 GMT"
},
{
"version": "v2",
"created": "Sun, 20 Apr 2025 22:15:07 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Nackan",
"Danny",
""
]
] |
2409.20006 | L\'eo Mangeolle | L\'eo Mangeolle, Johannes Knolle | Anomalous quantum oscillations from boson-mediated interband scattering | 6 + 10 pages (main text + appendices), 4 figures | null | 10.1103/PhysRevLett.134.146502 | null | cond-mat.str-el cond-mat.mes-hall | http://creativecommons.org/licenses/by/4.0/ | Quantum oscillations (QO) in metals refer to the periodic variation of
thermodynamic and transport properties as a function of inverse applied
magnetic field. QO frequencies are normally associated with semi-classical
trajectories of Fermi surface orbits but recent experiments challenge the
canonical description. We develop a theory of composite frequency quantum
oscillations (CFQO) in two-dimensional Fermi liquids with several Fermi
surfaces and interband scattering mediated by a dynamical boson, e.g. phonons
or spin fluctuations. Specifically, we show that CFQO arise from oscillations
in the fermionic self-energy with anomalous frequency splitting and distinct
strongly non-Lifshitz-Kosevich temperature dependencies. Our theory goes beyond
the framework of semi-classical Fermi surface trajectories highlighting the
role of many-body effects. We provide experimental predictions and discuss the
effect of non-equilibrium boson occupation in driven systems.
| [
{
"version": "v1",
"created": "Mon, 30 Sep 2024 07:00:05 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Mangeolle",
"Léo",
""
],
[
"Knolle",
"Johannes",
""
]
] |
2409.20309 | Suchira Sarkar | Suchira Sarkar, Kanak Saha | Finding massive double-exponential disk galaxies with extended low
surface brightness stellar disk -- an IllustrisTNG exploration | Accepted for publication in ApJ: accepted version; 18 pages, 7
Figures, 3 Tables | null | null | null | astro-ph.GA | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We study massive disk galaxies (stellar mass $>=10^{11}\mathrm{M_{\odot}}$)
at z=0 from IllustrisTNG simulation to detect galaxies that contain two
exponential stellar disks - a central high surface brightness (HSB) disk
surrounded by an extended low surface brightness (LSB) envelope. This is
motivated by the observation of several giant LSB galaxies (Malin 1, UGC 1378,
UGC 1382 etc), reported in the literature, showing such complex morphology.
Studying such systems can help us to understand the mass assembly process and
growth of stellar disks in massive galaxies and thus can shed light on massive
galaxy formation and evolution scenarios. We use the high-resolution
IllustrisTNG50 data and perform Sersic plus exponential profile modeling on the
idealized, synthetic SDSS g, r-band images of the massive disk galaxies using
GALFIT. We identify 7 disk galaxies (12% of the parent sample of disk galaxies)
that are best represented by a central S\'{e}rsic plus a central HSB disk
surrounded by an extended LSB disk. The radial scale lengths of the LSB disk
lie in the range of $\sim$ 9.7-31.7 kpc, in agreement with that found in the
literature. We study the star formation properties of these simulated
double-disk galaxies to understand the distribution of these from blue
star-forming to red-quenched region. Some of these double-disk galaxies display
a characteristic minima in their (g-r) color radial profiles. The double-disk
galaxies are found to lie within $\sim$ 1.5-$\mathrm{\sigma}$ region of the
Baryonic Tully-Fisher relation from observation.
| [
{
"version": "v1",
"created": "Mon, 30 Sep 2024 14:08:48 GMT"
},
{
"version": "v2",
"created": "Sat, 19 Apr 2025 10:05:10 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Sarkar",
"Suchira",
""
],
[
"Saha",
"Kanak",
""
]
] |
2410.01540 | Jente Vandersanden | Jente Vandersanden, Sascha Holl, Xingchang Huang, Gurprit Singh | Edge-preserving noise for diffusion models | null | null | null | null | cs.CV cs.AI cs.GR cs.LG | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Classical generative diffusion models learn an isotropic Gaussian denoising
process, treating all spatial regions uniformly, thus neglecting potentially
valuable structural information in the data. Inspired by the long-established
work on anisotropic diffusion in image processing, we present a novel
edge-preserving diffusion model that generalizes over existing isotropic models
by considering a hybrid noise scheme. In particular, we introduce an edge-aware
noise scheduler that varies between edge-preserving and isotropic Gaussian
noise. We show that our model's generative process converges faster to results
that more closely match the target distribution. We demonstrate its capability
to better learn the low-to-mid frequencies within the dataset, which plays a
crucial role in representing shapes and structural information. Our
edge-preserving diffusion process consistently outperforms state-of-the-art
baselines in unconditional image generation. It is also particularly more
robust for generative tasks guided by a shape-based prior, such as
stroke-to-image generation. We present qualitative and quantitative results
(FID and CLIP score) showing consistent improvements of up to 30% for both
tasks.
| [
{
"version": "v1",
"created": "Wed, 2 Oct 2024 13:29:52 GMT"
},
{
"version": "v2",
"created": "Fri, 25 Oct 2024 09:44:10 GMT"
},
{
"version": "v3",
"created": "Sat, 19 Apr 2025 12:43:13 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Vandersanden",
"Jente",
""
],
[
"Holl",
"Sascha",
""
],
[
"Huang",
"Xingchang",
""
],
[
"Singh",
"Gurprit",
""
]
] |
2410.01734 | Najib Khachiaa | Najib Khachiaa | Admissibility conditions for Multi-window super Gabor systems on
discrete periodic sets | arXiv admin note: text overlap with arXiv:2409.03423 | null | null | null | math.FA | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | In this paper, $\mathcal{G}(g,L,M,N,R)\subset
\ell^2(\mathbb{S},\mathbb{C}^R)$ denotes a $L-$window super Gabor system on a
periodic set $\mathbb{S}$, where $L,M,N,R\in \mathbb{N}$ and $g=\{g_l\}_{l\in
\mathbb{N}_L}\subset \ell^2(\mathbb{S},\mathbb{C}^R)$. We characterize which
$g$ generates a complete multi-window super Gabor system and a multi-window
super Gabor frame $\mathcal{G}(g,L,M,N,R)$ on $\mathbb{S}$ using the
vector-valued Zak transform. Admissibility conditions for a periodic set to
admit a complete multi-window super Gabor system, multi-window super Gabor
(Parseval) frame, and multi--window super Gabor (orthonormal) basis
$\mathcal{G}(g,L,M,N,R)$ are given with respect to the parameters $L$, $M$, $N$
and $R$.
| [
{
"version": "v1",
"created": "Wed, 2 Oct 2024 16:46:21 GMT"
},
{
"version": "v2",
"created": "Wed, 8 Jan 2025 20:53:04 GMT"
},
{
"version": "v3",
"created": "Mon, 21 Apr 2025 12:41:19 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Khachiaa",
"Najib",
""
]
] |
2410.02073 | Stephan R Richter | Aleksei Bochkovskii and Ama\"el Delaunoy and Hugo Germain and Marcel
Santos and Yichao Zhou and Stephan R. Richter and Vladlen Koltun | Depth Pro: Sharp Monocular Metric Depth in Less Than a Second | Published at ICLR 2025. Code and weights available at
https://github.com/apple/ml-depth-pro | null | null | null | cs.CV cs.LG | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We present a foundation model for zero-shot metric monocular depth
estimation. Our model, Depth Pro, synthesizes high-resolution depth maps with
unparalleled sharpness and high-frequency details. The predictions are metric,
with absolute scale, without relying on the availability of metadata such as
camera intrinsics. And the model is fast, producing a 2.25-megapixel depth map
in 0.3 seconds on a standard GPU. These characteristics are enabled by a number
of technical contributions, including an efficient multi-scale vision
transformer for dense prediction, a training protocol that combines real and
synthetic datasets to achieve high metric accuracy alongside fine boundary
tracing, dedicated evaluation metrics for boundary accuracy in estimated depth
maps, and state-of-the-art focal length estimation from a single image.
Extensive experiments analyze specific design choices and demonstrate that
Depth Pro outperforms prior work along multiple dimensions. We release code and
weights at https://github.com/apple/ml-depth-pro
| [
{
"version": "v1",
"created": "Wed, 2 Oct 2024 22:42:20 GMT"
},
{
"version": "v2",
"created": "Mon, 21 Apr 2025 12:09:08 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Bochkovskii",
"Aleksei",
""
],
[
"Delaunoy",
"Amaël",
""
],
[
"Germain",
"Hugo",
""
],
[
"Santos",
"Marcel",
""
],
[
"Zhou",
"Yichao",
""
],
[
"Richter",
"Stephan R.",
""
],
[
"Koltun",
"Vladlen",
""
]
] |
2410.02253 | Yueyuan Li | Yueyuan Li, Mingyang Jiang, Songan Zhang, Wei Yuan, Chunxiang Wang,
and Ming Yang | From Imitation to Exploration: End-to-end Autonomous Driving based on
World Model | 12 pages, 4 figures, 3 tables; T-ITS under review | null | null | null | cs.AI cs.LG cs.RO | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | In recent years, end-to-end autonomous driving architectures have gained
increasing attention due to their advantage in avoiding error accumulation.
Most existing end-to-end autonomous driving methods are based on Imitation
Learning (IL), which can quickly derive driving strategies by mimicking expert
behaviors. However, IL often struggles to handle scenarios outside the training
dataset, especially in high-dynamic and interaction-intensive traffic
environments. In contrast, Reinforcement Learning (RL)-based driving models can
optimize driving decisions through interaction with the environment, improving
adaptability and robustness.
To leverage the strengths of both IL and RL, we propose RAMBLE, an end-to-end
world model-based RL method for driving decision-making. RAMBLE extracts
environmental context information from RGB images and LiDAR data through an
asymmetrical variational autoencoder. A transformer-based architecture is then
used to capture the dynamic transitions of traffic participants. Next, an
actor-critic structure reinforcement learning algorithm is applied to derive
driving strategies based on the latent features of the current state and
dynamics. To accelerate policy convergence and ensure stable training, we
introduce a training scheme that initializes the policy network using IL, and
employs KL loss and soft update mechanisms to smoothly transition the model
from IL to RL.
RAMBLE achieves state-of-the-art performance in route completion rate on the
CARLA Leaderboard 1.0 and completes all 38 scenarios on the CARLA Leaderboard
2.0, demonstrating its effectiveness in handling complex and dynamic traffic
scenarios. The model will be open-sourced upon paper acceptance at
https://github.com/SCP-CN-001/ramble to support further research and
development in autonomous driving.
| [
{
"version": "v1",
"created": "Thu, 3 Oct 2024 06:45:59 GMT"
},
{
"version": "v2",
"created": "Sun, 20 Apr 2025 06:05:58 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Li",
"Yueyuan",
""
],
[
"Jiang",
"Mingyang",
""
],
[
"Zhang",
"Songan",
""
],
[
"Yuan",
"Wei",
""
],
[
"Wang",
"Chunxiang",
""
],
[
"Yang",
"Ming",
""
]
] |
2410.02291 | Chi-Thiem Hoang | Thiem Hoang (KASI & UST) | Spin-Polarized Electrons from Magnetically Aligned Grains and Chiral
Symmetry Breaking: Effects of Cosmic Rays in Protostellar Environments | 23 pages, 9 figures. Accepted to ApJ | null | null | null | astro-ph.GA astro-ph.EP astro-ph.HE astro-ph.SR | http://creativecommons.org/licenses/by/4.0/ | Low-energy spin-polarized electrons (SPEs) are thought to cause symmetry
breaking and could explain the origin of homochirality of prebiotic molecules
such as amino acids and sugars. Here we study the effect of cosmic rays (CRs)
on the emission of SPEs from aligned grains in dense protostellar environments
and explore their effects on chiral asymmetry of prebiotic molecules. We first
show that icy grains in protostellar environments can align with magnetic
fields due to magnetically enhanced radiative torque mechanism. We then study
the production of thermal electrons by CR ionization of H$_2$ and the
CR-induced UV radiation using the attenuated CR spectra in dense cores obtained
from a continuous slowing down model. Next, we show that thermal electrons with
initial random spins captured by aligned grains will become spin-polarized due
to the Barnett effect, converting unpolarized electrons into SPEs. We calculate
the rate of photoemission of such SPEs by CRs-induced UV radiation and
secondary electron emission from aligned grains and find that the photoemission
by CRs-induced UV radiation is dominant. Finally, we calculate the total
production rate of SPEs inside aligned dust grains by CRs. We estimate the
alignment degree of SPEs from superparmagnetic (SPM) grains and find that it is
only significant for SPM grains having large iron clusters and fast rotation.
We suggest that low-energy secondary SPEs from aligned superparamagnetic grains
with large iron inclusions induced by CRs might cause the chiral asymmetry of
chiral prebiotic molecules formed in the ice mantle of aligned grains, in
analogous to UV circularly polarized light.
| [
{
"version": "v1",
"created": "Thu, 3 Oct 2024 08:22:41 GMT"
},
{
"version": "v2",
"created": "Tue, 8 Oct 2024 06:55:12 GMT"
},
{
"version": "v3",
"created": "Mon, 21 Apr 2025 05:55:13 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Hoang",
"Thiem",
"",
"KASI & UST"
]
] |
2410.03052 | Siqi Zeng | Siqi Zeng, Sixian Du, Makoto Yamada, Han Zhao | Learning Structured Representations by Embedding Class Hierarchy with
Fast Optimal Transport | 31 pages, 12 figures. Published as a conference paper at ICLR '25 | null | null | null | cs.LG | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | To embed structured knowledge within labels into feature representations,
prior work [Zeng et al., 2022] proposed to use the Cophenetic Correlation
Coefficient (CPCC) as a regularizer during supervised learning. This
regularizer calculates pairwise Euclidean distances of class means and aligns
them with the corresponding shortest path distances derived from the label
hierarchy tree. However, class means may not be good representatives of the
class conditional distributions, especially when they are multi-mode in nature.
To address this limitation, under the CPCC framework, we propose to use the
Earth Mover's Distance (EMD) to measure the pairwise distances among classes in
the feature space. We show that our exact EMD method generalizes previous work,
and recovers the existing algorithm when class-conditional distributions are
Gaussian. To further improve the computational efficiency of our method, we
introduce the Optimal Transport-CPCC family by exploring four EMD approximation
variants. Our most efficient OT-CPCC variant, the proposed Fast FlowTree
algorithm, runs in linear time in the size of the dataset, while maintaining
competitive performance across datasets and tasks. The code is available at
https://github.com/uiuctml/OTCPCC.
| [
{
"version": "v1",
"created": "Fri, 4 Oct 2024 00:20:34 GMT"
},
{
"version": "v2",
"created": "Sun, 20 Apr 2025 01:00:27 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Zeng",
"Siqi",
""
],
[
"Du",
"Sixian",
""
],
[
"Yamada",
"Makoto",
""
],
[
"Zhao",
"Han",
""
]
] |
2410.03102 | Haein Kong | Eunbin Ha, Haein Kong, Shagun Jhaver | Examining Racial Stereotypes in YouTube Autocomplete Suggestions | null | null | null | null | cs.CY | http://creativecommons.org/licenses/by-nc-sa/4.0/ | Autocomplete is a popular search feature that predicts queries based on user
input and guides users to a set of potentially relevant suggestions. In this
study, we examine what YouTube autocompletes suggest to users seeking
information about race on the platform. Specifically, we perform an algorithm
output audit of autocomplete suggestions for input queries about four racial
groups and examine the stereotypes they embody. Using critical discourse
analysis, we identify five major sociocultural contexts in which racial
information appears -Appearance, Ability, Culture, Social Equity, and Manner.
We found that the participatory nature of YouTube produces a multifaceted
representation of race-related content in its search outputs, characterized by
enduring historical biases, aggregated discrimination, and interracial
tensions, while simultaneously depicting minority resistance and aspirations of
a post-racial society. We call for innovations in content moderation policy
design and enforcement to address existing racial harms in YouTube search
outputs.
| [
{
"version": "v1",
"created": "Fri, 4 Oct 2024 02:53:25 GMT"
},
{
"version": "v2",
"created": "Sun, 23 Feb 2025 23:45:43 GMT"
},
{
"version": "v3",
"created": "Sun, 20 Apr 2025 23:16:46 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Ha",
"Eunbin",
""
],
[
"Kong",
"Haein",
""
],
[
"Jhaver",
"Shagun",
""
]
] |
2410.03955 | Gang Li | Gang Li, Wendi Yu, Yao Yao, Wei Tong, Yingbin Liang, Qihang Lin,
Tianbao Yang | A Retention-Centric Framework for Continual Learning with Guaranteed
Model Developmental Safety | 44 pages, 7 figures | null | null | null | cs.LG cs.AI math.OC stat.ML | http://creativecommons.org/licenses/by/4.0/ | In real-world applications, learning-enabled systems often undergo iterative
model development to address challenging or emerging tasks, which involve
collecting new data, training a new model and validating the model. This
continual model development process raises a significant issue that acquiring
new or improving existing capabilities may inadvertently lose good capabilities
of the old model, also known as catastrophic forgetting. While existing
continual learning aims to mitigate catastrophic forgetting by trading off
performance on previous tasks and new tasks to ensure good average performance,
it often falls short in cost-sensitive applications, where failing to preserve
essential established capabilities introduces unforeseen costs and risks and
substantial expenses for re-improving these capabilities. To address this
issue, we impose a requirement on learning systems to ensure that a new model
strictly retains important capabilities of the old model while improving
target-task performance, which we term model developmental safety. To ensure
model developmental safety, we propose a retention-centric framework with
data-dependent constraints, and study how to continually develop a pretrained
CLIP model for acquiring new or improving existing capabilities of image
classification. We propose an efficient constrained optimization algorithm with
theoretical guarantees and use its insights to finetune the CLIP model with
task-dependent heads for promoting the model developmental safety. Experiments
on autonomous driving and scene recognition datasets validate the efficacy of
our method.
| [
{
"version": "v1",
"created": "Fri, 4 Oct 2024 22:34:58 GMT"
},
{
"version": "v2",
"created": "Sun, 13 Oct 2024 00:48:38 GMT"
},
{
"version": "v3",
"created": "Thu, 12 Dec 2024 15:43:14 GMT"
},
{
"version": "v4",
"created": "Sat, 19 Apr 2025 02:00:47 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Li",
"Gang",
""
],
[
"Yu",
"Wendi",
""
],
[
"Yao",
"Yao",
""
],
[
"Tong",
"Wei",
""
],
[
"Liang",
"Yingbin",
""
],
[
"Lin",
"Qihang",
""
],
[
"Yang",
"Tianbao",
""
]
] |
2410.04161 | Keda Tao | Keda Tao, Jinjin Gu, Yulun Zhang, Xiucheng Wang, Nan Cheng | Overcoming False Illusions in Real-World Face Restoration with
Multi-Modal Guided Diffusion Model | 23 Pages, 28 Figures, ICLR 2025 | null | null | null | cs.CV | http://creativecommons.org/licenses/by/4.0/ | We introduce a novel Multi-modal Guided Real-World Face Restoration (MGFR)
technique designed to improve the quality of facial image restoration from
low-quality inputs. Leveraging a blend of attribute text prompts, high-quality
reference images, and identity information, MGFR can mitigate the generation of
false facial attributes and identities often associated with generative face
restoration methods. By incorporating a dual-control adapter and a two-stage
training strategy, our method effectively utilizes multi-modal prior
information for targeted restoration tasks. We also present the Reface-HQ
dataset, comprising over 21,000 high-resolution facial images across 4800
identities, to address the need for reference face training images. Our
approach achieves superior visual quality in restoring facial details under
severe degradation and allows for controlled restoration processes, enhancing
the accuracy of identity preservation and attribute correction. Including
negative quality samples and attribute prompts in the training further refines
the model's ability to generate detailed and perceptually accurate images.
| [
{
"version": "v1",
"created": "Sat, 5 Oct 2024 13:46:56 GMT"
},
{
"version": "v2",
"created": "Mon, 21 Apr 2025 11:47:05 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Tao",
"Keda",
""
],
[
"Gu",
"Jinjin",
""
],
[
"Zhang",
"Yulun",
""
],
[
"Wang",
"Xiucheng",
""
],
[
"Cheng",
"Nan",
""
]
] |
2410.04178 | Chunlei Yang | Lili Du and Chunlei Yang | Proof of the Stokes conjecture for compressible gravity water waves | Any comments are welcome | null | null | null | math.AP | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | In 1880, Stokes examined an incompressible irrotational periodic traveling
water wave under the influence of gravity and conjectured the existence of an
extreme wave with a corner of $120^{\circ}$ at the crest. The first rigorous
proof of the conjecture was given by Amick, Fraenkel and Toland, as well as by
Plotnikov independently via the Nekrasov integral equation. In the early 2010s,
Weiss and Varvarucva revisited the conjecture by applying a new geometric
method, which provided an affirmative answer to the conjecture without
requiring structural assumptions such as the isolation of the stagnation
points, the symmetry and the monotonicity of the free surface that were
necessary in the previous works.
The main purpose of this paper is to establish the validity of the Stokes
conjecture in the context of compressible gravity water waves. More precisely,
we prove that a sharp crest forms near each stagnation point of a compressible
gravity water wave with an included angle of $120^{\circ}$, which gives a first
proof to the compressible counterpart of the classical conjecture by Stokes in
1880.
The central aspect of our approach is the discovery of a new monotonicity
formula for quasilinear free boundary problems of the Bernoulli-type. Another
observation is the introduction of a new nonlinear frequency formula, along
with a compensated compactness argument for the compressible Euler system. The
developed monotonicity formula enables us to do blow-up analysis at each
stagnation point and helps us obtain the singular profile of the free surface
near each stagnation points. The degenerate stagnation points can be further
analyzed with the help of the compensated compactness argument using the
frequency formula.
| [
{
"version": "v1",
"created": "Sat, 5 Oct 2024 14:50:35 GMT"
},
{
"version": "v2",
"created": "Mon, 21 Apr 2025 07:30:34 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Du",
"Lili",
""
],
[
"Yang",
"Chunlei",
""
]
] |
2410.04450 | Tony Huynh | Katie Clinch and Sean Dewar and Niloufar Fuladi and Maximilian Gorsky
and Tony Huynh and Eleftherios Kastis and Atsuhiro Nakamoto and Anthony Nixon
and Brigitte Servatius | Triangulated spheres with holes in triangulated surfaces | Paper has been completely rewritten to address an error in the
previous version. Lemma 2.1 (and all subsequent theorems which relied upon
it) is incorrect. We thank Atsuhiro Nakamoto for alerting us of the error.
The current version has a new title, two new results, and Atsuhiro Nakamoto
as a new co-author. 8 pages, 0 figures | null | null | null | math.CO cs.DM | http://creativecommons.org/licenses/by/4.0/ | Let $\mathbb{S}_h$ denote a sphere with $h$ holes. Given a triangulation $G$
of a surface $\mathbb{M}$, we consider the question of when $G$ contains a
spanning subgraph $H$ such that $H$ is a triangulated $\mathbb{S}_h$. We give a
new short proof of a theorem of Nevo and Tarabykin that every triangulation $G$
of the torus contains a spanning subgraph which is a triangulated cylinder. For
arbitrary surfaces, we prove that every high facewidth triangulation of a
surface with $h$ handles contains a spanning subgraph which is a triangulated
$\mathbb{S}_{2h}$. We also prove that for every $0 \leq g' < g$ and $w \in
\mathbb{N}$, there exists a triangulation of facewidth at least $w$ of a
surface of Euler genus $g$ that does not have a spanning subgraph which is a
triangulated $\mathbb{S}_{g'}$. Our results are motivated by, and have
applications for, rigidity questions in the plane.
| [
{
"version": "v1",
"created": "Sun, 6 Oct 2024 11:18:48 GMT"
},
{
"version": "v2",
"created": "Sun, 20 Apr 2025 22:04:53 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Clinch",
"Katie",
""
],
[
"Dewar",
"Sean",
""
],
[
"Fuladi",
"Niloufar",
""
],
[
"Gorsky",
"Maximilian",
""
],
[
"Huynh",
"Tony",
""
],
[
"Kastis",
"Eleftherios",
""
],
[
"Nakamoto",
"Atsuhiro",
""
],
[
"Nixon",
"Anthony",
""
],
[
"Servatius",
"Brigitte",
""
]
] |
2410.04585 | Pengcheng Jiang | Pengcheng Jiang, Cao Xiao, Minhao Jiang, Parminder Bhatia, Taha
Kass-Hout, Jimeng Sun, Jiawei Han | Reasoning-Enhanced Healthcare Predictions with Knowledge Graph Community
Retrieval | ICLR 2025 Camera-Ready | null | null | null | cs.CL | http://creativecommons.org/licenses/by/4.0/ | Large language models (LLMs) have demonstrated significant potential in
clinical decision support. Yet LLMs still suffer from hallucinations and lack
fine-grained contextual medical knowledge, limiting their high-stake healthcare
applications such as clinical diagnosis. Traditional retrieval-augmented
generation (RAG) methods attempt to address these limitations but frequently
retrieve sparse or irrelevant information, undermining prediction accuracy. We
introduce KARE, a novel framework that integrates knowledge graph (KG)
community-level retrieval with LLM reasoning to enhance healthcare predictions.
KARE constructs a comprehensive multi-source KG by integrating biomedical
databases, clinical literature, and LLM-generated insights, and organizes it
using hierarchical graph community detection and summarization for precise and
contextually relevant information retrieval. Our key innovations include: (1) a
dense medical knowledge structuring approach enabling accurate retrieval of
relevant information; (2) a dynamic knowledge retrieval mechanism that enriches
patient contexts with focused, multi-faceted medical insights; and (3) a
reasoning-enhanced prediction framework that leverages these enriched contexts
to produce both accurate and interpretable clinical predictions. Extensive
experiments demonstrate that KARE outperforms leading models by up to
10.8-15.0% on MIMIC-III and 12.6-12.7% on MIMIC-IV for mortality and
readmission predictions. In addition to its impressive prediction accuracy, our
framework leverages the reasoning capabilities of LLMs, enhancing the
trustworthiness of clinical predictions.
| [
{
"version": "v1",
"created": "Sun, 6 Oct 2024 18:46:28 GMT"
},
{
"version": "v2",
"created": "Sun, 20 Apr 2025 05:49:38 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Jiang",
"Pengcheng",
""
],
[
"Xiao",
"Cao",
""
],
[
"Jiang",
"Minhao",
""
],
[
"Bhatia",
"Parminder",
""
],
[
"Kass-Hout",
"Taha",
""
],
[
"Sun",
"Jimeng",
""
],
[
"Han",
"Jiawei",
""
]
] |
2410.05298 | Xinnan Dai | Xinnan Dai, Haohao Qu, Yifen Shen, Bohang Zhang, Qihao Wen, Wenqi Fan,
Dongsheng Li, Jiliang Tang, Caihua Shan | How Do Large Language Models Understand Graph Patterns? A Benchmark for
Graph Pattern Comprehension | The paper is published in ICLR 2025 | null | null | null | cs.LG cs.AI | http://creativecommons.org/licenses/by/4.0/ | Benchmarking the capabilities and limitations of large language models (LLMs)
in graph-related tasks is becoming an increasingly popular and crucial area of
research. Recent studies have shown that LLMs exhibit a preliminary ability to
understand graph structures and node features. However, the potential of LLMs
in graph pattern mining remains largely unexplored. This is a key component in
fields such as computational chemistry, biology, and social network analysis.
To bridge this gap, this work introduces a comprehensive benchmark to assess
LLMs' capabilities in graph pattern tasks. We have developed a benchmark that
evaluates whether LLMs can understand graph patterns based on either
terminological or topological descriptions. Additionally, our benchmark tests
the LLMs' capacity to autonomously discover graph patterns from data. The
benchmark encompasses both synthetic and real datasets, and a variety of
models, with a total of 11 tasks and 7 models. Our experimental framework is
designed for easy expansion to accommodate new models and datasets. Our
findings reveal that: (1) LLMs have preliminary abilities to understand graph
patterns, with O1-mini outperforming in the majority of tasks; (2) Formatting
input data to align with the knowledge acquired during pretraining can enhance
performance; (3) The strategies employed by LLMs may differ from those used in
conventional algorithms.
| [
{
"version": "v1",
"created": "Fri, 4 Oct 2024 04:48:33 GMT"
},
{
"version": "v2",
"created": "Sun, 20 Apr 2025 21:31:45 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Dai",
"Xinnan",
""
],
[
"Qu",
"Haohao",
""
],
[
"Shen",
"Yifen",
""
],
[
"Zhang",
"Bohang",
""
],
[
"Wen",
"Qihao",
""
],
[
"Fan",
"Wenqi",
""
],
[
"Li",
"Dongsheng",
""
],
[
"Tang",
"Jiliang",
""
],
[
"Shan",
"Caihua",
""
]
] |
2410.05413 | Aleksey Valouev | Aleksey Valouev | Implicitly Learned Neural Phase Functions for Basis-Free Point Spread
Function Engineering | 3 pages, 7 figures. To be published in ICVISP 2024
(https://www.icvisp.org/) | null | 10.1109/ICVISP64524.2024.10959478 | null | physics.optics cs.CV | http://creativecommons.org/licenses/by/4.0/ | Point spread function (PSF) engineering is vital for precisely controlling
the focus of light in computational imaging, with applications in neural
imaging, fluorescence microscopy, and biophotonics. The PSF is derived from the
magnitude of the Fourier transform of a phase function, making the construction
of the phase function given the PSF (PSF engineering) an ill-posed inverse
problem. Traditional PSF engineering methods rely on physical basis functions,
limiting their ability to generalize across the range of PSFs required for
imaging tasks. We introduce a novel approach leveraging implicit neural
representations that overcome the limitations of pixel-wise optimization
methods. Our approach achieves a median MSSIM of 0.8162 and a mean MSSIM of
0.5634, compared to a median MSSIM of 0.0 and a mean MSSIM of 0.1841 with
pixel-wise optimization when learning randomly generated phase functions. Our
approach also achieves a median PSNR of 10.38 dB and a mean PSNR of 8.672 dB,
compared to a median PSNR of 6.653 dB and a mean PSNR of 6.660 dB with
pixel-wise optimization for this task.
| [
{
"version": "v1",
"created": "Mon, 7 Oct 2024 18:26:30 GMT"
},
{
"version": "v2",
"created": "Fri, 11 Oct 2024 05:19:26 GMT"
},
{
"version": "v3",
"created": "Tue, 21 Jan 2025 05:43:51 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Valouev",
"Aleksey",
""
]
] |
2410.05520 | Roberto De Leo | Chirag Adwani and Roberto De Leo and James A. Yorke | What is the graph of a dynamical system? | 30 pages, 10 figures | null | null | null | math.DS nlin.CD | http://creativecommons.org/licenses/by-nc-sa/4.0/ | Some of the basic properties of any dynamical system can be summarized by a
graph. The dynamical systems in our theory run from maps like the logistic map
to ordinary differential equations to dissipative partial differential
equations. Our goal has been to define a meaningful concept of graph of any
dynamical system. As a result, we base our definition of ``chain graph'' on
``epsilon-chains'', defining both nodes and edges of the graph in terms of
chains. In particular, nodes are often maximal limit sets and there is an edge
between two nodes if there is a trajectory whose forward limit set is in one
node and its backward limit set is in the other. Our initial goal was to prove
that every ``chain graph'' of a dynamical system is, in some sense, connected,
and we prove connectedness under mild hypotheses.
| [
{
"version": "v1",
"created": "Mon, 7 Oct 2024 21:53:37 GMT"
},
{
"version": "v2",
"created": "Sat, 19 Apr 2025 13:56:55 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Adwani",
"Chirag",
""
],
[
"De Leo",
"Roberto",
""
],
[
"Yorke",
"James A.",
""
]
] |
2410.06090 | Hanxiao Wang | Bowen Ma and Hanxiao Wang | Solvability of Equilibrium Riccati Equations: A Direct Approach | Although we provide a new approach to the ERE associated
time-inconsistent stochastic LQ controls, the main result of this manuscript
has been obtained by Yong [29]. Thus, we just maintain it as a preprint for
the interested readers, and will not publish it officially | null | null | null | math.OC | http://creativecommons.org/licenses/by/4.0/ | The solvability of equilibrium Riccati equations (EREs) plays a central role
in the study of time-inconsistent stochastic linear-quadratic optimal control
problems, because it paves the way to constructing a closed-loop equilibrium
strategy. Under the standard conditions, Yong [29] established its
well-posedness by introducing the well-known multi-person differential game
method. However, this method depends on the dynamic programming principle (DPP)
of the sophisticated problems on every subinterval, and thus is essentially a
control theory approach. In this paper, we shall give a new and more direct
proof, in which the DPP is no longer needed. We first establish a priori
estimates for the ERE in the case of smooth coefficients. Using this estimate,
we then demonstrate both the local and global solvability of the ERE by
constructing an appropriate Picard iteration sequence, which actually provides
a numerical algorithm. Additionally, a mollification method is employed to
handle the case with non-smooth coefficients.
| [
{
"version": "v1",
"created": "Tue, 8 Oct 2024 14:47:58 GMT"
},
{
"version": "v2",
"created": "Sat, 19 Apr 2025 13:00:03 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Ma",
"Bowen",
""
],
[
"Wang",
"Hanxiao",
""
]
] |
2410.06159 | Shiv Singh | Manasa Manasa, Mohammad Azam, Tatiana Zajarniuk, Svitlana Stelmakh,
Taras Palasyuk, Jan Mizeracki, Tomasz Cetner, Andrzej Morawski, Cezariusz
Jastrz\k{e}bski, Micha{\l} Wierzbicki, Andrzej Wi\'sniewski, Shiv J. Singh | Praseodymium doping effect on the superconducting properties of
FeSe0.5Te0.5 bulks under ambient and high-pressure growth conditions | 50 pages, 1 table, 15 figures | Physica C: Superconductivity and its Applications (2025) | 10.1016/j.physc.2025.1354729 | null | cond-mat.supr-con cond-mat.mtrl-sci physics.app-ph | http://creativecommons.org/licenses/by/4.0/ | A series of Pr-doped FeSe0.5Te0.5 (Fe1-xPrxSe0.5Te0.5; x = 0 to 0.3) bulks
are prepared by conventional synthesis process at ambient pressure (CSP), and
high gas pressure and high temperature synthesis (HP-HTS) methods. These bulks
are well characterized by structural and microstructural analysis, Raman
spectroscopy, transport, and magnetic measurements. The HP-HTS process of the
parent bulks has enhanced the onset transition temperature (Tconset) by 1.5 K
and the critical current density (Jc) by two orders of magnitude compared to
the CSP method. Pr-doped FeSe0.5Te0.5 up to 10% doping content prepared, either
CSP or HP-HTS, slightly increases the unit cell volume, and high-pressure
growth produces an almost pure superconducting phase, which confirms the
successful Pr-doping at Fe sites. Raman spectroscopy measurements and DFT
calculations suggest the substitution of Pr-atoms in the interlayer spacing of
Fe(Se,Te) lattice. High-pressure growth of Fe1-xPrxSe0.5Te0.5 also makes the
sample less dense compared to the parent sample grown by HP-HTS. Transport and
magnetic measurements depict that Tconset is almost unaffected by Pr doping,
whereas Jc of Pr-doped FeSe0.5Te0.5 is enhanced by one order of magnitude
relative to the parent sample developed by CSP but lower than that of the
parent sample grown by HP-HTS. Hence, Pr doping at Fe sites preserves Tconset
and improves Jc of FeSe0.5Te0.5 regardless of the doping contents and growth
conditions. These results are promising for the practical application of
iron-based superconductors to improve Jc properties without affecting Tconset
through CSP process and congruent with discoveries from other superconductors,
like cuprates and MgB2.
| [
{
"version": "v1",
"created": "Tue, 8 Oct 2024 16:05:27 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Manasa",
"Manasa",
""
],
[
"Azam",
"Mohammad",
""
],
[
"Zajarniuk",
"Tatiana",
""
],
[
"Stelmakh",
"Svitlana",
""
],
[
"Palasyuk",
"Taras",
""
],
[
"Mizeracki",
"Jan",
""
],
[
"Cetner",
"Tomasz",
""
],
[
"Morawski",
"Andrzej",
""
],
[
"Jastrzębski",
"Cezariusz",
""
],
[
"Wierzbicki",
"Michał",
""
],
[
"Wiśniewski",
"Andrzej",
""
],
[
"Singh",
"Shiv J.",
""
]
] |
2410.06404 | Masataka Kuwamura | Hideo Ikeda, Masataka Kuwamura | Stability of Single Transition Layer in Mass-Conserving
Reaction-Diffusion Systems with Bistable Nonlinearity | 50 pages, 1 figure. arXiv admin note: text overlap with
arXiv:2305.00227 | null | null | null | math.AP | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Mass-conserving reaction-diffusion systems with bistable nonlinearity are
considered under general assumptions. The existence of stationary solutions
with a single internal transition layer in such reaction-diffusion systems is
shown using the analytical singular perturbation theory. Moreover, a stability
criterion for the stationary solutions is provided by calculating the Evans
function.
| [
{
"version": "v1",
"created": "Tue, 8 Oct 2024 22:25:32 GMT"
},
{
"version": "v2",
"created": "Mon, 21 Apr 2025 04:22:16 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Ikeda",
"Hideo",
""
],
[
"Kuwamura",
"Masataka",
""
]
] |
2410.06407 | Yingyu Lin | Yingyu Lin, Yuxing Huang, Wenqin Liu, Haoran Deng, Ignavier Ng, Kun
Zhang, Mingming Gong, Yi-An Ma, Biwei Huang | A Skewness-Based Criterion for Addressing Heteroscedastic Noise in
Causal Discovery | null | null | null | null | cs.LG stat.ME stat.ML | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Real-world data often violates the equal-variance assumption
(homoscedasticity), making it essential to account for heteroscedastic noise in
causal discovery. In this work, we explore heteroscedastic symmetric noise
models (HSNMs), where the effect $Y$ is modeled as $Y = f(X) + \sigma(X)N$,
with $X$ as the cause and $N$ as independent noise following a symmetric
distribution. We introduce a novel criterion for identifying HSNMs based on the
skewness of the score (i.e., the gradient of the log density) of the data
distribution. This criterion establishes a computationally tractable
measurement that is zero in the causal direction but nonzero in the anticausal
direction, enabling the causal direction discovery. We extend this
skewness-based criterion to the multivariate setting and propose SkewScore, an
algorithm that handles heteroscedastic noise without requiring the extraction
of exogenous noise. We also conduct a case study on the robustness of SkewScore
in a bivariate model with a latent confounder, providing theoretical insights
into its performance. Empirical studies further validate the effectiveness of
the proposed method.
| [
{
"version": "v1",
"created": "Tue, 8 Oct 2024 22:28:30 GMT"
},
{
"version": "v2",
"created": "Mon, 21 Apr 2025 00:04:26 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Lin",
"Yingyu",
""
],
[
"Huang",
"Yuxing",
""
],
[
"Liu",
"Wenqin",
""
],
[
"Deng",
"Haoran",
""
],
[
"Ng",
"Ignavier",
""
],
[
"Zhang",
"Kun",
""
],
[
"Gong",
"Mingming",
""
],
[
"Ma",
"Yi-An",
""
],
[
"Huang",
"Biwei",
""
]
] |
2410.07582 | Gyuwan Kim | Gyuwan Kim, Yang Li, Evangelia Spiliopoulou, Jie Ma, Miguel
Ballesteros, William Yang Wang | Detecting Training Data of Large Language Models via Expectation
Maximization | 15 pages | null | null | null | cs.CL cs.AI cs.CR cs.LG | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | The advancement of large language models has grown parallel to the opacity of
their training data. Membership inference attacks (MIAs) aim to determine
whether specific data was used to train a model. They offer valuable insights
into detecting data contamination and ensuring compliance with privacy and
copyright standards. However, MIA for LLMs is challenging due to the massive
scale of training data and the inherent ambiguity of membership in texts.
Moreover, creating realistic MIA evaluation benchmarks is difficult as training
and test data distributions are often unknown. We introduce EM-MIA, a novel
membership inference method that iteratively refines membership scores and
prefix scores via an expectation-maximization algorithm. Our approach leverages
the observation that these scores can improve each other: membership scores
help identify effective prefixes for detecting training data, while prefix
scores help determine membership. As a result, EM-MIA achieves state-of-the-art
results on WikiMIA. To enable comprehensive evaluation, we introduce OLMoMIA, a
benchmark built from OLMo resources, which allows controlling task difficulty
through varying degrees of overlap between training and test data
distributions. Our experiments demonstrate EM-MIA is robust across different
scenarios while also revealing fundamental limitations of current MIA
approaches when member and non-member distributions are nearly identical.
| [
{
"version": "v1",
"created": "Thu, 10 Oct 2024 03:31:16 GMT"
},
{
"version": "v2",
"created": "Mon, 21 Apr 2025 02:22:06 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Kim",
"Gyuwan",
""
],
[
"Li",
"Yang",
""
],
[
"Spiliopoulou",
"Evangelia",
""
],
[
"Ma",
"Jie",
""
],
[
"Ballesteros",
"Miguel",
""
],
[
"Wang",
"William Yang",
""
]
] |
2410.07848 | Roohan Ahmed Khan | Roohan Ahmed Khan, Malaika Zafar, Amber Batool, Aleksey Fedoseev,
Dzmitry Tsetserukou | SwarmPath: Drone Swarm Navigation through Cluttered Environments
Leveraging Artificial Potential Field and Impedance Control | Manuscript accepted in IEEE International Conference on Robotics and
Biomimetics (IEEE ROBIO 2024) | null | 10.1109/ROBIO64047.2024.10907517 | null | cs.RO | http://creativecommons.org/licenses/by-nc-nd/4.0/ | In the area of multi-drone systems, navigating through dynamic environments
from start to goal while providing collision-free trajectory and efficient path
planning is a significant challenge. To solve this problem, we propose a novel
SwarmPath technology that involves the integration of Artificial Potential
Field (APF) with Impedance Controller. The proposed approach provides a
solution based on collision free leader-follower behaviour where drones are
able to adapt themselves to the environment. Moreover, the leader is virtual
while drones are physical followers leveraging APF path planning approach to
find the smallest possible path to the target. Simultaneously, the drones
dynamically adjust impedance links, allowing themselves to create virtual links
with obstacles to avoid them. As compared to conventional APF, the proposed
SwarmPath system not only provides smooth collision-avoidance but also enable
agents to efficiently pass through narrow passages by reducing the total travel
time by 30% while ensuring safety in terms of drones connectivity. Lastly, the
results also illustrate that the discrepancies between simulated and real
environment, exhibit an average absolute percentage error (APE) of 6% of drone
trajectories. This underscores the reliability of our solution in real-world
scenarios.
| [
{
"version": "v1",
"created": "Thu, 10 Oct 2024 12:14:28 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Khan",
"Roohan Ahmed",
""
],
[
"Zafar",
"Malaika",
""
],
[
"Batool",
"Amber",
""
],
[
"Fedoseev",
"Aleksey",
""
],
[
"Tsetserukou",
"Dzmitry",
""
]
] |
2410.08448 | Zhongzheng Tang | Xujin Chen, Xiaodong Hu, Xinqi Jing, Zhongzheng Tang | When does additional information lead to longer travel time in
multi-origin-destination networks? | null | null | null | null | cs.GT | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | The Informational Braess' Paradox (IBP) illustrates a counterintuitive
scenario where revelation of additional roadway segments to some
self-interested travelers leads to increased travel times for these
individuals. IBP extends the original Braess' paradox by relaxing the
assumption that all travelers have identical and complete information about the
network. In this paper, we study the conditions under which IBP does not occur
in networks with non-atomic selfish travelers and multiple origin-destination
pairs. Our results completely characterize the network topologies immune to
IBP, thus resolving an open question proposed by Acemoglu et al.
| [
{
"version": "v1",
"created": "Fri, 11 Oct 2024 01:37:15 GMT"
},
{
"version": "v2",
"created": "Mon, 21 Apr 2025 09:01:02 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Chen",
"Xujin",
""
],
[
"Hu",
"Xiaodong",
""
],
[
"Jing",
"Xinqi",
""
],
[
"Tang",
"Zhongzheng",
""
]
] |
2410.08700 | Christoph D\"opmann | Christoph D\"opmann, Maximilian Weisenseel, Florian Tschorsch | Progressive Pruning: Analyzing the Impact of Intersection Attacks | null | null | null | null | cs.CR cs.NI | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Stream-based communication dominates today's Internet, posing unique
challenges for anonymous communication networks (ACNs). Traditionally designed
for independent messages, ACNs struggle to account for the inherent
vulnerabilities of streams, such as susceptibility to intersection attacks. In
this work, we address this gap and introduce progressive pruning, a novel
methodology for quantifying the susceptibility to intersection attacks.
Progressive pruning quantifies and monitors anonymity sets over time, providing
an assessment of an adversary's success in correlating senders and receivers.
We leverage this methodology to analyze synthetic scenarios and large-scale
simulations of the Tor network using our newly developed TorFS simulator. Our
findings reveal that anonymity is significantly influenced by stream length,
user population, and stream distribution across the network. These insights
highlight critical design challenges for future ACNs seeking to safeguard
stream-based communication against traffic analysis attacks.
| [
{
"version": "v1",
"created": "Fri, 11 Oct 2024 10:40:51 GMT"
},
{
"version": "v2",
"created": "Mon, 21 Apr 2025 17:06:04 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Döpmann",
"Christoph",
""
],
[
"Weisenseel",
"Maximilian",
""
],
[
"Tschorsch",
"Florian",
""
]
] |
2410.08775 | Ningji Wei | Ningji Wei | Convexoid: A Minimal Theory of Conjugate Convexity | null | null | null | null | math.OC | http://creativecommons.org/licenses/by-nc-nd/4.0/ | A key idea in convex optimization theory is to use well-structured affine
functions to approximate general functions, leading to impactful developments
in conjugate functions and convex duality theory. This raises the question:
what are the minimal requirements to establish these results? This paper aims
to address this inquiry through a carefully crafted system called the
convexoid. We demonstrate that fundamental constructs, such as conjugate
functions and subdifferentials, along with their relationships, can be derived
within this minimal system. Building on this, we define the associated duality
systems and develop conditions for weak and strong duality, generalizing the
classic results from conjugate duality and radial duality theories. Due to its
flexibility, our framework supports various approximation schemes, including
approximating general functions using symmetric-conic, bilinear, radial, or
piecewise constant functions, and representing general structures such as
graphs, set systems, fuzzy sets, or toposes using special membership functions.
The associated duality results for these systems also open new opportunities
for establishing bounds on objective values and verifying structural
properties.
| [
{
"version": "v1",
"created": "Fri, 11 Oct 2024 12:46:23 GMT"
},
{
"version": "v2",
"created": "Wed, 23 Oct 2024 17:08:00 GMT"
},
{
"version": "v3",
"created": "Sat, 19 Apr 2025 14:14:43 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Wei",
"Ningji",
""
]
] |
2410.08858 | YuQing Nie | Yuqing Nie, Chong Wang, Kailong Wang, Guoai Xu, Guosheng Xu and Haoyu
Wang | Decoding Secret Memorization in Code LLMs Through Token-Level
Characterization | 13 pages, 8 figures | null | null | null | cs.CR cs.SE | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Code Large Language Models (LLMs) have demonstrated remarkable capabilities
in generating, understanding, and manipulating programming code. However, their
training process inadvertently leads to the memorization of sensitive
information, posing severe privacy risks. Existing studies on memorization in
LLMs primarily rely on prompt engineering techniques, which suffer from
limitations such as widespread hallucination and inefficient extraction of the
target sensitive information. In this paper, we present a novel approach to
characterize real and fake secrets generated by Code LLMs based on token
probabilities. We identify four key characteristics that differentiate genuine
secrets from hallucinated ones, providing insights into distinguishing real and
fake secrets. To overcome the limitations of existing works, we propose DESEC,
a two-stage method that leverages token-level features derived from the
identified characteristics to guide the token decoding process. DESEC consists
of constructing an offline token scoring model using a proxy Code LLM and
employing the scoring model to guide the decoding process by reassigning token
likelihoods. Through extensive experiments on four state-of-the-art Code LLMs
using a diverse dataset, we demonstrate the superior performance of DESEC in
achieving a higher plausible rate and extracting more real secrets compared to
existing baselines. Our findings highlight the effectiveness of our token-level
approach in enabling an extensive assessment of the privacy leakage risks
associated with Code LLMs.
| [
{
"version": "v1",
"created": "Fri, 11 Oct 2024 14:39:24 GMT"
},
{
"version": "v2",
"created": "Sun, 20 Apr 2025 05:41:14 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Nie",
"Yuqing",
""
],
[
"Wang",
"Chong",
""
],
[
"Wang",
"Kailong",
""
],
[
"Xu",
"Guoai",
""
],
[
"Xu",
"Guosheng",
""
],
[
"Wang",
"Haoyu",
""
]
] |
2410.09300 | Yu Fei | Yu Fei, Yasaman Razeghi, Sameer Singh | Nudging: Inference-time Alignment of LLMs via Guided Decoding | null | null | null | null | cs.CL cs.AI cs.LG | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Large language models (LLMs) require alignment to effectively and safely
follow user instructions. This process necessitates training an aligned version
for every base model, resulting in significant computational overhead. In this
work, we propose nudging, a simple, plug-and-play, and training-free algorithm
that aligns any base model at inference time using a small aligned model.
Nudging is motivated by recent findings that alignment primarily alters the
model's behavior on a small subset of stylistic tokens (e.g., discourse
markers). We find that base models are significantly more uncertain when
generating these tokens. Building on this insight, nudging employs a small
aligned model to generate nudging tokens to guide the base model's output
during decoding when the base model's uncertainty is high. We evaluate nudging
across 3 model families on a diverse range of open-instruction tasks. Without
any training, nudging a large base model with a 7x-14x smaller aligned model
achieves zero-shot performance comparable to, and sometimes surpassing, that of
large aligned models. By operating at the token level, nudging enables
off-the-shelf collaboration between model families. For instance, nudging
Gemma-2-27b with Llama-2-7b-chat outperforms Llama-2-70b-chat on various tasks.
Overall, our work offers a modular and cost-efficient solution to LLM
alignment. Our project website: https://fywalter.github.io/nudging/ .
| [
{
"version": "v1",
"created": "Fri, 11 Oct 2024 23:24:38 GMT"
},
{
"version": "v2",
"created": "Tue, 15 Oct 2024 01:07:01 GMT"
},
{
"version": "v3",
"created": "Sun, 20 Apr 2025 00:16:18 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Fei",
"Yu",
""
],
[
"Razeghi",
"Yasaman",
""
],
[
"Singh",
"Sameer",
""
]
] |
2410.09344 | Wenlong Deng | Wenlong Deng, Yize Zhao, Vala Vakilian, Minghui Chen, Xiaoxiao Li,
Christos Thrampoulidis | DARE the Extreme: Revisiting Delta-Parameter Pruning For Fine-Tuned
Models | null | null | null | null | cs.LG cs.AI cs.CL | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Storing open-source fine-tuned models separately introduces redundancy and
increases response times in applications utilizing multiple models.
Delta-parameter pruning (DPP), particularly the random drop and rescale (DARE)
method proposed by Yu et al., addresses this by pruning the majority of delta
parameters--the differences between fine-tuned and pre-trained model
weights--while typically maintaining minimal performance loss. However, DARE
fails when either the pruning rate or the magnitude of the delta parameters is
large. We highlight two key reasons for this failure: (1) an excessively large
rescaling factor as pruning rates increase, and (2) high mean and variance in
the delta parameters. To push DARE's limits, we introduce DAREx (DARE the
eXtreme), which features two algorithmic improvements: (1) DAREx-q, a rescaling
factor modification that significantly boosts performance at high pruning rates
(e.g., >30 % on COLA and SST2 for encoder models, with even greater gains in
decoder models), and (2) DAREx-L2, which combines DARE with AdamR, an
in-training method that applies appropriate delta regularization before DPP. We
also demonstrate that DAREx-q can be seamlessly combined with vanilla
parameter-efficient fine-tuning techniques like LoRA and can facilitate
structural DPP. Additionally, we revisit the application of importance-based
pruning techniques within DPP, demonstrating that they outperform random-based
methods when delta parameters are large. Through this comprehensive study, we
develop a pipeline for selecting the most appropriate DPP method under various
practical scenarios.
| [
{
"version": "v1",
"created": "Sat, 12 Oct 2024 03:21:58 GMT"
},
{
"version": "v2",
"created": "Sun, 20 Apr 2025 20:53:39 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Deng",
"Wenlong",
""
],
[
"Zhao",
"Yize",
""
],
[
"Vakilian",
"Vala",
""
],
[
"Chen",
"Minghui",
""
],
[
"Li",
"Xiaoxiao",
""
],
[
"Thrampoulidis",
"Christos",
""
]
] |
2410.09511 | Cihan Pazarba\c{s}{\i} | Tatsuhiro Misumi, Cihan Pazarba\c{s}{\i} | Exact WKB in all sectors I: Potentials with degenerate saddles | 62 pages, 18 figures, v3: comments added | null | null | null | hep-th math-ph math.MP quant-ph | http://creativecommons.org/licenses/by/4.0/ | We explore the exact-WKB (EWKB) method through the analysis of Airy and Weber
types, with an emphasis on the exact quantization of locally harmonic
potentials in multiple sectors. The core innovation of our work lies in
introducing a novel complexification approach to the energy parameter $u$,
distinct from the common complexification of the (semi-classical) expansion
parameter used in Borel summability. This new technique allows for continuous
analytical continuation across different sectors of a potential while
maintaining the exact quantization condition, even before median summation. By
redefining the $A$-cycle above the potential barrier top, we ensure the
quantization condition remains real and, by use of the Stokes automorphism and
the median resummation, show that the resurgence structure is preserved across
transitions between sectors. Furthermore, we discuss the Weber-type exact-WKB
method, offering exact estimates for quantum actions around all types of saddle
points, generalizing previous results. Through the analysis of these quantum
actions, we reveal the presence of $S$-duality, facilitating the exchange
between perturbative and non-perturbative behaviors, and we conjecture the
mapping of the P-NP relations between dual theories. Our study encompasses
periodic and symmetric double-well potentials, demonstrating that the exact-WKB
method captures intricate structures in quantum systems in all sectors,
including multi-instanton contributions and the resurgence of quantum actions.
| [
{
"version": "v1",
"created": "Sat, 12 Oct 2024 12:20:47 GMT"
},
{
"version": "v2",
"created": "Sun, 17 Nov 2024 12:13:31 GMT"
},
{
"version": "v3",
"created": "Sat, 19 Apr 2025 12:14:02 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Misumi",
"Tatsuhiro",
""
],
[
"Pazarbaşı",
"Cihan",
""
]
] |
2410.09732 | Weijia Li | Junyan Ye, Baichuan Zhou, Zilong Huang, Junan Zhang, Tianyi Bai,
Hengrui Kang, Jun He, Honglin Lin, Zihao Wang, Tong Wu, Zhizheng Wu, Yiping
Chen, Dahua Lin, Conghui He, Weijia Li | LOKI: A Comprehensive Synthetic Data Detection Benchmark using Large
Multimodal Models | ICLR 2025 SPOTLIGHT, 83 pages, 63 figures | null | null | null | cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | With the rapid development of AI-generated content, the future internet may
be inundated with synthetic data, making the discrimination of authentic and
credible multimodal data increasingly challenging. Synthetic data detection has
thus garnered widespread attention, and the performance of large multimodal
models (LMMs) in this task has attracted significant interest. LMMs can provide
natural language explanations for their authenticity judgments, enhancing the
explainability of synthetic content detection. Simultaneously, the task of
distinguishing between real and synthetic data effectively tests the
perception, knowledge, and reasoning capabilities of LMMs. In response, we
introduce LOKI, a novel benchmark designed to evaluate the ability of LMMs to
detect synthetic data across multiple modalities. LOKI encompasses video,
image, 3D, text, and audio modalities, comprising 18K carefully curated
questions across 26 subcategories with clear difficulty levels. The benchmark
includes coarse-grained judgment and multiple-choice questions, as well as
fine-grained anomaly selection and explanation tasks, allowing for a
comprehensive analysis of LMMs. We evaluated 22 open-source LMMs and 6
closed-source models on LOKI, highlighting their potential as synthetic data
detectors and also revealing some limitations in the development of LMM
capabilities. More information about LOKI can be found at
https://opendatalab.github.io/LOKI/
| [
{
"version": "v1",
"created": "Sun, 13 Oct 2024 05:26:36 GMT"
},
{
"version": "v2",
"created": "Mon, 21 Apr 2025 02:36:09 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Ye",
"Junyan",
""
],
[
"Zhou",
"Baichuan",
""
],
[
"Huang",
"Zilong",
""
],
[
"Zhang",
"Junan",
""
],
[
"Bai",
"Tianyi",
""
],
[
"Kang",
"Hengrui",
""
],
[
"He",
"Jun",
""
],
[
"Lin",
"Honglin",
""
],
[
"Wang",
"Zihao",
""
],
[
"Wu",
"Tong",
""
],
[
"Wu",
"Zhizheng",
""
],
[
"Chen",
"Yiping",
""
],
[
"Lin",
"Dahua",
""
],
[
"He",
"Conghui",
""
],
[
"Li",
"Weijia",
""
]
] |
2410.10519 | Noemi B\"uhrer | Noemi B\"uhrer, Sa\'ul Alonso-Monsalve, Matthew Franks, Till
Dieminger, Davide Sgalaberna | AI-based particle track identification in scintillating fibres read out
with imaging sensors | 23 pages, 13 figures | null | null | null | cs.LG hep-ex physics.ins-det | http://creativecommons.org/licenses/by/4.0/ | This paper presents the development and application of an AI-based method for
particle track identification using scintillating fibres read out with imaging
sensors. We propose a variational autoencoder (VAE) to efficiently filter and
identify frames containing signal from the substantial data generated by SPAD
array sensors. Our VAE model, trained on purely background frames, demonstrated
a high capability to distinguish frames containing particle tracks from
background noise. The performance of the VAE-based anomaly detection was
validated with experimental data, demonstrating the method's ability to
efficiently identify relevant events with rapid processing time, suggesting a
solid prospect for deployment as a fast inference tool on hardware for
real-time anomaly detection. This work highlights the potential of combining
advanced sensor technology with machine learning techniques to enhance particle
detection and tracking.
| [
{
"version": "v1",
"created": "Mon, 14 Oct 2024 13:59:30 GMT"
},
{
"version": "v2",
"created": "Sat, 19 Apr 2025 09:04:10 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Bührer",
"Noemi",
""
],
[
"Alonso-Monsalve",
"Saúl",
""
],
[
"Franks",
"Matthew",
""
],
[
"Dieminger",
"Till",
""
],
[
"Sgalaberna",
"Davide",
""
]
] |
2410.10733 | Han Cai | Junyu Chen, Han Cai, Junsong Chen, Enze Xie, Shang Yang, Haotian Tang,
Muyang Li, Yao Lu, Song Han | Deep Compression Autoencoder for Efficient High-Resolution Diffusion
Models | ICLR 2025. The first two authors contributed equally to this work | null | null | null | cs.CV cs.AI | http://creativecommons.org/licenses/by/4.0/ | We present Deep Compression Autoencoder (DC-AE), a new family of autoencoder
models for accelerating high-resolution diffusion models. Existing autoencoder
models have demonstrated impressive results at a moderate spatial compression
ratio (e.g., 8x), but fail to maintain satisfactory reconstruction accuracy for
high spatial compression ratios (e.g., 64x). We address this challenge by
introducing two key techniques: (1) Residual Autoencoding, where we design our
models to learn residuals based on the space-to-channel transformed features to
alleviate the optimization difficulty of high spatial-compression autoencoders;
(2) Decoupled High-Resolution Adaptation, an efficient decoupled three-phases
training strategy for mitigating the generalization penalty of high
spatial-compression autoencoders. With these designs, we improve the
autoencoder's spatial compression ratio up to 128 while maintaining the
reconstruction quality. Applying our DC-AE to latent diffusion models, we
achieve significant speedup without accuracy drop. For example, on ImageNet
512x512, our DC-AE provides 19.1x inference speedup and 17.9x training speedup
on H100 GPU for UViT-H while achieving a better FID, compared with the widely
used SD-VAE-f8 autoencoder. Our code is available at
https://github.com/mit-han-lab/efficientvit.
| [
{
"version": "v1",
"created": "Mon, 14 Oct 2024 17:15:07 GMT"
},
{
"version": "v2",
"created": "Mon, 11 Nov 2024 17:42:37 GMT"
},
{
"version": "v3",
"created": "Tue, 10 Dec 2024 16:39:23 GMT"
},
{
"version": "v4",
"created": "Tue, 14 Jan 2025 16:47:44 GMT"
},
{
"version": "v5",
"created": "Fri, 17 Jan 2025 14:22:06 GMT"
},
{
"version": "v6",
"created": "Wed, 26 Feb 2025 17:56:06 GMT"
},
{
"version": "v7",
"created": "Mon, 21 Apr 2025 14:48:23 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Chen",
"Junyu",
""
],
[
"Cai",
"Han",
""
],
[
"Chen",
"Junsong",
""
],
[
"Xie",
"Enze",
""
],
[
"Yang",
"Shang",
""
],
[
"Tang",
"Haotian",
""
],
[
"Li",
"Muyang",
""
],
[
"Lu",
"Yao",
""
],
[
"Han",
"Song",
""
]
] |
2410.10744 | Hossein Mirzaei | Hossein Mirzaei, Mackenzie W. Mathis | Adversarially Robust Out-of-Distribution Detection Using
Lyapunov-Stabilized Embeddings | Accepted at the International Conference on Learning Representations
(ICLR) 2025. Code and pre-trained models are available at
https://github.com/AdaptiveMotorControlLab/AROS | ICLR 2025 | null | null | cs.LG cs.CV stat.ML | http://creativecommons.org/licenses/by/4.0/ | Despite significant advancements in out-of-distribution (OOD) detection,
existing methods still struggle to maintain robustness against adversarial
attacks, compromising their reliability in critical real-world applications.
Previous studies have attempted to address this challenge by exposing detectors
to auxiliary OOD datasets alongside adversarial training. However, the
increased data complexity inherent in adversarial training, and the myriad of
ways that OOD samples can arise during testing, often prevent these approaches
from establishing robust decision boundaries. To address these limitations, we
propose AROS, a novel approach leveraging neural ordinary differential
equations (NODEs) with Lyapunov stability theorem in order to obtain robust
embeddings for OOD detection. By incorporating a tailored loss function, we
apply Lyapunov stability theory to ensure that both in-distribution (ID) and
OOD data converge to stable equilibrium points within the dynamical system.
This approach encourages any perturbed input to return to its stable
equilibrium, thereby enhancing the model's robustness against adversarial
perturbations. To not use additional data, we generate fake OOD embeddings by
sampling from low-likelihood regions of the ID data feature space,
approximating the boundaries where OOD data are likely to reside. To then
further enhance robustness, we propose the use of an orthogonal binary layer
following the stable feature space, which maximizes the separation between the
equilibrium points of ID and OOD samples. We validate our method through
extensive experiments across several benchmarks, demonstrating superior
performance, particularly under adversarial attacks. Notably, our approach
improves robust detection performance from 37.8% to 80.1% on CIFAR-10 vs.
CIFAR-100 and from 29.0% to 67.0% on CIFAR-100 vs. CIFAR-10.
| [
{
"version": "v1",
"created": "Mon, 14 Oct 2024 17:22:12 GMT"
},
{
"version": "v2",
"created": "Sun, 26 Jan 2025 17:35:40 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Mirzaei",
"Hossein",
""
],
[
"Mathis",
"Mackenzie W.",
""
]
] |
2410.10796 | Sachin Goyal | Sachin Goyal, Christina Baek, J. Zico Kolter, Aditi Raghunathan | Context-Parametric Inversion: Why Instruction Finetuning Can Worsen
Context Reliance | Published at ICLR 2025 (Oral) | null | null | null | cs.LG cs.CL | http://creativecommons.org/licenses/by-nc-sa/4.0/ | A standard practice when using large language models is for users to
supplement their instruction with an input context containing new information
for the model to process. However, models struggle to reliably follow the input
context, especially when it conflicts with their parametric knowledge from
pretraining. In-principle, one would expect models to adapt to the user context
better after instruction finetuning, particularly when handling knowledge
conflicts. However, we observe a surprising failure mode: during instruction
tuning, the context reliance under knowledge conflicts initially increases as
expected, but then gradually decreases as instruction finetuning progresses.
This happens while the performance on standard benchmarks keeps on increasing
far after this drop. We call this phenomenon context-parametric inversion and
observe it across multiple general purpose instruction tuning datasets such as
TULU, Alpaca and Ultrachat, across different model families like Llama,
Mistral, and Pythia. We perform various controlled studies and theoretical
analysis to show that context-parametric inversion occurs due to examples in
the instruction finetuning data where the input context provides information
that aligns with model's parametric knowledge. Our analysis suggests some
natural mitigation strategies with limited but insightful gains, and serves as
a useful starting point in addressing this deficiency in instruction
finetuning.
| [
{
"version": "v1",
"created": "Mon, 14 Oct 2024 17:57:09 GMT"
},
{
"version": "v2",
"created": "Tue, 22 Oct 2024 17:35:03 GMT"
},
{
"version": "v3",
"created": "Mon, 21 Apr 2025 10:19:21 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Goyal",
"Sachin",
""
],
[
"Baek",
"Christina",
""
],
[
"Kolter",
"J. Zico",
""
],
[
"Raghunathan",
"Aditi",
""
]
] |
2410.10893 | Alexej Schelle Dr. | Alexej Schelle | Physical aspects of symmetry breaking in Bose gases at thermal
equilibrium | 10 pages, 11 figures | Whitepaper on arXiv.org (2025) | null | null | cond-mat.quant-gas quant-ph | http://creativecommons.org/licenses/by/4.0/ | The theory of non-interacting Bose gases is supplemented by a numerical
quantum field description with a two-dimensional non-local order parameter that
allows the modeling of wave-like atomic correlations and interference effects
in the limit of low atomic densities. From the present model, it is possible to
explain symmetry aspects of non-interacting and very weakly interacting Bose
gases in the limit of fluctuating particle numbers, like the forward
propagation of time and the relation to the breaking and preservation of phase
gauge symmetry in solids. In the present formalism, the propagation of
one-directional time arises from the pre-defined and equivalent convergence of
independent quantum fields towards the Boltzmann equilibrium, and it is shown
that Glauber coherent states are related to the definition of the quantized
field. Coherently coupling condensate and non-condensate parts as a direct
consequence of the increasing quantum coherence time between the different
quantum field components in the Bose gas from cooling to below the critical
temperature, the present model describes symmetry breaking, which is originally
known from the definition of a specific gauge field from Elitzur's theorem for
local gauge fields, as a global physical rather than a purely formal
mathematical process.
| [
{
"version": "v1",
"created": "Sat, 12 Oct 2024 08:55:34 GMT"
},
{
"version": "v2",
"created": "Wed, 12 Feb 2025 08:59:38 GMT"
},
{
"version": "v3",
"created": "Mon, 10 Mar 2025 16:22:11 GMT"
},
{
"version": "v4",
"created": "Mon, 21 Apr 2025 09:23:34 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Schelle",
"Alexej",
""
]
] |
2410.11129 | Parham Poureslami | Parham Poureslami, Mohammad Majidi, Javad Ranjbar Kermani, Mohamad Ali
Bijarchi | Deformation and breakup of a ferrofluid compound droplet migrating in a
microchannel under a magnetic field: A phase-field-based multiple-relaxation
time lattice Boltzmann study | 63 pages, 26 figures, 3 tables (Accepted for publication in the a
Journal of Fluid Mechanics on April 12 2025) | null | null | null | physics.flu-dyn | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Though ubiquitous in many engineering applications, including drug delivery,
the compound droplet hydrodynamics in confined geometries have been barely
surveyed. For the first time, this study thoroughly investigates the
hydrodynamics of a ferrofluid compound droplet (FCD) during its migration in a
microchannel under the presence of a pressure-driven flow and a uniform
external magnetic field (UEMF) to manipulate its morphology and retard its
breakup. Finite difference and phase-field multiple-relaxation time lattice
Boltzmann approaches are coupled to determine the magnetic field and ternary
flow system, respectively. Firstly, the influence of the magnetic Bond number
(Bo) on the FCD morphology is explored depending on whether the core or shell
is ferrofluid when the UEMF is applied along {\alpha}=0{\deg} and
{\alpha}=90{\deg} relative to the fluid flow. It is ascertained that imposing
the UEMF at {\alpha}=0{\deg} when the shell is ferrofluid can postpone the
breakup. Intriguingly, when the core is ferrofluid, strengthening the UEMF
enlarges the shell deformation. Afterward, the effects of the Capillary number
(Ca), density ratio, viscosity ratio, radius ratio, and surface tension
coefficients are scrutinized on the FCD deformation and breakup. The results
indicate that augmenting the core-to-shell viscosity and density ratios
accelerates the breakup process. Additionally, surface tension between the core
and shell suppresses the core deformation. Moreover, increasing the Ca number
intensifies the viscous drag force exerted on the shell, flattening its rear
side, which causes a triangular-like configuration. Ultimately, by varying Bo
and Ca numbers, five distinguished regimes are observed, whose regime map is
established.
| [
{
"version": "v1",
"created": "Mon, 14 Oct 2024 22:59:40 GMT"
},
{
"version": "v2",
"created": "Thu, 17 Oct 2024 04:32:57 GMT"
},
{
"version": "v3",
"created": "Fri, 18 Apr 2025 20:40:50 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Poureslami",
"Parham",
""
],
[
"Majidi",
"Mohammad",
""
],
[
"Kermani",
"Javad Ranjbar",
""
],
[
"Bijarchi",
"Mohamad Ali",
""
]
] |
2410.11201 | Tong Ding | Tong Ding, Wanhua Li, Zhongqi Miao, Hanspeter Pfister | Tree of Attributes Prompt Learning for Vision-Language Models | null | null | null | null | cs.CV cs.AI cs.LG | http://creativecommons.org/licenses/by/4.0/ | Prompt learning has proven effective in adapting vision language models for
downstream tasks. However, existing methods usually append learnable prompt
tokens solely with the category names to obtain textual features, which fails
to fully leverage the rich context indicated in the category name. To address
this issue, we propose the Tree of Attributes Prompt learning (TAP), which
first instructs LLMs to generate a tree of attributes with a "concept -
attribute - description" structure for each category, and then learn the
hierarchy with vision and text prompt tokens. Unlike existing methods that
merely augment category names with a set of unstructured descriptions, our
approach essentially distills structured knowledge graphs associated with class
names from LLMs. Furthermore, our approach introduces text and vision prompts
designed to explicitly learn the corresponding visual attributes, effectively
serving as domain experts. Additionally, the general and diverse descriptions
generated based on the class names may be wrong or absent in the specific given
images. To address this misalignment, we further introduce a vision-conditional
pooling module to extract instance-specific text features. Extensive
experimental results demonstrate that our approach outperforms state-of-the-art
methods on the zero-shot base-to-novel generalization, cross-dataset transfer,
as well as few-shot classification across 11 diverse datasets. Code is
available at https://github.com/HHenryD/TAP.
| [
{
"version": "v1",
"created": "Tue, 15 Oct 2024 02:37:39 GMT"
},
{
"version": "v2",
"created": "Mon, 21 Apr 2025 15:37:50 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Ding",
"Tong",
""
],
[
"Li",
"Wanhua",
""
],
[
"Miao",
"Zhongqi",
""
],
[
"Pfister",
"Hanspeter",
""
]
] |
2410.11673 | Jia Liu | Zhong Yangjie and Liu Jia and Luo Peng and Ke Yan and Cai Shen | INR-Based Generative Steganography by Point Cloud Representation | 10pages,13figures | null | null | null | cs.CR cs.CY | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Generative steganography (GS) directly generates stego-media through secret
message-driven generation. It makes the hiding capacity of GS higher than that
of traditional steganography, as well as more resistant to classical
steganalysis. However, the generators and extractors of existing GS methods can
only target specific formats and types of data and lack of universality.
Besides, the model size is usually related to the underlying grid resolution,
and the transmission behavior of the extractor is susceptible to suspicion of
steganalysis. Implicit neural representation(INR) is a technique for
representing data in a continuous manner. Inspired by this, we propose an
INR-based generative steganography by point cloud representation (INR-GSPC). By
using the function generator, the problem of the generator model size growing
exponentially with the increase of gridded data has been solved. That is able
to generate a wide range of data types and break through the limitation of
resolution. In order to unify the data formats of the generator and message
extractor, the data is converted to point cloud representation. We designed and
fixed a point cloud message extractor. By iterating over the point cloud with
adding small perturbations to generate stego-media. This method can avoid the
training and transmission process of the message extractor. To the best of our
knowledge, this is the first method to apply point cloud to generative
steganography. Experiments demonstrate that the stego-images generated by the
scheme have an average PSNR value of more than 65, and the accuracy of message
extraction reaches more than 99%.
| [
{
"version": "v1",
"created": "Tue, 15 Oct 2024 15:06:13 GMT"
},
{
"version": "v2",
"created": "Tue, 22 Oct 2024 09:10:00 GMT"
},
{
"version": "v3",
"created": "Sun, 20 Apr 2025 11:18:25 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Yangjie",
"Zhong",
""
],
[
"Jia",
"Liu",
""
],
[
"Peng",
"Luo",
""
],
[
"Yan",
"Ke",
""
],
[
"Shen",
"Cai",
""
]
] |
2410.11689 | Hikaru Shindo | Hikaru Shindo, Quentin Delfosse, Devendra Singh Dhami, Kristian
Kersting | BlendRL: A Framework for Merging Symbolic and Neural Policy Learning | ICLR 2025 (Spotlight) | null | null | null | cs.LG cs.AI | http://creativecommons.org/licenses/by/4.0/ | Humans can leverage both symbolic reasoning and intuitive reactions. In
contrast, reinforcement learning policies are typically encoded in either
opaque systems like neural networks or symbolic systems that rely on predefined
symbols and rules. This disjointed approach severely limits the agents'
capabilities, as they often lack either the flexible low-level reaction
characteristic of neural agents or the interpretable reasoning of symbolic
agents. To overcome this challenge, we introduce BlendRL, a neuro-symbolic RL
framework that harmoniously integrates both paradigms within RL agents that use
mixtures of both logic and neural policies. We empirically demonstrate that
BlendRL agents outperform both neural and symbolic baselines in standard Atari
environments, and showcase their robustness to environmental changes.
Additionally, we analyze the interaction between neural and symbolic policies,
illustrating how their hybrid use helps agents overcome each other's
limitations.
| [
{
"version": "v1",
"created": "Tue, 15 Oct 2024 15:24:20 GMT"
},
{
"version": "v2",
"created": "Mon, 21 Apr 2025 16:49:31 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Shindo",
"Hikaru",
""
],
[
"Delfosse",
"Quentin",
""
],
[
"Dhami",
"Devendra Singh",
""
],
[
"Kersting",
"Kristian",
""
]
] |
2410.11933 | Junjie Xu | Junjie Xu, Artem Moskalev, Tommaso Mansi, Mangal Prakash, Rui Liao | Beyond Sequence: Impact of Geometric Context for RNA Property Prediction | null | null | null | null | q-bio.QM cs.AI cs.LG q-bio.BM | http://creativecommons.org/licenses/by/4.0/ | Accurate prediction of RNA properties, such as stability and interactions, is
crucial for advancing our understanding of biological processes and developing
RNA-based therapeutics. RNA structures can be represented as 1D sequences, 2D
topological graphs, or 3D all-atom models, each offering different insights
into its function. Existing works predominantly focus on 1D sequence-based
models, which overlook the geometric context provided by 2D and 3D geometries.
This study presents the first systematic evaluation of incorporating explicit
2D and 3D geometric information into RNA property prediction, considering not
only performance but also real-world challenges such as limited data
availability, partial labeling, sequencing noise, and computational efficiency.
To this end, we introduce a newly curated set of RNA datasets with enhanced 2D
and 3D structural annotations, providing a resource for model evaluation on RNA
data. Our findings reveal that models with explicit geometry encoding generally
outperform sequence-based models, with an average prediction RMSE reduction of
around 12% across all various RNA tasks and excelling in low-data and partial
labeling regimes, underscoring the value of explicitly incorporating geometric
context. On the other hand, geometry-unaware sequence-based models are more
robust under sequencing noise but often require around $2-5\times$ training
data to match the performance of geometry-aware models. Our study offers
further insights into the trade-offs between different RNA representations in
practical applications and addresses a significant gap in evaluating deep
learning models for RNA tasks.
| [
{
"version": "v1",
"created": "Tue, 15 Oct 2024 17:09:34 GMT"
},
{
"version": "v2",
"created": "Mon, 21 Apr 2025 00:07:03 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Xu",
"Junjie",
""
],
[
"Moskalev",
"Artem",
""
],
[
"Mansi",
"Tommaso",
""
],
[
"Prakash",
"Mangal",
""
],
[
"Liao",
"Rui",
""
]
] |
2410.11944 | Nima Laal | Nima Laal, Stephen R. Taylor, Rutger van Haasteren, William G Lamb,
Xavier Siemens | Solving the PTA Data Analysis Problem with a Global Gibbs Scheme | Published at https://link.aps.org/doi/10.1103/PhysRevD.111.063067 | null | 10.1103/PhysRevD.111.063067 | null | astro-ph.IM gr-qc | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | The announcement in the summer of 2023 about the discovery of evidence for a
gravitational wave background (GWB) using pulsar timing arrays (PTAs) has
ignited both the PTA and the larger scientific community's interest in the
experiment and the scientific implications of its findings. As a result,
numerous scientific works have been published analyzing and further developing
various aspects of the experiment, from performing tests of gravity to
improving the efficiency of the current data analysis techniques. In this
regard, we contribute to the recent advancements in the field of PTAs by
presenting the most general, agnostic, per-frequency Bayesian search for a
low-frequency (red) noise process in these data. Our new method involves the
use of a conjugate Jeffrey's-like multivariate prior which allows one to model
all unique parameters of the global PTA-level red noise covariance matrix as a
separate model parameter for which a marginalized posterior-probability
distribution can be found using Gibbs sampling. Even though perfecting the
implementation of the Gibbs sampling and mitigating the numerical stability
challenges require further development, we show the power of this new method by
analyzing realistic and theoretical PTA simulated data sets. We show how our
technique is consistent with the more restricted standard techniques in
recovering both the auto and the cross-spectrum of pulsars' low-frequency (red)
noise. Furthermore, we highlight ways to approximately characterize a GWB (both
its auto- and cross-spectrum) using Fourier coefficient estimates from
single-pulsar and so-called CURN (common uncorrelated red noise) analyses via
analytic draws from a specific Inverse-Wishart distribution.
| [
{
"version": "v1",
"created": "Tue, 15 Oct 2024 18:00:01 GMT"
},
{
"version": "v2",
"created": "Mon, 21 Apr 2025 17:31:58 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Laal",
"Nima",
""
],
[
"Taylor",
"Stephen R.",
""
],
[
"van Haasteren",
"Rutger",
""
],
[
"Lamb",
"William G",
""
],
[
"Siemens",
"Xavier",
""
]
] |
2410.12184 | Arvind Renganathan | Kshitij Tayal, Arvind Renganathan, Xiaowei Jia, Vipin Kumar, Dan Lu | ExoTST: Exogenous-Aware Temporal Sequence Transformer for Time Series
Prediction | Accepted at ICDM 2024. in 2024 IEEE International Conference on Data
Mining (ICDM) 2024 | null | 10.1109/ICDM59182.2024.00105 | null | cs.LG | http://creativecommons.org/licenses/by/4.0/ | Accurate long-term predictions are the foundations for many machine learning
applications and decision-making processes. Traditional time series approaches
for prediction often focus on either autoregressive modeling, which relies
solely on past observations of the target ``endogenous variables'', or forward
modeling, which considers only current covariate drivers ``exogenous
variables''. However, effectively integrating past endogenous and past
exogenous with current exogenous variables remains a significant challenge. In
this paper, we propose ExoTST, a novel transformer-based framework that
effectively incorporates current exogenous variables alongside past context for
improved time series prediction. To integrate exogenous information
efficiently, ExoTST leverages the strengths of attention mechanisms and
introduces a novel cross-temporal modality fusion module. This module enables
the model to jointly learn from both past and current exogenous series,
treating them as distinct modalities. By considering these series separately,
ExoTST provides robustness and flexibility in handling data uncertainties that
arise from the inherent distribution shift between historical and current
exogenous variables. Extensive experiments on real-world carbon flux datasets
and time series benchmarks demonstrate ExoTST's superior performance compared
to state-of-the-art baselines, with improvements of up to 10\% in prediction
accuracy. Moreover, ExoTST exhibits strong robustness against missing values
and noise in exogenous drivers, maintaining consistent performance in
real-world situations where these imperfections are common.
| [
{
"version": "v1",
"created": "Wed, 16 Oct 2024 03:04:37 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Tayal",
"Kshitij",
""
],
[
"Renganathan",
"Arvind",
""
],
[
"Jia",
"Xiaowei",
""
],
[
"Kumar",
"Vipin",
""
],
[
"Lu",
"Dan",
""
]
] |
2410.12443 | Zhigang Lu | Shuchao Pang, Zhigang Lu, Haichen Wang, Peng Fu, Yongbin Zhou, Minhui
Xue | Reconstruction of Differentially Private Text Sanitization via Large
Language Models | null | null | null | null | cs.CR cs.AI cs.LG | http://creativecommons.org/licenses/by/4.0/ | Differential privacy (DP) is the de facto privacy standard against privacy
leakage attacks, including many recently discovered ones against large language
models (LLMs). However, we discovered that LLMs could reconstruct the
altered/removed privacy from given DP-sanitized prompts. We propose two attacks
(black-box and white-box) based on the accessibility to LLMs and show that LLMs
could connect the pair of DP-sanitized text and the corresponding private
training data of LLMs by giving sample text pairs as instructions (in the
black-box attacks) or fine-tuning data (in the white-box attacks). To
illustrate our findings, we conduct comprehensive experiments on modern LLMs
(e.g., LLaMA-2, LLaMA-3, ChatGPT-3.5, ChatGPT-4, ChatGPT-4o, Claude-3,
Claude-3.5, OPT, GPT-Neo, GPT-J, Gemma-2, and Pythia) using commonly used
datasets (such as WikiMIA, Pile-CC, and Pile-Wiki) against both word-level and
sentence-level DP. The experimental results show promising recovery rates,
e.g., the black-box attacks against the word-level DP over WikiMIA dataset gave
72.18% on LLaMA-2 (70B), 82.39% on LLaMA-3 (70B), 75.35% on Gemma-2, 91.2% on
ChatGPT-4o, and 94.01% on Claude-3.5 (Sonnet). More urgently, this study
indicates that these well-known LLMs have emerged as a new security risk for
existing DP text sanitization approaches in the current environment.
| [
{
"version": "v1",
"created": "Wed, 16 Oct 2024 10:41:17 GMT"
},
{
"version": "v2",
"created": "Sun, 20 Apr 2025 03:22:57 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Pang",
"Shuchao",
""
],
[
"Lu",
"Zhigang",
""
],
[
"Wang",
"Haichen",
""
],
[
"Fu",
"Peng",
""
],
[
"Zhou",
"Yongbin",
""
],
[
"Xue",
"Minhui",
""
]
] |
2410.12851 | Lisa Dunlap | Lisa Dunlap, Krishna Mandal, Trevor Darrell, Jacob Steinhardt, Joseph
E Gonzalez | VibeCheck: Discover and Quantify Qualitative Differences in Large
Language Models | unironic use of the word 'vibe', added more analysis and cooler
graphs. added website link | null | null | null | cs.CL cs.AI | http://creativecommons.org/licenses/by/4.0/ | Large language models (LLMs) often exhibit subtle yet distinctive
characteristics in their outputs that users intuitively recognize, but struggle
to quantify. These "vibes" -- such as tone, formatting, or writing style --
influence user preferences, yet traditional evaluations focus primarily on the
singular axis of correctness. We introduce VibeCheck, a system for
automatically comparing a pair of LLMs by discovering identifying traits of a
model (vibes) that are well-defined, differentiating, and user-aligned.
VibeCheck iteratively discovers vibes from model outputs and then utilizes a
panel of LLM judges to quantitatively measure the utility of each vibe. We
validate that the vibes generated by VibeCheck align with those found in human
discovery and run VibeCheck on pairwise preference data from real-world user
conversations with Llama-3-70b vs GPT-4. VibeCheck reveals that Llama has a
friendly, funny, and somewhat controversial vibe. These vibes predict model
identity with 80% accuracy and human preference with 61% accuracy. Lastly, we
run VibeCheck on a variety of models and tasks including summarization, math,
and captioning to provide insight into differences in model behavior. VibeCheck
discovers vibes like Command X prefers to add concrete intros and conclusions
when summarizing in comparison to TNGL, Llama-405b often overexplains its
thought process on math problems compared to GPT-4o, and GPT-4 prefers to focus
on the mood and emotions of the scene when captioning compared to
Gemini-1.5-Flash. Code and vibe visualizer found at https://bench-mark.org/
| [
{
"version": "v1",
"created": "Thu, 10 Oct 2024 17:59:17 GMT"
},
{
"version": "v2",
"created": "Thu, 24 Oct 2024 20:01:12 GMT"
},
{
"version": "v3",
"created": "Mon, 28 Oct 2024 06:11:31 GMT"
},
{
"version": "v4",
"created": "Mon, 2 Dec 2024 20:27:39 GMT"
},
{
"version": "v5",
"created": "Fri, 13 Dec 2024 22:09:12 GMT"
},
{
"version": "v6",
"created": "Sat, 15 Feb 2025 01:05:49 GMT"
},
{
"version": "v7",
"created": "Sat, 19 Apr 2025 23:14:54 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Dunlap",
"Lisa",
""
],
[
"Mandal",
"Krishna",
""
],
[
"Darrell",
"Trevor",
""
],
[
"Steinhardt",
"Jacob",
""
],
[
"Gonzalez",
"Joseph E",
""
]
] |
2410.13117 | Shuo Liu | Shuo Liu, An Zhang, Guoqing Hu, Hong Qian, Tat-seng Chua | Preference Diffusion for Recommendation | Accepted by ICLR 2025 | ICLR 2025 | null | null | cs.IR cs.AI | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Recommender systems predict personalized item rankings based on user
preference distributions derived from historical behavior data. Recently,
diffusion models (DMs) have gained attention in recommendation for their
ability to model complex distributions, yet current DM-based recommenders often
rely on traditional objectives like mean squared error (MSE) or recommendation
objectives, which are not optimized for personalized ranking tasks or fail to
fully leverage DM's generative potential. To address this, we propose
PreferDiff, a tailored optimization objective for DM-based recommenders.
PreferDiff transforms BPR into a log-likelihood ranking objective and
integrates multiple negative samples to better capture user preferences.
Specifically, we employ variational inference to handle the intractability
through minimizing the variational upper bound and replaces MSE with cosine
error to improve alignment with recommendation tasks. Finally, we balance
learning generation and preference to enhance the training stability of DMs.
PreferDiff offers three key benefits: it is the first personalized ranking loss
designed specifically for DM-based recommenders and it improves ranking and
faster convergence by addressing hard negatives. We also prove that it is
theoretically connected to Direct Preference Optimization which indicates that
it has the potential to align user preferences in DM-based recommenders via
generative modeling. Extensive experiments across three benchmarks validate its
superior recommendation performance and commendable general sequential
recommendation capabilities. Our codes are available at
https://github.com/lswhim/PreferDiff.
| [
{
"version": "v1",
"created": "Thu, 17 Oct 2024 01:02:04 GMT"
},
{
"version": "v2",
"created": "Sun, 20 Apr 2025 10:42:19 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Liu",
"Shuo",
""
],
[
"Zhang",
"An",
""
],
[
"Hu",
"Guoqing",
""
],
[
"Qian",
"Hong",
""
],
[
"Chua",
"Tat-seng",
""
]
] |
2410.14054 | Yufeng Yang | Yufeng Yang and Erin Tripp and Yifan Sun and Shaofeng Zou and Yi Zhou | Adaptive Gradient Normalization and Independent Sampling for
(Stochastic) Generalized-Smooth Optimization | 40 pages, 1 tables | null | null | null | math.OC stat.ML | http://creativecommons.org/licenses/by/4.0/ | Recent studies have shown that many nonconvex machine learning problems
satisfy a generalized-smooth condition that extends beyond traditional smooth
nonconvex optimization. However, the existing algorithms are not fully adapted
to such generalized-smooth nonconvex geometry and encounter significant
technical limitations on their convergence analysis. In this work, we first
analyze the convergence of adaptively normalized gradient descent under
function geometries characterized by generalized-smoothness and generalized
P{\L} condition, revealing the advantage of adaptive gradient normalization.
Our results provide theoretical insights into adaptive normalization across
various scenarios.For stochastic generalized-smooth nonconvex optimization, we
propose \textbf{I}ndependent-\textbf{A}daptively \textbf{N}ormalized
\textbf{S}tochastic \textbf{G}radient \textbf{D}escent, which leverages
adaptive gradient normalization, independent sampling, and gradient clipping to
achieve an $\mathcal{O}(\epsilon^{-4})$ sample complexity under relaxed noise
assumptions. Experiments on large-scale nonconvex generalized-smooth problems
demonstrate the fast convergence of our algorithm.
| [
{
"version": "v1",
"created": "Thu, 17 Oct 2024 21:52:00 GMT"
},
{
"version": "v2",
"created": "Mon, 21 Apr 2025 16:57:16 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Yang",
"Yufeng",
""
],
[
"Tripp",
"Erin",
""
],
[
"Sun",
"Yifan",
""
],
[
"Zou",
"Shaofeng",
""
],
[
"Zhou",
"Yi",
""
]
] |
2410.14148 | Chenhang Cui | Chenhang Cui, An Zhang, Yiyang Zhou, Zhaorun Chen, Gelei Deng, Huaxiu
Yao, Tat-Seng Chua | Fine-Grained Verifiers: Preference Modeling as Next-token Prediction in
Vision-Language Alignment | 23 pages; Published as a conference paper at ICLR 2025 | null | null | null | cs.CV cs.CL | http://creativecommons.org/licenses/by-nc-nd/4.0/ | The recent advancements in large language models (LLMs) and pre-trained
vision models have accelerated the development of vision-language large models
(VLLMs), enhancing the interaction between visual and linguistic modalities.
Despite their notable success across various domains, VLLMs face challenges in
modality alignment, which can lead to issues like hallucinations and unsafe
content generation. Current alignment techniques often rely on coarse feedback
and external datasets, limiting scalability and performance. In this paper, we
propose FiSAO (Fine-Grained Self-Alignment Optimization), a novel
self-alignment method that utilizes the model's own visual encoder as a
fine-grained verifier to improve vision-language alignment without the need for
additional data. By leveraging token-level feedback from the vision encoder,
FiSAO significantly improves vision-language alignment, even surpassing
traditional preference tuning methods that require additional data. Through
both theoretical analysis and experimental validation, we demonstrate that
FiSAO effectively addresses the misalignment problem in VLLMs, marking the
first instance of token-level rewards being applied to such models.
| [
{
"version": "v1",
"created": "Fri, 18 Oct 2024 03:34:32 GMT"
},
{
"version": "v2",
"created": "Mon, 18 Nov 2024 11:58:16 GMT"
},
{
"version": "v3",
"created": "Tue, 19 Nov 2024 03:08:34 GMT"
},
{
"version": "v4",
"created": "Mon, 21 Apr 2025 04:04:53 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Cui",
"Chenhang",
""
],
[
"Zhang",
"An",
""
],
[
"Zhou",
"Yiyang",
""
],
[
"Chen",
"Zhaorun",
""
],
[
"Deng",
"Gelei",
""
],
[
"Yao",
"Huaxiu",
""
],
[
"Chua",
"Tat-Seng",
""
]
] |
2410.14286 | Zibo Miao | Zigui Zhang, Zibo Miao, and Xiu-Hao Deng | Smolyak algorithm assisted robust control for quantum systems with
uncertainties | null | null | null | null | quant-ph | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Efficient and systematic numerical methods for robust control design are
crucial in quantum systems due to inevitable uncertainties or disturbances. We
propose a novel approach that models uncertainties as random variables and
quantifies robustness using the expectation of infidelity by reformulating it
as a weighted tensor product quadrature. We employ the Smolyak algorithm to
develop a parametric robust quantum control scheme, which balances the
reduction of computational cost with the enhancement of estimation accuracy. We
demonstrate the effectiveness of our proposed algorithm by incorporating the
Smolyak sparse grids into conventional gradient-based quantum optimal control
methods such as GRAPE and GOAT. In robust control problems concerning quantum
gate realization, low infidelity and strong robustness can be achieved. These
results contribute to improving the reliability and security of quantum
computing and communication systems in the presence of real-world
imperfections.
| [
{
"version": "v1",
"created": "Fri, 18 Oct 2024 08:40:59 GMT"
},
{
"version": "v2",
"created": "Sat, 2 Nov 2024 04:03:54 GMT"
},
{
"version": "v3",
"created": "Mon, 21 Apr 2025 00:49:20 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Zhang",
"Zigui",
""
],
[
"Miao",
"Zibo",
""
],
[
"Deng",
"Xiu-Hao",
""
]
] |
2410.14586 | Baran Atalar | Baran Atalar, Carlee Joe-Wong | Neural Combinatorial Clustered Bandits for Recommendation Systems | null | null | 10.1609/aaai.v39i15.33692 | null | cs.LG cs.AI | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We consider the contextual combinatorial bandit setting where in each round,
the learning agent, e.g., a recommender system, selects a subset of "arms,"
e.g., products, and observes rewards for both the individual base arms, which
are a function of known features (called "context"), and the super arm (the
subset of arms), which is a function of the base arm rewards. The agent's goal
is to simultaneously learn the unknown reward functions and choose the
highest-reward arms. For example, the "reward" may represent a user's
probability of clicking on one of the recommended products. Conventional bandit
models, however, employ restrictive reward function models in order to obtain
performance guarantees. We make use of deep neural networks to estimate and
learn the unknown reward functions and propose Neural UCB Clustering
(NeUClust), which adopts a clustering approach to select the super arm in every
round by exploiting underlying structure in the context space. Unlike prior
neural bandit works, NeUClust uses a neural network to estimate the super arm
reward and select the super arm, thus eliminating the need for a known
optimization oracle. We non-trivially extend prior neural combinatorial bandit
works to prove that NeUClust achieves
$\widetilde{O}\left(\widetilde{d}\sqrt{T}\right)$ regret, where $\widetilde{d}$
is the effective dimension of a neural tangent kernel matrix, $T$ the number of
rounds. Experiments on real world recommendation datasets show that NeUClust
achieves better regret and reward than other contextual combinatorial and
neural bandit algorithms.
| [
{
"version": "v1",
"created": "Fri, 18 Oct 2024 16:37:28 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Atalar",
"Baran",
""
],
[
"Joe-Wong",
"Carlee",
""
]
] |
2410.14592 | Mattia Bianchi | Colin Dirren, Mattia Bianchi, Panagiotis D. Grontas, John Lygeros,
Florian D\"orfler | Contractivity and linear convergence in bilinear saddle-point problems:
An operator-theoretic approach | AISTATS 2025 | null | null | null | math.OC cs.LG | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We study the convex-concave bilinear saddle-point problem $\min_x \max_y f(x)
+ y^\top Ax - g(y)$, where both, only one, or none of the functions $f$ and $g$
are strongly convex, and suitable rank conditions on the matrix $A$ hold. The
solution of this problem is at the core of many machine learning tasks. By
employing tools from monotone operator theory, we systematically prove the
contractivity (in turn, the linear convergence) of several first-order
primal-dual algorithms, including the Chambolle-Pock method. Our approach
results in concise proofs, and it yields new convergence guarantees and tighter
bounds compared to known results.
| [
{
"version": "v1",
"created": "Fri, 18 Oct 2024 16:43:10 GMT"
},
{
"version": "v2",
"created": "Mon, 21 Apr 2025 11:13:32 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Dirren",
"Colin",
""
],
[
"Bianchi",
"Mattia",
""
],
[
"Grontas",
"Panagiotis D.",
""
],
[
"Lygeros",
"John",
""
],
[
"Dörfler",
"Florian",
""
]
] |
2410.14701 | Esteban Le\'on | E. Le\'on, A. Li, M. A. Bahena Schott, B. Bos, M. Busch, J. R.
Chapman, G. L. Duran, J. Gruszko, R. Henning, E. L. Martin, J. F. Wilkerson | Machine learning-powered data cleaning for LEGEND: a semi-supervised
approach using affinity propagation and support vector machines | 17 pages, 13 figures | 2025 Mach. Learn.: Sci. Technol. 6 015064 | 10.1088/2632-2153/adbb37 | null | physics.data-an nucl-ex physics.ins-det | http://creativecommons.org/licenses/by/4.0/ | Neutrinoless double-beta decay ($0\nu\beta\beta$) is a rare nuclear process
that, if observed, will provide insight into the nature of neutrinos and help
explain the matter-antimatter asymmetry in the universe. The Large Enriched
Germanium Experiment for Neutrinoless Double-Beta Decay (LEGEND) will operate
in two phases to search for $0\nu\beta\beta$. The first (second) stage will
employ 200 (1000) kg of High-Purity Germanium (HPGe) enriched in $^{76}$Ge to
achieve a half-life sensitivity of 10$^{27}$ (10$^{28}$) years. In this study,
we present a semi-supervised data-driven approach to remove non-physical events
captured by HPGe detectors powered by a novel artificial intelligence model. We
utilize Affinity Propagation to cluster waveform signals based on their shape
and a Support Vector Machine to classify them into different categories. We
train, optimize, test our model on data taken from a natural abundance HPGe
detector installed in the Full Chain Test experimental stand at the University
of North Carolina at Chapel Hill. We demonstrate that our model yields a
maximum sacrifice of physics events of $0.024 ^{+0.004}_{-0.003} \%$. Our model
is being used to accelerate data cleaning development for LEGEND-200 and will
serve to improve data cleaning procedures for LEGEND-1000.
| [
{
"version": "v1",
"created": "Sat, 5 Oct 2024 16:40:34 GMT"
},
{
"version": "v2",
"created": "Sat, 19 Apr 2025 17:58:48 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"León",
"E.",
""
],
[
"Li",
"A.",
""
],
[
"Schott",
"M. A. Bahena",
""
],
[
"Bos",
"B.",
""
],
[
"Busch",
"M.",
""
],
[
"Chapman",
"J. R.",
""
],
[
"Duran",
"G. L.",
""
],
[
"Gruszko",
"J.",
""
],
[
"Henning",
"R.",
""
],
[
"Martin",
"E. L.",
""
],
[
"Wilkerson",
"J. F.",
""
]
] |
2410.14815 | Raviraj Joshi | Raviraj Joshi, Kanishk Singla, Anusha Kamath, Raunak Kalani, Rakesh
Paul, Utkarsh Vaidya, Sanjay Singh Chauhan, Niranjan Wartikar, Eileen Long | Adapting Multilingual LLMs to Low-Resource Languages using Continued
Pre-training and Synthetic Corpus | null | null | null | null | cs.CL cs.LG | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Multilingual LLMs support a variety of languages; however, their performance
is suboptimal for low-resource languages. In this work, we emphasize the
importance of continued pre-training of multilingual LLMs and the use of
translation-based synthetic pre-training corpora for improving LLMs in
low-resource languages. We conduct our study in the context of the low-resource
Indic language Hindi. We introduce Nemotron-Mini-Hindi 4B, a bilingual SLM
supporting both Hindi and English, based on Nemotron-Mini 4B. The model is
trained using a mix of real and synthetic Hindi + English tokens, with
continuous pre-training performed on 400B tokens. We demonstrate that both the
base and instruct models achieve state-of-the-art results on Hindi benchmarks
while remaining competitive on English tasks. Additionally, we observe that the
continued pre-training approach enhances the model's overall factual accuracy.
We perform an ablation study to highlight the impact of Hindi pre-training,
showing significant improvements in Hindi chat capabilities and factual
accuracy, which cannot be achieved through Hindi alignment alone.
| [
{
"version": "v1",
"created": "Fri, 18 Oct 2024 18:35:19 GMT"
},
{
"version": "v2",
"created": "Mon, 21 Apr 2025 05:29:01 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Joshi",
"Raviraj",
""
],
[
"Singla",
"Kanishk",
""
],
[
"Kamath",
"Anusha",
""
],
[
"Kalani",
"Raunak",
""
],
[
"Paul",
"Rakesh",
""
],
[
"Vaidya",
"Utkarsh",
""
],
[
"Chauhan",
"Sanjay Singh",
""
],
[
"Wartikar",
"Niranjan",
""
],
[
"Long",
"Eileen",
""
]
] |
2410.15085 | Alonso Beaumont | Alonso Beaumont | On the Nilpotency of Locally Pro-p Contraction Groups | null | null | 10.5802/crmath.728 | null | math.GR | http://creativecommons.org/licenses/by/4.0/ | H. Gl\"ockner and G. A. Willis have recently shown that locally pro-p
contraction groups are nilpotent. The proof hinges on a fixed-point result: if
the local field $\mathbb{F}_{p}(\!(t)\!)$ acts on its $d$-th power
$\mathbb{F}_{p}(\!(t)\!)^{d}$ additively, continuously, and in an appropriately
equivariant manner, then the action has a non-zero fixed point. We provide a
short proof of this theorem.
| [
{
"version": "v1",
"created": "Sat, 19 Oct 2024 12:21:38 GMT"
},
{
"version": "v2",
"created": "Wed, 8 Jan 2025 16:04:47 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Beaumont",
"Alonso",
""
]
] |
2410.15282 | Shihoko Ishii | Shihoko Ishii and Ken-ichi Yoshida | On vanishing of higher direct images of the structure sheaf | We strengthened the main theorem and the corollary. We added some
references. We corrected some typos. to appear in Michigan Mathematical
Journal | null | null | null | math.AG math.AC | http://creativecommons.org/licenses/by/4.0/ | We show the vanishing of the first direct image of the structure sheaf of a
normal scheme $X$ which is mapped properly and birationally over a regular
scheme of any dimension. On the other hand, for any dimension greater than two,
we show examples of a proper birational morphism from a normal and
Cohen-Macaulay scheme to a regular scheme such that the second direct image
does not vanish and has an isolated support.
| [
{
"version": "v1",
"created": "Sun, 20 Oct 2024 04:41:35 GMT"
},
{
"version": "v2",
"created": "Mon, 11 Nov 2024 04:38:13 GMT"
},
{
"version": "v3",
"created": "Sat, 19 Apr 2025 09:09:37 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Ishii",
"Shihoko",
""
],
[
"Yoshida",
"Ken-ichi",
""
]
] |
2410.15643 | Chunbo Hua | Chunbo Hua and Dong-Hui Xu | Higher-order topology in twisted multilayer systems: a review | Invited review | Chin. Phys. B 34 037301 (2025) | 10.1088/1674-1056/ada885 | null | cond-mat.mes-hall | http://creativecommons.org/licenses/by/4.0/ | In recent years, there has been a surge of interest in higher-order
topological phases (HOTPs) across various disciplines within the field of
physics. These unique phases are characterized by their ability to harbor
topological protected boundary states at lower-dimensional boundaries, a
distinguishing feature that sets them apart from conventional topological
phases and is attributed to the higher-order bulk-boundary correspondence.
Two-dimensional (2D) twisted systems offer an optimal platform for
investigating HOTPs, owing to their strong controllability and experimental
feasibility. Here, we provide a comprehensive overview of the latest research
advancements on HOTPs in 2D twisted multilayer systems. We will mainly review
the HOTPs in electronic, magnonic, acoustic, photonic and mechanical twisted
systems, and finally provide a perspective of this topic.
| [
{
"version": "v1",
"created": "Mon, 21 Oct 2024 05:09:22 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Hua",
"Chunbo",
""
],
[
"Xu",
"Dong-Hui",
""
]
] |
2410.16398 | Baris Askin | Baris Askin, Pranay Sharma, Gauri Joshi, Carlee Joe-Wong | Federated Communication-Efficient Multi-Objective Optimization | Accepted to AISTATS 2025 | null | null | null | cs.LG cs.DC | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We study a federated version of multi-objective optimization (MOO), where a
single model is trained to optimize multiple objective functions. MOO has been
extensively studied in the centralized setting but is less explored in
federated or distributed settings. We propose FedCMOO, a novel
communication-efficient federated multi-objective optimization (FMOO) algorithm
that improves the error convergence performance of the model compared to
existing approaches. Unlike prior works, the communication cost of FedCMOO does
not scale with the number of objectives, as each client sends a single
aggregated gradient to the central server. We provide a convergence analysis of
the proposed method for smooth and non-convex objective functions under milder
assumptions than in prior work. In addition, we introduce a variant of FedCMOO
that allows users to specify a preference over the objectives in terms of a
desired ratio of the final objective values. Through extensive experiments, we
demonstrate the superiority of our proposed method over baseline approaches.
| [
{
"version": "v1",
"created": "Mon, 21 Oct 2024 18:09:22 GMT"
},
{
"version": "v2",
"created": "Sat, 19 Apr 2025 22:07:07 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Askin",
"Baris",
""
],
[
"Sharma",
"Pranay",
""
],
[
"Joshi",
"Gauri",
""
],
[
"Joe-Wong",
"Carlee",
""
]
] |
2410.16507 | Mamtaj Akter | Ozioma C. Oguine, Jinkyung Katie Park, Mamtaj Akter, Johanna Olesk,
Abdulmalik Alluhidan, Pamela Wisniewski, Karla Badillo-Urquiola | How the Internet Facilitates Adverse Childhood Experiences for Youth Who
Self-Identify as in Need of Services | null | Proceedings of the ACM on Human-Computer Interaction, CSCW 2025 | null | null | cs.HC cs.CY | http://creativecommons.org/licenses/by/4.0/ | Youth implicated in the child welfare and juvenile justice systems, as well
as those with an incarcerated parent, are considered the most vulnerable
Children in Need of Services (CHINS). We identified 1,160 of these at-risk
youth (ages 13-17) who sought support via an online peer support platform to
understand their adverse childhood experiences and explore how the internet
played a role in providing an outlet for support, as well as potentially
facilitating risks. We first analyzed posts from 1,160 youth who
self-identified as CHINS while sharing about their adverse experiences. Then,
we retrieved all 239,929 posts by these users to identify salient topics within
their support-seeking posts: 1) Urges to self-harm due to social drama, 2)
desire for social connection, 3) struggles with family, and 4) substance use
and sexual risks. We found that the internet often helped facilitate these
problems; for example, the desperation for social connection often led to
meeting unsafe people online, causing additional trauma. Family members and
other unsafe people used the internet to perpetrate cyberabuse, while CHINS
themselves leveraged online channels to engage in illegal and risky behavior.
Our study calls for tailored support systems that address the unique needs of
CHINS to promote safe online spaces and foster resilience to break the cycle of
adversity. Empowering CHINS requires amplifying their voices and acknowledging
the challenges they face as a result of their adverse childhood experiences.
| [
{
"version": "v1",
"created": "Mon, 21 Oct 2024 20:59:11 GMT"
},
{
"version": "v2",
"created": "Sat, 19 Apr 2025 02:57:42 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Oguine",
"Ozioma C.",
""
],
[
"Park",
"Jinkyung Katie",
""
],
[
"Akter",
"Mamtaj",
""
],
[
"Olesk",
"Johanna",
""
],
[
"Alluhidan",
"Abdulmalik",
""
],
[
"Wisniewski",
"Pamela",
""
],
[
"Badillo-Urquiola",
"Karla",
""
]
] |
2410.16559 | Hui Hu | Hui Hu, Jia Wang, Han Pu, and Xia-Ji Liu | Breakdown of the single-mode description of ultradilute quantum droplets
in binary Bose mixtures: A perspective from a microscopic bosonic pairing
theory | 12 pages, 9 figures | Physical Review A 111, 023309 (2025) | 10.1103/PhysRevA.111.023309 | null | cond-mat.quant-gas | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | In his seminal proposal of quantum droplets in binary Bose mixtures {[}Phys.
Rev. Lett. \textbf{115}, 155302 (2015){]}, Dmitry Petrov suggested that the
density ratio $n_{2}/n_{1}$ of the two bosonic components are locked to an
optimal value, which is given by the square root of the ratio of the two
intra-species scattering lengths, i.e., $\sqrt{a_{11}/a_{22}}$. Due to such a
density locking, quantum droplets can be efficiently described by using an
extended Gross--Pitaevskii equation within the single-mode approximation. Here,
we find that this single-mode description necessarily breaks down in the deep
quantum droplet regime, when the attractive inter-species scattering length
$a_{12}$ significantly deviates away from the threshold of mean-field collapse
(i.e., $-\sqrt{a_{11}a_{22}}$). By applying a bosonic pairing theory, we show
that the density ratio is allowed to fluctuate in a sizable interval. Most
importantly, the optimal density ratio would be very different from
$\sqrt{a_{11}/a_{22}}$, in the case of unequal intra-species scattering lengths
($a_{11}\neq a_{22}$). Our finding might provide a plausible microscopic
explanation of the puzzling low critical particle number of quantum droplets,
as experimentally observed. Our predicted interval of the density ratio, as a
function of the inter-species scattering length, could also be experimentally
examined in cold-atom laboratories in the near future.
| [
{
"version": "v1",
"created": "Mon, 21 Oct 2024 22:39:46 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Hu",
"Hui",
""
],
[
"Wang",
"Jia",
""
],
[
"Pu",
"Han",
""
],
[
"Liu",
"Xia-Ji",
""
]
] |
2410.16714 | Mingzhi Wang | Mingzhi Wang, Chengdong Ma, Qizhi Chen, Linjian Meng, Yang Han,
Jiancong Xiao, Zhaowei Zhang, Jing Huo, Weijie J. Su, Yaodong Yang | Magnetic Preference Optimization: Achieving Last-iterate Convergence for
Language Model Alignment | ICLR 2025 | null | null | null | cs.CL | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Self-play methods have demonstrated remarkable success in enhancing model
capabilities across various domains. In the context of Reinforcement Learning
from Human Feedback (RLHF), self-play not only boosts Large Language Model
(LLM) performance but also overcomes the limitations of traditional
Bradley-Terry (BT) model assumptions by finding the Nash equilibrium (NE) of a
preference-based, two-player constant-sum game. However, existing methods
either guarantee only average-iterate convergence, incurring high storage and
inference costs, or converge to the NE of a regularized game, failing to
accurately reflect true human preferences. In this paper, we introduce Magnetic
Preference Optimization (MPO), a novel approach capable of achieving
last-iterate convergence to the NE of the original game, effectively overcoming
the limitations of existing methods. Building upon Magnetic Mirror Descent
(MMD), MPO attains a linear convergence rate, making it particularly suitable
for fine-tuning LLMs. To ensure our algorithm is both theoretically sound and
practically viable, we present a simple yet effective implementation that
adapts the theoretical insights to the RLHF setting. Empirical results
demonstrate that MPO can significantly enhance the performance of LLMs,
highlighting the potential of self-play methods in alignment.
| [
{
"version": "v1",
"created": "Tue, 22 Oct 2024 05:51:34 GMT"
},
{
"version": "v2",
"created": "Fri, 20 Dec 2024 16:26:58 GMT"
},
{
"version": "v3",
"created": "Sat, 19 Apr 2025 15:59:33 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Wang",
"Mingzhi",
""
],
[
"Ma",
"Chengdong",
""
],
[
"Chen",
"Qizhi",
""
],
[
"Meng",
"Linjian",
""
],
[
"Han",
"Yang",
""
],
[
"Xiao",
"Jiancong",
""
],
[
"Zhang",
"Zhaowei",
""
],
[
"Huo",
"Jing",
""
],
[
"Su",
"Weijie J.",
""
],
[
"Yang",
"Yaodong",
""
]
] |
2410.17231 | Mads Christensen | Mads Bjerge Christensen | Linking numbers and non-holomorphic Siegel modular forms | Added a new theorem, fixed various typos, and improved the
exposition. 46 pages, 2 figures | null | null | null | math.NT math.DG math.GT | http://creativecommons.org/licenses/by/4.0/ | We study generating series encoding linking numbers between geodesics in
arithmetic hyperbolic $3$-folds. We show that the series converge to functions
on genus $2$ Siegel space and that certain explicit modifications have the
transformation properties of genus $2$ Siegel modular forms of weight $2$. This
is done by carefully analyzing the integral of the Kudla--Millson theta series
over a Seifert surface with geodesic boundary. As a corollary, we deduce a
polynomial bound on the linking numbers.
| [
{
"version": "v1",
"created": "Tue, 22 Oct 2024 17:51:36 GMT"
},
{
"version": "v2",
"created": "Sat, 19 Apr 2025 09:49:37 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Christensen",
"Mads Bjerge",
""
]
] |
2410.17353 | Teimour Hosseinalizadeh | Teimour Hosseinalizadeh and Nima Monshizadeh | Preserving Privacy in Cloud-based Data-Driven Stabilization | null | null | null | null | eess.SY cs.SY | http://creativecommons.org/licenses/by/4.0/ | In the recent years, we have observed three significant trends in control
systems: a renewed interest in data-driven control design, the abundance of
cloud computational services and the importance of preserving privacy for the
system under control. Motivated by these factors, this work investigates
privacy-preserving outsourcing for the design of a stabilizing controller for
unknown linear time-invariant systems.The main objective of this research is to
preserve the privacy for the system dynamics by designing an outsourcing
mechanism. To achieve this goal, we propose a scheme that combines
transformation-based techniques and robust data-driven control design methods.
The scheme preserves the privacy of both the open-loop and closed-loop system
matrices while stabilizing the system under control.The scheme is applicable to
both data with and without disturbance and is lightweight in terms of
computational overhead. Numerical investigations for a case study demonstrate
the impacts of our mechanism and its role in hindering malicious adversaries
from achieving their goals.
| [
{
"version": "v1",
"created": "Tue, 22 Oct 2024 18:44:46 GMT"
},
{
"version": "v2",
"created": "Mon, 21 Apr 2025 15:23:55 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Hosseinalizadeh",
"Teimour",
""
],
[
"Monshizadeh",
"Nima",
""
]
] |
2410.17516 | Virginia Feldman | Virginia Feldman and Ariel Bendersky | Selective continuous-variable quantum process tomography | null | Physical Review A 111.4 (2025): 042409 | 10.1103/PhysRevA.111.042409 | null | quant-ph | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Quantum process tomography is a useful tool for characterizing quantum
processes. This task is essential for the development of different areas, such
as quantum information processing. In this work, we present a protocol for
selective continuous-variable quantum process tomography. Our proposal allows
one to selectively estimate any element of an unknown continuous-variable
quantum process in the position representation, without requiring the complete
reconstruction of the process. By resorting to controlled squeezing and
translation operations, and adaptatively discretizing the process, a direct
measure of an estimate of any process element can be obtained. Furthermore, we
show, supported by numerical simulations, how the protocol can be used to
partially reconstruct on a region a continuous-variable quantum process.
| [
{
"version": "v1",
"created": "Wed, 23 Oct 2024 02:49:32 GMT"
},
{
"version": "v2",
"created": "Mon, 21 Apr 2025 14:58:07 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Feldman",
"Virginia",
""
],
[
"Bendersky",
"Ariel",
""
]
] |
2410.18301 | Harish Kumar Dureppagari | Harish K. Dureppagari, Chiranjib Saha, Harikumar Krishnamurthy, Xiao
Feng Wang, Alberto Rico-Alvari\~no, R. Michael Buehrer, Harpreet S. Dhillon | LEO-based Positioning: Foundations, Signal Design, and Receiver
Enhancements for 6G NTN | Accepted for Publication in IEEE Communications Magazine | null | null | null | cs.IT eess.SP math.IT | http://creativecommons.org/licenses/by/4.0/ | The integration of non-terrestrial networks (NTN) into 5G new radio (NR) has
opened up the possibility of developing a new positioning infrastructure using
NR signals from Low-Earth Orbit (LEO) satellites. Compared to existing Global
Navigation Satellite Systems (GNSS), LEO-based cellular positioning offers
several advantages, such as a superior link budget, higher operating bandwidth,
and large forthcoming constellations. Due to these factors, LEO-based
positioning, navigation, and timing (PNT) is a potential enhancement for NTN in
6G cellular networks. However, extending the existing terrestrial cellular
positioning methods to LEO-based NTN positioning requires key fundamental
enhancements. These include creating broad positioning beams orthogonal to
conventional communication beams, time-domain processing at the user equipment
(UE) to resolve large delay and Doppler uncertainties, and efficiently
accommodating positioning reference signals (PRS) from multiple satellites
within the communication resource grid. In this paper, we present the first set
of design insights by incorporating these enhancements and thoroughly
evaluating LEO-based positioning, considering the constraints and capabilities
of the NR-NTN physical layer. To evaluate the performance of LEO-based NTN
positioning, we develop a comprehensive NR-compliant simulation framework,
including LEO orbit simulation, transmission (Tx) and receiver (Rx)
architectures, and a positioning engine incorporating the necessary
enhancements. Our findings suggest that LEO-based NTN positioning could serve
as a complementary infrastructure to GNSS and, with appropriate enhancements,
may also offer a viable alternative.
| [
{
"version": "v1",
"created": "Wed, 23 Oct 2024 22:19:48 GMT"
},
{
"version": "v2",
"created": "Mon, 21 Apr 2025 03:25:53 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Dureppagari",
"Harish K.",
""
],
[
"Saha",
"Chiranjib",
""
],
[
"Krishnamurthy",
"Harikumar",
""
],
[
"Wang",
"Xiao Feng",
""
],
[
"Rico-Alvariño",
"Alberto",
""
],
[
"Buehrer",
"R. Michael",
""
],
[
"Dhillon",
"Harpreet S.",
""
]
] |
2410.18443 | Thierry Marchant | Denis Bouyssou and Thierry Marchant and Marc Pirlot | ELECTRE TRI-nB, pseudo-disjunctive: axiomatic and combinatorial results | Companion paper: About maximal antichains in a product of two
chains:A catch-all note (arXiv:2410.16243) | Theory and Decision, published online:15 April 2025 | 10.1007/s11238-025-10034-2 | null | cs.DM | http://creativecommons.org/licenses/by-nc-nd/4.0/ | ELECTRE TRI-nB is a method designed to sort alternatives evaluated on several
attributes into ordered categories. It is an extension of ELECTRE TRI-B, using
several limiting profiles, instead of just one, to delimit each category.
ELECTRE TRI-nB comes in two flavours: pseudo-conjunctive and
pseudo-disjunctive. In a previous paper we have characterized the ordered
partitions that can be obtained with ELECTRE TRI-nB, pseudo-conjunctive, using
a simple axiom called linearity. The present paper is dedicated to the
axiomatic analysis of ELECTRE TRI-nB, pseudo-disjunctive. It also provides some
combinatorial results.
| [
{
"version": "v1",
"created": "Thu, 24 Oct 2024 05:39:41 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Bouyssou",
"Denis",
""
],
[
"Marchant",
"Thierry",
""
],
[
"Pirlot",
"Marc",
""
]
] |
2410.18713 | Yixu Wang | Yixu Wang, Yijia Xu, Zi-Wen Liu | Tessellation codes: encoded quantum gates by geometric rotation | 16 pages, 8 figures | null | null | null | quant-ph math-ph math.MP | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We utilize the symmetry groups of regular tessellations on two-dimensional
surfaces of different constant curvatures, including spheres, Euclidean planes
and hyperbolic planes, to encode a qubit or qudit into the physical degrees of
freedom on these surfaces, which we call tessellation codes. We show that
tessellation codes exhibit decent error correction properties by analysis via
geometric considerations and the representation theory of the isometry groups
on the corresponding surfaces. Interestingly, we demonstrate how this formalism
enables the implementation of certain logical operations through geometric
rotations of surfaces in real space, opening a new approach to logical quantum
computation. We provide a variety of concrete constructions of such codes
associated with different tessellations, which give rise to different logical
groups.
| [
{
"version": "v1",
"created": "Thu, 24 Oct 2024 13:17:56 GMT"
},
{
"version": "v2",
"created": "Mon, 21 Apr 2025 05:41:58 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Wang",
"Yixu",
""
],
[
"Xu",
"Yijia",
""
],
[
"Liu",
"Zi-Wen",
""
]
] |
2410.18820 | Tatsuya Terao | Tatsuya Terao | Deterministic $(2/3-\varepsilon)$-Approximation of Matroid Intersection
Using Nearly-Linear Independence-Oracle Queries | 18 pages, to appear in WADS 2025; Fix typo (v2) | null | null | null | cs.DS | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | In the matroid intersection problem, we are given two matroids $\mathcal{M}_1
= (V, \mathcal{I}_1)$ and $\mathcal{M}_2 = (V, \mathcal{I}_2)$ defined on the
same ground set $V$ of $n$ elements, and the objective is to find a common
independent set $S \in \mathcal{I}_1 \cap \mathcal{I}_2$ of largest possible
cardinality, denoted by $r$. In this paper, we consider a deterministic matroid
intersection algorithm with only a nearly linear number of independence oracle
queries. Our contribution is to present a deterministic
$O(\frac{n}{\varepsilon} + r \log r)$-independence-query
$(2/3-\varepsilon)$-approximation algorithm for any $\varepsilon > 0$. Our idea
is very simple: we apply a recent $\tilde{O}(n
\sqrt{r}/\varepsilon)$-independence-query $(1 - \varepsilon)$-approximation
algorithm of Blikstad [ICALP 2021], but terminate it before completion.
Moreover, we also present a semi-streaming algorithm for $(2/3
-\varepsilon)$-approximation of matroid intersection in $O(1/\varepsilon)$
passes.
| [
{
"version": "v1",
"created": "Thu, 24 Oct 2024 15:08:38 GMT"
},
{
"version": "v2",
"created": "Thu, 17 Apr 2025 09:24:20 GMT"
},
{
"version": "v3",
"created": "Sat, 19 Apr 2025 19:04:58 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Terao",
"Tatsuya",
""
]
] |
2410.18908 | Jing Peng | Jing Peng and Yucheng Wang and Yu Xi and Xu Li and Xizhuo Zhang and
Kai Yu | A Survey on Speech Large Language Models | This version has been updated to incorporate recent work in the field
and includes revised illustrations and textual descriptions | null | null | null | eess.AS | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Large Language Models (LLMs) exhibit strong contextual understanding and
remarkable multitask performance. As a result, researchers have been actively
exploring the integration of LLMs into the domain of speech understanding, with
a primary focus on a broad range of speech-to-text tasks. These include
automatic speech recognition (ASR), speech-to-text translation (ST), speech
emotion recognition (SER), and others. We refer to such models as Speech LLMs,
which are typically built on a unified architecture that follows the pipeline
of Audio Feature Extraction -> Multimodal Information Fusion -> LLM Inference.
This approach enables richer audio feature extraction while facilitating
end-to-end fusion of audio and text modalities, thereby achieving deeper
understanding and reasoning from audio data. This paper elucidates the
development of Speech LLMs, offering an in-depth analysis of system
architectures. Through extensive research and a series of targeted experiments,
the paper assesses the advancements in Speech LLMs and their potential for
cross-task integration within the speech understanding field. Furthermore, it
highlights key challenges identified through experimentation, such as the
dormancy of LLMs under certain conditions. The paper further explores training
strategies for Speech LLMs, proposes potential solutions based on these
findings, and offers valuable insights and references for future research.
| [
{
"version": "v1",
"created": "Thu, 24 Oct 2024 16:59:28 GMT"
},
{
"version": "v2",
"created": "Fri, 25 Oct 2024 08:30:21 GMT"
},
{
"version": "v3",
"created": "Mon, 21 Apr 2025 07:17:53 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Peng",
"Jing",
""
],
[
"Wang",
"Yucheng",
""
],
[
"Xi",
"Yu",
""
],
[
"Li",
"Xu",
""
],
[
"Zhang",
"Xizhuo",
""
],
[
"Yu",
"Kai",
""
]
] |
2410.19051 | Tal Schwartzman | Tal Schwartzman | The complexity of entanglement embezzlement | 9 pages | null | null | null | quant-ph hep-th | http://creativecommons.org/licenses/by-sa/4.0/ | Embezzlement of entanglement is the counterintuitive process in which
entanglement is extracted from a resource system using local unitary
operations, with almost no detectable change in the resource's state. It has
recently been argued that any state of a relativistic quantum field theory can
serve as a resource for perfect embezzlement. We study the circuit complexity
of embezzlement, using sequences of states that enable arbitrary precision for
the process, commonly called universal embezzling families. In addition, we
argue that this approach provides a well-defined model for the complexity of
embezzlement from quantum field theories. Under fairly general assumptions, we
establish a generic lower bound on the complexity, which increases with the
precision of the process or embezzled entanglement, and diverges as these
become infinite. As an example, we consider a $1d$ critical system as the
resource and derive an exponentially growing lower bound on the complexity.
Consequently, the findings imply that circuit complexity acts as a physical
obstruction to perfect embezzlement. Supplementary to the main results, we
derive lower bounds for common models of circuit complexity for state
preparation, based on the difference between the Schatten norms of the initial
and final states.
| [
{
"version": "v1",
"created": "Thu, 24 Oct 2024 18:00:33 GMT"
},
{
"version": "v2",
"created": "Sun, 20 Apr 2025 11:16:55 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Schwartzman",
"Tal",
""
]
] |
2410.19245 | Zixiao Zhao | Zixiao Zhao, Jing Sun, Zhe Hou, Zhiyuan Wei, Cheng-Hao Cai, Miao Qiao,
Jin Song Dong | MaCTG: Multi-Agent Collaborative Thought Graph for Automatic Programming | null | null | null | null | cs.SE cs.CV cs.MA | http://creativecommons.org/licenses/by/4.0/ | With the rapid advancement of Large Language Models (LLMs), LLM-based
approaches have demonstrated strong problem-solving capabilities across various
domains. However, in automatic programming, a single LLM is typically limited
to function-level code generation, while multi-agent systems composed of
multiple LLMs often suffer from inefficient task planning. This lack of
structured coordination can lead to cascading hallucinations, where accumulated
errors across agents result in suboptimal workflows and excessive computational
costs. To overcome these challenges, we introduce MaCTG (Multi-Agent
Collaborative Thought Graph), a novel multi-agent framework that employs a
dynamic graph structure to facilitate precise task allocation and controlled
collaboration among LLM agents. MaCTG autonomously assigns agent roles based on
programming requirements, dynamically refines task distribution through
context-aware adjustments, and systematically verifies and integrates
project-level code, effectively reducing hallucination errors and improving
overall accuracy. MaCTG enhances cost-effectiveness by implementing a hybrid
LLM deployment, where proprietary models handle complex reasoning, while
open-source models are used for routine coding and validation tasks. To
evaluate MaCTG's effectiveness, we applied it to traditional image processing
auto-programming tasks, achieving a state-of-the-art accuracy of 83.33%.
Additionally, by leveraging its hybrid LLM configuration, MaCTG significantly
reduced operational costs by 89.09% compared to existing multi-agent
frameworks, demonstrating its efficiency, scalability, and real-world
applicability.
| [
{
"version": "v1",
"created": "Fri, 25 Oct 2024 01:52:15 GMT"
},
{
"version": "v2",
"created": "Mon, 21 Apr 2025 04:40:38 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Zhao",
"Zixiao",
""
],
[
"Sun",
"Jing",
""
],
[
"Hou",
"Zhe",
""
],
[
"Wei",
"Zhiyuan",
""
],
[
"Cai",
"Cheng-Hao",
""
],
[
"Qiao",
"Miao",
""
],
[
"Dong",
"Jin Song",
""
]
] |
2410.19746 | Nils Wandel | Nils Wandel, Stefan Schulz, Reinhard Klein | Metamizer: a versatile neural optimizer for fast and accurate physics
simulations | to be published at International Conference on Learning
Representations (ICLR) 2025 | null | null | null | physics.comp-ph cs.AI | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Efficient physics simulations are essential for numerous applications,
ranging from realistic cloth animations or smoke effects in video games, to
analyzing pollutant dispersion in environmental sciences, to calculating
vehicle drag coefficients in engineering applications. Unfortunately,
analytical solutions to the underlying physical equations are rarely available,
and numerical solutions require high computational resources. Latest
developments in the field of physics-based Deep Learning have led to promising
efficiency improvements but still suffer from limited generalization
capabilities and low accuracy compared to numerical solvers.
In this work, we introduce Metamizer, a novel neural optimizer that
iteratively solves a wide range of physical systems with high accuracy by
minimizing a physics-based loss function. To this end, our approach leverages a
scale-invariant architecture that enhances gradient descent updates to
accelerate convergence. Since the neural network itself acts as an optimizer,
training this neural optimizer falls into the category of meta-optimization
approaches.
We demonstrate that Metamizer achieves unprecedented accuracy for deep
learning based approaches - sometimes approaching machine precision - across
multiple PDEs after training on the Laplace, advection-diffusion and
incompressible Navier-Stokes equation as well as on cloth simulations.
Remarkably, the model also generalizes to PDEs that were not covered during
training such as the Poisson, wave and Burgers equation.
Our results suggest that Metamizer could have a profound impact on future
numerical solvers, paving the way for fast and accurate neural physics
simulations without the need for retraining.
| [
{
"version": "v1",
"created": "Thu, 10 Oct 2024 11:54:31 GMT"
},
{
"version": "v2",
"created": "Sun, 20 Apr 2025 19:42:42 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Wandel",
"Nils",
""
],
[
"Schulz",
"Stefan",
""
],
[
"Klein",
"Reinhard",
""
]
] |
2410.20197 | Song Xia | Song Xia and Wenhan Yang and Yi Yu and Xun Lin and Henghui Ding and
Ling-Yu Duan and Xudong Jiang | Transferable Adversarial Attacks on SAM and Its Downstream Models | update fig 1 | null | null | null | cs.LG cs.CR cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | The utilization of large foundational models has a dilemma: while fine-tuning
downstream tasks from them holds promise for making use of the well-generalized
knowledge in practical applications, their open accessibility also poses
threats of adverse usage. This paper, for the first time, explores the
feasibility of adversarial attacking various downstream models fine-tuned from
the segment anything model (SAM), by solely utilizing the information from the
open-sourced SAM. In contrast to prevailing transfer-based adversarial attacks,
we demonstrate the existence of adversarial dangers even without accessing the
downstream task and dataset to train a similar surrogate model. To enhance the
effectiveness of the adversarial attack towards models fine-tuned on unknown
datasets, we propose a universal meta-initialization (UMI) algorithm to extract
the intrinsic vulnerability inherent in the foundation model, which is then
utilized as the prior knowledge to guide the generation of adversarial
perturbations. Moreover, by formulating the gradient difference in the
attacking process between the open-sourced SAM and its fine-tuned downstream
models, we theoretically demonstrate that a deviation occurs in the adversarial
update direction by directly maximizing the distance of encoded feature
embeddings in the open-sourced SAM. Consequently, we propose a gradient robust
loss that simulates the associated uncertainty with gradient-based noise
augmentation to enhance the robustness of generated adversarial examples (AEs)
towards this deviation, thus improving the transferability. Extensive
experiments demonstrate the effectiveness of the proposed universal
meta-initialized and gradient robust adversarial attack (UMI-GRAT) toward SAMs
and their downstream models. Code is available at
https://github.com/xiasong0501/GRAT.
| [
{
"version": "v1",
"created": "Sat, 26 Oct 2024 15:04:04 GMT"
},
{
"version": "v2",
"created": "Tue, 29 Oct 2024 05:40:19 GMT"
},
{
"version": "v3",
"created": "Mon, 21 Apr 2025 10:57:32 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Xia",
"Song",
""
],
[
"Yang",
"Wenhan",
""
],
[
"Yu",
"Yi",
""
],
[
"Lin",
"Xun",
""
],
[
"Ding",
"Henghui",
""
],
[
"Duan",
"Ling-Yu",
""
],
[
"Jiang",
"Xudong",
""
]
] |
2410.20941 | Yirong Sun | Yirong Sun, Dawei Zhu, Yanjun Chen, Erjia Xiao, Xinghao Chen, Xiaoyu
Shen | Fine-Grained and Multi-Dimensional Metrics for Document-Level Machine
Translation | Accepted at NAACL 2025 Student Research Workshop | null | null | null | cs.CL cs.AI | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Large language models (LLMs) have excelled in various NLP tasks, including
machine translation (MT), yet most studies focus on sentence-level translation.
This work investigates the inherent capability of instruction-tuned LLMs for
document-level translation (docMT). Unlike prior approaches that require
specialized techniques, we evaluate LLMs by directly prompting them to
translate entire documents in a single pass. Our results show that this method
improves translation quality compared to translating sentences separately, even
without document-level fine-tuning. However, this advantage is not reflected in
BLEU scores, which often favor sentence-based translations. We propose using
the LLM-as-a-judge paradigm for evaluation, where GPT-4 is used to assess
document coherence, accuracy, and fluency in a more nuanced way than
n-gram-based metrics. Overall, our work demonstrates that instruction-tuned
LLMs can effectively leverage document context for translation. However, we
caution against using BLEU scores for evaluating docMT, as they often provide
misleading outcomes, failing to capture the quality of document-level
translation. Code and the outputs from GPT4-as-a-judge are available at
https://github.com/EIT-NLP/BLEUless_DocMT
| [
{
"version": "v1",
"created": "Mon, 28 Oct 2024 11:49:58 GMT"
},
{
"version": "v2",
"created": "Tue, 29 Oct 2024 09:08:19 GMT"
},
{
"version": "v3",
"created": "Fri, 14 Mar 2025 13:12:38 GMT"
},
{
"version": "v4",
"created": "Sun, 20 Apr 2025 12:03:36 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Sun",
"Yirong",
""
],
[
"Zhu",
"Dawei",
""
],
[
"Chen",
"Yanjun",
""
],
[
"Xiao",
"Erjia",
""
],
[
"Chen",
"Xinghao",
""
],
[
"Shen",
"Xiaoyu",
""
]
] |
2410.21060 | Yutong Cheng | Yutong Cheng, Osama Bajaber, Saimon Amanuel Tsegai, Dawn Song, Peng
Gao | CTINexus: Automatic Cyber Threat Intelligence Knowledge Graph
Construction Using Large Language Models | Accepted at 2025 IEEE European Symposium on Security and Privacy
(Euro S&P) | null | null | null | cs.CR cs.AI cs.LG | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Textual descriptions in cyber threat intelligence (CTI) reports, such as
security articles and news, are rich sources of knowledge about cyber threats,
crucial for organizations to stay informed about the rapidly evolving threat
landscape. However, current CTI knowledge extraction methods lack flexibility
and generalizability, often resulting in inaccurate and incomplete knowledge
extraction. Syntax parsing relies on fixed rules and dictionaries, while model
fine-tuning requires large annotated datasets, making both paradigms
challenging to adapt to new threats and ontologies. To bridge the gap, we
propose CTINexus, a novel framework leveraging optimized in-context learning
(ICL) of large language models (LLMs) for data-efficient CTI knowledge
extraction and high-quality cybersecurity knowledge graph (CSKG) construction.
Unlike existing methods, CTINexus requires neither extensive data nor parameter
tuning and can adapt to various ontologies with minimal annotated examples.
This is achieved through: (1) a carefully designed automatic prompt
construction strategy with optimal demonstration retrieval for extracting a
wide range of cybersecurity entities and relations; (2) a hierarchical entity
alignment technique that canonicalizes the extracted knowledge and removes
redundancy; (3) an long-distance relation prediction technique to further
complete the CSKG with missing links. Our extensive evaluations using 150
real-world CTI reports collected from 10 platforms demonstrate that CTINexus
significantly outperforms existing methods in constructing accurate and
complete CSKG, highlighting its potential to transform CTI analysis with an
efficient and adaptable solution for the dynamic threat landscape.
| [
{
"version": "v1",
"created": "Mon, 28 Oct 2024 14:18:32 GMT"
},
{
"version": "v2",
"created": "Mon, 21 Apr 2025 14:37:40 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Cheng",
"Yutong",
""
],
[
"Bajaber",
"Osama",
""
],
[
"Tsegai",
"Saimon Amanuel",
""
],
[
"Song",
"Dawn",
""
],
[
"Gao",
"Peng",
""
]
] |
2410.21676 | Hanlin Zhang | Hanlin Zhang, Depen Morwani, Nikhil Vyas, Jingfeng Wu, Difan Zou,
Udaya Ghai, Dean Foster, and Sham Kakade | How Does Critical Batch Size Scale in Pre-training? | ICLR 2025, Blog post:
https://kempnerinstitute.harvard.edu/research/deeper-learning/how-does-critical-batch-size-scale-in-pre-training-decoupling-data-and-model-size | null | null | null | cs.LG cs.AI math.OC stat.ML | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Training large-scale models under given resources requires careful design of
parallelism strategies. In particular, the efficiency notion of critical batch
size (CBS), concerning the compromise between time and compute, marks the
threshold beyond which greater data parallelism leads to diminishing returns.
To operationalize it, we propose a measure of CBS and pre-train a series of
auto-regressive language models, ranging from 85 million to 1.2 billion
parameters, on the C4 dataset. Through extensive hyper-parameter sweeps and
careful control of factors such as batch size, momentum, and learning rate
along with its scheduling, we systematically investigate the impact of scale on
CBS. Then we fit scaling laws with respect to model and data sizes to decouple
their effects. Overall, our results demonstrate that CBS scales primarily with
data size rather than model size, a finding we justify theoretically through
the analysis of infinite-width limits of neural networks and
infinite-dimensional least squares regression. Of independent interest, we
highlight the importance of common hyper-parameter choices and strategies for
studying large-scale pre-training beyond fixed training durations.
| [
{
"version": "v1",
"created": "Tue, 29 Oct 2024 02:54:06 GMT"
},
{
"version": "v2",
"created": "Fri, 8 Nov 2024 19:04:32 GMT"
},
{
"version": "v3",
"created": "Thu, 20 Feb 2025 01:24:46 GMT"
},
{
"version": "v4",
"created": "Mon, 21 Apr 2025 04:19:56 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Zhang",
"Hanlin",
""
],
[
"Morwani",
"Depen",
""
],
[
"Vyas",
"Nikhil",
""
],
[
"Wu",
"Jingfeng",
""
],
[
"Zou",
"Difan",
""
],
[
"Ghai",
"Udaya",
""
],
[
"Foster",
"Dean",
""
],
[
"Kakade",
"Sham",
""
]
] |
2410.21687 | Hao-Xuan Gao | Hao-Xuan Gao, Jin-Jun Geng, Yi-Fang Liang, Hui Sun, Fan Xu, Xue-Feng
Wu, Yong-Feng Huang, Zi-Gao Dai, Wei-Min Yuan | The Soft X-ray Aspect of Gamma-ray Bursts in the Einstein Probe Era | 21 pages, 11 figures, 3 tables; Accepted for publication in ApJ | null | null | null | astro-ph.HE | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | The Einstein Probe (EP) satellite, dedicated at time-domain high-energy
astrophysics and multi-messenger astronomy, was recently launched and
successfully put into operation. The wide-field X-ray telescope (WXT, 0.5-4
keV) onboard has identified multiple gamma-ray burst (GRB) events, with an
average duration of several hundred seconds. This duration is several times
longer than the average duration of long gamma-ray bursts (LGRBs) detected by
the Neil Gehrels Swift Observatory, which typically stands at several tens of
seconds. Additionally, EP has detected some unknown X-ray transients whose
connection to GRBs is uncertain, due to the absence of gamma-ray counterparts
and efficient follow-up observation at multi-wavelengths. Several main factors
could account for the longer time, including the Doppler effect of off-axis
viewing, the spectral lag effect of the synchrotron spectrum of cooling
electrons, and some unknown prolonged intrinsic X-ray activities. Our studies
indicate that EP GRBs may primarily consist of off-axis viewed bursts, forming
a unique population among the GRB zoo, yet the intrinsic origin for the
specific bursts could not be excluded. By analyzing the statistical properties
of the historical LGRB samples, we explored observable properties of on-axis
and off-axis LGRBs in the soft X-ray band. The predicted characteristics of
off-axis viewed GRBs, including the duration, energy fluence, low-energy
spectral index, and the slopes of Amati and Yonetoku relations, could be tested
with a larger sample of GRB events detected by EP in the future.
| [
{
"version": "v1",
"created": "Tue, 29 Oct 2024 03:12:08 GMT"
},
{
"version": "v2",
"created": "Thu, 31 Oct 2024 02:08:03 GMT"
},
{
"version": "v3",
"created": "Sat, 19 Apr 2025 02:44:54 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Gao",
"Hao-Xuan",
""
],
[
"Geng",
"Jin-Jun",
""
],
[
"Liang",
"Yi-Fang",
""
],
[
"Sun",
"Hui",
""
],
[
"Xu",
"Fan",
""
],
[
"Wu",
"Xue-Feng",
""
],
[
"Huang",
"Yong-Feng",
""
],
[
"Dai",
"Zi-Gao",
""
],
[
"Yuan",
"Wei-Min",
""
]
] |
2410.21822 | Ming Kang | Ming Kang, Fung Fung Ting, Rapha\"el C.-W. Phan, Chee-Ming Ting | PK-YOLO: Pretrained Knowledge Guided YOLO for Brain Tumor Detection in
Multiplanar MRI Slices | References updated; for example, papers in NeurIPS 2024 proceedings
appeared on 6 Feb 2025 and AAAI 2025 one on 11 Apr 2025 | In WACV (2025) 3732--3741 | 10.1109/WACV61041.2025.00367 | null | cs.CV eess.IV eess.SP stat.AP | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Brain tumor detection in multiplane Magnetic Resonance Imaging (MRI) slices
is a challenging task due to the various appearances and relationships in the
structure of the multiplane images. In this paper, we propose a new You Only
Look Once (YOLO)-based detection model that incorporates Pretrained Knowledge
(PK), called PK-YOLO, to improve the performance for brain tumor detection in
multiplane MRI slices. To our best knowledge, PK-YOLO is the first pretrained
knowledge guided YOLO-based object detector. The main components of the new
method are a pretrained pure lightweight convolutional neural network-based
backbone via sparse masked modeling, a YOLO architecture with the pretrained
backbone, and a regression loss function for improving small object detection.
The pretrained backbone allows for feature transferability of object queries on
individual plane MRI slices into the model encoders, and the learned domain
knowledge base can improve in-domain detection. The improved loss function can
further boost detection performance on small-size brain tumors in multiplanar
two-dimensional MRI slices. Experimental results show that the proposed PK-YOLO
achieves competitive performance on the multiplanar MRI brain tumor detection
datasets compared to state-of-the-art YOLO-like and DETR-like object detectors.
The code is available at https://github.com/mkang315/PK-YOLO.
| [
{
"version": "v1",
"created": "Tue, 29 Oct 2024 07:45:59 GMT"
},
{
"version": "v2",
"created": "Mon, 21 Apr 2025 06:30:30 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Kang",
"Ming",
""
],
[
"Ting",
"Fung Fung",
""
],
[
"Phan",
"Raphaël C. -W.",
""
],
[
"Ting",
"Chee-Ming",
""
]
] |
2410.21863 | Huaiqiang Yu | Bao-Zhu Guo, Huaiqiang Yu, Meixuan Zhang | On invariance of observability for BSDEs and its applications to
stochastic control systems | 26 Pages | null | null | null | math.OC | http://creativecommons.org/licenses/by/4.0/ | In this paper, we establish the invariance of observability for the observed
backward stochastic differential equations (BSDEs) with constant coefficients,
relative to the filtered probability space. This signifies that the
observability of these observed BSDEs with constant coefficients remains
unaffected by the selection of the filtered probability space. As an
illustrative application, we demonstrate that for stochastic control systems
with constant coefficients, weak observability, approximate null
controllability with cost, and stabilizability are equivalent across some or
any filtered probability spaces.
| [
{
"version": "v1",
"created": "Tue, 29 Oct 2024 08:56:34 GMT"
},
{
"version": "v2",
"created": "Thu, 7 Nov 2024 00:33:28 GMT"
},
{
"version": "v3",
"created": "Tue, 1 Apr 2025 21:59:51 GMT"
},
{
"version": "v4",
"created": "Mon, 21 Apr 2025 01:55:13 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Guo",
"Bao-Zhu",
""
],
[
"Yu",
"Huaiqiang",
""
],
[
"Zhang",
"Meixuan",
""
]
] |
2410.22098 | Yuxin Huang | Yuxin Huang, Khee-Gan Lee, Noam I. Libeskind, Sunil Simha, Aur\'elien
Valade and J. Xavier Prochaska | Modeling the Cosmic Dispersion Measure in the D < 120 Mpc Local Universe | 15 pages, 15 figures, 5 tables, Submitted to MNRAS | Mon. Not. R. Astron. Soc., 538, 2785-2799 (2025) | 10.1093/mnras/staf417 | null | astro-ph.CO astro-ph.GA | http://creativecommons.org/licenses/by/4.0/ | The Local Universe (D < 120 Mpc) has been intensely studied for decades, with
highly complete galaxy redshift surveys now publicly available. These data have
driven density reconstructions of the underlying matter density field, as well
as constrained simulations that aim to reproduce the observed structures. In
this paper, we introduce a dispersion measure (DM) model that makes use of this
detailed knowledge of our Local Universe within D < 120 Mpc. The model
comprises three key components: (i) the DM from the Milky Way halo and the
intra-group medium (up to 3.4 Mpc), derived from the HESTIA simulations, a
series of constrained hydrodynamic simulations designed to reproduce our Local
Group; (ii) the DM contribution from the large-scale intergalactic medium
beyond the Local Group (3.4 Mpc < D < 120 Mpc), calculated using the HAMLET
reconstructed matter density field; and (iii) the individual DM contributions
from Local Universe galaxy halos and clusters based on data from the 2MASS
Galaxy Group Catalog and the NASA/IPAC Extragalactic Database. This
comprehensive model will be made available as a Python package. As the most
realistic model to date for DM in the local volume, it promises to improve the
constraints of DM contributions from the Intergalactic Medium and
Circumgalactic Medium of FRBs, thereby enhancing the accuracy of cosmic baryon
distribution calculations based on DM analysis of FRBs.
| [
{
"version": "v1",
"created": "Tue, 29 Oct 2024 14:51:39 GMT"
},
{
"version": "v2",
"created": "Wed, 30 Oct 2024 07:17:21 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Huang",
"Yuxin",
""
],
[
"Lee",
"Khee-Gan",
""
],
[
"Libeskind",
"Noam I.",
""
],
[
"Simha",
"Sunil",
""
],
[
"Valade",
"Aurélien",
""
],
[
"Prochaska",
"J. Xavier",
""
]
] |
2410.22867 | Jianxiong Li | Jianxiong Li, Boyang Li, Zhuoqiang Guo, Mingzhen Li, Enji Li, Lijun
Liu, Guojun Yuan, Zhan Wang, Guangming Tan, Weile Jia | Scaling Molecular Dynamics with ab initio Accuracy to 149 Nanoseconds
per Day | 11 pages, 11 figures, 3 tables, SC'24 | SC24.(2024) 1-15 | 10.1109/SC41406.2024.00036 | null | cs.DC | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Physical phenomena such as chemical reactions, bond breaking, and phase
transition require molecular dynamics (MD) simulation with ab initio accuracy
ranging from milliseconds to microseconds. However, previous state-of-the-art
neural network based MD packages such as DeePMD-kit can only reach 4.7
nanoseconds per day on the Fugaku supercomputer. In this paper, we present a
novel node-based parallelization scheme to reduce communication by 81%, then
optimize the computationally intensive kernels with sve-gemm and mixed
precision. Finally, we implement intra-node load balance to further improve the
scalability. Numerical results on the Fugaku supercomputer show that our work
has significantly improved the time-to-solution of the DeePMD-kit by a factor
of 31.7x, reaching 149 nanoseconds per day on 12,000 computing nodes. This work
has opened the door for millisecond simulation with ab initio accuracy within
one week for the first time.
| [
{
"version": "v1",
"created": "Wed, 30 Oct 2024 10:04:31 GMT"
}
] | 2025-04-22T00:00:00 | [
[
"Li",
"Jianxiong",
""
],
[
"Li",
"Boyang",
""
],
[
"Guo",
"Zhuoqiang",
""
],
[
"Li",
"Mingzhen",
""
],
[
"Li",
"Enji",
""
],
[
"Liu",
"Lijun",
""
],
[
"Yuan",
"Guojun",
""
],
[
"Wang",
"Zhan",
""
],
[
"Tan",
"Guangming",
""
],
[
"Jia",
"Weile",
""
]
] |
Subsets and Splits
No saved queries yet
Save your SQL queries to embed, download, and access them later. Queries will appear here once saved.