Sanghyun Lee

Integrated MS-PhD Student · KAIST CVLAB

I am an Integrated MS-PhD Student at KAIST CVLAB, advised by Seungryong Kim. I am broadly interested in diffusion models, with a current focus on discrete diffusion models and unified multimodal models. On the inference side, I am particularly interested in test-time scaling, efficiently scaling compute at inference time. Separately, on the training side, I am drawn to modeling, designing architectures and formulations that make these capabilities possible.

I received my BS from KAIST in 2025.

Email Scholar LinkedIn CV

News

Oct 2026
ReAnchor, a training-free method that speeds up diffusion speculative decoding, is now on arXiv with code.
Sep 2026
Looped Diffusion Language Models has been accepted to NeurIPS 2026.
May 2026
Two papers accepted to ICML 2026: Lookahead Unmasking and Unifying Masked Diffusion Models (Spotlight).
Jan 2026
A Noise Is Worth Diffusion Guidance has been accepted to ICLR 2026.
Sep 2025
Started my Integrated M.S.-Ph.D. at KAIST CVLAB. Fine-Grained Perturbation Guidance has been accepted to NeurIPS 2025.

Publications

* Equal contribution

ReAnchor: Reusing Rejected Verifier States for Diffusion Drafting

Sanghyun Lee^*, Wonjin Nam^*, Seungho Jang, Jaeho Lee, Jiho Park, Jaewoon Lee, Seungryong Kim

arXiv, 2026

PDF arXiv Code

Looped Diffusion Language Models

Sanghyun Lee, Chunsan Hong, Seungryong Kim, Jonghyun Lee, Jongho Park, Dongmin Park

NeurIPS, 2026 COLM workshop, 2026 (Spotlight)

PDF arXiv

Lookahead Unmasking Elicits Accurate Decoding in Diffusion Language Models

Sanghyun Lee, Seungryong Kim, Jongho Park, Dongmin Park

ICML, 2026

PDF arXiv

Unifying Masked Diffusion Models with Various Generation Orders and Beyond

Chunsan Hong, Sanghyun Lee, Jong Chul Ye

ICML, 2026 (Spotlight)

PDF arXiv

Effective Test-Time Scaling of Discrete Diffusion through Iterative Refinement

Sanghyun Lee, Sunwoo Kim, Seungryong Kim, Jongho Park, Dongmin Park

ICML workshop, 2026

PDF arXiv

Understanding and Accelerating the Training of Masked Diffusion Language Models

Chunsan Hong^*, Sanghyun Lee^*, Chieh-Hsin Lai, Satoshi Hayakawa, Yuhta Takida, Yuki Mitsufuji, Seungryong Kim, Jong Chul Ye

ICML workshop, 2026

PDF arXiv

A Noise Is Worth Diffusion Guidance

Donghoon Ahn^*, Jiwon Kang^*, Sanghyun Lee, Jaewon Min, Minjae Kim, Wooseok Jang, Hyoungwon Cho, Sayak Paul, SeonHwa Kim, Eunju Cha, Kyong Hwan Jin, Seungryong Kim

ICLR, 2026

PDF arXiv

Fine-Grained Perturbation Guidance via Attention Head Selection

Donghoon Ahn^*, Jiwon Kang^*, Sanghyun Lee, Minjae Kim, Jaewon Min, Wooseok Jang, Sangwu Lee, Sayak Paul, Susung Hong, Seungryong Kim

NeurIPS, 2025

PDF arXiv

Experience

Jun 2025 - Apr 2026 · Seoul, Korea

Krafton

Research Intern, Foundation Research Team.

Education

Sep 2025 - Present

Korea Advanced Institute of Science and Technology (KAIST)

Integrated M.S.-Ph.D., Kim Jaechul Graduate School of AI. Advised by Seungryong Kim.
Mar 2019 - Sep 2025

Korea Advanced Institute of Science and Technology (KAIST)

B.S. in Electronic Engineering, Minor in Mathematical Sciences.