arXiv preprints from January 1, 2026 through September 25, 2026 — 00:35:13 EST
Composed Image Retrieval (CIR) enables users to search for target images using both a reference image and manipulation text, offering substantial advantages over single-modality retrieval systems. However, existing CIR methods suffer from representation space fragmentation: queries and targets comprise heterogeneous modalities and are...
Automatic speech recognition systems have achieved remarkable performance on fluent speech but continue to degrade significantly when processing stuttered speech, a limitation that is particularly acute for low-resource languages like Indonesian where specialized datasets are virtually non-existent. To overcome this scarcity, we...
We provide a self-contained geometric description of the geodesic flow in the three-dimensional Lie group $\mathrm{Sol}$, one of Thurston's eight model geometries. The geometry of geodesics is governed by a single invariant $k\in[0,1]$, its modulus. Generic geodesics spiral around an axis, with well-defined amplitude $A(k)$, period...
Domain-specific large language models (LLMs), typically developed by fine-tuning a pre-trained general-purpose LLM on specialized datasets, represent a significant advancement in applied AI. A common strategy in LLM fine-tuning is curriculum learning, which pre-orders training samples based on metrics like difficulty to improve...
Turbulence transition often arises from a subcritical transition between bistable states characterized by invariant sets of deterministic dynamical systems, and such transitions can be triggered by system noise as rare events. In this study, we employ the Onsager-Machlup (OM) formulation of stochastic dynamics to examine the Hamilton...
Reinforcement Learning with Verifiable Rewards (RLVR) has emerged as an important paradigm for unlocking reasoning capabilities in large language models, exemplified by the success of OpenAI o1 and DeepSeek-R1. Currently, Group Relative Policy Optimization (GRPO) stands as the dominant algorithm in this domain due to its stable...
I describe my activities in Flavour Physics from 1976 to 2026. However, this 50th anniversary is not the only motivation for this writing. The second reason is the 350th anniversary of the discovery of the first animalcula by van Leeuvanhoek in 1676. Flavour physics makes it possible to search for new animalcula at distance scales far...
In this paper we establish gradient estimates for positive solutions to the nonlinear elliptic equation $$Δ_{V}u^{m}+μ(x)u+p(x)u^α=0 , \quad m>1$$on any smooth metric measure space whose $k$-Bakry-Émery curvature is bounded from below by $-(k-1)K$ with $K \geq 0$. Additionally, we obtain related Liouville theorems and Harnack...
A space $X$ is called a generalized Namioka space (g$\mathcal{N}$-space), if for every compact space $Y$ and every separately continuous function $f\colon X\times Y\rightarrow\mathbb{R}$, there exists at least one point $x\in X$ such that $f$ is jointly continuous at each point of $\{x\}\times Y$. We principally prove the following...
We develop a Bayesian nonparametric framework for inference in spatio-temporal Hawkes processes, extending existing theoretical results beyond the purely temporal setting. Our framework encompasses modelling both the background and triggering components of the Hawkes process through Gaussian Processes priors. Under appropriate...
Active Alignment (AA) is a key technology for the large-scale automated assembly of high-precision optical systems. Compared with labor-intensive per-model on-device calibration, a digital-twin pipeline built on optical simulation offers a substantial advantage in generating large-scale labeled data. However, complex imaging...
Cloud computing has been shown to be an essential enabling technology for public sector organizations PSOs and offers numerous potential benefits, including reduced information technology infrastructure costs, increased innovation potential, and improved resource resilience and scalability. Despite governments' intensifying efforts to...
The possibility to constrain cosmological parameters from galaxy surveys using field-level machine learning methods that bypass traditional summary statistics analyses, depends crucially on our ability to generate simulated training sets. The latter need to be both realistic, as to reproduce the key features of the real data, and...
Sampling-based model predictive control (MPC) is effective for nonlinear systems but often produces non-smooth control inputs due to random sampling. To address this issue, we extend the model predictive path integral (MPPI) framework with deterministic sampling and improvements from cross-entropy method (CEM)--MPC, such as iterative...
Electro-laryngeal (EL) speech is characterized by constant pitch, limited prosody, and mechanical noise, reducing naturalness and intelligibility. We propose a lightweight adaptation of the state-of-the-art StreamVC framework to this setting by removing pitch and energy modules and combining self-supervised pretraining with supervised...
This paper introduces and studies the stability of the strong domination number of a graph, denoted $\operatorname{st}_{γ_{st}}(G)$, defined as the minimum number of vertices whose removal changes the strong domination number $γ_{st}(G)$. We determine exact values of this stability parameter for several fundamental graph classes,...
Active colloid clustering is central to understanding non-equilibrium self-organization, with implications for programmable active materials and synthetic or biological assemblies. While most prior studies have focused on dimers or small aggregates, the dynamics of larger clusters remain relatively unexplored. Here, we experimentally...
Mixture of Experts (MoE) architectures enable efficient scaling of neural networks but suffer from expert collapse, where routing converges to a few dominant experts. This reduces model capacity and causes catastrophic interference during adaptation. We propose the Spectrally-Regularized Mixture of Experts (SR-MoE), which imposes...
The disclosure of the Log4Shell vulnerability in December 2021 led to an unprecedented wave of global scanning and exploitation activity. A recent study provided important initial insights, but was largely limited in duration and geography, focusing primarily on European and U.S. network telescope deployments and covering the...
In prior work, we introduced IndexTTS 2, a zero-shot neural text-to-speech foundation model comprising two core components: a transformer-based Text-to-Semantic (T2S) module and a non-autoregressive Semantic-to-Mel (S2M) module, which together enable faithful emotion replication and establish the first autoregressive...
The microscopic and fundamental origin of slide electrification, where droplets of water move across insulating surfaces accumulating and depositing electrical charges, is still debated. Charge transfer is often attributed to ion transfer at the receding contact line. However, it is still unclear whether ion transfer alone can fully...
Griffith thermodynamic energy balance is employed to analyze cleavage phenomenon from atomic level. Results show that the cleavage toughness, the strain energy release rate, and the surface energy can be defined by the bond strength (the appropriate elastic modulus ) and the bond density. Such simple definition of fracture parameters...
Both tensor trains (TTs) and quantum states provide compressed representations of grid-structured data with potentially exponential compression power. We present a unified framework for upsampling data encoded in vector amplitudes, with efficient realizations in both classical TT and quantum settings. Starting from an \(n\)-core TT or...
Distributing government relief efforts after a flood is challenging. In India, the crops are widely affected by floods; therefore, making rapid and accurate crop damage assessment is crucial for effective post-disaster agricultural management. Traditional manual surveys are slow and biased, while current satellite-based methods face...
The slitless spectroscopic method employed by missions such as Euclid and the Chinese Space-station Survey Telescope (CSST) faces a fundamental challenge: spectroscopic redshifts derived from their data are susceptible to emission-line misidentification due to the limited spectral resolution and signal-to-noise ratio. This effect...