Jisu Han

I obtained my Bachelor Degree in Computer Science from Ewha Womans University, Master Degree in Graduate School of AI at KAIST, in Intelligent Mobile-Manipulation (IM^2) lab directed by Beomjoon Kim.

Research

	Hierarchical and Modular Network on Non-prehensile Manipulation in General Environments Yoonyoung Cho, Junhyek Han, Jisu Han, Beomjoon Kim Robotics: Science and Systems (RSS), 2025
	Adaptive visual abstraction via object token merging and pruning for efficient robot manipulation Jisu Han CVPR Workshop (Causal and Object-Centric Representations for Robotics) 2024 Oral [Paper][Github]
	Preference learning for guiding the tree search in continuous POMDPs Jiyong Ahn, Sanghyeon Son, Dongryung Lee, Jisu Han, Dongwon Son, and Beomjoon Kim. Conference on Robot Learnining (CoRL), 2023. [Paper][Video][Project][Github]

	Hybrid-Space Diffusion for Task and Motion Planning with Offline RL Jisu Han, Yoonyoung Cho, and Haewon Jung (AI707: Advanced Topics in Deep Learning (Prof. Kimin Lee) Class project) TL;DR To resolve the downward refinement issues from bi-level planning in Task and Motion planning (TAMP), we adopt hybrid-space diffusion model. [Project]
	Clarifying the task: Identifying task from human videos as a representation Jisu Han and Doohyun Lee (AI611: Machine Learning for Robotics (Prof. Joseph Lim) Class project) TL;DR To learn a generalized reward function that can be utilized on reinforcement learning, we devise a representation that can effectively disentangle environment information and task information. [Project]
	Compositional Meta Reinforcement Learning Yoonyoung Cho and Jisu Han (AI614: Advanced deep learning (Prof. Sungju Hwang) Class project) TL;DR We inject compositional reasoning for robots to quickly learn novel tasks by leveraging reusable elements from previous experiments. [Paper]