Hyunjin Kim 김현진
I'm a first-year PhD student in CS at University of Maryland, College Park, where I'm advised by Prof. Dinesh Manocha.

I received my MS in CS from UC San Diego, advised by Prof. Xiaolong Wang. Before that, I earned my BS from Yonsei University, during which I worked with Prof. Jaesik Park.

Email:
Google Scholar GitHub LinkedIn X
Portrait of Hyunjin Kim in Times Square

News
Oct. 2026 I was selected as a Top Reviewer for NeurIPS 2026.
Aug. 2026 I joined UMD CS as a PhD student!
June 2026 I joined Nokia Bell Labs as a research intern this summer!
Sep 2025 One paper accepted to NeurIPS 2025.

Research
My research interests lie at the intersection of robotics and computer vision, where I aim to build computational systems capable of human-like scene understanding. These systems perceive complex environments and construct world models that capture both visual appearance and physical dynamics. My current research focus is on physics-grounded world models and their applications to robotics.

EgoPhys: Physics-Informed Digital Twins for Deformable Objects from Egocentric Video
Hyunjin Kim, Ri-Zhao Qiu, Rocktim Jyoti Das, Guangqi Jiang, Dinesh Manocha, Xiaolong Wang
arXiv preprint, 2026
Paper | Project Page | bibtex 
@article{kim2026egophys,
  title={EgoPhys: Learning Generalizable Physics Models of Deformable Objects from Egocentric Video},
  author={Kim, Hyunjin and Qiu, Ri-Zhao and Jiang, Guangqi and Wang, Xiaolong},
  journal={arXiv preprint arXiv:2606.16202},
  year={2026}
}
Metropolis-Hastings Sampling for 3D Gaussian Reconstruction
Hyunjin Kim, Haebeom Jung, Jaesik Park
Conference on Neural Information Processing Systems (NeurIPS), 2025
Paper | Project Page | Code | bibtex 
@inproceedings{
  kim2025metropolishastings,
  title={Metropolis-Hastings Sampling for 3D Gaussian Reconstruction},
  author={Hyunjin Kim and Haebeom Jung and Jaesik Park},
  booktitle={The Thirty-ninth Annual Conference on Neural Information Processing Systems},
  year={2025}
}
EgoPhys: Physics-Informed Digital Twins for Deformable Objects from Egocentric Video
Hyunjin Kim, Ri-Zhao Qiu, Rocktim Jyoti Das, Guangqi Jiang, Dinesh Manocha, Xiaolong Wang
arXiv preprint, 2026
Paper | Project Page | bibtex 
@article{kim2026egophys,
  title={EgoPhys: Learning Generalizable Physics Models of Deformable Objects from Egocentric Video},
  author={Kim, Hyunjin and Qiu, Ri-Zhao and Jiang, Guangqi and Wang, Xiaolong},
  journal={arXiv preprint arXiv:2606.16202},
  year={2026}
}
Metropolis-Hastings Sampling for 3D Gaussian Reconstruction
Hyunjin Kim, Haebeom Jung, Jaesik Park
Conference on Neural Information Processing Systems (NeurIPS), 2025
Paper | Project Page | Code | bibtex 
@inproceedings{
  kim2025metropolishastings,
  title={Metropolis-Hastings Sampling for 3D Gaussian Reconstruction},
  author={Hyunjin Kim and Haebeom Jung and Jaesik Park},
  booktitle={The Thirty-ninth Annual Conference on Neural Information Processing Systems},
  year={2025}
}
Visual Acoustic Fields
Yuelei Li, Hyunjin Kim, Fangneng Zhan, Ri-Zhao Qiu, Mazeyu Ji, Xiaojun Shan, Xueyan Zou, Paul Liang, Hanspeter Pfister, Xiaolong Wang
CVPR 3rd Workshop on Generative Models for Computer Vision, 2025
Paper | Project Page | Dataset | bibtex 
@article{li2025visual,
  title={Visual acoustic fields},
  author={Li, Yuelei and Kim, Hyunjin and Zhan, Fangneng and Qiu, Ri-Zhao and Ji, Mazeyu and Shan, Xiaojun and Zou, Xueyan and Liang, Paul and Pfister, Hanspeter and Wang, Xiaolong},
  journal={arXiv preprint arXiv:2503.24270},
  year={2025}
}
Chain-of-Thought separators paper preview
Can Separators Improve Chain-of-Thought Prompting?
Yoonjeong Park*, Hyunjin Kim*, Chanyeol Choi, Junseong Kim, Jy-yong Sohn
IEEE International Conference on Foundation and Large Language Models, 2024
(Oral Presentation)
Paper | Code | bibtex 
@inproceedings{park2024can,
  title={Can separators improve chain-of-thought prompting?},
  author={Park, Yoonjeong and Kim, Hyunjin and Choi, Chanyeol and Kim, Junseong and Sohn, Jy-Yong},
  booktitle={2024 2nd International Conference on Foundation and Large Language Models (FLLM)},
  pages={493--500},
  year={2024},
  organization={IEEE}
}

Academic Activities
Reviewer Conference Reviewer
CoRL 2026, NeurIPS 2026

Journal Reviewer
IEEE Transactions on Image Processing (TIP)


Template based on this amazing website.