Ponnurangam Kumaraguru

A new method to improve language model behavior by reducing bias and toxic output.

An analysis of bias and incivility in Indian television debates.

A new method to improve machine learning models affected by poor data.

A new benchmark aims to measure and mitigate AI-related dangers.

Examining code-mixing and its impact on language acceptability in multilingual settings.

HOGE enhances explanations of Graph Neural Networks using cell complexes.

Examining the effects of leaks on smartphone companies and trustworthiness of sources.

Reducing language model size without losing intelligence.

Thinking Tokens fail to improve AI reasoning compared to Chain-of-Thought.

Understanding how language models tackle proportional analogies.

Learn how to improve Graph Neural Networks by correcting harmful data.