arXiv subjects

Kaushal Sharma

Publications and source records attributed to Kaushal Sharma.

15 recordsLinked to original sources

arXiv2026

SectEval: Evaluating the Latent Sectarian Preferences of Large Language Models

As Large Language Models (LLMs) becomes a popular source for religious knowledge, it is important to know if it treats different groups fairly. This study is the first to measure how LLMs handle the differences between the two main sects of Islam: Sunni and Shia. We present a test called SectEval, available in both English and Hindi, consisting of 88 questions, to check the bias-ness of 15 top LLM models, both proprietary and open-weights. Our results show a major inconsistency based on language. In English, many powerful models DeepSeek-v3 and GPT-4o often favored Shia answers. However, when asked the exact same questions in Hindi, these models switched to favoring Sunni answers. This means a user could get completely different religious advice just by changing languages. We also looked at how models react to location. Advanced models Claude-3.5 changed their answers to match the user's country-giving Shia answers to a user from Iran and Sunni answers to a user from Saudi Arabia. In contrast, smaller models (especially in Hindi) ignored the user's location and stuck to a Sunni viewpoint. These findings show that AI is not neutral; its religious ``truth'' changes depending on the language you speak and the country you claim to be from. The data set is available at https://github.com/secteval/SectEval/

cs.CL↗

arXiv2026

IndicParam: Benchmark to evaluate LLMs on low-resource Indic Languages

While large language models excel on high-resource multilingual tasks, low- and extremely low-resource Indic languages remain severely under-evaluated. We present IndicParam, a human-curated benchmark of over 13,000 multiple-choice questions covering 11 such languages (Nepali, Gujarati, Marathi, Odia as low-resource; Dogri, Maithili, Rajasthani, Sanskrit, Bodo, Santali, Konkani as extremely low-resource) plus Sanskrit-English code-mixed set. We evaluated 20 LLMs, both proprietary and open-weights, which reveals that even the top-performing \texttt{Gemini-2.5} reaches 58\% average accuracy, followed by \texttt{GPT-5} (45) and \texttt{DeepSeek-3.2} (43.1). We additionally label each question as knowledge-oriented or purely linguistic to discriminate factual recall from grammatical proficiency. Further, we assess the ability of LLMs to handle diverse question formats-such as list-based matching, assertion-reason pairs, and sequence ordering-alongside conventional multiple-choice questions. \benchmark\ provides insights into limitations of cross-lingual transfer and establishes a challenging benchmark for Indic languages. The dataset is available at https://huggingface.co/datasets/bharatgenai/IndicParam. Scripts to run benchmark are present at https://github.com/ayushbits/IndicParam.

cs.CL↗

arXiv2025

ParamBench: A Graduate-Level Benchmark for Evaluating LLM Understanding on Indic Subjects

Large language models have been widely evaluated on tasks such as comprehension, summarization, code generation, etc. However, their performance on graduate-level, culturally grounded questions in the Indian context remains largely unexplored. Existing Indian benchmarks emphasise basic fact-orientated queries that offer limited assessment of a deeper disciplinary understanding tailored to the Indian setting. In this paper, we present ParamBench, consisting of more than 17K questions in the Hindi language, comprising questionnaires from 21 diverse subjects. These questions are primarily derived from a nationwide graduate-level entrance examination covering topics such as history, music, instruments, yoga, literature, philosophy, law, etc.~ specifically for the Indian context. Additionally, we assess the ability of LLMs to handle diverse question formats - such as list-based matching, assertion-reason pairs, and sequence ordering - alongside conventional multiple-choice questions. We evaluated the performance of more than 16 open source LLMs on this benchmark, observing that Gemma3-27B attains the highest overall accuracy of 56.4\%. Furthermore, subject-wise analysis indicates that even for the best-performing LLMs, performance remains weak on topics such as music, classical instruments, and law, underscoring persistent challenges in culturally grounded reasoning. The dataset and source code is present at https://github.com/ayushbits/ParamBench.

cs.CL↗

arXiv2025

Exploring the habitability and interior composition of exoplanets lying within the extended habitable zone

Studying the habitability, internal structure and composition of exoplanets is crucial for understanding their potential to sustain life beyond our solar system. Characterizing planetary structures and atmospheric evolution provides valuable insights into surface conditions and the long-term habitability of these planets. In this study, we present a comprehensive analysis of exoplanets spanning from super-Earths to mini-Neptunes ($R_{\textrm{p}}$ $\leq$ 4 $R_{\oplus}$ and $M_{\textrm{p}}$ $\leq$ 15 $M_{\oplus}$) located within the extended habitable zone, along with parameterization of their host stars. We find that the planets in our sample orbit M dwarf stars and are tidally locked to them. Using archival photometric data from Gaia, Pan-STARRS1, 2MASS, and WISE, we estimate the atmospheric and physical parameters of the host stars. We also model the interior structure of these planets to infer their possible compositions. Additionally, under the assumption that these exoplanets can accrete a gaseous layer, we model the envelope fraction of the habitable exoplanets. With an Earth-like rocky composition, LHS 1140 b and TOI-1452 b can hold onto negligible amount of their initial gas layer. However, sustaining a sufficient amount of atmosphere over time, the planets LP 791-18 c, LTT 3780 c and K2-18 b are likely to be water worlds. The models suggest a water rich composition for TOI-1266 c without any significant amount of atmosphere. Modeling interior compositions and atmospheric escape scenarios allow us to assess the potential habitability of these planets by evaluating the likelihood of surface liquid water and the retention of stable atmospheres.

astro-ph.EP↗

arXiv2024

Light Curve Properties of Gamma-Ray Burst Associated Supernovae

A rapidly spinning, millisecond magnetar is widely considered one of the most plausible power sources for gamma-ray burst-associated supernovae (GRB-SNe). Recent studies have demonstrated that the magnetar model can effectively explain the bolometric light curves of most GRB-SNe. In this work, we investigate the bolometric light curves of 13 GRB-SNe, focusing on key observational parameters such as peak luminosity, rise time, and decay time, estimated using Gaussian Process (GP) regression for light curve fitting. We also apply Principal Component Analysis to all the light curve parameters to reduce the dimensionality of the dataset and visualize the distribution of SNe in lower-dimensional space. Our findings indicate that while most GRB-SNe share common physical characteristics, a few outliers, notably SNe 2010ma and 2011kl, exhibit distinct features. These events suggest potential differences in progenitor properties or explosion mechanisms, offering deeper insight into the diversity of GRB-SNe and their central engines.

astro-ph.HE↗

arXiv2024

Magnetars as Powering Sources of Gamma-Ray Burst Associated Supernovae, and Unsupervised Clustering of Cosmic Explosions

We present the semi-analytical light curve modelling of 13 supernovae associated with gamma-ray bursts (GRB-SNe) along with two relativistic broad-lined (Ic-BL) SNe without GRBs association (SNe 2009bb and 2012ap), considering millisecond magnetars as central-engine-based power sources for these events. The bolometric light curves of all 15 SNe in our sample are well-regenerated utilising a $χ^2-$minimisation code, $\texttt{MINIM}$, and numerous parameters are constrained. The median values of ejecta mass ($M_{\textrm{ej}}$), magnetar's initial spin period ($P_\textrm{i}$) and magnetic field ($B$) for GRB-SNe are determined to be $\approx$ 5.2 M$_\odot$, 20.5 ms and 20.1 $\times$ 10$^{14}$ G, respectively. We leverage machine learning (ML) algorithms to comprehensively compare the 3-dimensional parameter space encompassing $M_{\textrm{ej}}$, $P_\textrm{i}$, and $B$ for GRB-SNe determined herein to those of H-deficient superluminous SNe (SLSNe-I), fast blue optical transients (FBOTs), long GRBs (LGRBs), and short GRBs (SGRBs) obtained from the literature. The application of unsupervised ML clustering algorithms on the parameters $M_{\textrm{ej}}$, $P_\textrm{i}$, and $B$ for GRB-SNe, SLSNe-I, and FBOTs yields a classification accuracy of $\sim$95%. Extending these methods to classify GRB-SNe, SLSNe-I, LGRBs, and SGRBs based on $P_\textrm{i}$ and $B$ values results in an accuracy of $\sim$84%. Our investigations show that GRB-SNe and relativistic Ic-BL SNe presented in this study occupy different parameter spaces for $M_{\textrm{ej}}$, $P_\textrm{i}$, and $B$ than those of SLSNe-I, FBOTs, LGRBs and SGRBs. This indicates that magnetars with different $P_\textrm{i}$ and $B$ can give birth to distinct types of transients.

astro-ph.HE↗

arXiv2022

Photometric identification of compact galaxies, stars and quasars using multiple neural networks

We present MargNet, a deep learning-based classifier for identifying stars, quasars and compact galaxies using photometric parameters and images from the Sloan Digital Sky Survey (SDSS) Data Release 16 (DR16) catalogue. MargNet consists of a combination of Convolutional Neural Network (CNN) and Artificial Neural Network (ANN) architectures. Using a carefully curated dataset consisting of 240,000 compact objects and an additional 150,000 faint objects, the machine learns classification directly from the data, minimising the need for human intervention. MargNet is the first classifier focusing exclusively on compact galaxies and performs better than other methods to classify compact galaxies from stars and quasars, even at fainter magnitudes. This model and feature engineering in such deep learning architectures will provide greater success in identifying objects in the ongoing and upcoming surveys, such as Dark Energy Survey (DES) and images from the Vera C. Rubin Observatory.

astro-ph.GA↗

arXiv2020

Stellar Spectral Interpolation using Machine Learning

Theoretical stellar spectra rely on model stellar atmospheres computed based on our understanding of the physical laws at play in the stellar interiors. These models, coupled with atomic and molecular line databases, are used to generate theoretical stellar spectral libraries (SSLs) comprising of stellar spectra over a regular grid of atmospheric parameters (temperature, surface gravity, abundances) at any desired resolution. Another class of SSLs is referred to as empirical spectral libraries; these contain observed spectra at limited resolution. SSLs play an essential role in deriving the properties of stars and stellar populations. Both theoretical and empirical libraries suffer from limited coverage over the parameter space. This limitation is overcome to some extent by generating spectra for specific sets of atmospheric parameters by interpolating within the grid of available parameter space. In this work, we present a method for spectral interpolation in the optical region using machine learning algorithms that are generic, easily adaptable for any SSL without much change in the model parameters, and computationally inexpensive. We use two machine learning techniques, Random Forest (RF) and Artificial Neural Networks (ANN), and train the models on the MILES library. We apply the trained models to spectra from the CFLIB for testing and show that the performance of the two models is comparable. We show that both the models achieve better accuracy than the existing methods of polynomial based interpolation and the Gaussian radial basis function (RBF) interpolation.

astro-ph.SR↗

arXiv2019

Application of Convolutional Neural Networks for Stellar Spectral Classification

Due to the ever-expanding volume of observed spectroscopic data from surveys such as SDSS and LAMOST, it has become important to apply artificial intelligence (AI) techniques for analysing stellar spectra to solve spectral classification and regression problems like the determination of stellar atmospheric parameters Teff, log g, [Fe/H]. We propose an automated approach for the classification of stellar spectra in the optical region using Convolutional Neural Networks. Traditional machine learning (ML) methods with "shallow" architecture (usually up to 2 hidden layers) have been trained for these purposes in the past. However, deep learning methods with a larger number of hidden layers allow the use of finer details in the spectrum which results in improved accuracy and better generalisation. Studying finer spectral signatures also enables us to determine accurate differential stellar parameters and find rare objects. We examine various machine and deep learning algorithms like Artificial Neural Networks (ANN), Random Forest (RF), and Convolutional Neural Network (CNN) to classify stellar spectra using the Jacoby Atlas, ELODIE and MILES spectral libraries as training samples. We test the performance of the trained networks on the Indo-U.S. Library of Coude Feed Stellar Spectra (CFLIB). We show that using convolutional neural networks, we are able to lower the error up to 1.23 spectral sub-classes as compared to that of 2 sub-classes achieved in the past studies with ML approach. We further apply the trained model to classify stellar spectra retrieved from the SDSS database with SNR>20.

astro-ph.SR↗

arXiv2019

A Long-term photometric variability and spectroscopic study of luminous blue variable AF And in M31

We present photometric and spectroscopic analysis of the Hubble Sandage variable AF And in M31. The data has been taken under the Nainital Microlensing Survey during 1998-2002 and follow-up observations were carried out until 2011. During this period, photometric observations in Cousins R and I bands were obtained for 169 nights spanning over about 5000 days. AF And has shown a prominent outburst around mid-January in 1999 followed by a gradual decrease in brightness of about 1.5 mag in the next 3 years with a declining rate of ~0.0015 mag/day leading to a quiescent phase at the end of 2001. After lying low for about 9 years, AF And again went through a secondary outburst phase in late 2010 with an amplitude of 0.44 mag where it lasted for one year before fading back to its quiescence phase. The spectroscopic observations of AF And show prominent Balmer and He I emission lines along with the comparatively weaker FeII and [FeII] emissions. Asymmetric emission line profiles in its spectrum imply the mass loss rate of about 2.2x10^{-4} solar mass per yr through the stellar winds in the photosphere. Using SED fitting, we find the photospheric temperature of 33,000+/-3000 K during the visual minimum. Using a weak P Cygni profile of HeI emission line, the wind terminal velocity for AF And is found to be around 280-300 km/s.

astro-ph.SR↗

arXiv2019

Multiwavelength Period-Luminosity and Period-Luminosity-Color relations at maximum-light for Mira variables in the Magellanic Clouds

We present Period-Luminosity and Period-Luminosity-Color relations at maximum-light for Mira variables in the Magellanic Clouds using time-series data from the Optical Gravitational Lensing Experiment (OGLE-III) and {\it Gaia} data release 2. The maximum-light relations exhibit a scatter typically up to $\sim 30\%$ smaller than their mean-light counterparts. The apparent magnitudes of Oxygen-rich Miras at maximum-light display significantly smaller cycle-to-cycle variations than at minimum-light. High-precision photometric data for Kepler Mira candidates also exhibit stable magnitude variations at the brightest epochs while their multi-epoch spectra display strong Balmer emission lines and weak molecular absorption at maximum-light. The stability of maximum-light magnitudes for Miras possibly occurs due to the decrease in the sensitivity to molecular bands at their warmest phase. At near-infrared wavelengths, the Period-Luminosity relations of Miras display similar dispersion at mean and maximum-light with limited time-series data in the Magellanic Clouds. A kink in the Oxygen-rich Mira Period-Luminosity relations is found at 300 days in the $VI$-bands which shifts to longer-periods ($\sim 350$~days) at near-infrared wavelengths. Oxygen-rich Mira Period-Luminosity relations at maximum-light provide a relative distance modulus, $Δμ= 0.48\pm0.08$~mag, between the Magellanic Clouds with a smaller statistical uncertainty than the mean-light relations. The maximum-light properties of Miras can be very useful for stellar atmosphere modeling and distance scale studies provided their stability and the universality can be established in other stellar environments in the era of extremely large telescopes.

astro-ph.SR↗

arXiv2018

Estimating Stellar Atmospheric Parameters by Automated Methods Using SSLs

Libraries of stellar spectra, such as ELODIE (Prugniel & Soubiran 2001), CFLIB (Valdes et al. 2004), or MILES (Sánchez-Blázquez et al. 2006), are used for a variety of applications, and especially in modelling stellar populations (e. g. Le Borgne et al. (2004)). In that context, apart from the completeness and quality of these spectral databases (Singh et al. 2006), the accurate calibration of stellar atmospheric parameters, temperature (Teff), surface gravity (log g), and metallicity ([Fe/H]), is known to be critical (Prugniel et al. 2007; Percival & Salaris 2009). We discuss the technique of determining stellar atmospheric parameters accurately by `full spectrum fitting'.

astro-ph.SR↗

arXiv2018

Low resolution spectroscopic investigation of Am stars using Automated method

Automated method of full spectrum fitting gives reliable estimates of stellar atmospheric parameters (Teff, logg and [Fe/H]) for late A, F, G and early K type stars. Recently, the technique was further improved in the cooler regime and the validity range was extended up to M6 - M7 spectral type( Teff $\sim\,2900$ K). The present study aims to explore the application of this method on the low-resolution spectra of Am stars, a class of chemically peculiar (CP) stars, to examine its robustness for these objects. We use ULySS with MILES (Medium-resolution INT Library of Empirical Spectra) V2 spectral interpolator for parameter determination. Determined Teff and logg are found to be in good agreement with those obtained from high-resolution spectroscopy.

astro-ph.SR↗

arXiv2017

High-resolution Spectroscopy and Spectropolarimetry of Selected Delta Scuti Pulsating Variables

The combination of photometry, spectroscopy and spectropolarimetry of the chemically peculiar stars often aims to study the complex physical phenomena such as stellar pulsation, chemical inhomogeneity, magnetic field and their interplay with stellar atmosphere and circumstellar environment. The prime objective of the present study is to determine the atmospheric parameters of a set of Am stars to understand their evolutionary status. Atmospheric abundances and basic parameters are determined using full spectrum fitting technique by comparing the high-resolution spectra to the synthetic spectra. To know the evolutionary status we derive the effective temperature and luminosity from different methods and compare them with the literature. The location of these stars in the H-R diagram demonstrate that all the sample stars are evolved from the Zero-Age-Main-Sequence towards Terminal-Age-Main-Sequence and occupy the region of $δ$ Sct instability strip. The abundance analysis shows that the light elements e.g. Ca and Sc are underabundant while iron peak elements such as Ba, Ce etc. are overabundant and these chemical properties are typical for Am stars. The results obtained from the spectropolarimetric analysis shows that the longitudinal magnetic fields in all the studied stars are negligible that gives further support their Am class of peculiarity.

astro-ph.SR↗

arXiv2015

New atmospheric parameters and spectral interpolator for the MILES cool stars

Context: The full spectrum fitting of stellar spectra against a library of empirical spectra is a well-established approach to measure the atmospheric parameters of FGK stars with a high internal consistency. Extending it towards cooler stars still remains a challenge. Aims: We address this question by improving the interpolator of the MILES (Medium-resolution INT Library of Empirical Spectra) library in the low effective temperature regime (Tefff < 4800 K), and we refine the determination of the parameters of the cool MILES stars. Methods: We use the ULySS package to determine the atmospheric parameters (Teff, logg and [Fe/H]), and measure the biases of the results with respect to our updated compilation of parameters calibrated against theoretical spectra. After correcting some systematic effects, we compute a new interpolator that we finally use to redetermine the atmospheric parameters homogeneously and assess the biases. Results: Based on an updated literature compilation, we determine Teff in a more accurate and unbiased manner compared to those determined with the original interpolator. The validity range is extended downwards to about Teff = 2900 K compared to 3500 K previously. The mean residual biases on Teff, logg, and [Fe/H], with respect to the literature compilation for the coolest stars (Teff <= 3800 K) computed using the new interpolator, are -15 K, -0.02 dex, and 0.02 dex respectively. The corresponding estimations of the external precision are 63 K, 0.23 dex, and 0.15 dex respectively. For the stars with Teff in the range 3800 - 4200 K, the determinations of Teff and [Fe/H] have been slightly improved. At higher temperatures, the new interpolator is comparable to the original one. The new version of the interpolator is publicly available.

astro-ph.SR↗