I am a Research Officer at the Agency for Defense Development (ADD), working on LLM serving in heterogeneous clusters and efficient vision inference.

My current research is advised by Prof. Olivia Hsu (CMU) and Prof. Jongse Park (KAIST). Previously, I worked on memory bug detection and fuzzing at UNIST S2Lab under Prof. Yuseok Jeon (Korea University).

Publications

Tortoise and Hare Guidance project visual

NeurIPS 2025 acceptance rate: 24.52%

Tortoise and Hare Guidance: Accelerating Diffusion Model Inference with Multirate Integration

Yunghee Lee, Byeonghyun Pak, Junwha Hong, Hoseong Kim

A training-free strategy for accelerating diffusion sampling by treating classifier-free guidance as a multirate system.

IISWC 2026 acceptance rate: 40.16%

Understanding and Exploiting Diagonal Attention Sparsity in Autoregressive Image Generation

Daeun Kim, Junwha Hong, Changhun Oh, Yoonsung Kim, Yoonhyeong Lee, Jongse Park

IISWC 2026 Short Paper

How Far Can PIMs Replace GPUs in Energy-Efficient Multi-Turn Agent Serving?

Junwha Hong, Guseul Heo, Hyunsu Ye, Jongse Park, Olivia Hsu

Experience

Education

Ulsan National Institute of Science and Technology (UNIST)

Mar 2020 - Feb 2024

B.S. in Computer Science and Engineering

GPA: 4.13/4.3, summa cum laude (ranked 4th out of 296 students).

Activities: 11th President of HeXA, Hacker's eXciting Academy.

Honors

  • Top Prize (Minister of Science and ICT Award), 2022 Millitech Challenge, Republic of Korea.
  • Excellence Prize, 2021 UNIST-POSTECH-KAIST Data Science Competition.
  • National Science & Technology Scholarship, 2022.