Junyoung Sung

M.S. Student @KAIST AI, AGI-LAB

jys7451@korea.ac.kr

About

Hello! I’m Junyoung Sung. I am currently an M.S. student at AGI-LAB, KAIST, advised by Prof. Jinwoo Shin. My research focuses on robotic foundation models that can generalize effectively to diverse and dynamic real-world situations. Recently, I have been particularly interested in data-centric approaches to robot learning, including synthetic data generation and leveraging large-scale human video data to address the scarcity of robot interaction data.

News

Publications

Most recent publications on Google Scholar.
‡ indicates equal contribution.

GOAT: A Training Framework for Goal-Oriented Agent with Tools

Hyunji Min, Sangwon Jung, Junyoung Sung, Dosung Lee, Leekyeung Han, Paul Hongsuck Seo

Annual Meeting of the Association for Computational Linguistics (ACL) 2026 (Findings)

CRIT: Graph-Based Automatic Data Synthesis to Enhance Cross-Modal Multi-Hop Reasoning

Junyoung Sung, Seungwoo Lyu, Minjun Kim, Sumin An, Arsha Nagrani, Paul Hongsuck Seo

IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2026

LCIRC: A Recurrent Compression Approach for Efficient Long-form Context and Query Dependent Modeling in LLMs

Sumin An, Junyoung Sung, Wonpyo Park, Chanjun Park, Paul Hongsuck Seo

Annual Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics (NAACL) 2025, Oral Presentation

RLDX-1 Technical Report

RLDX Team (Core contributor)

GOAT: A Training Framework for Goal-Oriented Agent with Tools

Hyunji Min, Sangwon Jung, Junyoung Sung, Dosung Lee, Leekyeung Han, Paul Hongsuck Seo

Annual Meeting of the Association for Computational Linguistics (ACL) 2026 (Findings)

CRIT: Graph-Based Automatic Data Synthesis to Enhance Cross-Modal Multi-Hop Reasoning

Junyoung Sung, Seungwoo Lyu, Minjun Kim, Sumin An, Arsha Nagrani, Paul Hongsuck Seo

IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2026

Breaking the Visual Shortcuts in Multimodal Knowledge-Based Visual Question Answering

Dosung Lee, Sangwon Jung, Boyoung Kim, Minyoung Kim, Sungyeon Kim, Junyoung Sung, Paul Hongsuck Seo

Preprint

LCIRC: A Recurrent Compression Approach for Efficient Long-form Context and Query Dependent Modeling in LLMs

Sumin An, Junyoung Sung, Wonpyo Park, Chanjun Park, Paul Hongsuck Seo

Annual Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics (NAACL) 2025, Oral Presentation

Education

KAIST Sep. 2026 – Present
M.S. in Artificial Intelligence
Algorithmic General Intelligence Laboratory (AGI-LAB) (Advisor: Jinwoo Shin)

Korea University Mar. 2020 – Aug. 2026
B.S. in Computing Science and Engineering
GPA: 4.41/4.5

Experience

Algorithmic General Intelligence Laboratory (AGI-LAB) Feb. 2026 – Aug. 2026
Research Intern (Advisor: Jinwoo Shin)

Multimodal Interactive Intelligence Laboratory (MIIL) Jan. 2024 – Jan. 2026
Research Intern (Advisor: Paul Hongsuck Seo)