Kyuhwan Shim
Research Scientist at LG AI Research
I am a Research Scientist in the Physical Intelligence Lab at LG AI Research. I received my M.S. in Artificial Intelligence from the Graduate School of Artificial Intelligence at Seoul National University, where I was advised by Prof. Byoung-Tak Zhang.
Previously, I was a visiting student researcher in the
Autonomous Intelligent Systems Group
at University of Bonn and a
machine learning engineer at
. I received my B.S. in Computer Science
from
Sogang University and co-founded
OCONNECT.
Hover, focus, or tap a circle.
Generalizable Physical Intelligence
I study robot learning systems that connect multimodal perception, policy learning, and data-centric engineering for robust real-world manipulation.
Vision-language-action models · imitation learning · task generalization · interpretable policies
Multimodal Perception & Grounding
I am interested in grounding manipulation policies with target-object and target-place masks, dual-view correspondence, and explicit 3D gripper–object relationships.
Visual prompting · DINOv2 / CLIP · Grounded-SAM2 · depth estimation · camera calibration
Policy Learning & Generalization
I study language-conditioned and goal-conditioned policies that generalize beyond fixed task distributions through visual plans, progress-aware execution, and reusable action primitives.
VLA · diffusion policies · flow matching · action chunking · zero-shot task composition
Robot Data & Learning Systems
I build multi-embodiment pipelines spanning synchronized robot-data collection, ETL and quality analysis, model training, and real-time inference with external robots and model servers.
ROS 2 / rosbridge · LeRobot · HDF5 / RLDS · trajectory retargeting · selective downsampling
Related work: PRISM AI Tools · ESPADA
Selected Publications
- RAIN: Robotic Task Generalization with Region-Aware Interaction Networks X-Reason Workshop @ ECCV 2026
-
Making Your Action Policies Interpretable: Mixture of Action Queries
Foundation Models Meet Embodied Agents Workshop @ CVPR 2026
AI4GOOD Workshop @ ICML 2026 - ESPADA: Execution Speedup via Semantics Aware Demonstration Data Downsampling for Imitation Learning IEEE Robotics and Automation Letters (RA-L) 2026
-
CLIP-RT : Learning Language-Conditioned Robotic Policies from Natural Language Supervision
Robotics: Science and Systems (RSS) 2025 (Oral Presentation)
3rd Workshop on Language and Robot Learning (LangRob) @ CoRL 2024
Education
Experience
Research Experience
Teaching Experience
Seoul National University
Work Experience
Yeongdong Fire StationAssisted in developing ONESTEP, an easy-insert socket.
Awards
@ Team TIDYBOY
RoboCup@Home
- 🥉 3rd Place in DSPL League
- 🥈 2nd Place in OPL League
@ SNU GSDS & Google ExploreCSR
SNU GSDS · Google Research
Samsung · DACONCamera Invariant Domain Adaptation for Semantic Segmentation
Smart Factory Product Quality Status Classification
Gangseo-gu OfficeRisk Rating Model for Small Business Owners using the Credit Rating Model
Sogang People
BIG CONTESTJeju Island Travel Route Recommender System for Eco-Friendly & Gen-MZ @ BIGCONTEST Innovation League
Recent projects
Camera Invariant Domain Adaptation
Risk Rating Model for Small Business Owners (Winner)
Jeju Island Travel Route Recommender System for Eco-Friendly & Gen-MZ (Winner)
LG Aimers 2 : Data Intelligence
- Sep. 2026 RAIN Accepted to X-Reason Workshop at ECCV 2026 Our paper RAIN: Robotic Task Generalization with Region-Aware Interaction Networks was accepted to the X-Reason Workshop at ECCV 2026.
- Aug. 2026 Research Scientist @ LG AI Research I joined the Physical Intelligence Lab at LG AI Research as a Research Scientist, working on embodied intelligence and real-world agents.
- May 2026 MoAQ Accepted to CVPR & ICML 2026 Workshops 'Making Your Action Policies Interpretable: Mixture of Action Queries' was accepted to the Foundation Models Meet Embodied Agents Workshop @ CVPR 2026 and the AI4GOOD Workshop @ ICML 2026.
- Dec 2025 ESPADA Accepted to IEEE RA-L 'ESPADA: Execution Speedup via Semantics Aware Demonstration Data Downsampling for Imitation Learning' was accepted to IEEE Robotics and Automation Letters (RA-L). Preprint available on arXiv.
- Nov 2025 Research Scientist Intern @ LG AI Research Joining LG AI Research as a Research Scientist Intern in the Physical Intelligence Lab.
- Apr 2025 CLIP-RT Accepted to RSS 2025 (Oral Presentation) Our paper CLIP-RT was accepted to Robotics: Science and Systems (RSS) 2025 as an oral presentation. The paper introduces a novel vision-language-action framework that enables robotic policies to be trained using natural language supervision.
Contact Information
Email: kyuhwan.shim@snu.ac.kr
Location: Bldg. #303, Gwanak-ro, Gwanak-gu, Seoul, Republic of Korea (08826)