|
Jinuk Kim (Justin Jinuk Kim)
Hello!
I am a 4th year PhD student in the Department of Computer Science and Engineering at Seoul National University, Machine Learning Lab, advised by Hyun Oh Song. I recently interned as an Applied Scientist at Amazon on the Annapurna Science team (2026), and previously interned at Google (2024) and Samsung Advanced Institute of Technology (SAIT) (2023). My PhD research is supported by the Apple Scholars in AIML PhD Fellowship and the Presidential Science Scholarship for Graduate Students.
I work on resource-aware LLM development, spanning pretraining, post-training, and agentic systems, with an emphasis on building capable models and tools under practical compute, memory, and deployment constraints. Previously, I worked on mixed precision scaling laws for quantized LLM pretraining and LLM agents for semiconductor design rule checking, as well as KV-cache compaction, post-training LLM quantization, neural network depth compression, and dataset condensation.
Refer to my Research Statement for more details.
CV   / 
Scholar   / 
Github   / 
Twitter
  / 
Blog
  / 
LinkedIn
|
|
|
Highlights
|
- Mixed Precision Scaling Laws proposed a scaling law for mixed precision quantized pretraining of LLMs that identifies the compute-optimal training recipe, including per-layer forward and backward precision.
- Rule2DRC introduced a benchmark for domain-specific DRC script coding agents in chip design and SplitTester, an execution-guided test generation agent for program selection.
- KVzip introduced a query-agnostic KV cache compaction method that keeps the KV pairs needed to reconstruct the original context.
- GuidedQuant proposed an improved objective function and quantization method for large language models.
|
|
Mixed Precision Scaling Laws for Quantized Pretraining
Jinuk Kim, Rajarshi Saha, Muhammed Fatih Balin, Youngsuk Park
Preprint
Paper |
Blog post |
Bibtex
|
|
Rule2DRC: Benchmarking LLM Agents for DRC Script Synthesis with Execution-Guided Test Generation
Jinuk Kim, Junsoo Byun, Donghwi Hwang, Seong-Jin Park, Hyun Oh Song
ICML 2026
Paper |
Code |
Blog post |
Poster |
Bibtex
|
|
KVzip: Query-Agnostic KV Cache Compression with Context Reconstruction
Jang-Hyun Kim, Jinuk Kim, Sangwoo Kwon, Jae W. Lee, Sangdoo Yun, Hyun Oh Song
NeurIPS 2025 Oral Presentation (77/21575=0.35%)
ICML 2025 ES-FoMo-III Workshop
Paper |
Code |
Blog post |
Bibtex
|
|
GuidedQuant: Large Language Model Quantization via Exploiting End Loss Guidance
Jinuk Kim, Marwa El Halabi, Wonpyo Park, Clemens JS Schaefer, Deokjae Lee, Yeonhong Park, Jae W. Lee, Hyun Oh Song
ICML 2025
Paper |
Code |
Blog post |
Poster |
Bibtex
|
|
LayerMerge: Neural Network Depth Compression through Layer Pruning and Merging
Jinuk Kim, Marwa El Halabi, Mingi Ji, Hyun Oh Song
ICML 2024
Paper |
Code |
Blog post |
Poster |
Bibtex
|
|
Efficient Latency-Aware CNN Depth Compression via Two-Stage Dynamic Programming
Jinuk Kim*, Yeonwoo Jeong*, Deokjae Lee, Hyun Oh Song
ICML 2023
Paper |
Code |
Blog post |
Bibtex
|
|
Dataset Condensation via Efficient Synthetic-Data Parameterization
Jang-Hyun Kim, Jinuk Kim, Seong Joon Oh, Sangdoo Yun, Hwanjun Song, Joonhyun
Jeong, Jung-Woo Ha, Hyun Oh Song
ICML 2022
Paper |
Code |
Bibtex
|
|
|
I own a SO100 robot, which I sometimes play with and teach to do things.
|
|
Talk-to-President
is a service that allows you to chat with AI personas of the presidential candidates in the voice of the candidates.
|
|
VibeCite
uses your Claude Code to search and generate bibtex entries from your high-level paper descritions.
|
|
SNU Board
is an Android/iOS service which collects notices from website of SNU departments and gather them
(Android / Aug 2021 / 100+ MAU / 80+ WAU / 1000+ Downloads).
|
|