- Auditing Language Model Unlearning via Information Decomposition
- Differentially Private Steering for Large Language Model Alignment
- Socratic Reasoning Improves Positive Text Rewriting
These papers are under review or have been presented at conferences such as ICLR, TPDP, and CLPsych Workshop.
Research Experience
PhD student in the ELLIS project, focusing on the privacy and safety of large language models.
Education
ELLIS PhD student, supervised by Prof. Iryna Gurevych (UKP Lab, TU Darmstadt) and co-supervised by Prof. Amartya Sanyal (University of Copenhagen); previously spent two years at IIIT Hyderabad, India, with Prof. Ponnurangam Kumaraguru.
Background
Research interests: privacy and safety of large language models; recent work focuses on developing methods for protecting sensitive information when working with billion-scale parameter models.
Miscellany
Outside of research, enjoys cricket and traveling.