arXiv subjects

Tarunima Prabhakar

Publications and source records attributed to Tarunima Prabhakar.

7 recordsLinked to original sources

arXiv2025

AILuminate: Introducing v1.0 of the AI Risk and Reliability Benchmark from MLCommons

The rapid advancement and deployment of AI systems have created an urgent need for standard safety-evaluation frameworks. This paper introduces AILuminate v1.0, the first comprehensive industry-standard benchmark for assessing AI-product risk and reliability. Its development employed an open process that included participants from multiple fields. The benchmark evaluates an AI system's resistance to prompts designed to elicit dangerous, illegal, or undesirable behavior in 12 hazard categories, including violent crimes, nonviolent crimes, sex-related crimes, child sexual exploitation, indiscriminate weapons, suicide and self-harm, intellectual property, privacy, defamation, hate, sexual content, and specialized advice (election, financial, health, legal). Our method incorporates a complete assessment standard, extensive prompt datasets, a novel evaluation framework, a grading and reporting system, and the technical as well as organizational infrastructure for long-term support and evolution. In particular, the benchmark employs an understandable five-tier grading scale (Poor to Excellent) and incorporates an innovative entropy-based system-response evaluation. In addition to unveiling the benchmark, this report also identifies limitations of our method and of building safety benchmarks generally, including evaluator uncertainty and the constraints of single-turn interactions. This work represents a crucial step toward establishing global standards for AI risk and reliability evaluation while acknowledging the need for continued development in areas such as multiturn interactions, multimodal understanding, coverage of additional languages, and emerging hazard categories. Our findings provide valuable insights for model developers, system integrators, and policymakers working to promote safer AI deployment.

cs.CY↗

arXiv2025

Analysis of Indic Language Capabilities in LLMs

This report evaluates the performance of text-in text-out Large Language Models (LLMs) to understand and generate Indic languages. This evaluation is used to identify and prioritize Indic languages suited for inclusion in safety benchmarks. We conduct this study by reviewing existing evaluation studies and datasets; and a set of twenty-eight LLMs that support Indic languages. We analyze the LLMs on the basis of the training data, license for model and data, type of access and model developers. We also compare Indic language performance across evaluation datasets and find that significant performance disparities in performance across Indic languages. Hindi is the most widely represented language in models. While model performance roughly correlates with number of speakers for the top five languages, the assessment after that varies.

cs.CL↗

arXiv2024

The Uli Dataset: An Exercise in Experience Led Annotation of oGBV

Online gender based violence has grown concomitantly with adoption of the internet and social media. Its effects are worse in the Global majority where many users use social media in languages other than English. The scale and volume of conversations on the internet has necessitated the need for automated detection of hate speech, and more specifically gendered abuse. There is, however, a lack of language specific and contextual data to build such automated tools. In this paper we present a dataset on gendered abuse in three languages- Hindi, Tamil and Indian English. The dataset comprises of tweets annotated along three questions pertaining to the experience of gender abuse, by experts who identify as women or a member of the LGBTQIA community in South Asia. Through this dataset we demonstrate a participatory approach to creating datasets that drive AI systems.

cs.CL↗

arXiv2024

Misinformation as a harm: structured approaches for fact-checking prioritization

In this work, we examine how fact-checkers prioritize which claims to fact-check and what tools may assist them in their efforts. Through a series of interviews with 23 professional fact-checkers from around the world, we validate that harm assessment is a central component of how fact-checkers triage their work. We also clarify the processes behind fact-checking prioritization, finding that they are typically ad hoc, and gather suggestions for tools that could help with these processes. To address the needs articulated by fact-checkers, we present a structured framework of questions to help fact-checkers negotiate the priority of claims through assessing potential harms. Our FABLE Framework of Misinformation Harms incorporates five dimensions of magnitude -- (social) Fragmentation, Actionability, Believability, Likelihood of spread, and Exploitativeness -- that can help determine the potential urgency of a specific message or claim when considering misinformation as harm. The result is a practical and conceptual tool to support fact-checkers and others as they make strategic decisions to prioritize their efforts. We conclude with a discussion of computational approaches to support structured prioritization, as well as applications beyond fact-checking to content moderation and curation.

cs.HC↗

arXiv2024

Overview of the 2023 ICON Shared Task on Gendered Abuse Detection in Indic Languages

This paper reports the findings of the ICON 2023 on Gendered Abuse Detection in Indic Languages. The shared task deals with the detection of gendered abuse in online text. The shared task was conducted as a part of ICON 2023, based on a novel dataset in Hindi, Tamil and the Indian dialect of English. The participants were given three subtasks with the train dataset consisting of approximately 6500 posts sourced from Twitter. For the test set, approximately 1200 posts were provided. The shared task received a total of 9 registrations. The best F-1 scores are 0.616 for subtask 1, 0.572 for subtask 2 and, 0.616 and 0.582 for subtask 3. The paper contains examples of hateful content owing to its topic.

cs.CL↗

arXiv2021

Check Mate: Prioritizing User Generated Multi-Media Content for Fact-Checking

Volume of content and misinformation on social media is rapidly increasing. There is a need for systems that can support fact checkers by prioritizing content that needs to be fact checked. Prior research on prioritizing content for fact-checking has focused on news media articles, predominantly in English language. Increasingly, misinformation is found in user-generated content. In this paper we present a novel dataset that can be used to prioritize check-worthy posts from multi-media content in Hindi. It is unique in its 1) focus on user generated content, 2) language and 3) accommodation of multi-modality in social media posts. In addition, we also provide metadata for each post such as number of shares and likes of the post on ShareChat, a popular Indian social media platform, that allows for correlative analysis around virality and misinformation. The data is accessible on Zenodo (https://zenodo.org/record/4032629) under Creative Commons Attribution License (CC BY 4.0).

cs.SI↗

arXiv2015

Plagiarism Detection in Polyphonic Music using Monaural Signal Separation

Given the large number of new musical tracks released each year, automated approaches to plagiarism detection are essential to help us track potential violations of copyright. Most current approaches to plagiarism detection are based on musical similarity measures, which typically ignore the issue of polyphony in music. We present a novel feature space for audio derived from compositional modelling techniques, commonly used in signal separation, that provides a mechanism to account for polyphony without incurring an inordinate amount of computational overhead. We employ this feature representation in conjunction with traditional audio feature representations in a classification framework which uses an ensemble of distance features to characterize pairs of songs as being plagiarized or not. Our experiments on a database of about 3000 musical track pairs show that the new feature space characterization produces significant improvements over standard baselines.

cs.SD↗