Skip to content

Aditya (Jack) Sarkar - ジャック

Office: 8125 Paint Branch Dr, College Park, MD 20742

Work Email: asarkar6 [at] umd [dot] edu

Personal Email: asnov2k [at] gmail [dot] com


I am a second-year Ph.D. student at University of Maryland, advised by Prof. Ang Li. Further, I closely collaborate with Shlok Mishra, Profs. Nuno Vasconcelos and David Jacobs. I research computer vision models, specifically video-centric multimodal LLMs and diffusion models.

In my free time, I document my readings on Jacklog — a running log of what I’m learning. Feel free to check it out here.

Before my Ph.D., I obtained my bachelor’s degree in Electrical Engineering from Indian Institute of Technology (IIT) Mandi, where I was advised by Prof. Sreelakshmi Manjunath and received the President of India Gold Medal and Institute Silver Medal.

News

[01/2026] MA-PaPSP got accepted at ICLR '26. See you in Rio de Janeiro, Brazil.

[09/2024] 🎉 Excited to start my Ph.D. at University of Maryland advised by Prof. Ang Li.

[09/2023] 🎉 Graduated from IIT Mandi with a B.Tech. in Electrical Engineering.

Publications & Preprints

* denotes equal contribution

Quantifying symbol binding capability of Audio-Visual LLMs

Aditya Sarkar, Shwai He, David Jacobs, Ang Li.

Preprint   arXiv   Project Page

Keeping Them HONEST: A Framework for Detailed Understanding of How Video Complexity Affects Video-LLM Performance

Aditya Sarkar, Yi Li, Jiacheng Cheng, David Jacobs, Nuno Vasconcelos.

Preprint   arXiv   Project Page

Leveraging Data to Say No: Memory Augmented Plug-and-Play Selective PredictionImage title

Aditya Sarkar, Yi Li, Jiacheng Cheng, Shlok Mishra, Nuno Vasconcelos.

ICLR 2026    arXiv   Project Page

Education

University of Maryland College ParkImage title

Aug. 2024 -- Present

Ph.D. in Computer Vision, advised by Prof. Ang Li.

Indian Institute of Technology MandiImage title

Sept. 2019 -- June 2023

B.Tech. in Electrical Engineering (President of India Gold Medal Awardee), advised by Prof. Sreelakshmi Manjunath

Academic Service

ICLR'26