Shubham Kumar | Personal Website
Welcome to my website! I am a fourth-year PhD student in the Computer Vision and Robotics Laboratory at the University of Illinois at Urbana-Champaign, advised by Prof. Narendra Ahuja.
I want to understand how and why AI models (for any modality) fail. I believe the key to getting there is by making sense of a model's intermediate representations. I just spent a summer at IBM, under the mentorship of Saurabh Jha, where I worked on world models for decision-making and planning.
I obtained my B.S. from UCSD, where I did research with Prof. Truong Nguyen and Prof. Pamela Cosman.
Recent News
Aug 2026
See our pre-print on belief-based world models for LLM decision-making under partial observability.
July 2026
Our paper, LOCA, has been accepted to COLM 2026! LOCA explains jailbreak attack success in LLMs.
June 2026
Attended CVPR 2026 in Denver! Read my reflection on the experience.
May 2026
Joined IBM as an extern for the summer.
Feb 2026
Our paper, SURF, was accepted to CVPR 2026!
June 2023
Graduated Summa Cum Laude from UCSD.
Apr 2023
Recieved the NSF Graduate Research Fellowship.
Mar 2023
Commited to UIUC for a MS/PhD program. Recieved the Illinois Distinguished Fellowship, ECE Distinguished Research Fellowship, and ECE Promise of Excellence Fellowship.
