Hi, I'm Mansi Phute

CS PhD at Georgia Tech
My research focuses on the security and explainability of multimodal foundation models. I work on developing explanations for ML systems, analyzing them to identify vulnerabilities, and finding solutions to mitigate these issues. My work spans a wide range of application areas, including multi-object robust tracking in computer vision, developing defenses against attacks on large language models, and understanding large language models and the insights they can give us into human interactions. I have also worked with 3D simulation software like CARLA and Unreal Engine.
I am currently a PhD student at Georgia Tech advised by Polo Chau as a part of the Polo Club of Data Science.
I have collaborated with designers, developers, and scientists at Intel Labs, Nanyang Technological University, and Dassault Systems.

Featured Publications

A Large-Scale Dataset for Testing Robustness of Image Classifiers
NeurIPS, 2024
Interactive Visual Attribution for LLM Generation
ACL demo, 2024
By Self Examination, LLMs Know They Are Being Tricked!
ICLR Tiny Paper, 2024
Architectural Design Principles for Adversarially Robust CNNs
BMVC, 2023