Searcharxiv⌕ Search

arXiv subjects

Deepak Kumar

Publications and source records attributed to Deepak Kumar.

At least 55 records · Page 3Linked to original sources
arXiv2026

The Effect of Corneal Topography and Mucins on Tear Film Rupture

Tear film rupture on the corneal surface plays a critical role in ocular health and visual comfort. Conventional theoretical approaches often idealize the cornea as a perfectly smooth surface, ignoring the surface roughness that are characteristic of healthy as well as diseased eyes. In this study, we develop a comprehensive mathematical model to investigate tear film dynamics over the corneal surface incorporating the effects of surface roughness, slip, van der Waals forces, and lipid transport at the film-air interface. The corneal surface is represented by a small-amplitude periodic modulation. Steady-state solutions obtained using asymptotics reveal nonlinear corrections to the base profile at $O(η^2)$, which are confirmed numerically. Linear stability analysis performed using the Floquet theory demonstrates that an increase in the amplitude of roughness destabilizes the film. Specifically, both the dominant growth rate and the most unstable wavenumber increase with the roughness amplitude. Nonlinear simulations show that surface roughness significantly accelerates tear-film rupture. The slip coefficient, amplitude of roughness of the corneal surface and the initial film profile are found to significantly influence the rupture time. Moreover, the location of the rupture is sensitive to the initial disturbance. These results highlight the crucial role of surface topography and slip in determining tear film stability. The predicted rupture times are consistent with the experimental observations. The proposed model provides a realistic and accurate prediction of tear film dynamics and rupture over the corneal surface. This study offers a new perspective on tear film instability and will help address challenges such as contact lens failure which is related to tear film behavior.

physics.flu-dyn↗
arXiv2026

First ex-vivo positronium imaging of tissues with modular J-PET scanner using $^{44}$Sc radionuclide

This study presents the first ex-vivo positronium imaging of human tissues using the modular J-PET scanner with the $^{44}$Sc radionuclide. The $^{44}$Sc isotope was produced via the $^{44}$Ca(p, n)$^{44}$Sc nuclear reaction and used to perform positronium imaging of phantom composed of human adipose tissue, cardiac myxoma tissue, thrombi blood clot, and also porous polymer XAD4, and a certified reference material (CRM) made from fused silica. The experiment demonstrates the suitability of $^{44}$Sc as a positron source for positronium imaging. The performance of J-PET for positronium imaging with $^{44}$Sc was validated by proper reconstruction of the mean orthopositronium lifetime for CRM material and XAD-4 polymer. The mean ortho-positronium (oPs) lifetimes determined for adipose tissue, cardiac myxoma tissues and thrombi were consistent with results of previous experiments. The study highlights the potential $^{44}$Sc radionuclide for positronium lifetime imaging (PLI).

physics.med-ph↗
arXiv2026

How Far Can Pretrained LLMs Go in Symbolic Music? Controlled Comparisons of Supervised and Preference-based Adaptation

Music often shares notable parallels with language, motivating the use of pretrained large language models (LLMs) for symbolic music understanding and generation. Despite growing interest, the practical effectiveness of adapting instruction-tuned LLMs to symbolic music remains insufficiently characterized. We present a controlled comparative study of finetuning strategies for ABC-based generation and understanding, comparing an off-the-shelf instruction-tuned backbone to domain-adapted variants and a music-specialized LLM baseline. Across multiple symbolic music corpora and evaluation signals, we provide some insights into adaptation choices for symbolic music applications. We highlight the domain adaptation vs.~preserving prior information tradeoff as well as the distinct behaviour of metrics used to measure the domain adaptation for symbolic music.

cs.SD↗
arXiv2026

Taming Toxic Talk: Using chatbots to intervene with users posting toxic comments

Generative AI chatbots have proven surprisingly effective at persuading people to change their beliefs and attitudes in lab settings. However, the practical implications of these findings are not yet clear. In this work, we explore the impact of rehabilitative conversations with generative AI chatbots on users who share toxic content online. Toxic behaviors -- like insults or threats of violence, are widespread in online communities. Strategies to deal with toxic behavior are typically punitive, such as removing content or banning users. Rehabilitative approaches are rarely attempted, in part due to the emotional and psychological cost of engaging with aggressive users. In collaboration with seven large Reddit communities, we conducted a large-scale field experiment (N=893) to invite people who had recently posted toxic content to participate in conversations with AI chatbots. A qualitative analysis of the conversations shows that many participants engaged in good faith and even expressed remorse or a desire to change. However, we did not observe a significant change in toxic behavior in the following month compared to a control group. We discuss possible explanations for our findings, as well as theoretical and practical implications based on our results.

cs.HC↗
arXiv2026

Feasibility study of the positronium lifetime imaging with the Biograph Vision Quadra and J-PET tomographs

Background: After its first ex-vivo and in-vivo demonstration, Positronium Lifetime Imaging (PLI) has received considerable interest as a potential new diagnostic biomarker. High sensitivity Positron Emission Tomography (PET) systems are needed for PLI since it requires simultaneous registration of annihilation photons and prompt gamma. In this simulation-based study, a~feasibility of PLI with the long axial field-of-view Biograph Vision Quadra (Quadra) and the Total Body J-PET scanner was investigated. Methods: The study was performed using the GATE software. Background radiation, present within the Quadra tomograph, was added to the simulation. First, the optimal placement of the energy window for the registration of the prompt gamma was investigated. Next, the organ-wise sensitivity of Quadra was calculated for the $^{68}$Ga, $^{44}$Sc, $^{22}$Na and $^{124}$I radioisotopes. Finally, the sensitivity for the scandium isotope was compared to the sensitivities obtainable with the Total Body J-PET scanner, as well as with the modular J-PET prototype. Results: The PLI sensitivities for the Quadra with the background radiation are estimated to 9.22(3), 10.46(4), 5.91(3), and 15.39(4) cps/kBq for the $^{44}$Sc, $^{68}$Ga, $^{22}$Na and $^{124}$I radioisotopes, respectively. The highest sensitivity was obtained when the energy window for the deexcitation photon is adjacent to the energy window for the annihilation photons. The determined PLI sensitivities with Quadra and the Total Body J-PET are in the order of sensitivities of standard PET imaging with the short axial field-of-view ($\sim$20 cm) PET scanners. Conclusion: The organ-wise PLI sensitivity of Quadra has been computed for the $^{68}$Ga, $^{44}$Sc, $^{22}$Na and $^{124}$I radioisotopes. A sensitivity gain by a factor of 150 was estimated relative to the modular J-PET system previously used for the first in-vivo PLI.

physics.med-ph↗
arXiv2025

What Comes After Harm? Mapping Reparative Actions in AI through Justice Frameworks

As Artificial Intelligence (AI) systems are integrated into more aspects of society, they offer new capabilities but also cause a range of harms that are drawing increasing scrutiny. A large body of work in the Responsible AI community has focused on identifying and auditing these harms. However, much less is understood about what happens after harm occurs: what constitutes reparation, who initiates it, and how effective these reparations are. In this paper, we develop a taxonomy of AI harm reparation based on a thematic analysis of real-world incidents. The taxonomy organizes reparative actions into four overarching goals: acknowledging harm, attributing responsibility, providing remedies, and enabling systemic change. We apply this framework to a dataset of 1,060 AI-related incidents, analyzing the prevalence of each action and the distribution of stakeholder involvement. Our findings show that reparation efforts are concentrated in early, symbolic stages, with limited actions toward accountability or structural reform. Drawing on theories of justice, we argue that existing responses fall short of delivering meaningful redress. This work contributes a foundation for advancing more accountable and reparative approaches to Responsible AI.

cs.HC↗
arXiv2025

Surface energy-driven crumpling transition in a thin sheet under compression

In our common experience, crumpling a sheet requires external compressive force and leads to a random network of folds. However, thin sheets have been theoretically predicted to spontaneously transition from a flat to a crumpled state driven by thermal fluctuations, a phenomenon that has been elusive in experiments. We report the first observation of a similar crumpling transition driven instead by surface energy. Using a sensitive experimental protocol, when we gently compress a thin polymer sheet weakly adhered to a hydrogel substrate it transitions to a self-crumpling state at a well defined critical compression independent of system details. The transition is marked by the percolation of a fold network, and a power law increase in fold density. Most remarkably, the crumpled state shows a tunable order of folds establishing the phenomenon's potential as a simple and scalable technique to do origami with extremely thin sheets.

cond-mat.soft↗
arXiv2025

FT-ARM: Fine-Tuned Agentic Reflection Multimodal Language Model for Pressure Ulcer Severity Classification with Reasoning

Pressure ulcers (PUs) are a serious and prevalent healthcare concern. Accurate classification of PU severity (Stages I-IV) is essential for proper treatment but remains challenging due to subtle visual distinctions and subjective interpretation, leading to variability among clinicians. Prior AI-based approaches using Convolutional Neural Networks (CNNs) and Vision Transformers (ViTs) achieved promising accuracy but offered limited interpretability. We present FT-ARM (Fine-Tuned Agentic Reflection Multimodal model), a fine-tuned multimodal large language model (MLLM) with an agentic self-reflection mechanism for pressure ulcer severity classification. Inspired by clinician-style diagnostic reassessment, FT-ARM iteratively refines its predictions by reasoning over visual features and encoded clinical knowledge from text, enhancing both accuracy and consistency. On the publicly available Pressure Injury Image Dataset (PIID), FT-ARM, fine-tuned from LLaMA 3.2 90B, achieved 85% accuracy in classifying PU stages I-IV, surpassing prior CNN-based models by +4%. Unlike earlier CNN/ViT studies that relied solely on offline evaluations, FT-ARM is designed and tested for live inference, reflecting real-time deployment conditions. Furthermore, it produces clinically grounded natural-language explanations, improving interpretability and trust. By integrating fine-tuning and reflective reasoning across multimodal inputs, FT-ARM advances the reliability, transparency, and clinical applicability of automated wound assessment systems, addressing the critical need for consistent and explainable PU staging to support improved patient care.

cs.CV↗
arXiv2025

Stop the Nonconsensual Use of Nude Images in Research

In order to train, test, and evaluate nudity detection models, machine learning researchers typically rely on nude images scraped from the Internet. Our research finds that this content is collected and, in some cases, subsequently distributed by researchers without consent, leading to potential misuse and exacerbating harm against the subjects depicted. This position paper argues that the distribution of nonconsensually collected nude images by researchers perpetuates image-based sexual abuse and that the machine learning community should stop the nonconsensual use of nude images in research. To characterize the scope and nature of this problem, we conducted a systematic review of papers published in computing venues that collect and use nude images. Our results paint a grim reality: norms around the usage of nude images are sparse, leading to a litany of problematic practices like distributing and publishing nude images with uncensored faces, and intentionally collecting and sharing abusive content. We conclude with a call-to-action for publishing venues and a vision for research in nudity detection that balances user agency with concrete research objectives.

cs.CY↗
arXiv2025

First positronium imaging using $^{44}$Sc with the J-PET scanner: a case study on the NEMA-Image Quality phantom

Positronium Lifetime Imaging (PLI), an emerging extension of conventional positron emission tomography (PET) imaging, offers a novel window for probing the submolecular properties of biological tissues by imaging the mean lifetime of the positronium atom. Currently, the method is under rapid development in terms of reconstruction and detection systems. Recently, the first in vivo PLI of the human brain was performed using the J-PET scanner utilizing the $^{68}$Ga isotope. However, this isotope has limitations due to its comparatively low prompt gamma yields, which is crucial for positronium lifetime measurement. Among alternative radionuclides, $^{44}$Sc stands out as a promising isotope for PLI, characterized by a clinically suitable half-life (4.04 hours) emitting 1157 keV prompt gamma in 100% cases after the emission of the positron. This study reports the first experimental demonstration of PLI with $^{44}$Sc, carried out on a NEMA-Image Quality (IQ) phantom using the Modular J-PET tomograph-the first plastic scintillators-based PET scanner.

physics.med-ph↗
arXiv2025

GPT-OSS-20B: A Comprehensive Deployment-Centric Analysis of OpenAI's Open-Weight Mixture of Experts Model

We present a single-GPU (H100, bf16) evaluation of GPT-OSS-20B (Mixture-of-Experts; 20.9B total, approx. 3.61B active) against dense baselines Qwen3-32B and Yi-34B across multiple dimensions. We measure true time-to-first-token (TTFT), full-decode throughput (TPOT), end-to-end latency percentiles, peak VRAM with past key values (PKV) held, and energy via a consistent nvidia-smi-based sampler. At a 2048-token context with 64-token decode, GPT-OSS-20B delivers higher decode throughput and tokens per Joule than dense baselines Qwen3-32B and Yi-34B, while substantially reducing peak VRAM and energy per 1000 generated tokens; its TTFT is higher due to MoE routing overhead. With only 17.3% of parameters active (3.61B of 20.9B), GPT-OSS-20B provides about 31.8% higher decode throughput and 25.8% lower energy per 1000 generated tokens than Qwen3-32B at 2048/64, while using 31.7% less peak VRAM. Normalized by active parameters, GPT-OSS-20B shows markedly stronger per-active-parameter efficiency (APE), underscoring MoE's deployment advantages. We do not evaluate accuracy; this is a deployment-focused study. We release code and consolidated results to enable replication and extension.

cs.AR↗
arXiv2025

Gate Voltage-Controlled Magnetic Anisotropy Effect on Pt-Porphyrin functionalized single-layer graphene

We report a novel approach to engineering large voltage-controlled magnetic anisotropy (VCMA) and enhanced spin-orbit coupling (SOC) at the interface of single-layer graphene (SLG) and NiFe (Py) through non-covalent functionalization with Platinum (II) 5,10,15,20-tetraphenyl porphyrin (Pt-porphyrin). Using chemical vapor deposition (CVD)-grown SLG, we demonstrate that Pt-porphyrin functionalization significantly increases the SOC and enables robust voltage modulation of interfacial magnetic anisotropy, as confirmed by spin-torque ferromagnetic resonance (ST-FMR) measurements. A substantial VCMA coefficient of 375.6 (fJ/(V-m)) is achieved, accompanied by an order-of-magnitude enhancement in spin torque efficiency (θsh) compared to pristine SLG. The resonance field exhibits a clear, reversible shift under applied gate voltage, confirming robust electric-field modulation of interfacial magnetic anisotropy. Raman spectroscopy and X-ray photoelectron spectroscopy (XPS) confirm the structural integrity and effective charge transfer at the functionalized interface. Electrical characterization of back-gated graphene field-effect transistors (GFETs) further reveals tunable electronic properties upon functionalization. Our results establish functionalized graphene/ferromagnet interfaces as a promising platform for low-power, voltage-controlled spintronic devices, paving the way for scalable, energy-efficient memory and logic technologies

cond-mat.mes-hall↗
arXiv2025

Constraints on maximum neutron star mass from proto-neutron star evolution

A proto-neutron star (PNS) gets formed after a successful supernova when the stellar remnant decouples from the ejecta. In this study, we explore a relativistic framework for the finite-temperature $β$-equilibrium limit of equation of state (EOS), constrained via a Bayesian inference methodology. The EOS is constrained by minimal approximations on a few nuclear saturation properties, low-density pure neutron matter constraints from chiral effective field theory, and a neutron star (NS) maximum mass greater than 2.0 $M_{\odot}$. Two sets of EOS derived from the relativistic mean field model for nucleonic and hyperonic matter constrained by a Bayesian inference calculation at the zero temperature limit are used. The thermal adiabatic index ($Γ_{\rm Th}$) is calculated as a function of the baryonic density across several temperatures for both the sets. Our results suggest that the maximum NS mass is of the order of 2.15 $M_\odot$ if hyperons are present. In addition, the present study suggests that an observation of NS with mass larger than $2.2\ M_{\odot}$ can indirectly indicates the absence of hyperons in its core. The deleptonization of hyperonic PNS reduces the stellar maximum mass rendering the PNS exceeding the zero temperature maximum stellar (baryonic) mass limit becomes metastable which is prone to collapse into a black hole while PNS below such a mass threshold evolves to a stable NS.

nucl-th↗
arXiv2025

Modification of the universal relation between mass, radius and nonradial $f$-mode oscillation in proto-neutron stars

Neutron stars are usually assumed to be cold; however, in certain dynamical astrophysical scenarios such as newly born neutron stars or binary star mergers, the temperature effects play a non-negligible role. We systematically derive the equation of state at finite-temperature within a relativistic mean-field hadronic model applicable to such proto-neutron stars. The equation of state so derived considerably affects the mass-radius curve, thereby affecting the nonradial quadruple $f$-mode oscillation frequencies.} Temperature effectively makes the equation of state stiffer at relatively low and intermediate densities, thereby making the star less compact and flattening the mass-radius curve. The $f$-mode frequency for low and intermediate-mass neutron stars decreases with temperature and thus should be easier to detect. The universal relation (connecting $f$-mode frequency, mass, and radius) changes nonlinearly with temperature. The parameters defining the universal relation [$ωM = a(T) \left(\frac{M}{R}\right) + b(T)$] becomes temperature dependent with the coefficients following a parabolic relation with temperature.

astro-ph.HE↗
arXiv2025

Some Compact Generalization of Bernstein-Type Inequalities Preserved by Modified Smirnov Operator

Let $P(z)$ be a polynomial of degree $n$. In $2004$, Aziz and Rather \cite{aziz2004some} investigated the dependence of \[\bigg|P(Rz)-αP(z)+β\biggl\{\biggl(\frac{R+1}{2}\biggr)^n-|α|\biggr\}P(z)\bigg|, \ \text{for} \ z \in B(\mathbb{D}),\] on $\max_{z\in B(\mathbb{D})}|P(z)|$, for every real and complex number $α, β$ satisfying $|α| \leq 1$, $|β| \leq 1$, and $R \geq 1$. This paper presents a compact generalization of several well-known polynomial inequalities using modified Smirnov operator, demonstrating that the operator preserves inequalities between polynomials.

math.CV↗
arXiv2025

Analyzing the dense matter equation of states in the light of the compact object HESS J1731-347

The recent mass ($0.77 \pm ^{0.20}_{0.17}M_{\odot}$) and radius ($10.4\pm^{0.86}_{0.78} \text{km}$) measurement of HESS J1731-347 made it one of the most fascinating object if it is indeed a neutron star. In this work, we examine the current status of the dense matter equation of states in the context of this compact object being a neutron star. We use three sets of equation of states corresponding to the three classes - neutron stars, strange stars, and hybrid stars and perform Bayesian model selection on them. Our results show that for hadronic models, the EoS is preferred to be stiff at the intermediate densities. This makes the Brueckner-Hartree-Fock approximation and models based on effective interactions deviate from current astrophysical observations on the inclusion of HESS J1731-347. Furthermore, for the strange star family, the equation of states composed of three flavor quarks prefers relatively smaller bag parameters. Analyzing the hybrid family of equation of states consisting of a first-order phase transition revealed preferences for early first-order phase transition. Comparing all the preferred equations of state among each family, it was found that the current astrophysical constraints prefer the hybrid equation of states the most.

astro-ph.HE↗
arXiv2025

Enhancing Contrast and Resolution for Electron-beam Lithography on Insulating Substrates

We report on the effect of ambient gas on the contrast and the resolution of electron beam lithography (EBL) in gaseous environments on insulating substrates. Poly(methyl methacrylate) (PMMA) films were exposed in an environmental scanning electron microscope using a 30 keV electron-beam under 1 mbar pressure of helium, water, nitrogen and argon. We found that the choice of ambient gas results in significant variations in contrast, and the clearing dose increases with the gases molecular weight and proton number, consistent with the increase in scattering cross-section. Significantly higher contrast values are obtained for exposure under helium and are accompanied by improved sensitivity. Despite higher sensitivity, helium exhibited the best resolution with 20-nm half-pitch dense lines and spaces. However, water vapor offered a larger process window, particularly on fused silica substrates. We also demonstrate that higher sensitivity results from effective charge dissipation. Thus, for EBL on insulating substrates, helium and water vapor may be desirable choices for charge dissipation depending on the substrate and process conditions.

cond-mat.mes-hall↗