Donghu Kim

logo_uni_tue          logo_uni_tue

Hi there, nice to meet you!

My name is Donghu Kim. Currently I am a research scientist at Holiday Robotics, where we are developing dexterous manipulation skills with RL in simulation.

I received my Master's Degree in KAIST (advised by Jaegul Choo). My study is based on the belief that RL in simulated environments is inevitable, let it be low-level data generation [1, 2] or building a library of atomic skillsets [3]. In this direction, I am invested in pushing the absolute limit of efficiency in RL for control: Can we make RL work with only 1K samples? Can we do it within an hour? As far fetched as the goal may seem, there are so many exciting components we can tackle, including feature learning, exploration, architecture design, optimizer.

I still have a long long way to go; if you want to discuss anything research related, I'd be more than happy to be engaged!

Email  /  CV  /  Google Scholar  /  LinkedIn  /  Github

profile photo

Publications

[ Selected / All ]
flashsac

FlashSAC: Fast and Stable Off-Policy RL for High-Dimensional Robot Control
Donghu Kim*, Youngdo Lee*, Minho Park, Kinam Kim, Takuma Seno, Aswin Nahrendra, Sehee Min, Daniel Palenicek, Florian Vogt, Danica Kragic, Jan Peters, Jaegul Choo, Hojoon Lee.
Preprint.
arXiv / project page / github (⭐100+)

fire

FIRE: Frobenius-Isometry Reinitialization for Balancing Stability-Plasticity Tradeoff
Isaac Han. Sangyeon Park, Seungwon Oh, Donghu Kim, Hojoon Lee, Kyungjoon Kim†.
ICLR'26, Oral.
arXiv / project page / github

dynamicmoe

Dynamic Mixture of Experts Against Severe Distribution Shifts
Donghu Kim
Preprint.
arXiv

toxchat

Building Resource-Constrained Language Agents: A Korean Case Study on Chemical Toxicity Information
Hojun Cho, Donghu Kim, Soyoung Yang, Chan Lee, Hunjoo Lee, Jaegul Choo.
EMNLP'25 (Industry Track).
arXiv

simbav2

SimBaV2: Hyperspherical Normalization for Scalable Deep Reinforcement Learning
Hojoon Lee*, Youngdo Lee*, Takuma Seno, Donghu Kim, Peter Stone, Jaegul Choo.
NeurIPS'25, Spotlight.
arXiv / project page / github

simba

SimBa: Simplicity Bias for Scaling Up Parameters in Deep Reinforcement Learning
Hojoon Lee*, Dongyoon Hwang*, Donghu Kim, Hyunseung Kim, Jun Jet Tai, Kaushik Subramanian, Peter R. Wurman, Jaegul Choo, Peter Stone, Takuma Seno.
ICLR'25, Spotlight.
arXiv / project page / github

preprint2024dodont

Do’s and Don’ts: Learning Desirable Skills with Instruction Videos
Hyunseung Kim, Byungkun Lee, Hojoon Lee, Dongyoon Hwang, Donghu Kim, Jaegul Choo
NeurIPS'24.
arXiv / project page

icml2024atari-pb

ATARI-PB: Investigating Pre-Training Objectives for Generalization in Pixel-Based RL
Donghu Kim*, Hojoon Lee*, Kyungmin Lee*, Dongyoon Hwang, Jaegul Choo.
ICML'24.
arXiv / project page / github / poster

icml2024hnt

Slow and Steady Wins the Race: Maintaining Plasticity with Hare and Tortoise Networks
Hojoon Lee, Hyeonseo Cho, Hyunseung Kim, Donghu Kim, Dugki Min, Jaegul Choo, Clare Lyle.
ICML'24.
arXiv / github


Work Experience

Holiday Robotics
Holiday Robotics Jan 2026 - Present
AI Research Scientist
  • Dexterous manipulation via reinforcement learning
Krafton
Krafton AI Jun 2025 - Sep 2025
Research Intern (Physical Intelligence Team)
  • Recreated company workplace in IsaacLab simulator
  • Developed a locomotion RL policy with Unitree G1 for navigation

Study Materials

Note: These slides are made for studying purposes only, and likely have got something wrong here and there. If you happen to find some, feel free to make fun of me via e-mail :).

Honors & Awards

Academic Services

Template based on Jon Barron's website.