Labs & researchers
Explore the Korean research community by topic, researcher, and institution, alongside global labs.
Follow the people behind the papers
Labs & Korean Academia5 resources
Anthropic Research
Interpretability, alignment, and frontier-model research
Google DeepMind Research
Publications across LLMs, science AI, and RL
LG AI Research
Korea's corporate frontier lab — EXAONE papers and open releases
KAIST Kim Jaechul Graduate School of AI
Korea's leading AI graduate school
SNU AI Institute (AIIS)
Seoul National University's interdisciplinary AI research institute
Korean AI research map
Source review ·
Select a topic or researcher to explore connections. Lines indicate shared interests, not co-authorship. The directory is a dated selection; use the linked lab pages for current affiliations and rosters.
Who works on what
Topics in the center, 23 researchers around them, 425 lab members on the outer ring. A lab's slice scales with its size.
Hover or tap a researcher to see their topics, who shares them, and their lab members; tap a topic to see everyone in that area. Connections mean shared research topics, not necessarily co-authorship; outer-ring dots are current grad students from each lab's own page.
How these topics connect to the AI waves
Does academia fit the Next Waves?
Each researcher's work mapped onto the five waves of AI value creation — our editorial framework from Next Waves.
The center of gravity is unmistakable: 19 of 23 researchers here work on Wave 2 — frontier models, their evaluation, and their economics. That is the wave Korea already missed commercially and where Chinese open weights are commoditizing value. But look closer and much of this W2 work is really about making models cheap — distillation (DistiLLM-2), KV-cache compression (KVzip), adaptive recursion depth — which is exactly the research a sovereign-AI program running on constrained compute needs.
Wave 3, agentic labor — the wave industry is in right now — has thinner but real coverage (6 researchers): web agents that generalize out of domain, search-reasoning agents, taxonomies for steering how reasoning models think, RLHF and alignment. Wave 4 has genuine seeds (5): simulated patients and hospital workflows winning best-paper awards, fusion-plasma forecasting, brain imaging. And Wave 5 — Korea's structural opening — is where the academic bench lines up best with the industrial thesis (4 labs): humanoid loco-manipulation VLAs, language-driven dexterous grasping, 3D affordance grounding, with roughly 130 grad students in those labs' rosters.
The honest gaps: no one in this AI-faculty set works on Wave 1 — the compute-and-memory wave Korea actually won lives in EE departments and industry fabs, not AI graduate schools. And the W3 gap matters most: industry's current wave is agents shipped as products, and Korea has no academic cluster the size of its model-efficiency cluster working on agentic systems. The verdict: academia is not "off" so much as one wave behind the industrial front — deep where models get cheap, positioned well for embodiment, and thin exactly where the money is moving this year.
Institutions & labs
KAIST — Kim Jaechul Graduate School of AI
Korea's largest dedicated AI graduate school. Faculty labs span LLMs and generative models, ML theory, NLP, and healthcare AI — and seed a notable share of Korea's AI startups.
University
KAIST — Kim Jaechul Graduate School of AI
Korea's largest dedicated AI graduate school. Faculty labs span LLMs and generative models, ML theory, NLP, and healthcare AI — and seed a notable share of Korea's AI startups.
The largest single-institution contributor in the NAIRL network's ICML 2026 cohort — 13 papers from 7 PIs — with KAIST labs publishing across NeurIPS, ICML, and ICLR every cycle.
Sung Ju Hwang
MLAI Lab
Efficient and long-context generative models, LLMs, self-evolving AI
Jinwoo Shin
ALIN Lab
Algorithmic foundations of ML, robustness, self-supervised learning
Alice Oh
NLP·CSS group
Multilingual and multicultural LLM evaluation, AI ethics and safety
Minjoon Seo
LK Lab
Retrieval and knowledge in LLMs; co-founder/CEO of Config Intelligence
Edward Choi
Representation learning on electronic health records, medical AI
Se-Young Yun
OSI Lab
Distributed and federated learning, LLM efficiency and reasoning
Jaegul Choo
DAVIAN Lab
Generative models, visual computing, human-AI interaction
Kimin Lee
Reinforcement learning, RLHF and alignment for generative models
Juho Lee
SIML Lab
Bayesian deep learning, probabilistic and generative modeling
Steven Euijong Whang
Data Intelligence Lab
Data-centric AI, responsible and robust machine learning
Seoul National University — AI Institute (AIIS)
SNU's interdisciplinary AI institute, connecting labs across engineering, natural sciences, and medicine — deepest in computer vision, deep learning systems, and biointelligence.
University
Seoul National University — AI Institute (AIIS)
SNU's interdisciplinary AI institute, connecting labs across engineering, natural sciences, and medicine — deepest in computer vision, deep learning systems, and biointelligence.
SNU vision and ML labs place papers at NeurIPS, ICML, and CVPR every cycle; the Vision & Learning Lab alone reported NeurIPS 2025 and ICML 2025 acceptances.
Kyoung Mu Lee
Computer Vision Lab
Computer vision, image restoration, 3D understanding
Sungroh Yoon
Data Science & AI Lab
Deep learning, on-device AI, bioinformatics
Byoung-Tak Zhang
Biointelligence Lab
Embodied and cognitive AI, human-level learning
Gunhee Kim
Vision & Learning Lab
Multimodal learning, vision-language models
Hyun Oh Song
Machine Learning Lab
Efficient deep learning, discrete optimization, reinforcement learning
Songhwai Oh
Robot Learning Lab
Robot learning, motion planning, safe autonomy
Nojun Kwak
MIPAL
Computer vision, object detection, efficient on-device models
NAIRL — National AI Research Lab Network
A cross-institution network of AI labs spanning KAIST, Yonsei, Korea University, and POSTECH, pooling frontier research across the member universities.
Research Network
NAIRL — National AI Research Lab Network
A cross-institution network of AI labs spanning KAIST, Yonsei, Korea University, and POSTECH, pooling frontier research across the member universities.
Announced 31 accepted papers at ICML 2026 across the network — the clearest single number for how Korean university AI research is showing up at top venues.
Youngjae Yu
Yonsei
Multimodal learning, commonsense reasoning, robot learning
Minsu Cho
POSTECH · Computer Vision Lab
Visual correspondence, object discovery, minimally-supervised learning
Seong-Whan Lee
Korea Univ
Pattern recognition, brain-computer interfaces; head of the AI department
Heuiseok Lim
Korea Univ · NLP & AI Lab
Korean NLP, large language models, dialogue systems
NAVER AI Lab (TEAM NAVER)
NAVER's research arm behind HyperCLOVA X — publishing on efficient large-scale AI, generative-AI safety, and robotics/physical AI alongside the product work. Former lab head Jung-Woo Ha became Korea's first senior presidential secretary for AI in 2025.
Corporate Lab
NAVER AI Lab (TEAM NAVER)
NAVER's research arm behind HyperCLOVA X — publishing on efficient large-scale AI, generative-AI safety, and robotics/physical AI alongside the product work. Former lab head Jung-Woo Ha became Korea's first senior presidential secretary for AI in 2025.
10 papers accepted at NeurIPS 2025 — the strongest disclosed top-venue showing by a Korean corporate lab that year.
Sangdoo Yun
NAVER AI Lab
Vision and multimodal foundation models, data-centric training methods
LG AI Research
Korea's corporate frontier lab: the EXAONE model family (including open weights), AI for science and materials discovery, and one of the five MSIT sovereign-AI mandates. Former president Kyunghoon Bae became Korea's Minister of Science and ICT.
Corporate Lab
LG AI Research
Korea's corporate frontier lab: the EXAONE model family (including open weights), AI for science and materials discovery, and one of the five MSIT sovereign-AI mandates. Former president Kyunghoon Bae became Korea's Minister of Science and ICT.
Publishes EXAONE technical reports and top-venue papers alongside open-weight releases on Hugging Face.
Honglak Lee
LG AI Research
Deep representation and generative learning; chief scientist (also U. Michigan)