Jehyun Park

JEHYUN PARK

Currently a Research Assistant in PI-Lab at Seoul National University. I am working on Optimizing Test-Time-Compute and Reinforcement Learning Algorithms for Physical AI and AI4Science.

I’ve conducted research in few-shot learning and model-compression, focusing on solving Out-Of-Distribution problems and improving performance in low-paramter situations.
I’ve done my B.S in Computer Science at Sungkyunkwan University.

My research aims to develop foundation models for better generalization through optimization. I am particularly interested in Offline-RL, Transductive-Learning, Physical Simulations.

Always feel free to reach out to me with things you find exciting. Always open for new ideas :)

Links: jaheon555@g.skku.edu | GitHub | LinkedIn | X | Google Scholar

Announcements

Jan, 2026 Received Excellence Award in Undergraduate Paper Competition at KSC 2025 as a Sole Author for the paper Adaptation in Out-Of-Distribution via Sparse Autoencoders in Vision Transformers.
Dec, 2025 Joined PI-Lab at Seoul National University as a Research Intern, focusing on optimizing Transformer architecture and solving Reasoning tasks in Vision-Language-Action Models.
Sep, 2025 1st Place in the 2025 Samsung AI Challenge. Competed individually and received 1st Prize Award for Optimizing Large Language Model. Awarded 7,500$
Feb, 2025 2nd Place in the 20th Korean Economic Securities Derivatives Competition. Received Korea Exchange Chairman's Award for ETF design. Awarded 4,500$
Jan, 2025 Received Engineering Innovation Award from SungKyunKwan University for LLM-based Recommendation System project with iKooB Healthcare.

Selected Papers

Learning to Transfer Mechanics from Simulation to 3D Generation via Image and Text

Learning to Transfer Mechanics from Simulation to 3D Generation via Image and Text

Jehyun Park

Under review

ActionDiT: Action Dynamics Pretraining for Data-Efficient Robot Adaptation

ActionDiT: Action Dynamics Pretraining for Data-Efficient Robot Adaptation

Jehyun Park

Under review

MUART: Mobile-Use Agent Red-Teaming Framework

MUART: Mobile-Use Agent Red-Teaming Framework

Jehyun Park

Under review

Self-Improving CAD Generation Agents with Finite Element Analysis as Feedback

Self-Improving CAD Generation Agents with Finite Element Analysis as Feedback

Guijin Son†, Jehyun Park†, Seyeon Park, Sunghee Ahn, Youngjae Yu

Under review

Adaptation in Out-Of-Distribution via Sparse Autoencoders in Vision Transformers

Adaptation in Out-Of-Distribution via Sparse Autoencoders in Vision Transformers

Je Hyun Park*

KSC 2025