Ponnurangam Kumaraguru
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Ponnurangam Kumaraguru has a total of 224 co-authors. Here are the top co-authors by number of shared publications:
- Rajiv Ratn Shah
Indraprastha Institute of Information Technology Delhi
8 shared publications - Ritwik Mishra
Indraprastha Institute of Information Technology Delhi
5 shared publications - Vamshi Krishna Bonagiri
International Institute of Information Technology Hyderabad (IIITH
5 shared publications - Anmol Agarwal
International Institute of Information Technology, Hyderabad, India
4 shared publications - Manas Gaur
CSEE, University of Maryland, Baltimore County, Baltimore, USA
4 shared publications - Shrey Gupta
International Institute of Information Technology, Hyderabad, India
4 shared publications
Latest Publications
- 11 September 2026
- Published by Springer Nature
- p. 398-429
GridVQA-X: A Diagnostic Framework for Evaluating Multimodal Explainability Methods
- 11 September 2026
- Published by Springer Nature
- p. 398-429
GridVQA-X: A Diagnostic Framework for Evaluating Multimodal Explainability Methods
- 18 August 2026
- Published by Springer Nature in International Journal of Data Science and Analytics
- Vol. 22 (1) , 279
A multifaceted GCN-SBERT hybrid model for political polarization detection in online social network
- 18 August 2026
- Published by Springer Nature in International Journal of Data Science and Analytics
- Vol. 22 (1) , 279
A multifaceted GCN-SBERT hybrid model for political polarization detection in online social network
- 14 August 2026
- Published by World Scientific Pub Co Pte Ltd in International Journal of Computational Intelligence and Applications
From Noise to Knowledge: Distill Social Media Opinions on Rumor Detection Exploiting LLMs
The majority of social media research relies on datasets gathered from social media websites such as Reddit and Twitter. However, their intrinsic high noisy content reduces the performance of data-driven models. This restriction of noisy data has made data preparation techniques necessary. Current systems usually need to pay more attention to data quality and their performance heavily depends on noisy social media datasets. In this work, we concentrate on creating high-quality social media data for rumor detection tasks on the widely popular PHEME-9 dataset. Eliminating noisy responses is essential for misinformation tasks since it guarantees that the model has been trained on precise and pertinent data, improving its capacity to identify and validate rumors successfully. Large language models (LLMs) are used in this work to filter out irrelevant comments prior to the application of machine and deep learning techniques. This bifurcated approach helps to improve model accuracy and lower computing burden. We assume that this will further help in accurate rumor identification and can support environmental sustainability using less computational resources. Our proposed methodology shows an average improvement in the trained filtered models’ performance in terms of accuracy and F1-scores on six events in the PHEME-9 dataset. Further, to validate the effectiveness of the trained model, we performed interpretability and error analysis.
- 14 August 2026
- Published by World Scientific Pub Co Pte Ltd in International Journal of Computational Intelligence and Applications
From Noise to Knowledge: Distill Social Media Opinions on Rumor Detection Exploiting LLMs
The majority of social media research relies on datasets gathered from social media websites such as Reddit and Twitter. However, their intrinsic high noisy content reduces the performance of data-driven models. This restriction of noisy data has made data preparation techniques necessary. Current systems usually need to pay more attention to data quality and their performance heavily depends on noisy social media datasets. In this work, we concentrate on creating high-quality social media data for rumor detection tasks on the widely popular PHEME-9 dataset. Eliminating noisy responses is essential for misinformation tasks since it guarantees that the model has been trained on precise and pertinent data, improving its capacity to identify and validate rumors successfully. Large language models (LLMs) are used in this work to filter out irrelevant comments prior to the application of machine and deep learning techniques. This bifurcated approach helps to improve model accuracy and lower computing burden. We assume that this will further help in accurate rumor identification and can support environmental sustainability using less computational resources. Our proposed methodology shows an average improvement in the trained filtered models’ performance in terms of accuracy and F1-scores on six events in the PHEME-9 dataset. Further, to validate the effectiveness of the trained model, we performed interpretability and error analysis.
- 9 June 2026
- Published by Springer Nature
- p. 551-564
MACA: A Framework for Distilling Trustworthy LLMs into Efficient Retrievers
- 9 June 2026
- Published by Springer Nature
- p. 551-564
MACA: A Framework for Distilling Trustworthy LLMs into Efficient Retrievers
- 21 February 2026
- Published by Springer Nature
- p. 295-304
Structured Definitions and Segmentations for Legal Reasoning in LLMs: A Study on Indian Legal Data
- 21 February 2026
- Published by Springer Nature
- p. 295-304
Structured Definitions and Segmentations for Legal Reasoning in LLMs: A Study on Indian Legal Data
- 21 February 2026
- Published by Springer Nature
- p. 3-18
Long-Context Non-factoid Question Answering in Indic Languages
- 21 February 2026
- Published by Springer Nature
- p. 3-18
Long-Context Non-factoid Question Answering in Indic Languages
- Open Access
MACA: A Framework for Distilling Trustworthy LLMs into Efficient Retrievers
Modern enterprise retrieval systems must handle short, underspecified queries such as ``foreign transaction fee refund'' and ``recent check status''. In these cases, semantic nuance and metadata matter but per-query large language model (LLM) re-ranking and manual labeling are costly. We present Metadata-Aware Cross-Model Alignment (MACA), which distills a calibrated metadata aware LLM re-ranker into a compact student retriever, avoiding online LLM calls. A metadata-aware prompt verifies the teacher's trustworthiness by checking consistency under permutations and robustness to paraphrases, then supplies listwise scores, hard negatives, and calibrated relevance margins. The student trains with MACA's MetaFusion objective, which combines a metadata conditioned ranking loss with a cross model margin loss so it learns to push the correct answer above semantically similar candidates with mismatched topic, sub-topic, or entity. On a proprietary consumer banking FAQ corpus and BankFAQs, the MACA teacher surpasses a MAFA baseline at Accuracy@1 by five points on the proprietary set and three points on BankFAQs. MACA students substantially outperform pretrained encoders; e.g., on the proprietary corpus MiniLM Accuracy@1 improves from 0.23 to 0.48, while keeping inference free of LLM calls and supporting retrieval-augmented generation.
Open AccessMACA: A Framework for Distilling Trustworthy LLMs into Efficient Retrievers
Modern enterprise retrieval systems must handle short, underspecified queries such as ``foreign transaction fee refund'' and ``recent check status''. In these cases, semantic nuance and metadata matter but per-query large language model (LLM) re-ranking and manual labeling are costly. We present Metadata-Aware Cross-Model Alignment (MACA), which distills a calibrated metadata aware LLM re-ranker into a compact student retriever, avoiding online LLM calls. A metadata-aware prompt verifies the teacher's trustworthiness by checking consistency under permutations and robustness to paraphrases, then supplies listwise scores, hard negatives, and calibrated relevance margins. The student trains with MACA's MetaFusion objective, which combines a metadata conditioned ranking loss with a cross model margin loss so it learns to push the correct answer above semantically similar candidates with mismatched topic, sub-topic, or entity. On a proprietary consumer banking FAQ corpus and BankFAQs, the MACA teacher surpasses a MAFA baseline at Accuracy@1 by five points on the proprietary set and three points on BankFAQs. MACA students substantially outperform pretrained encoders; e.g., on the proprietary corpus MiniLM Accuracy@1 improves from 0.23 to 0.48, while keeping inference free of LLM calls and supporting retrieval-augmented generation.
- Open Access
PrivacyBench: A Conversational Benchmark for Evaluating Privacy in Personalized AI
Personalized AI agents rely on access to a user's digital footprint, which often includes sensitive data from private emails, chats and purchase histories. Yet this access creates a fundamental societal and privacy risk: systems lacking social-context awareness can unintentionally expose user secrets, threatening digital well-being. We introduce PrivacyBench, a benchmark with socially grounded datasets containing embedded secrets and a multi-turn conversational evaluation to measure secret preservation. Testing Retrieval-Augmented Generation (RAG) assistants reveals that they leak secrets in up to 26.56% of interactions. A privacy-aware prompt lowers leakage to 5.12%, yet this measure offers only partial mitigation. The retrieval mechanism continues to access sensitive data indiscriminately, which shifts the entire burden of privacy preservation onto the generator. This creates a single point of failure, rendering current architectures unsafe for wide-scale deployment. Our findings underscore the urgent need for structural, privacy-by-design safeguards to ensure an ethical and inclusive web for everyone.
Open AccessPrivacyBench: A Conversational Benchmark for Evaluating Privacy in Personalized AI
Personalized AI agents rely on access to a user's digital footprint, which often includes sensitive data from private emails, chats and purchase histories. Yet this access creates a fundamental societal and privacy risk: systems lacking social-context awareness can unintentionally expose user secrets, threatening digital well-being. We introduce PrivacyBench, a benchmark with socially grounded datasets containing embedded secrets and a multi-turn conversational evaluation to measure secret preservation. Testing Retrieval-Augmented Generation (RAG) assistants reveals that they leak secrets in up to 26.56% of interactions. A privacy-aware prompt lowers leakage to 5.12%, yet this measure offers only partial mitigation. The retrieval mechanism continues to access sensitive data indiscriminately, which shifts the entire burden of privacy preservation onto the generator. This creates a single point of failure, rendering current architectures unsafe for wide-scale deployment. Our findings underscore the urgent need for structural, privacy-by-design safeguards to ensure an ethical and inclusive web for everyone.
- Open Access
- 17 December 2025
- Published by Association for Computing Machinery (ACM)
- p. 199-207
Adapting Multilingual Models to Code-Mixed Tasks via Model Merging
- 17 December 2025
- Published by Association for Computing Machinery (ACM)
- p. 199-207
Open AccessAdapting Multilingual Models to Code-Mixed Tasks via Model Merging
- Open Access
Causal Reasoning Favors Encoders: On The Limits of Decoder-Only Models
In context learning (ICL) underpins recent advances in large language models (LLMs), although its role and performance in causal reasoning remains unclear. Causal reasoning demands multihop composition and strict conjunctive control, and reliance on spurious lexical relations of the input could provide misleading results. We hypothesize that, due to their ability to project the input into a latent space, encoder and encoder decoder architectures are better suited for said multihop conjunctive reasoning versus decoder only models. To do this, we compare fine-tuned versions of all the aforementioned architectures with zero and few shot ICL in both natural language and non natural language scenarios. We find that ICL alone is insufficient for reliable causal reasoning, often overfocusing on irrelevant input features. In particular, decoder only models are noticeably brittle to distributional shifts, while finetuned encoder and encoder decoder models can generalize more robustly across our tests, including the non natural language split. Both architectures are only matched or surpassed by decoder only architectures at large scales. We conclude by noting that for cost effective, short horizon robust causal reasoning, encoder or encoder decoder architectures with targeted finetuning are preferable.
Open AccessCausal Reasoning Favors Encoders: On The Limits of Decoder-Only Models
In context learning (ICL) underpins recent advances in large language models (LLMs), although its role and performance in causal reasoning remains unclear. Causal reasoning demands multihop composition and strict conjunctive control, and reliance on spurious lexical relations of the input could provide misleading results. We hypothesize that, due to their ability to project the input into a latent space, encoder and encoder decoder architectures are better suited for said multihop conjunctive reasoning versus decoder only models. To do this, we compare fine-tuned versions of all the aforementioned architectures with zero and few shot ICL in both natural language and non natural language scenarios. We find that ICL alone is insufficient for reliable causal reasoning, often overfocusing on irrelevant input features. In particular, decoder only models are noticeably brittle to distributional shifts, while finetuned encoder and encoder decoder models can generalize more robustly across our tests, including the non natural language split. Both architectures are only matched or surpassed by decoder only architectures at large scales. We conclude by noting that for cost effective, short horizon robust causal reasoning, encoder or encoder decoder architectures with targeted finetuning are preferable.