Research Interests

Adversarial Machine Learning

Investigating vulnerabilities in LLMs and Reinforcement Learning agents against intentional perturbations.

Trustworthy & Safe AI

Building alignment protocols and safety guardrails to ensure AI systems behave reliably in edge cases.

NLP Robustness

Analyzing the linguistic limits of foundational models and improving their generalization in medical and legal domains.

Publications

Equity-Aware Geospatial AI for Forecasting Demand-Driven Hospital Locations in Germany
arXiv Pre-print | 2nd best project award at DS course at Saarland University2025

Equity-Aware Geospatial AI for Forecasting Demand-Driven Hospital Locations in Germany

Piyush Pant, M.W. Suntoro, A. Siddiqua, M.S. Sharif, D. Ahmed

PDF arXiv Code
GREAT: Generalizable Backdoor Attacks in RLHF via Emotion-Aware Trigger Synthesis
arXiv Pre-print2025

GREAT: Generalizable Backdoor Attacks in RLHF via Emotion-Aware Trigger Synthesis

S.K. Dutta, Yuelin Xu, Piyush Pant, Xiao Zhang

PDF arXiv Code
Improving LLM Safety and Helpfulness using SFT and DPO: A Study on OPT-350M
arXiv Pre-print2025

Improving LLM Safety and Helpfulness using SFT and DPO: A Study on OPT-350M

Piyush Pant

PDF arXiv Code