arXiv subjects

Atharva Kulkarni

Publications and source records attributed to Atharva Kulkarni.

At least 19 recordsLinked to original sources

arXiv2026

SurakshaEval: An Indic Safety Benchmark for Multilingual LLMs

Existing safety evaluation datasets for large language models (LLMs) predominantly focus on English and Western contexts, often overlooking the linguistic diversity and culturally grounded safety risks present in other languages. To address this gap, we introduce SurakshaEval, a novel safety benchmark composed of human-written prompts spanning real-world scenarios, explicitly designed for ten major Indian languages - Assamese, Bengali, Gujarati, Hindi, Kannada, Malayalam, Marathi, Punjabi, Tamil, and Telugu, along with English. SurakshaEval includes both generic prompts common across India and region- and language-specific prompts that capture localized sociocultural sensitivities. We benchmark a broad range of state-of-the-art LLMs on SurakshaEval, establish baseline safety performance, and identify recurring failure modes, including over-refusal, missed detection of implicit bias, and insufficient contextual awareness in regionally sensitive settings. Our results show that even strong multilingual LLMs struggle to reliably meet nuanced safety requirements when operating in Indic languages, particularly in native scripts. These findings highlight the urgent need for safety evaluation frameworks that incorporate region-specific data and structured assessment protocols, enabling the development and deployment of AI systems that operate securely, ethically, and in alignment with diverse societal values. Our code and data are available at https://github.com/debobanerjee/SurakshaEval. Warning: This paper contains text that may be offensive or unsafe.

cs.CL↗

arXiv2026

High temporal resolution THz streaking of high brightness relativistic electron beams

We report a systematic experimental study of terahertz (THz) streaking structures for ultrafast characterization of relativistic, high-brightness electron beams. Horn-coupled waveguide geometries are investigated, enabling a comparative characterization of streaking strength, dispersion, transmission, and temporal fidelity. Analytical models and electromagnetic simulations are used to describe the dependence of streaking power on the waveguide dimensions and the drive frequency. Experimentally, the structures are characterized using compressed electron beam from an RF photoinjector over a range of THz field strengths, beam energies, and bunch durations. These results establish general design principles and performance limits for THz streaking structures applicable to ultrafast electron beam diagnostics.

physics.acc-ph↗

arXiv2026

Do Compact SSL Backbones Matter for Audio Deepfake Detection? A Controlled Study with RAPTOR

Self-supervised learning (SSL) underpins modern audio deepfake detection, yet most prior work centers on a single large wav2vec2-XLSR backbone, leaving compact under studied. We present RAPTOR, Representation Aware Pairwise-gated Transformer for Out-of-domain Recognition a controlled study of compact SSL backbones from the HuBERT and WavLM within a unified pairwise-gated fusion detector, evaluated across 14 cross-domain benchmarks. We show that multilingual HuBERT pre-training is the primary driver of cross-domain robustness, enabling 100M models to match larger and commercial systems. Beyond EER, we introduce a test-time augmentation protocol with perturbation-based aleatoric uncertainty to expose calibration differences invisible to standard metrics: WavLM variants exhibit overconfident miscalibration under perturbation, whereas iterative mHuBERT remains stable. These findings indicate that SSL pre-training trajectory, not model scale, drives reliable audio deepfake detection.

cs.SD↗

arXiv2026

Disentangling Geometry, Performance, and Training in Language Models

Geometric properties of Transformer weights, particularly the unembedding matrix, have been widely useful in language model interpretability research. Yet, their utility for estimating downstream performance remains unclear. In this work, we systematically investigate the relationship between model performance and the unembedding matrix geometry, particularly its effective rank. Our experiments, involving a suite of 108 OLMo-style language models trained under controlled variation, reveal several key findings. While the best-performing models often exhibit a high effective rank, this trend is not universal across tasks and training setups. Contrary to prior work, we find that low effective rank does not cause late-stage performance degradation in small models, but instead co-occurs with it; we find adversarial cases where low-rank models do not exhibit saturation. Moreover, we show that effective rank is strongly influenced by pre-training hyperparameters, such as batch size and weight decay, which in-turn affect the model's performance. Lastly, extending our analysis to other geometric metrics and final-layer representation, we find that these metrics are largely aligned, but none can reliably predict downstream performance. Overall, our findings suggest that the model's geometry, as captured by existing metrics, primarily reflects training choices rather than performance.

cs.CL↗

arXiv2025

Economies of Open Intelligence: Tracing Power & Participation in the Model Ecosystem

Since 2019, the Hugging Face Model Hub has been the primary global platform for sharing open weight AI models. By releasing a dataset of the complete history of weekly model downloads (June 2020-August 2025) alongside model metadata, we provide the most rigorous examination to-date of concentration dynamics and evolving characteristics in the open model economy. Our analysis spans 851,000 models, over 200 aggregated attributes per model, and 2.2B downloads. We document a fundamental rebalancing of economic power: US open-weight industry dominance by Google, Meta, and OpenAI has declined sharply in favor of unaffiliated developers, community organizations, and, as of 2025, Chinese industry, with DeepSeek and Qwen models potentially heralding a new consolidation of market power. We identify statistically significant shifts in model properties, a 17X increase in average model size, rapid growth in multimodal generation (3.4X), quantization (5X), and mixture-of-experts architectures (7X), alongside concerning declines in data transparency, with open weights models surpassing truly open source models for the first time in 2025. We expose a new layer of developer intermediaries that has emerged, focused on quantizing and adapting base models for both efficiency and artistic expression. To enable continued research and oversight, we release the complete dataset with an interactive dashboard for real-time monitoring of concentration dynamics and evolving properties in the open model economy.

cs.CY↗

arXiv2025

Experimental verification of space-charge saturation scaling laws in high-gradient photocathode RF guns

We investigate the limits of photoemission yield in a high-gradient S-band radiofrequency photoinjector in the space-charge-dominated regime. Using an RF phase-scan technique, where the emitted charge is measured as a function of the RF-field phase in the gun, we directly monitor photoemission over a range of launch fields and laser parameters, enabling quantitative characterization of space-charge saturation. Measurements, supported by simulations and analytic modeling, confirm the characteristic charge-field scaling laws for pancake beams and provide the first experimental verification of cigar-regime scaling in an RF photogun. These results establish a predictive framework for identifying the onset of space-charge saturation and guide the optimization of photoinjectors for ultrafast electron diffraction, microscopy, and high-brightness light sources operating at ultra-high gradients.

physics.acc-ph↗

arXiv2025

Speech DF Arena: A Leaderboard for Speech DeepFake Detection Models

Parallel to the development of advanced deepfake audio generation, audio deepfake detection has also seen significant progress. However, a standardized and comprehensive benchmark is still missing. To address this, we introduce Speech DeepFake (DF) Arena, the first comprehensive benchmark for audio deepfake detection. Speech DF Arena provides a toolkit to uniformly evaluate detection systems, currently across 14 diverse datasets and attack scenarios, standardized evaluation metrics and protocols for reproducibility and transparency. It also includes a leaderboard to compare and rank the systems to help researchers and developers enhance their reliability and robustness. We include 14 evaluation sets, 12 state-of-the-art open-source and 3 proprietary detection systems. Our study presents many systems exhibiting high EER in out-of-domain scenarios, highlighting the need for extensive cross-domain evaluation. The leaderboard is hosted on Huggingface1 and a toolkit for reproducing results across the listed datasets is available on GitHub.

cs.SD↗

arXiv2025

Code-Switching in End-to-End Automatic Speech Recognition: A Systematic Literature Review

Motivated by a growing research interest into automatic speech recognition (ASR), and the growing body of work for languages in which code-switching (CS) often occurs, we present a systematic literature review of code-switching in end-to-end ASR models. We collect and manually annotate papers published in peer reviewed venues. We document the languages considered, datasets, metrics, model choices, and performance, and present a discussion of challenges in end-to-end ASR for code-switching. Our analysis thus provides insights on current research efforts and available resources as well as opportunities and gaps to guide future research.

cs.CL↗

arXiv2025

Music for All: Representational Bias and Cross-Cultural Adaptability of Music Generation Models

The advent of Music-Language Models has greatly enhanced the automatic music generation capability of AI systems, but they are also limited in their coverage of the musical genres and cultures of the world. We present a study of the datasets and research papers for music generation and quantify the bias and under-representation of genres. We find that only 5.7% of the total hours of existing music datasets come from non-Western genres, which naturally leads to disparate performance of the models across genres. We then investigate the efficacy of Parameter-Efficient Fine-Tuning (PEFT) techniques in mitigating this bias. Our experiments with two popular models -- MusicGen and Mustango, for two underrepresented non-Western music traditions -- Hindustani Classical and Turkish Makam music, highlight the promises as well as the non-triviality of cross-genre adaptation of music through small datasets, implying the need for more equitable baseline music-language models that are designed for cross-cultural transfer learning.

cs.SD↗

arXiv2025

Evaluating Evaluation Metrics -- The Mirage of Hallucination Detection

Hallucinations pose a significant obstacle to the reliability and widespread adoption of language models, yet their accurate measurement remains a persistent challenge. While many task- and domain-specific metrics have been proposed to assess faithfulness and factuality concerns, the robustness and generalization of these metrics are still untested. In this paper, we conduct a large-scale empirical evaluation of 6 diverse sets of hallucination detection metrics across 4 datasets, 37 language models from 5 families, and 5 decoding methods. Our extensive investigation reveals concerning gaps in current hallucination evaluation: metrics often fail to align with human judgments, take an overtly myopic view of the problem, and show inconsistent gains with parameter scaling. Encouragingly, LLM-based evaluation, particularly with GPT-4, yields the best overall results, and mode-seeking decoding methods seem to reduce hallucinations, especially in knowledge-grounded settings. These findings underscore the need for more robust metrics to understand and quantify hallucinations, and better strategies to mitigate them.

cs.CL↗

arXiv2025

Unveiling Biases while Embracing Sustainability: Assessing the Dual Challenges of Automatic Speech Recognition Systems

In this paper, we present a bias and sustainability focused investigation of Automatic Speech Recognition (ASR) systems, namely Whisper and Massively Multilingual Speech (MMS), which have achieved state-of-the-art (SOTA) performances. Despite their improved performance in controlled settings, there remains a critical gap in understanding their efficacy and equity in real-world scenarios. We analyze ASR biases w.r.t. gender, accent, and age group, as well as their effect on downstream tasks. In addition, we examine the environmental impact of ASR systems, scrutinizing the use of large acoustic models on carbon emission and energy consumption. We also provide insights into our empirical analyses, offering a valuable contribution to the claims surrounding bias and sustainability in ASR systems.

cs.CL↗

arXiv2024

Still Not Quite There! Evaluating Large Language Models for Comorbid Mental Health Diagnosis

In this study, we introduce ANGST, a novel, first-of-its kind benchmark for depression-anxiety comorbidity classification from social media posts. Unlike contemporary datasets that often oversimplify the intricate interplay between different mental health disorders by treating them as isolated conditions, ANGST enables multi-label classification, allowing each post to be simultaneously identified as indicating depression and/or anxiety. Comprising 2876 meticulously annotated posts by expert psychologists and an additional 7667 silver-labeled posts, ANGST posits a more representative sample of online mental health discourse. Moreover, we benchmark ANGST using various state-of-the-art language models, ranging from Mental-BERT to GPT-4. Our results provide significant insights into the capabilities and limitations of these models in complex diagnostic scenarios. While GPT-4 generally outperforms other models, none achieve an F1 score exceeding 72% in multi-class comorbid classification, underscoring the ongoing challenges in applying language models to mental health diagnostics.

cs.CL↗

arXiv2024

Vector-based feedback of continuous wave radiofrequency compression cavity for ultrafast electron diffraction

The temporal resolution of ultrafast electron diffraction (UED) at weakly relativistic beam energies ($\lesssim$ 100 keV) suffers from space-charge induced electron pulse broadening. We describe the implementation of a radio frequency (RF) cavity operating in the continuous wave regime to compress high repetition rate electron bunches from a 40.4 kV DC photoinjector for ultrafast electron diffraction applications. Active stabilization of the RF amplitude and phase through a feedback loop based on the demodulated in-phase and quadrature components of the RF signal is demonstrated. This scheme yields 144$\pm$19 fs RMS temporal resolution in pump-probe studies.

physics.acc-ph↗

arXiv2024

Multitask Learning Can Improve Worst-Group Outcomes

In order to create machine learning systems that serve a variety of users well, it is vital to not only achieve high average performance but also ensure equitable outcomes across diverse groups. However, most machine learning methods are designed to improve a model's average performance on a chosen end task without consideration for their impact on worst group error. Multitask learning (MTL) is one such widely used technique. In this paper, we seek not only to understand the impact of MTL on worst-group accuracy but also to explore its potential as a tool to address the challenge of group-wise fairness. We primarily consider the standard setting of fine-tuning a pre-trained model, where, following recent work \citep{gururangan2020don, dery2023aang}, we multitask the end task with the pre-training objective constructed from the end task data itself. In settings with few or no group annotations, we find that multitasking often, but not consistently, achieves better worst-group accuracy than Just-Train-Twice (JTT; \citet{pmlr-v139-liu21f}) -- a representative distributionally robust optimization (DRO) method. Leveraging insights from synthetic data experiments, we propose to modify standard MTL by regularizing the joint multitask representation space. We run a large number of fine-tuning experiments across computer vision and natural language processing datasets and find that our regularized MTL approach \emph{consistently} outperforms JTT on both average and worst-group outcomes. Our official code can be found here: \href{https://github.com/atharvajk98/MTL-group-robustness.git}{\url{https://github.com/atharvajk98/MTL-group-robustness}}.

cs.LG↗

arXiv2024

SynthDST: Synthetic Data is All You Need for Few-Shot Dialog State Tracking

In-context learning with Large Language Models (LLMs) has emerged as a promising avenue of research in Dialog State Tracking (DST). However, the best-performing in-context learning methods involve retrieving and adding similar examples to the prompt, requiring access to labeled training data. Procuring such training data for a wide range of domains and applications is time-consuming, expensive, and, at times, infeasible. While zero-shot learning requires no training data, it significantly lags behind the few-shot setup. Thus, `\textit{Can we efficiently generate synthetic data for any dialogue schema to enable few-shot prompting?}' Addressing this question, we propose \method, a data generation framework tailored for DST, utilizing LLMs. Our approach only requires the dialogue schema and a few hand-crafted dialogue templates to synthesize natural, coherent, and free-flowing dialogues with DST annotations. Few-shot learning using data from {\method} results in $4-5%$ improvement in Joint Goal Accuracy over the zero-shot baseline on MultiWOZ 2.1 and 2.4. Remarkably, our few-shot learning approach recovers nearly $98%$ of the performance compared to the few-shot setup using human-annotated training data. Our synthetic data and code can be accessed at https://github.com/apple/ml-synthdst

cs.CL↗

arXiv2023

Counting the Bugs in ChatGPT's Wugs: A Multilingual Investigation into the Morphological Capabilities of a Large Language Model

Large language models (LLMs) have recently reached an impressive level of linguistic capability, prompting comparisons with human language skills. However, there have been relatively few systematic inquiries into the linguistic capabilities of the latest generation of LLMs, and those studies that do exist (i) ignore the remarkable ability of humans to generalize, (ii) focus only on English, and (iii) investigate syntax or semantics and overlook other capabilities that lie at the heart of human language, like morphology. Here, we close these gaps by conducting the first rigorous analysis of the morphological capabilities of ChatGPT in four typologically varied languages (specifically, English, German, Tamil, and Turkish). We apply a version of Berko's (1958) wug test to ChatGPT, using novel, uncontaminated datasets for the four examined languages. We find that ChatGPT massively underperforms purpose-built systems, particularly in English. Overall, our results -- through the lens of morphology -- cast a new light on the linguistic capabilities of ChatGPT, suggesting that claims of human-like language skills are premature and misleading.

cs.CL↗

arXiv2023

The Parkes Pulsar Timing Array Third Data Release

We present the third data release from the Parkes Pulsar Timing Array (PPTA) project. The release contains observations of 32 pulsars obtained using the 64-m Parkes "Murriyang" radio telescope. The data span is up to 18 years with a typical cadence of 3 weeks. This data release is formed by combining an updated version of our second data release with $\sim 3$ years of more recent data primarily obtained using an ultra-wide-bandwidth receiver system that operates between 704 and 4032 MHz. We provide calibrated pulse profiles, flux-density dynamic spectra, pulse times of arrival, and initial pulsar timing models. We describe methods for processing such wide-bandwidth observations, and compare this data release with our previous release.

astro-ph.HE↗

arXiv2023

SciFix: Outperforming GPT3 on Scientific Factual Error Correction

Due to the prohibitively high cost of creating error correction datasets, most Factual Claim Correction methods rely on a powerful verification model to guide the correction process. This leads to a significant drop in performance in domains like scientific claims, where good verification models do not always exist. In this work, we introduce SciFix, a scientific claim correction system that does not require a verifier but can outperform existing methods by a considerable margin -- achieving correction accuracy of 84% on the SciFact dataset, 77% on SciFact-Open and 72% on the CovidFact dataset, compared to next best accuracies of 7%, 5%, and 15% on the same datasets respectively. Our method leverages the power of prompting with LLMs during training to create a richly annotated dataset that can be used for fully supervised training and regularization. We additionally use a claim-aware decoding procedure to improve the quality of corrected claims. Our method outperforms the very LLM that was used to generate the annotated dataset -- with Few-Shot Prompting on GPT3.5 achieving 58%, 61%, and 64% on the respective datasets, a consistently lower correction accuracy, despite using nearly 800 times as many parameters as our model.

cs.CL↗