profile photo

Sihyun Yu

I am a first-year Ph.D. student at Algorithmic Intelligence Laboratory @ KAIST, advised by professor Jinwoo Shin.

Prior to this, I received B.S. in Computer Science and Mathematics (double major) at KAIST, and I did my research about physics-inspired distribution sampling methods under the supervision of professor Hongseok Yang.

Email  /  CV  /  Google Scholar  /  Twitter  /  Github


My research focuses on designing deep generative models of extremely high-dimensional, complicated structural inputs. Specifically, I am interested in structural assumptions of latent representations to effectively synthesize these data, e.g., propositional logical embeddings from images (ICMLW'21a), continuous relaxations of discrete proteins and molecules (NeurIPS'21), and implicit neural representations of videos. Recently, I have been actively focusing on bridging implicit neural representations to deep generative models, such as generative adversarial networks or diffusion models. I also have an interest in adversarial robustness (ICMLW'21b).

RoMA: Robust Model Adaptation for Offline Model-based Optimization
Sihyun Yu, Sungsoo Ahn, Le Song, Jinwoo Shin
Neural Information Processing Systems (NeurIPS), 2021
arXiv  /  slides  /  code
Abstract Reasoning via Logic-guided Generation
Sihyun Yu, Sangwoo Mo, Sungsoo Ahn, Jinwoo Shin
ICML Workshop on Self-Supervised Learning for Reasoning and Perception, 2021   (Oral presentation)
arXiv  /  slides  /  poster
Consistency Regularization for Adversarial Robustness
Jihoon Tack, Sihyun Yu, Jongheon Jeong, Minseon Kim, Sung Ju Hwang, Jinwoo Shin
ICML Workshop on Adversarial Machine Learning (AdvML), 2021   (Oral presentation)
arXiv  /  slides  /  code
Invited Talks

Abstract Reasoning via Logic-guided Generation
ICML Workshop on Self-Supervised Learning for Reasoning and Perception 2021 (Jul 2021; remote)

Work Experience

Collaborate with Dr. Jung-Woo Ha (Aug 2021 - present; Jeongja, Korea (remote))

Mohamed bin Zayed University of Artificial Intelligence (MBZUAI)
Collaborate with Prof. Le Song (May 2021 - Aug 2021; Abu Dhabi, UAE (remote))

Korea Advanced Institute of Science and Technology (KAIST)
Collaborate with Prof. Hongseok Yang (April 2018 - June 2019; Daejeon, Korea)

Selective Honors

Best Paper Awards, Korean Artificial Intelligence Association, 2021

Qualcomm-KAIST Innovation Awards, 2019

KAIST Presidental Fellowship, 2018-2020

National Presidental Scholarship for Science, 2017-2020

Hansung Scholarship for Gifted Students, 2015-2016


TA, Deep Learning (AI502), Fall 2020

TA, Samsung Electronics Device Solutions, AI-Expert Program, Summer 2020

Peer Tutor, System Programming (CS230), Spring 2019

TA, Linear Algebra (MAS212), Spring 2019

TA, Calculus 1 (MAS101), Spring 2018

The template is from this website.