skip to content
Site header image Jinsung Lee’s Personal Homepage
Email copied to clipboard

Hi, I’m Jinsung

I am currently a Ph.D candidate at POSTECH CVLab (Advisor: Prof. Suha Kwak), where I am having fun doing research!

I am interested in designing a compressive neural networks that extract the most informative features within a fixed space of limited complexity. I believe that state-space models (SSMs) offer valuable properties for achieving this goal, and I am actively exploring their application in my work. Previously, I worked on video understanding tasks, such as human action recognition and long-term context modeling.




Updates

Sep 9, 2025
Planning in 16 Tokens: A Compact Discrete Tokenizer for Latent World Model
A paper on compact tokenizer for world model has been accepted to the Conference on Robot Learning Workshop (CoRLW) 2025.

Aug 17, 2025
From HiPPO to Mamba: A Beginner’s Guide to State-Space Models

A new blog post that shares my understanding of State-Space Models (SSMs) has been released!


Publications

Featured image for Planning in 16 Tokens: A Compact Discrete Tokenizer for Latent World Model

Planning in 16 Tokens: A Compact Discrete Tokenizer for Latent World Model

Dongwon Kim, Jinsung Lee, Gawon Seo, Minsu Cho, Suha Kwak

The Conference on Robot Learning Workshop (CoRLW), 2025 - Learning to Simulate Robot Worlds (LSRW)

Featured image for Classification Matters: Improving Video Action Detection with Class-Specific Attention

Classification Matters: Improving Video Action Detection with Class-Specific Attention

Jinsung Lee, Taeoh Kim, Inwoong Lee, Minho Shim, Dongyoon Wee, Minsu Cho, Suha Kwak

European Conference on Computer Vision (ECCV), 2024 (oral)

Featured image for Detector-Free Weakly Supervised Group Activity Recognition

Detector-Free Weakly Supervised Group Activity Recognition

Dongkeun Kim, Jinsung Lee, Minsu Cho, Suha Kwak

IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) 2022