arXiv preprints from January 1, 2026 through July 20, 2026 — 01:21:44 EST
Long-context inference enhances the reasoning capability of Large Language Models (LLMs), but incurs significant computational overhead. Token-oriented methods, such as pruning and skipping, have shown great promise in reducing inference latency, yet still suffer from inherently insufficient structure optimization, outdated selection...
The crystal structure of Rb$_2$Ru$_2$O$_7$.H$_2$O was determined by three-dimensional electron diffraction from the individual crystallites of a solid-state powder product. Rb$_2$Ru$_2$O$_7$.H$_2$O crystallizes in space group \textit{C}2/\textit{c} ($a=7.841(3)$ Å, $b=12.500(3)$ Å, $c=8.392(2)$ Å, $β=93.57(4)^\circ$, Z=4). The...
The high-energy proton-proton collisions at the Large Hadron Collider provide the ideal conditions to study the rare processes predicted by the Standard Model (SM) such as the production of multiple electroweak bosons. These processes involve the self-interactions of the gauge bosons through triple and quartic gauge couplings (TGCs...
Let $p$ be an odd prime. We show that for sufficiently large $n$, every $2$-block of $\mathfrak{S}_n$ and $\mathfrak{A}_n$ contains an ordinary irreducible character of degree divisible by $p$. For almost all $2$-blocks of $\mathfrak{A}_n$, we classify whether it contains a rational valued ordinary irreducible character of degree...
For a normal projective variety $X$, the $\bf Q$-factoriality defect $σ(X)$ is defined to be the rank of the quotient of the group of Weil divisors by the subgroup of Cartier ones. We prove a slight improvement of a topological formula of S.G. Park and M. Popa asserting that $σ(X)=h^{2n-2}(X)-h^2(X)$ by assuming only...
Condition-Based Maintenance (CBM) signifies a paradigm shift from reactive to proactive equipment management strategies in modern industrial systems. Conventional time-based maintenance schedules frequently engender superfluous expenditures and unanticipated equipment failures. In contrast, CBM utilizes real-time equipment condition...
Design-based inference, also known as randomization-based or finite-population inference, provides a principled framework for trustworthy statistical inference. It attributes randomness solely to the design mechanism, such as treatment assignment, survey sampling, or missingness, without imposing super-population distributional or...
We study optimal design problems for stationary diffusion involving one or more state equations and mixtures of an arbitrary number of anisotropic materials. Since such problems typically do not admit classical solutions, we adopt a homogenization-based relaxation framework. The objective considered is the maximization of a weighted...
This work presents Interactive Conversational 3D Virtual Human (ICo3D), a method for generating an interactive, conversational, and photorealistic 3D human avatar. Based on multi-view captures of a subject, we create an animatable 3D face model and a dynamic 3D body model, both rendered by splatting Gaussian primitives. Once merged...
We investigate a minimal extension of the Standard Model with a real singlet scalar and a singlet Dirac fermion acting as dark matter. Unlike a conventional singlet scalar setup, we assume that the singlet scalar does not acquire a vacuum expectation value at zero temperature. This decouples the scalar mixing angle from the...
This paper introduces OPTIMUM-DERAM, a highly consistent, scalable, secure, and decentralized shared memory solution. Traditional distributed shared memory implementations offer multi-object support by multi-threading a single object memory instance over the same set of data hosts. While theoretically sound, the amount of resources...
The development of accurate forecasts of solar eruptive activity has become increasingly important for preventing potential impacts on space technologies and exploration. Therefore, it is crucial to detect Active Regions (ARs) before they start forming on the solar surface. This will enable the development of early-warning...
Early and accurate prediction of solar active region (AR) emergence is crucial for space weather forecasting. Building on established Long Short-Term Memory (LSTM) based approaches for forecasting the continuum intensity decrease associated with AR emergence, this work expands the modeling with new architectures and targets. We...
In this work, we present FastAV, the first token pruning framework tailored for audio-visual large language models (AV-LLMs). While token pruning has been actively explored in standard large language models (LLMs) and vision-language models (LVLMs), its application to AV-LLMs has received little attention, even though multimodal...
Recent large vision-language models (LVLMs) have demonstrated strong potential for device control. However, existing research has primarily focused on point-and-click (PnC) interaction, while remote-control (RC) interaction commonly encountered in everyday TV usage remains largely underexplored. To fill this gap, we introduce...
Carbon monoxide is thought to be the second most abundant molecule in the Universe. This makes observation of both its parent isotopologue ($^{12}$C$^{16}$O) and its stable isotopologues, $^{13}$C$^{16}$O, $^{12}$C$^{18}$O, $^{12}$C$^{17}$O, $^{13}$C$^{18}$O and $^{13}$C$^{17}$O, important in variety of objects. Here the MARVEL...
Diffusion models have recently achieved impressive results in reconstructing images from noisy inputs, and similar ideas have been applied to speech enhancement by treating time-frequency representations as images. With the ubiquity of multi-microphone devices, we extend state-of-the-art diffusion-based methods to exploit multichannel...
Refactoring is a software engineering practice that aims to improve code quality without altering program behavior. Although automated refactoring tools have been extensively studied, their practical applicability remains limited. Recent advances in Large Language Models (LLMs) have introduced new opportunities for automated code...
We investigate the impact of flickering variability in jet power on the luminosity and morphology of radio galaxies. We use a Lagrangian particle method together with relativistic hydrodynamics simulations using the PLUTO code to track the evolution of electron spectra through particle acceleration at shocks and cooling processes. We...
With the wide application of large language models (LLMs), the problems of bias and value inconsistency in sensitive domains have gradually emerged, especially in terms of race, society and politics. In this paper, we propose an adversarial alignment framework, which enhances the value consistency of the model in sensitive domains...
In this paper, we consider a discrete-time Markov Decision Process (MDP) on a finite state-action space with a long-run risk-sensitive criterion used as the objective function. We discuss the concept of Blackwell optimality and comment on intricacies which arise when the risk-neutral expectation is replaced by the risk-sensitive...
We compute the exact complexity of the set of all arc-connected compact subsets of $\boldmath R^2$, which turns out to be strictly higher than the classical $\boldmath Σ^1_1$ and $\boldmath Π^1_1$ classes of analytic and coanalytic sets, but stricly lower than the class $\boldmath Π^1_2$ which is the exact descriptive class of the set...
Geospatial Foundation Models (GFMs) provide powerful representations, but high compute costs hinder their widespread use. Pre-computed embedding data products offer a practical "frozen" alternative, yet they currently exist in a fragmented ecosystem of incompatible formats and resolutions. This lack of standardization creates an...
Human-centric visual analysis plays a pivotal role in diverse applications, including surveillance, healthcare, and human-computer interaction. With the emergence of large-scale unlabeled human image datasets, there is an increasing need for a general unsupervised pre-training model capable of supporting diverse human-centric...
Vision-Language Models (VLMs) have demonstrated strong reasoning capabilities over images and videos, yet their application to embodied scene understanding often constrained by the fixed viewpoints stored in episodic RGB-D memories. These observations may fail to capture query-relevant evidence due to occlusions, object truncation,...