Searcharxiv⌕ Search

arXiv subjects

Pooja Yadav

Publications and source records attributed to Pooja Yadav.

18 recordsLinked to original sources
arXiv2026

Assessing Reliability of BERT-Based Models on Question Answering Tasks

Reliability estimation of large language models is in many cases as crucial as their accuracy, as reliable models are more trustworthy, robust, and suitable for practical applications. Recent advancements in natural language processing (NLP), particularly those based on transformer architectures, have significantly accelerated progress across various NLP tasks. This study focuses on the reliability of transformer-based question answering (QA) models, specifically BERT models and its variants (RoBERTa, ALBERT, DistilBERT). These encoder-only pretrained transformers have demonstrated remarkable accuracy in QA tasks that can be treated as classification tasks. However, their reliability remains underexplored. This study evaluates the reliability of four BERT-based models by assessing response stability under two conditions: (1) internal model variations induced via Monte Carlo Dropout (MCD) and (2) input perturbations through paraphrasing. Using the SQuAD and QuAC datasets, we investigate how dropout rates affect prediction consistency and whether lexical changes impact answer stability. Our findings reveal that RoBERTa maintains higher reliability, whereas AlBERT and DistilBERT exhibit significant inconsistencies. Statistical analyses confirm that enabling MCD during prediction does not disrupt inference dynamics, validating its effectiveness as a reliability metric. These findings underscore the importance of evaluating both accuracy and stability in QA models to ensure stability in real-world applications.

cs.CL↗
arXiv2025

Persistence diagrams as morphological signatures of cells: A method to measure and compare cells within a population

Cell biologists study in parallel the morphology of cells with the regulation mechanisms that modify this morphology. Such studies are complicated by the inherent heterogeneity present in the cell population. It remains difficult to define the morphology of a cell with parameters that can quantify this heterogeneity, leaving the cell biologist to rely on manual inspection of cell images. We propose an alternative to this manual inspection that is based on topological data analysis. We characterise the shape of a cell by its contour and nucleus. We build a filtering of the edges defining the contour using a radial distance function initiated from the nucleus. This filtering is then used to construct a persistence diagram that serves as a signature of the cell shape. Two cells can then be compared by computing the Wasserstein distance between their persistence diagrams. Given a cell population, we then compute a distance matrix that includes all pairwise distances between its members. We analyse this distance matrix using hierarchical clustering with different linkage schemes and define a purity score that quantifies consistency between those different schemes, which can then be used to assess homogeneity within the cell population. We illustrate and validate our approach to identify sub-populations in human mesenchymal stem cell populations.

q-bio.QM↗
arXiv2025

Likelihood Geometry of the Gumbel's Type-I Bivariate Exponential Distribution

In algebraic statistics, the maximum likelihood degree of a statistical model refers to the number of solutions (counted with multiplicity) of the score equations over the complex field. In this paper, the maximum likelihood degree of the association parameter of Gumbel's Type-I bivariate exponential distribution is investigated using algebraic techniques.

math.ST↗
arXiv2025

Maximum Likelihood Estimation in the Multivariate and Matrix Variate Symmetric Laplace Distributions through Group Actions

In this paper, we study the maximum likelihood estimation of the parameters of the multivariate and matrix variate symmetric Laplace distributions through group actions. The multivariate and matrix variate symmetric Laplace distributions are not in the exponential family of distributions. We relate the maximum likelihood estimation problems of these distributions to norm minimization over a group and build a correspondence between stability of data with respect to the group action and the properties of the likelihood function.

math.ST↗
arXiv2025

Maximum Likelihood Estimation of the Parameters of Matrix Variate Symmetric Laplace Distribution

This paper considers an extension of the multivariate symmetric Laplace distribution to matrix variate case. The symmetric Laplace distribution is a scale mixture of normal distribution. The maximum likelihood estimators (MLE) of the parameters of multivariate and matrix variate symmetric Laplace distribution are proposed, which are not explicitly obtainable, as the density function involves the modified Bessel function of the third kind. Thus, the EM algorithm is applied to find the maximum likelihood estimators. The parameters and their maximum likelihood estimators of matrix variate symmetric Laplace distribution are defined up to a positive multiplicative constant with their Kronecker product uniquely defined. The condition for the existence of the MLE is given, and the stability of the estimators is discussed. The empirical bias and the dispersion of the Kronecker product of the estimators for different sample sizes are discussed using simulated data.

math.ST↗
arXiv2024

A new differential subordination technique for a subclass of starlike functions

In the present investigation, we employ a new technique to find several first and second order differential subordination implications involving the following starlike class associated with a bean shaped domain: \begin{equation*} \mathcal{S}^*_{\mathfrak{B}}:=\left\{f\in\mathcal{S}:\dfrac{zf'(z)}{f(z)}\prec\sqrt{1+\tanh{z}}=:\mathfrak{B}(z)\right\}. \end{equation*} Also, we give several applications stemming from our derived results.

math.CV↗
arXiv2024

On Applications of Extended Jack's Lemma

We introduce and study the class ${\bf\mathcal{G}}(α,β)$ comprising analytic functions associated with a sector domain, where $α,β\in(0,1]$. Using the extended version of Jack's lemma, we deduce Open-Door lemma type sufficient conditions for functions to be in ${\bf\mathcal{G}}(α,β)$. Furthermore, we point out special cases of our results that align with known results.

math.CV↗
arXiv2024

On Starlike Functions Associated with a Bean Shaped Domain

In this paper, we introduce and explore a new class of starlike functions denoted by $\mathcal{S}^*_{\mathfrak{B}}$, defined as follows: $$\mathcal{S}^*_{\mathfrak{B}}=\{f\in \mathcal{A}:zf'(z)/f(z)\prec \sqrt{1+\tanh{z}}=:\mathfrak{B}(z)\}.$$ Here, $\mathfrak{B}(z)$ represents a mapping from the unit disk onto a bean-shaped domain. Our study focuses on understanding the characteristic properties of both $\mathfrak{B}(z)$ and the functions in $\mathcal{S}^*_{\mathfrak{B}}$. We derive sharp conditions under which $ψ(p)\prec\sqrt{1+\tanh(z)}$ implies $p(z)\prec ((1+A z)/(1+B z))^γ$, where $ψ(p)$ is defined as: \begin{equation*} (1-α)p(z)+αp^2(z)+β\frac{zp'(z)}{p^k(z)}\quad \text{and}\quad (p(z))^δ+β\frac{zp'(z)}{(p(z))^k}. \end{equation*} Additionally, we establish inclusion relations involving $\mathcal{S}^*_{\mathfrak{B}}$ and derive precise estimates for the sharp radii constants of $\mathcal{S}^*_{\mathfrak{B}}$.

math.CV↗
arXiv2024

LLM-based Smart Reply (LSR): Enhancing Collaborative Performance with ChatGPT-mediated Smart Reply System

Interactive user interfaces have increasingly explored AI's role in enhancing communication efficiency and productivity in collaborative tasks. The emergence of Large Language Models (LLMs) such as ChatGPT has revolutionized conversational agents, employing advanced deep learning techniques to generate context-aware, coherent, and personalized responses. Consequently, LLM-based AI assistants provide a more natural and efficient user experience across various scenarios. In this paper, we study how LLM models can be used to improve work efficiency in collaborative workplaces. Specifically, we present an LLM-based Smart Reply (LSR) system utilizing the ChatGPT to generate personalized responses in professional collaborative scenarios while adapting to context and communication style based on prior responses. Our two-step process involves generating a preliminary response type (e.g., Agree, Disagree) to provide a generalized direction for message generation, thus reducing response drafting time. We conducted an experiment where participants completed simulated work tasks involving a Dual N-back test and subtask scheduling through Google Calendar while interacting with co-workers. Our findings indicate that the proposed LSR reduces overall workload, as measured by the NASA TLX, and improves work performance and productivity in the N-back task. We also provide qualitative analysis based on participants' experiences, as well as design considerations to provide future directions for improving such implementations.

cs.HC↗
arXiv2023

Nitrogen in Silicon for Room Temperature Single Electron Tunneling Devices

Single electron transistor (SET) is an advanced tool to exploit in quantum devices. Working of such devices at room-temperature is essential for practical utilization. Dopant based single-electron devices are well studied at low-temperature although a few devices are developed for high-temperature operation with certain limitations. Here, we propose and theoretically exhibit that nitrogen (N) donor in silicon is an important candidate for effective designing of such devices. Theoretical calculation of density-of-states using semi-empirical DFT method indicates that N-donor in silicon has deep ground state compared to a phosphorus (P) donor. N-donor spectrum is explored in nano-silicon along with the P-donor. Comparative data of Bohr radius of N-donor and P-donor is also reported. The simulated current-voltage characteristics confirm that N-doped device is better suited for SET operation at room-temperature.

cond-mat.mes-hall↗
arXiv2022

On a Class of certain Non-Univalent Functions

In this paper, we introduce a family of analytic functions given by $$ψ_{A,B}(z):= \dfrac{1}{A-B}\log{\dfrac{1+Az}{1+Bz}},$$ which maps univalently the unit disk onto either elliptical or strip domains, where either $A=-B=α$ or $A=αe^{iγ}$ and $B=αe^{-iγ}$ ($α\in(0,1]$ and $γ\in(0,π/2]$). We study a class of non-univalent analytic functions defined by \begin{equation*} \mathcal{F}[A,B]:=\left\{f\in\mathcal{A}:\left( \dfrac{zf'(z)}{f(z)}-1\right)\precψ_{A,B}(z)\right \}. \end{equation*} Further, we investigate various characteristic properties of $ψ_{A,B}(z)$ as well as functions in the class $\mathcal{F}[A,B]$ and obtain the sharp radius of starlikeness of order $δ$ and univalence for the functions in $\mathcal{F}[A,B]$. Also, we find the sharp radii for functions in $\mathcal{BS}(α):=\{f\in\mathcal{A}:zf'(z)/f(z)-1\prec z/(1-αz^2),\;α\in(0,1)\}$, $\mathcal{S}_{cs}(α):=\{f\in\mathcal{A}:zf'(z)/f(z)-1\prec z/((1-z)(1+αz)),\;α\in(0,1)\}$ and others to be in the class $\mathcal{F}[A,B].$

math.CV↗
arXiv2022

Common Fixed Point Theorems for Four Transformations in Vector S-metric Spaces

We establish some common fixed point results for four transformations in vector S-metric spaces by using the notion of weakly compatibility (WC) and occasionally weakly compatibility (OWC). The first theorem is proved by using the concept of CLR (p,q) property (common limit range property w.r.t. transformations p and q) and weakly compatiblity whereas second theorem is proved by using the concept of occasionally weakly compatiblity

math.GM↗
arXiv2022

Inelastic Cotunneling Resonances in the Coulomb-Blockade Transport in Donor-Atom Transistors

We report finite-bias characteristics of electrical transport through phosphorus donors in silicon nanoscale transistors, in which we observe inelastic-cotunneling current in the Coulomb blockade region. The cotunneling current appears like a resonant-tunneling current peak emerging from the excited state at the crossover between blockade and non-blockade regions. These cotunneling features are unique, since the inelastic-cotunneling currents have so far been reported either as a broader hump or as a continuous increment of current. This finding is ascribed purely due to excitation-related inelastic cotunneling involving the ground and excited states. Theoretical calculations were performed for a two-level quantum dot, supporting our experimental observation.

cond-mat.mes-hall↗
arXiv2021

On Oblique Domains of Janowski Function

We investigate certain properties of tilted (oblique) domains, associated with the Janowski function $(1+Az)/(1+Bz),$ where $A,B\in\mathbb{C}$ with $A\neq B$ and $|B|\leq 1$. We find several bounds for these oblique domains. Further, we extend the various known argument, radius, and subordination results involving Janowski functions with complex parameters.

math.CV↗
arXiv2016

Role of Metal Nanoparticles on porosification of silicon by metal induced etching (MIE)

Porosification of silicon (Si) by metal induced etching (MIE) process have been studies here to understand the etching mechanism. The etching mechanism has been discussed on the basis of electron transfer from Si to metal ion (Ag$^+$) and metal to H$_2$O$_2$. Role of silver nanoparticles (AgNPs) in the etching process has been investigated by studying the effect of AgNPs coverage on surface porosity. A quantitative analysis of SEM images, done using Image J, shows a direct correlation between AgNPs coverage and surface porosity after the porosification. Density of Si nanowires (NWs) also varies as a function of AgNPs fractional coverage which reasserts the fact that AgNPs governs the porosification process during MIE.

cond-mat.mes-hall↗